You are on page 1of 13

Transportation Letters

The International Journal of Transportation Research

ISSN: 1942-7867 (Print) 1942-7875 (Online) Journal homepage: https://www.tandfonline.com/loi/ytrl20

The effect of soft variables on travel mode choices

Seungil Yum

To cite this article: Seungil Yum (2019): The effect of soft variables on travel mode choices,
Transportation Letters, DOI: 10.1080/19427867.2019.1687196

To link to this article: https://doi.org/10.1080/19427867.2019.1687196

Published online: 05 Nov 2019.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=ytrl20
TRANSPORTATION LETTERS
https://doi.org/10.1080/19427867.2019.1687196

The effect of soft variables on travel mode choices


Seungil Yum
Design, Construction, and Planning, University of Florida, Gainesville, FL, USA

ABSTRACT KEYWORDS
This paper aims to highlight how soft variables play an important role in travel mode choices in the New York Soft variables; racial/ethnic
metropolitan area. This study focuses on the effects of racial/ethnic affinities and travel purposes on travel affinities; travel mode
mode choices. This study finds that racial/ethnic affinities play a different role in travel mode choices choices; travel purposes;
according to racial/ethnic groups. For instance, racial/ethnic affinities play a negative role in carpooling of carpooling; New York;
multinomial logit model;
whites for leisure, unlike other racial/ethnic groups. This result shows that the general hypothesis is wrong, transportation planning
that is, living in racial/ethnic neighborhoods has a positive impact on carpooling. This study highlights that
travel mode choices are differentiated by travel purposes across racial/ethnic groups. For example, Hispanics
are positively related to the household carpool, whereas they show a negative probability of those for
school. The results show that urban planners should take soft variables into account for travel mode choices.

Introduction Therefore, this study explores the effect of soft variables on travel
mode choices, focusing on two variables, which have been barely
Reducing car use has been one of the important topics for urban
highlighted in the prior research: racial/ethnic affinities and travel
planners to pursue sustainable development and relieve traffic con-
purposes. First, racial/ethnic affinities can play a pivotal role in
gestion. For instance, according to the Census Bureau’s American
people’s travel behavior. For instance, racial/ethnic affinities in
Community Survey (see https://factfinder.census.gov/faces/tableser
neighborhoods can provide a better environment for carpooling
v i c es / j s f / p a g e s / p r o d u c t v i e w . x h t m l ? p i d = A C S _ 1 7 _ 5 Y R _
with a rapport based on the cultural, racial, linguistical bond.
S0802&prodType=table), solo driving to work accounts for 76.4% of
Blumenberg and Smart (2009) find a significant relationship
all commuters in the US in 2017. Scholars have focused on the travel
between the residential location in ethnic clusters and travel beha-
mode choices to encourage other alternative means of transportation
vior. They show that residence in an ethnic neighborhood is posi-
against car use. They have widely highlighted the effects of socio-
tively associated with carpools and public transit in urban areas.
demographic factors, such as gender and age, and built environment
With respect to the suburbs, residence in ethnic neighborhoods is
factors, such as connectivity and bus stops, on the travel mode
related to a greater propensity to carpool but it is negatively asso-
choices. These variables are referred to as ‘hard variables’
ciated with public transit. Smart (2015) reveals that living in an
(Sztompka 2000). However, even though ‘soft variables’, such as
immigrant neighborhood has a strong effect on travel mode choices
subjective and psychological factors, can play an important role in
for immigrant residents and a much weaker power for non-
travel mode choices as much as hard variables, they have been largely
immigrant residents in immigrant neighborhoods.
unexamined in the existing literature. For instance, Stern and
Second, travel purposes can be one of the important determi-
Richardson (2005) argue that most of the travel behavior models
nants for travel mode choices. The fact is easily understood by
lack cognitive explanatory variables explaining travel mode choices.
a question, that is, ‘why do people travel?’ The answer is that people
This is because soft variables are problematic to include the variables
travel because of purposes to do something at the destinations. In
into a mathematical model (Ross 1975).
fact, people travel more for other purposes than commuting to
Recent research has emphasized that soft variables should also
work in their daily lives. For instance, the frequency of trips for
be included in the travel behavior analysis (Angueira et al. 2019;
other purposes is about 58%, which is higher than one half, whereas
Chorus 2012; Smart and Klein 2013; Van Acker, Mokhtarian, and
that of commuting trips is about 42% in the study area of this
Witlox 2010; Van Wee, Holwerda, and Van Baren 2002). This is
article. In other words, trips for other purposes are also one of the
because travel mode choices would be influenced by subjective and
main drivers for understanding people’s travel behavior and gen-
psychological factors as well as objective and physical factors (Fujii
erating traffic flows.
and Kitamura 2003). According to existing literature, soft variables
In addition, while ridesharing, such as carpool, vanpool, and other
have an important impact on travel behavior (see e.g. Kamargianni
group rides, is emerging as a new travel mode against car use and
et al. 2015; Mosa and Esawey 2013; Spears, Houston, and Boarnet
plays an important role in racial/ethnic groups who earn a low
2013; Van Acker, Mokhtarian, and Witlox 2010; Vredin Johansson,
income, many prior studies simply categorize travel modes into
Heldt, and Johansson 2005). For instance, Van Acker, Mokhtarian,
solo driving, carpool, and transit (see e.g. Bhat 2000; Hess 2001; van
and Witlox (2010) show that subjective characteristics, such as
der Waerden, Lem, and Schaefer 2015; Wang and Chen 2012;
lifestyles and attitudes, play important decisive factors for model
Washbrook, Haider, and Jaccard 2006). In contrast, this study speci-
choices in leisure travel at various model levels. Vredin Johansson,
fically divides ridesharing into household carpool and inter-
Heldt, and Johansson (2005) exhibit that soft variables enriched
household carpool to shed more light on ridesharing modes and
discrete choice model outperforms the traditional discrete choice
provide deep analysis of travel behavior of individuals. This study
model.

CONTACT Seungil Yum yumseungil@ufl.edu Design, Construction, and Planning, University of Florida, Gainesville, FL, USA
© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 S. YUM

explores the relationship between four racial/ethnic groups (whites, Rey (1999) find that blacks show the most significant race/ethnicity-
blacks, Asians, Hispanics) and four travel modes: solo driving (SD), based difference for greater use of public transit even though other
household carpools (HCP), inter-household carpools (ICP), and socioeconomic variables of individuals are controlled. Giuliano
public transit (PT). (2003) demonstrates that racial/ethnic differences are clearly not
In this background, the main purposes of this article are three- limited to results caused by different location patterns and socio-
fold as follows: first, this study aims to provide empirical models economic variables, but rather by primary differences in what
explaining how racial/ethnic affinities in neighborhoods play an motivates travel behavior and mode choices by analyzing the 1995
important role in travel mode choices with controlling other hard US Nationwide Personal Transportation Survey data. Bento et al.
variables. To be specific, this study employs a race/ethnicity Index (2003) highlight that blacks are more apt to walk and take public
(REI) to measure the effects of racial/ethnic affinities in neighbor- transit to commute, whereas whites are less inclined to choose a bus
hoods for travel mode choices according to race/ethnicity (whites, in 114 urbanized areas in the US based on the data of 1990
blacks, Asians, Hispanics). Second, this study highlights the rela- Nationwide Personal Transportation Survey.
tionship between the four racial/ethnic groups and travel behavior Some authors have highlighted that travel purposes are highly
according to six travel purposes (work, school, shopping, leisure, related to travel behavior. For instance, McFadden (1978) argues
home, and others). Third, this study analyzes the relationship that travel behavior would differ by what people do at travel desti-
between the four racial/ethnic groups and four travel mode cate- nations, that is, travel purposes. Dieleman, Dijst, and Burghouwt
gories to suggest more implications for improving ridesharing, the (2002) find that the purpose of a trip, such as work, shopping, and
mobility of different racial/ethnic groups, and equity. To the best of leisure, is related to travel mode and distance by using the
my knowledge, this is the first article exploring the effects of soft Netherlands National Travel Survey. Scheepers et al. (2013) report
variables on travel behavior according to the multitude of racial/ that trip purposes are an important determinant for the use of
ethnic groups, travel purposes, and travel modes (four, six, and active (walking or cycling) transport versus passive (car) transport
four, respectively). For the purposes of this articles, this study uses modes based on a dataset of household members (n = about 53,000
the 2010/2011 Regional Household Travel Survey (RHTS) in the respondents annually) in Mobility Research Netherlands from 2004
New York/New Jersey/Connecticut metropolitan area and the US to 2009. For example, they show that taking or bringing persons and
Census Bureau. sports are more inclined to be made by active transport modes than
by car. Klöckner and Friedrichsmeier (2011) indicate that the pur-
pose of a trip is another significant predictor for travel mode
Literature review
choices. They find that both trips to work and leisure activities
While soft variables have not been fully highlighted as much as hard show a higher likelihood of car use by analyzing 26,865 individual
variables as they are tricky to be included into econometric models, trips of the Ruhr-University students in Bochum.
some authors have tried to highlight the effects of soft variables on However, while travel purposes are highly related to travel mode
travel mode choices (see e.g. Chorus 2012; Kamargianni et al. 2015; choices given that people travel based on the purpose, most studies
Spears, Houston, and Boarnet 2013). Recent empirical studies show have heavily focused on commuting to work (see e.g. Blumenberg
that soft variables are important determinants for travel behavior and Pierce 2014; Chatman 2003; Eluru, Chakour, and El-Geneidy
and mode choices (Chorus 2012). For example, Kamargianni et al. 2012; Hess 2001; Pinjari et al. 2007; Van Vugt, Van Lange, and
(2015) find that soft variables play a significant role in school mode Meertens 1996; Washbrook et al. 2006). Therefore, understanding
choices by employing multinomial probit-based integrated choice non-work travel is becoming increasingly important because of its
latent variable models in Cyprus in 2012. Spears, Houston, and growing impact on the lives of people and transportation infra-
Boarnet (2013) highlight that soft variables have a significant and structures (Polzin, Chu, and Rey 1999). Some studies explore the
consistent impact on the decision for using public transit in Los relationship between travel purposes and travel mode choices,
Angeles. whereas they only divide travel purposes into binary categories,
Previous studies also have indicated that racial/ethnic groups that is, commuting trips and non-work trips (see e.g. Hensher and
have different preferences for same-race neighbors. For instance, Rose 2007; Kroesen and Handy 2014; LaMondia and Bhat 2011).
Charles (2000) finds that whites prefer a higher percentage of same- Other studies design their research for a single purpose, such as
race neighbors on average and are the most likely to specify all- shopping, leisure, and school (see e.g. Boarnet and Sarmiento 1998;
same-race neighborhoods. In contrast, blacks are the least likely to LaMondia and Bhat 2011; Schwanen, Dijst, and Dieleman 2001). In
specify all-same-race neighborhoods in a multiracial sample of other words, there are few studies, which highlight the relationship
adults (N = 4,025) in Los Angeles. Emerson, Chai, and Yancey between travel mode choices and various travel purposes in empiri-
(2001) explain that while black neighborhood composition does cal models.
matter to whites, Asian and Hispanic neighborhood composition Lastly, many prior studies analyze travel mode choices among
do not matter based on a national, random-digit-dial survey of driving alone, carpooling, transit to highlight the travel patterns of
1,663 white Americans. These racial/ethnic affinities would affect individuals by employing a empirical model (see e.g. Bhat 2000; Hess
travel behavior because ethnic-based social networks can play an 2001; van der Waerden, Lem, and Schaefer 2015; Wang and Chen
important role in travel mode choices. For instance, Blumenberg 2012; Washbrook, Haider, and Jaccard 2006). For example,
and Smart (2009) show that individuals in ethnic neighborhoods Washbrook, Haider, and Jaccard (2006) analyze commuter mode
tend to commute by carpool and public transit more than other choice among driving alone, carpooling, and taking express bus
workers in Los Angeles based on the 2000 U.S. Census. Liu and based on the effect of road pricing and parking charges. Hess
Painter (2012) also find that people in ethnic neighborhoods are (2001) examines the probabilities of solo driving, carpool, and transit
more likely to choose carpooling and public transit. for work when they receive free parking in Oregon, Portland. Bhat
Also, it is a well-known fact that race/ethnicity plays a crucial (2000) highlights that individual characteristics for mode preferences
role in travel behavior (see e.g. Bento et al. 2003; Blumenberg 2009; and responsiveness among sold driving, carpool, and transit in San
Blumenberg and Smart 2009; Charles 2000; Emerson, Chai, and Francisco Bay area. While some studies highlight other travel mode
Yancey 2001; Giuliano 2003; Liu and Painter 2012; Polzin, Chu, and choices, such as inter-household carpool (see e.g. Blumenberg and
Rey 1999; Schlossberg et al. 2006). For instance, Polzin, Chu, and Smart 2010; Li et al. 2007; Neoh, Chipulu, and Marshall 2017), they
TRANSPORTATION LETTERS 3

are not related to the effect of soft variables on travel mode choices. where j = (1, . . ., n) refers to the census tract, Eij indicates the
Therefore, this study analyzes how soft variables play a pivotal role in number of the racial/ethnic group in the census tract, Ej means
various travel modes (SD, HCP, ICP, and PT). the total population in the census tract, Eie is the number of the
racial/ethnic group in the reference area; and Ee refers to the total
population in the reference area. The value of the REI index is
Research design calculated by the 2009 American Community Survey (ACS) 5 year
estimates and 2010/2011 RHTS. An REI greater than 1 indicates
This study highlights the effects of soft variables, as well as hard
that the neighborhood has a greater level of the ethnic group than
variables, on travel mode choices in the New York/New Jersey/
the case in the reference area. If an REI is equal to 1, then the
Connecticut metropolitan area, using the 2010/2011 Regional
neighborhood has the same level of the ethnic group as it does in
Household Travel Survey (RHTS). The RHTS was sponsored by the
the reference area. This study defines high racial/ethnic affinities
New York Metropolitan Transportation Council (NYMTC) and the
based on the REI value higher than 1.5 as authors recommend the
North Jersey Transportation Planning Authority (NJTPA). In total,
value to identify a concentration. For instance, Ellis, Wright, and
31,156 households were recruited to participate in the survey. Of
Parks (2006) use a value of 1.5 to distinguish clustered tracts and
these, 18,965 households completed travel diaries. All households
non-clustered tracts to highlight the immigrant household and
within the 28-counties constituting the New York/New Jersey/
spatial assimilation. Liu and Painter (2012) define Latino immigrant
Connecticut metropolitan area were eligible for inclusion in the
enclaves with cutoff 1.5 given that the value greater than 1 means
survey through a random sampling process (NYMTC and NJTPA
a greater level of Latino immigrant concentration.
2013). Figure 1 demonstrates the 28-counties in the study area.
Figure 2 shows that racial/ethnic neighborhoods are differen-
This study assumes that higher racial/ethnic affinities can be
tiated by racial/ethnic groups and regions. For instance, whites are
measured by higher values of racial/ethnic neighborhoods. Racial/
concentrated in the Northwest side of New York and the Northeast
ethnic groups tend to cluster together in neighborhoods based on
side of New Jersey, and blacks are clustered in the East side of
racial/ethnic affinities. This is because they provide a pool of tangi-
New York and the North side of New Jersey. Asians are heavily
ble and intangible assets that promote social, cultural, economic
located near the boundary of New York and New Jersey, and
devilment of its members. In other words, racial/ethnic affinities in
Hispanics are largely placed in the East side of New York and the
neighborhoods help people get valuable knowledge and resources,
Northeast side of New Jersey.
transportation networks, and social capital. For instance, they have
This article divides samples into four categories: whites, blacks,
better opportunities of getting information and networks for vehi-
Asians, and Hispanics older than 16. After ruling out cases, which
cles, carpooling, and public transportation systems. This study
have missing information, the final sample includes 108,718 trips.
employs an REI index to measure the racial/ethnic affinities in
This study categorizes travel purposes into six groups: work, school,
neighborhoods. The REI is calculated as follows:
shopping, leisure, home, and other. This study divides travel mode
choices as follows: SD, HCP, ICP, and PT.
Eij Eie
REI ¼ = (1) After analyzing travel patterns according to racial/ethnic groups
Ej Ee and travel purposes, racial/ethnic groups show different results for

Figure 1. Study area.


Source: https://slideplayer.com/slide/10537992/
4 S. YUM

Figure 2. The value of REI (whites, blacks, Asians, Hispanics).

travel mode choices (see Table 1). For example, for the purpose of For the purpose of school, while all racial/ethnic groups exhibit
work, whites, blacks, and Asians show similar patterns, whereas the highest share of HCP, there are some differences in detail.
Hispanics reveal a different pattern. For instance, the three racial/ Blacks, Asians, and Hispanics have a share of it higher than 50%
ethnic groups have the highest share of PT (64.1%, 41.2%, and (57.1%, 50.2%, and 54.3%, respectively), whereas whites show a low
41.0%, respectively), ahead of HCP (20.3%, 39.7%, and 36.1%, share of 35.8% in HCP. Also, whites and Asians show a second
respectively), whereas Hispanics demonstrate the highest share of highest preference for PT (35.1% and 25.7%, respectively), but
HCP (40.3%), followed by PT(38.4%). blacks and Hispanics demonstrate the second highest preference
TRANSPORTATION LETTERS 5

Table 1. The shares of travel modes according to ravel purposes.


Travel purpose
Work School
Whites Blacks Asians Hispanics Whites Blacks Asians Hispanics
SD 6.4% 7.2% 9.4% 7.4% 14.8% 7.6% 12.5% 9.4%
HCP 20.3% 39.7% 36.1% 40.3% 35.8% 57.1% 50.2% 54.3%
ICP 9.2% 11.9% 13.5% 13.9% 14.4% 21.3% 11.6% 21.3%
PT 64.1% 41.2% 41.0% 38.4% 35.1% 14.0% 25.7% 15.0%
Shopping Leisure
Whites Blacks Asians Hispanics Whites Blacks Asians Hispanics
SD 23.4% 15.5% 25.2% 19.0% 23.8% 12.8% 30.5% 14.6%
HCP 10.6% 25.3% 14.8% 26.4% 18.7% 35.8% 19.5% 37.7%
ICP 14.1% 30.0% 20.4% 34.4% 14.7% 27.9% 23.3% 28.6%
PT 51.9% 29.2% 39.6% 20.2% 42.9% 23.5% 26.7% 19.1%
Home Other
Whites Blacks Asians Hispanics Whites Blacks Asians Hispanics

SD 53.6% 6.4% 10.1% 9.8% 30.0% 16.6% 38.2% 20.9%


HCP 11.2% 30.4% 26.1% 23.0% 15.3% 38.7% 14.6% 35.0%
ICP 12.9% 25.6% 21.7% 32.8% 10.3% 18.8% 13.9% 22.6%
PT 22.2% 37.6% 42.0% 34.4% 44.4% 26.0% 33.3% 21.5%

for ICP (21.3%). This result shows that blacks and Hispanics are To sum up, this study finds that blacks and Hispanics are more
more likely to carpool for school given that they have the highest likely to carpool than Whites and Asians and travel mode choices
and second highest preference for HCP and ICP. are differentiated by travel purposes. This study highlights how soft
For the purpose of shopping, whites and Asians show similar variables play an important role in travel mode choices when hard
patterns, and blacks and Hispanics are analogous. The former variables are controlled by employing multinomial logit models in
groups have the highest proportion of PT (51.9% and 39.6%, the next section.
respectively), followed by SD (23.4% and 25.2%, respectively). In
contrast, the latter groups show the highest proportion of ICP
(30.0% and 34.4%, respectively), ahead of HCP (25.3% and 26.4%, Multinomial logit results
respectively), meaning that blacks and Hispanics also prefer car-
pooling for shopping. This study employs 24 multinomial logit models to explore how soft
For the purpose of leisure, blacks and Hispanics reveal simi- variables, as well as hard variables, play an important role in travel
lar preferences, and other racial/ethnic groups have unique mode choices according to racial/ethnic groups and travel pur-
patterns. Blacks and Hispanics exhibit the highest share of poses. The multinomial logit models provide a convenient closed
HCP (35.8% and 37.7%, respectively) and the second highest form for the underlying choice probabilities without any require-
share of ICP (27.9% and 28.6%, respectively). In contrast, ment of multivariate integration (Hausman and Mcfadden 1984).
whites have the highest share of PT (42.9%), followed by SD The multinomial logit models are one of the most commonly used
(23.8%), and Asians show the highest share of SD (30.5%), regression models for nominal outcomes in academic fields (see e.g.
ahead of PT (26.7%). The result shows that blacks and Adams, Wilson, and Wang 1997; Cheng and Long 2007; Fiebig et al.
Hispanics tend to carpool more than whites and Asians for 2010; McFadden 1978; Yaylali, Çelik, and Dilek 2016). The multi-
leisure consistent with the result of shopping. nomial logit models in this article are as follows:
For the purpose of home, whites have distinguishable character- Y = f (R, G, A, E, L, H, I, N, D)
istics from other racial/ethnic groups. For instance, whites exhibit
the highest proportion of SD (53.6%) more than one half and show where R = vector of the REI index, G = vector of gender, A = vector of
the second highest proportion of PT (22.2%). In contrast, the others age, E = vector of employment, L = vector of license, H = vector of
show the lowest proportion of SD (6.4%, 10.1%, and 9.8%, respec- household size, I = vector of income, N = vector of the number of
tively) and reveal the highest share of PT (37.6%, 42.0%, and 34.4%, vehicles, and D = vector of regional dummies. The dependent vari-
respectively). Also, blacks and Asians have the second highest able of this study is a four-category variable: SD, HCP, ICP, and PT.
proportion of HCP (30.4% and 26.1%, respectively), and After running 24 multinomial logit models, the travel mode
Hispanics demonstrate the second highest proportion of choices are differentiated by travel purposes (see Table 2–7). To
ICP (32.8%). be specific, for the purpose of work, each racial/ethnic group shows
For other purposes, the racial/ethnic groups show biggest dif- different characteristics for travel mode choices. whites are less
ferent patterns among purposes. First, whites show the highest likely to choose ICP (−0.273), blacks and Asians are more apt to
share of PT with a percentage of 44.4, followed by SD (30.0%), use it (0.218 and 0.523, respectively), and Hispanics do not show
and blacks have the highest share of HCP with 38.7%, ahead of PT a preference for it in high racial/ethnic affinities. This result shows
(26.0%). Asians reveal the highest share of SD (38.2%), followed by that racial/ethnic affinities play a different role in ICP according to
PT (33.3%), and Hispanics show the highest share of HCP (35.0%), racial/ethnic groups. Blacks are also less inclined to choose HCP
ahead of ICP (22.6%). The findings show that Hispanics tend to (−0.297). Whites and Hispanics tend not to use PT (−0.428 and
carpool, whereas whites and Asians are reluctant to do carpooling −0.493, respectively), whereas Asians prefer PT (0.322) in high
for other purposes. racial/ethnic affinities. White males are less prone to take HCP
6 S. YUM

Table 2. Travel mode choices for work.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***16.809 ***15.717 ***15.01 ***13.109 ***17.964 ***16.86 −3.941 −4.750 7.176 0.364 10.941 −19.242
REI > 1.5 0.067 ***-0.273 ***-0.428 **-0.297 **0.218 −0.119 −0.151 ***0.523 **0.322 −0.177 0.094 ***-0.493
Male ***-0.492 ***0.299 ***0.301 ***-0.435 0.111 −0.175 ***-0.483 −0.091 0.120 ***-0.789 0.162 0.103
Age (reference: 65+)
16–18 −0.380 −0.396 −0.077 ***4.376 −2.152 −2.081 0.000 0.000 0.000 −16.096 *-1.91 **-2.697
19–24 *-0.303 ***0.361 ***0.661 **1.865 **-0.797 −0.487 −0.724 **-1.26 1.325 −0.666 0.191 −0.787
25–34 −0.119 ***0.294 ***0.464 ***2.567 −0.297 −0.525 −0.457 −0.379 *1.74 −0.139 0.433 −0.042
35–54 −0.155 0.036 **0.256 *1.906 **-0.57 ***-0.759 *-1.072 −0.699 1.336 −0.479 0.206 −0.127
55–64 ***-0.423 −0.042 0.162 *1.744 **-0.541 ***-0.845 −0.685 **-0.984 0.988 −0.953 0.111 −0.240
Employment ***-1.023 ***-0.397 ***-0.496 1.244 −0.252 ***-0.944 14.476 14.875 −0.714 12.796 1.001 15.807
License ***-17.567 ***-17.884 −17.824 ***-17.287 ***-19.134 −18.975 −12.927 −13.478 −13.424 −13.195 −14.129 −13.823
Household size
(reference: 4+)
1 ***-5.061 ***-1.002 ***-1.192 ***-4.995 −0.211 0.029 ***-4.696 ***-1.53 ***-1.037 −14.757 ***-1.066 ***-1.459
2 ***-1.126 ***-0.197 ***-0.21 ***-0.895 ***-0.37 −0.247 ***-1.02 *-0.279 *-0.38 ***-0.877 0.203 0.018
3 ***-0.287 **-0.097 −0.093 **-0.465 −0.225 **-0.531 ***-0.584 ***-0.547 **-0.468 −0.453 −0.044 −0.316
Income (reference:
$100K+)
Below $30K 0.176 ***-1.086 ***-1.108 −0.258 ***-0.87 ***-0.651 ***-1.461 ***-1.223 ***-1.128 −0.091 −0.052 0.375
$30K – $74.9K ***-0.196 ***-0.808 ***-0.908 −0.074 *-0.217 −0.180 ***-0.6 ***-0.663 ***-0.535 0.040 ***-0.673 **-0.518
$75K – $99.9K *-0.121 ***-0.443 ***-0.504 0.100 ***-0.389 ***-0.88 ***-0.592 ***-0.438 ***-0.708 0.339 −0.257 0.071
Number of Household
Vehicles (reference:
3+)
1 ***2.622 ***6.619 ***7.499 **2.267 ***6.827 ***6.766 ***4.861 ***7.009 ***7.368 0.571 ***4.957 ***5.051
2 ***1.772 ***2.159 ***2.661 ***0.69 ***1.089 ***0.934 ***1.911 ***2.178 ***2.324 ***1.664 ***1.747 ***1.635
3 ***0.717 ***0.501 ***0.419 0.163 **0.348 −0.133 ***0.963 ***0.503 0.409 *0.457 **0.509 −0.478
Regional dummy
(reference:
Connecticut)
New York **-0.187 ***0.918 ***0.396 0.078 ***0.972 ***2.056 **2.03 ***3.146 ***3.728 ***-1.051 *0.838 15.838
New Jersey −0.048 ***0.415 ***-0.307 −0.352 −0.032 0.666 **1.83 ***2.589 **2.255 ***-1.525 −0.055 15.008
N 37,313 3664 2712 2252
−2 Log Likelihood 14,965 3675 2863 2580
Significance 0.000 0.000 0.000 0.000
Pseudo R2 0.402 0.535 0.475 0.544
Note: *** <0.01 **<0.05 *< 0.1

(−0.492) and are more likely to use ICP and PT (0.299 and 0.301, PT. Black and Hispanic workers are less apt to use ICP and PT,
respectively). Black and Hispanic males indicate a lower likelihood whereas Asians tend to use them. Whites and blacks who possess
of HCP (−0.435 and −0.789, respectively). Age groups are largely a license are more likely to choose SD over other means of transpor-
associated with blacks and barely related to Hispanics. For instance, tation. Whites are associated with household size more than other
younger blacks have a higher possibility of HCP, whereas Hispanics groups, and those who have a higher number of the household
aged between 19–64 do not show any preference for travel mode generally show a negative relationship with HCP, ICP, and PT.
choices. White workers are less apt to use carpooling and PT, and Whites and blacks earning a higher income tend to choose SD,
black workers are less inclined to take PT. Whites and blacks with whereas Asians and Hispanics are less affected by the income.
a driver’s license decrease the likelihood of carpooling. Generally Whites prefer to use other means of transportation, whereas Asians
speaking, all racial/ethnic groups with a bigger family have a higher reveal a higher likelihood of SD in New York and New Jersey.
probability of SD over other means of transportation, and those For the purpose of shopping, in high racial/ethnic affinities, only
who earn a higher income are more inclined to use SD than Hispanics show a positive coefficient value for HCP and ICP (0.882
carpooling and PT. The number of household vehicles is relatively and 0.590, respectively). This result shows that racial/ethnic affinities
related to whites, and all racial/ethnic groups with a smaller number play an important role in carpooling for Hispanics. Whites have
of vehicles tend to choose other means of transportation. Asians are a negative predisposition toward PT (−0.739), whereas Asians and
more likely to carpool and use PT in New York and New Jersey. Hispanics prefer to take PT (1.255 and 0.770, respectively) in high
For the purpose of school, blacks and Hispanics show a negative racial/ethnic affinities. All males regardless of their racial/ethnic groups
coefficient value for HCP (−0.925 and −3.087, respectively), whites are more likely to use PT, and younger whites, blacks, and Asians also
and Asians are irrelevant to travel mode choices in high racial/ethnic tend to take PT. Black workers are more inclined to choose SD. Whites
affinities. Hispanics are negatively associated with ICP (−2.613), and Hispanics who possess a license are more apt to choose SD rather
meaning that Hispanics living in high racial/ethnic affinities tend to than carpooling. A smaller number of families are less prone to ride
choose SD rather than carpooling. Younger whites prefer to use HCP ICP across racial/ethnic groups. Whites prefer to take HCP, and blacks
and are mostly reluctant to use ICP and PT, whereas younger blacks tend to use ICP when their income decreases. Whites and blacks with
are more apt to choose HCP, ICP, and PT. Asians and Hispanics do a smaller number of vehicles have a more positive predisposition
not show any preference for the travel modes according to age toward PT. Whites, blacks, and Asians have a higher possibility of
variables. White workers are less prone to choose HCP, ICP, and taking PT in New York.
TRANSPORTATION LETTERS 7

Table 3. Travel mode choices for school.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***-14.819 ***3.689 **2.309 ***-22.725 ***-34.325 ***-34.311 15.339 13.985 −10.296 −1.772 2.542 −0.943
REI > 1.5 0.012 −0.023 −0.228 *-0.925 −0.350 −0.230 −0.700 0.751 0.731 ***-3.087 **-2.613 −2.731
Male **-0.28 −0.137 ***0.474 −0.748 0.071 0.158 −0.883 **-1.913 −0.641 0.267 *1.591 1.272
Age* (reference: 65+)
16–18 ***18.837 0.675 −0.267 ***24.118 ***36.281 ***35.677 17.702 16.011 27.831 −1.120 −3.441 −2.996
19–24 ***17.102 *-0.728 **-1.013 ***22.606 ***36.065 ***34.991 2.392 10.958 22.188 −3.518 −2.817 −2.857
25–34 ***17.448 **-0.969 **-1.055 ***23.465 ***35.722 ***34.782 1.949 12.350 26.655 −2.718 −3.826 −3.870
35–54 ***17.458 −0.621 ***-1.402 ***23.543 ***35.616 ***34.379 14.483 8.139 20.256 −0.719 −1.577 −1.911
55–64 17.902 ***-1.556 ***-1.591 23.325 36.165 33.749
Employment ***-0.976 **-0.292 ***-0.499 −0.113 *-0.936 ***-1.558 1.475 **1.888 ***3.681 −0.782 ***-2.503 **-2.226
License ***-5.276 ***-5.623 ***-5.482 ***-2.191 ***-3.932 ***-3.869 −15.674 −12.857 −12.869 −12.039 −14.132 −13.744
Household size
(reference: 4+)
1 −19.780 ***-1.22 ***-2.302 −0.644 1.328 0.812 −0.241 −16.636 −19.185 −16.379 −12.835 −14.528
2 ***-1.347 ***-0.788 ***-1.429 **-1.845 *-1.166 **-1.403 −0.756 **-3.095 **-3.329 **-3.674 ***-4.136 ***-4.16
3 ***-0.623 ***-0.368 *-0.292 0.514 −0.493 −0.113 ***3.119 1.180 −0.642 0.820 1.522 0.580
Income (reference:
$100K+)
Below $30K **-0.525 ***-1.014 ***-1.007 −0.742 **-2.015 *-1.552 −9.927 −1.485 −1.595 −0.825 *2.641 2.119
$30K – $74.9K ***-0.442 ***-0.38 **-0.442 −0.748 **-1.343 −0.482 **2.429 0.724 0.887 0.397 0.700 1.169
$75K – $99.9K −0.238 −0.092 0.075 **-2.075 ***-1.913 *-1.821 0.287 −0.344 0.131 1.676 *2.874 2.695
Number of Household
Vehicles (reference:
3+)
1 2.913 ***22.632 24.526 1.176 ***5.736 ***4.917 16.460 32.835 41.869 12.466 24.187 27.380
2 ***2.262 ***2.439 ***3.887 ***2.372 ***3.98 ***3.229 **5.071 ***8.312 16.287 ***7.683 ***7.36 ***9.766
3 ***1.299 ***1.155 ***1.768 *1.273 ***1.803 0.737 ***5.967 ***5.851 10.530 *2.437 1.743 4.698
Regional dummy
(reference:
Connecticut)
New York **0.841 ***1.404 ***1.253 1.248 ***2.443 **2.175 ***-22.125 ***-17.438 ***-15.915 ***13.17 ***11.466 ***10.972
New Jersey ***1.013 ***0.968 **0.969 1.467 *1.51 ***2.281 ***-23.244 ***-19.477 −15.466 13.912 11.863 12.831
N 3381 571 280 400
−2 Log Likelihood 3807 692 221 416
Significance 0.000 0.000 0.000 0.000
Pseudo R2 0.570 0.586 0.825 0.713
Note: *** <0.01 **<0.05 *< 0.1. The age reference for Asians and Hispanics is 55–64 given no sample in 65 +

For the purpose of leisure, whites and Hispanics living in high racial/ less than two vehicles tend to use other means of transportation.
ethnic affinities show a negative preference for HCP (−0.168 and Whites living in New Jersey are more likely to use HCP.
−1.110, respectively), whites also reveal a negative coefficient value for For other purposes, only whites in high racial/ethnic affinities tend
ICP (−0.106). In contrast, blacks and Asians do not have any preference not to use HCP and ICP (−0.082 and −0.193, respectively) as well as
for the travel mode. While whites are negatively related to PT, other PT (−0.638). In contrast, Asians and Hispanics prefer to ride PT
racial/ethnic groups are unrelated to it. White males tend not to car- (0.474 and 0.573, respectively). White males are less likely to choose
pool, and black males demonstrate a negative coefficient value for HCP. other means of transportation, all males regardless of racial/ethnic
Younger whites and Asians are less likely to choose HCP. Hispanic groups are less inclined to use PT. Younger whites are more prone to
workers also show a negative probability of HCP. Whites, blacks, and use ICP and PT. Black and Hispanic workers are more apt to take PT.
Hispanics with a license are less apt to carpool. Whites are not inclined Whites who possess a license reveal a lower likelihood of other means
to carpool, and Asians have a lower probability of choosing ICP and PT of transportation, and Hispanics with the license are less inclined to
when they have a small number of the household. All racial/ethnic carpool. Blacks and Hispanics in a smaller number of families are less
groups are less related to the income, but they are more associated with likely to choose HCP, whites and Hispanics in that condition
the number of household vehicles. Whites, blacks, and Asians living in decrease the likelihood of PT. Whites prefer to use PT, and
New York tend to use PT. Hispanics tend to use HCP when they have a small number of
For the purpose of home, only whites show some significant household vehicles. Whites, blacks, and Hispanics indicate a higher
results, whereas other racial/ethnic groups are not significant, likelihood of ICP, Asians tend not to take PT, but Hispanics are more
which may be because of the small sample size. Whites living in inclined to ride PT in New York and New Jersey.
high racial/ethnic affinities are not associated with travel mode One of the most important findings in this research is that while
choices. White males are less inclined to choose ICP and PT. Liu and Painter (2012) argue that living in ethnic neighborhoods
Younger whites indicate a lower likelihood of HCP and PT, and increases both the likelihood of carpooling and of taking public
white workers also are less likely to choose them. Whites who have transit, this study finds that living in ethnic neighborhoods plays
a license are less prone to take carpool. A number of households are a different role in them across race/ethnicity and travel purposes.
not related to carpool, but a smaller number of households are This may be because their research is only designed for Latino
generally associated with the negative preference of PT. Income immigrants and commuting to work, whereas this study considers
variables do not display particular tendencies, and whites who have all racial/ethnic groups and travel purposes. This finding suggests
8 S. YUM

Table 4. Travel mode choices for shopping.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***16.481 ***15.4 ***12.668 15.929 15.471 −3.527 17.723 −1.837 12.687 ***18.903 ***16.65 ***15.111
REI > 1.5 0.036 −0.034 ***-0.739 0.010 −0.249 **-0.498 0.088 −0.111 ***1.255 ***0.882 *0.59 **0.77
Male −0.014 ***-0.352 **0.178 0.245 −0.090 ***0.662 0.381 −0.197 **0.846 −0.170 −0.283 ***0.951
Age (reference: 65+)
16–18 **0.689 ***1.679 0.644 1.429 1.980 0.091 19.644 1.308 0.484 ***9.915 ***11.944 12.605
19–24 0.105 ***1.235 ***0.928 ***1.848 **1.475 ***1.698 *1.844 ***3.821 ***3.96 −1.045 0.024 −0.772
25–34 ***0.642 ***0.91 ***1.404 0.350 0.455 ***2.173 0.371 **1.547 ***3.278 **-0.995 *-0.992 −0.904
35–54 −0.116 0.143 ***0.664 −0.312 0.037 **0.718 −0.225 **1.49 **1.411 **-1.123 −0.463 *-0.838
55–64 *-0.114 −0.040 ***0.392 −0.076 *0.425 ***0.914 0.480 0.927 **1.66 ***-1.198 −0.416 −0.725
Employment ***-0.205 0.005 0.056 ***-0.711 ***-0.679 ***-0.736 0.374 −0.281 −0.471 −0.029 *0.493 **0.705
License ***-17.008 ***-17.58 −17.552 −16.621 −18.137 −18.281 −17.243 −17.225 −18.909 ***-18.265 ***-18.397 −19.127
Household size (reference:
4+)
1 −18.347 ***-0.862 ***-1.019 −19.271 ***-1.146 **-0.671 −18.219 **-1.63 −0.990 −19.273 *-0.995 −0.607
2 ***-0.444 −0.168 −0.117 ***-0.905 ***-0.812 −0.321 ***-1.667 0.061 **-1.051 −0.200 *-0.751 −0.494
3 ***-0.373 ***-0.334 0.192 −0.049 −0.525 **0.718 ***-0.89 **-1.229 ***-1.998 0.238 −0.229 0.144
Income (reference: $100K
+)
Below $30K ***0.331 **0.265 −0.174 ***-1.219 **1.001 0.101 *0.885 −0.036 0.625 ***-1.16 ***1.116 ***1.41
$30K – $74.9K ***0.398 *0.149 ***-0.281 ***-0.917 ***1.172 −0.276 −0.082 **0.923 0.563 −0.291 0.546 0.560
$75K – $99.9K ***0.272 0.136 ***0.382 −0.385 **1.042 −0.317 *-0.692 −0.654 −0.629 **1.037 **-2.558 0.558
Number of Household
Vehicles (reference: 3+)
1 0.241 ***5.548 ***8.218 ***1.78 ***4.39 ***5.451 −0.201 18.987 19.893 −15.391 ***4.269 ***5.911
2 ***0.611 ***0.977 ***3.169 ***1.42 ***1.645 ***2.603 0.148 −0.061 1.163 **0.797 0.117 ***1.885
3 ***0.297 −0.009 ***1.115 0.303 0.206 *1.154 **-0.74 *-0.995 0.061 **0.93 0.413 0.737
Regional dummy
(reference: Connecticut)
New York −0.072 0.236 ***1.197 ***1.246 0.429 ***18.934 −0.850 ***17.129 1.810 −0.654 0.891 0.635
New Jersey −0.093 −0.026 0.262 0.630 −0.196 17.109 −0.278 16.471 0.601 −0.479 0.011 −0.678
N 12,340 1667 591.000 915
−2 Log Likelihood 11,490 2352 865 1135
Significance 0.000 0.000 0.000 0.000
Pseudo R2 0.496 0.656 0.675 0.705
Note: *** <0.01 **<0.05 *< 0.1

that scholars should explore the effect of enclaves on travel mode whereas they show a negative possibility of those for school. They
choices according to race/ethnicity and travel purposes. also have a negative preference for HCP for leisure.
The results mean that racial/ethnic affinities play the most sig-
nificant role in whites with a negative effect for carpooling, whereas
Conclusions they exert the least impact on Asians except for ICP in the purpose
of work. This result shows that the general assumption is wrong,
This study explores how soft variables play an important role in
that is, living in racial/ethnic neighborhoods has a positive impact
travel mode choices in the New York/New Jersey/Connecticut
on carpooling because of their racial/ethnic affinities. This study
metropolitan area, using the 2010/2011 Regional Household
also finds that hard variables, such as the age, job, license, and so on,
Travel Survey (RHTS) and the US Census Bureau. In order to
play a different role in travel behavior according to travel purposes.
highlight the relationship between soft variables and travel mode
One of the interesting findings of this research is that racial/ethnic
choices, the study analyzes the patterns of 108,718 trips by employ-
affinities play a negative role in carpooling for whites, unlike other
ing 24 multinomial logit models.
racial/ethnic groups. Future research should highlight how and why
This article finds some important implications for transporta-
they are differently associated with whites and minority groups.
tion literature. First, the results of this study highlight that racial/
ethnic affinities play a different role in travel mode choices accord-
ing to racial/ethnic groups and travel purposes. For instance, in
Discussion
high racial/ethnic affinities, whites especially show negative prefer-
ences for carpooling. They are less likely to use ICP for work as well This study provides some specific recommendations for decision
as HCP and ICP for leisure and other purposes, meaning that and policy makers. First, given that the findings of this article are
whites prefer to drive alone more than other racial/ethnic groups. only based on the New York/New Jersey/Connecticut metropolitan
Also, they are the only group showing a negative coefficient value area, they should further explore how soft variables play an impor-
for carpooling in other purposes among racial/ethnic groups. tant role in travel modes in other metropolitan areas. The more
Blacks in high racial/ethnic affinities are less apt to choose HCP findings of soft variables would provide a solid foundation for
and more likely to use ICP for work, and they are less inclined to developing policies. For instance, decision and policy makers in
take HCP for school. Asians tend to use ICP for work and are not other metropolitan areas, such as Caltrans (California Department
related to travel mode choices in other five purposes. Also, Asians of Transportation) and ARC (Atlanta Regional Commission),
are the least affected group according to racial/ethnic affinities. should explore the effects of soft variables on travel mode choices
Hispanics are positively related to HCP and ICP for shopping, and compare their results with the findings of this article.
TRANSPORTATION LETTERS 9

Table 5. Travel mode choices for leisure.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***17.293 ***16.007 ***13.836 ***17.577 ***17.107 ***16.142 16.459 −4.540 −17.755 ***18.424 ***15.823 ***16.034
REI > 1.5 ***-0.168 *-0.106 ***-0.337 −0.100 −0.275 *-0.376 −0.241 −0.157 −0.222 ***-1.11 0.392 0.461
Male ***-0.158 ***-0.336 0.072 ***-0.633 −0.260 −0.310 0.011 0.148 0.268 0.017 0.205 −0.200
Age (reference: 65+)
16–18 ***-0.401 ***0.547 0.126 0.177 −1.092 0.023 ***-2.105 −0.487 −0.770 −0.596 0.064 −0.585
19–24 ***-0.681 ***0.679 ***0.73 −0.121 0.745 **1.051 ***-2.54 *-2.071 ***-3.203 −18.637 1.096 0.743
25–34 ***-0.488 −0.073 ***0.536 −0.080 0.201 0.283 *-0.806 ***-1.516 ***-1.551 −0.416 0.730 0.672
35–54 −0.070 0.084 ***0.363 −0.214 0.118 0.460 **-1.185 −0.513 **-1.081 0.497 0.420 0.011
55–64 ***-0.227 −0.073 0.100 −0.269 0.176 0.222 −0.038 −0.564 −0.055 **-1.203 −0.029 −0.413
Employment −0.066 −0.051 0.077 −0.001 −0.029 −0.044 0.270 ***0.967 0.440 *-0.616 0.263 −0.376
License ***-17.257 ***-17.081 −17.189 ***-17.937 ***-19.133 −18.630 −15.368 −16.233 −16.381 ***-17.258 ***-18.712 −18.684
Household size
(reference: 4+)
1 ***-5.895 ***-0.988 ***-1.564 −19.529 −0.344 −0.270 −18.436 ***-2.507 ***-3.798 −20.027 ***-2.25 ***-2.361
2 ***-0.771 ***-0.256 ***-0.478 **-0.61 **-0.537 −0.404 ***-1.274 **-1.015 ***-1.746 0.547 −0.179 −0.136
3 ***-0.345 ***-0.322 −0.091 −0.378 0.242 −0.007 −0.485 ***-1.3 ***-1.661 −0.513 ***-1.22 **-1.026
Income (reference:
$100K+)
Below $30K 0.058 0.021 ***-0.313 −0.572 ***1.247 *0.619 −0.825 0.377 0.145 −0.178 0.662 0.412
$30K – $74.9K −0.004 −0.067 ***-0.378 −0.382 0.477 0.081 0.366 0.559 −0.073 −0.057 **0.871 *0.755
$75K – $99.9K ***0.231 −0.096 −0.141 0.271 0.400 0.085 −0.275 **0.901 0.320 −0.018 0.423 0.605
Number of Household
Vehicles (reference:
3+)
1 ***2.28 ***5.437 ***7.632 **2.824 ***5.771 ***6.2 *1.85 ***5.975 21.749 **3.285 ***6.933 ***6.725
2 ***0.624 ***0.904 ***2.781 ***1.052 ***1.212 ***1.097 ***1.224 ***2.896 18.115 ***1.523 ***2.816 ***2.925
3 ***0.427 0.029 ***0.969 0.103 0.391 0.243 0.034 0.605 16.214 *0.752 **1.535 0.387
Regional dummy
(reference:
Connecticut)
New York ***-0.204 ***0.313 ***0.956 0.340 0.493 *1.113 0.407 ***19.083 ***18.768 −1.005 0.201 0.714
New Jersey 0.054 0.134 ***0.572 0.480 −0.115 0.484 0.220 18.887 17.685 *-1.694 −0.489 −0.175
N 16235 1450 671 876
−2 Log Likelihood 15656 2121 1077 1251
Significance 0.000 0.000 0.000 0.000
Pseudo R2 0.4 0.562 0.579 0.604
Note: *** <0.01 **<0.05 *< 0.1

Second, while most local governments have their own transpor- females for school, and Hispanic females for other purposes(males:
tation institute, develop data sets, and publish transportation -0.352, -1.913, and -0.541, respectively).
reports, they do not seriously take soft variables into account for Fourth, this study finds that racial/ethnic affinities play a different
their analysis. For instance, when this article investigates their role in travel mode choices according to race/ethnicity. Therefore,
transportation reports (21 reports in the Transportation Security decision and policy makers should first research the characteristics of
Data Center in the US), They heavily focus on hard variables, such neighborhoods for community planning to encourage sustainable
as gender, age, or income, to report travel patterns of people in the development. For instance, they should design ICP programs for
region. Decision and policy makers should strongly argue the commuting to work in black and Asian communities (REI>1.5)
importance of soft variables given that prior transportation litera- since the two groups show a positive preference for ICP in the
ture as well as this study find that soft variables play an important purpose of work (0.218 and 0.523, respectively).
role in travel behavior of individuals. Fifth, while prior studies simply divide explanatory variables,
Third, this study provides specific implications for understanding such as racial/ethnic groups (whites, blacks, and Asians), travel
the travel mode choices by dividing various explanatory categories. modes (solo driving, carpool, and transit), travel purposes (com-
Especially, the findings of this article play a crucial role in improving muting and non-commuting), this study delicately designs inde-
the mobility of different racial/ethnic, social groups, and equity. For pendent variables, such as racial/ethnic groups (whites, blacks,
instance, the multinomial logit models in this article show the relation- Asians, and Hispanics living in racial/ethnic affinities and those
ship between low-income minority groups and travel mode choices. living outside racial/ethnic affinities), travel modes (solo driving,
To be specific, low-income black families (below $30k) are more apt to HCP, ICP, and transit), and travel purposes (work, school, shop-
use ICP for leisure, and low-income Hispanic families are more like to ping, leisure, home, and others). The findings of this article allow
choose ICP for school (2.641) and shopping (1.116). In contrast, low- decision and policy makers to understand complex situations of the
income Asian families do not show any preference for ICP. Also, this race/ethnicity, travel modes, and travel purposes to encourage
study provides some important implications for different social groups, ridesharing and public transit for sustainable development against
such as females, elderly, or unemployed people by presenting 24 multi- car use. For instance, this study could highlight that low-income
nomial logit models. Urban planners and practitioners could develop Hispanic families are less likely to use HCP (−1.160) and are more
transportation policy for different social groups and communities to apt to choose ICP (1.116) for shopping by employing a multitude of
encourage other means of transportation against car use. For instance, racial/ethnic groups, travel modes, and travel purposes, whereas
they could design ICP programs for white females for shopping, Asian prior studies, which use a carpool category and binary travel
10 S. YUM

Table 6. Travel mode choices for home.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***15.181 ***17.021 ***17.489
REI > 1.5 0.099 −0.057 0.414
Male −0.282 **-0.502 ***-0.84
Age (reference: 65+)
16–18 0.461 0.378 *-2.168
19–24 **-1.681 −0.623 −0.921
25–34 *-0.928 −0.800 −0.394
35–54 −0.492 −0.363 **-0.775
55–64 **-0.885 −0.639 ***-0.893
Employment **-0.704 0.044 **-0.572
License ***-17.08 ***-18.928 −18.312
Household size (reference: 4+)
1 −18.235 −0.412 **-1.002
2 −0.358 −0.209 0.195
3 −0.469 0.556 **-0.853
Income (reference: $100K+)
Below $30K 0.050 −0.177 0.253
$30K – $74.9K −0.240 **0.681 *0.478
$75K – $99.9K ***-1.519 −0.708 *-0.703
Number of Household Vehicles (reference: 3+)
1 ***3.244 ***4.093 ***5.305
2 ***1.134 ***1.416 ***1.752
3 0.283 0.103 0.439
Regional dummy (reference: Connecticut)
New York 1.358 0.725 0.215
New Jersey *1.917 −0.766 −0.066
N 811
−2 Log Likelihood 1262
Significance 0.000
Pseudo R2   0.485
Note: *** <0.01 **<0.05 *< 0.1. The results of the three groups are not reported since they are insignificant.

Table 7. Travel mode choices for other purposes.


Whites Blacks Asians Hispanics
Reference (SD) HCP ICP PT HCP ICP PT HCP ICP PT HCP ICP PT
Intercept ***5.364 ***4.038 ***3.086 15.983 15.349 12.605 16.966 1.188 14.901 ***17.991 ***15.81 ***13.893
REI > 1.5 *-0.082 ***-0.193 ***-0.638 0.160 0.040 0.003 0.023 0.274 *0.474 −0.126 0.152 **0.573
Male ***-0.158 ***-0.195 ***-0.163 −0.039 −0.013 *-0.283 *-0.315 0.291 ***-1.131 −0.146 ***-0.541 ***-0.672
Age (reference: 65+)
16–18 *0.342 ***0.658 ***0.789 0.105 **1.274 3.283 −0.218 −1.487 1.166 −0.515 −1.083 −1.133
19–24 **-0.297 ***0.461 0.011 0.261 ***1.216 **1.085 −17.387 0.669 **2.787 0.140 0.332 −0.194
25–34 0.076 ***0.569 ***0.921 −0.361 ***1.077 ***2.095 *-0.758 *-1.05 −0.181 0.330 0.245 0.567
35–54 −0.082 ***0.251 ***0.572 −0.065 *0.4 ***0.827 **-0.813 −0.321 −0.231 0.152 −0.176 −0.405
55–64 ***-0.322 *0.128 ***0.317 ***-0.588 **0.464 ***0.942 ***-1.434 −0.412 −0.301 −0.347 −0.008 −0.574
Employment −0.051 −0.036 0.074 −0.008 0.169 **0.436 0.062 0.058 0.150 −0.081 0.211 ***0.686
License ***-5.393 ***-6.019 ***-6.258 −16.928 −18.230 −17.993 −15.519 −17.421 −16.939 ***-17.636 ***-19.031 −18.925
Household size (reference: 4+)
1 ***-6.55 ***-0.463 ***-1.429 ***-4.275 **0.591 −0.332 −16.612 0.105 *-0.854 −19.394 −0.162 ***-1.086
2 ***-0.934 0.055 ***-0.883 ***-0.553 0.250 −0.336 **-0.634 −0.512 ***-1.082 ***-1.416 0.034 ***-1.635
3 ***-0.42 ***-0.241 ***-0.534 **-0.485 **0.448 **0.567 **-0.373 −0.476 ***-1.113 ***-0.603 −0.013 *-0.501
Income (reference: $100K+)
Below $30K **0.199 0.056 ***-0.335 *-0.415 0.403 ***1.003 0.441 0.456 0.658 0.344 ***1.162 ***1.4
$30K – $74.9K **0.097 **0.125 0.013 0.150 *0.393 **0.736 0.176 0.084 −0.128 −0.015 *0.588 0.617
$75K – $99.9K **0.115 ***0.199 **0.172 *0.37 0.113 0.115 0.068 **-0.989 **-0.817 −0.168 −0.410 0.722
Number of Household
Vehicles (reference: 3+)
1 0.586 ***4.446 ***5.881 ***2.963 ***5.342 ***7.09 0.539 18.393 21.989 **2.069 ***4.601 ***6.721
2 ***0.792 ***0.572 ***2.291 ***0.802 0.368 **1.232 *0.495 ***1.003 ***4.218 ***0.831 0.593 ***2.335
3 ***0.318 −0.060 ***0.532 *0.352 **-0.541 0.764 0.218 −0.471 ***2.474 *0.547 0.105 0.875
Regional dummy (reference:
Connecticut)
New York 0.068 ***0.531 ***0.82 0.255 **0.921 **1.104 −0.564 ***14.971 **-1.399 −0.398 ***1.928 ***1.846
New Jersey 0.049 **0.312 0.166 **0.722 **0.773 0.464 −0.481 15.232 ***-2.598 −0.490 ***1.595 **1.296
N 17219 2513 1164 1447
−2 Log Likelihood 14034 3004 1398 1875
Significance 0.000 0.000 0.000 0.000
Pseudo R2 0.317 0.601 0.486 0.608
Note: *** <0.01 **<0.05 *< 0.1
TRANSPORTATION LETTERS 11

purposes (commuting vs non-commuting), cannot highlight the Hausman, J., and D. Mcfadden. 1984. “Specification Tests for the Multinomial
findings. The results and findings suggest that decision and policy Logit Model.” Econometrica 52 (5): 1219–1240. doi:10.2307/1910997.
Hensher, D. A., and J. M. Rose. 2007. “Development of Commuter and
makers should develop their transportation program by consider-
Non-commuter Mode Choice Models for the Assessment of New Public
ing various racial/ethnic groups, travel modes, and travel purposes. Transport Infrastructure Projects: A Case Study.” Transportation Research
Part A: Policy and Practice 41 (5): 428–443.
Hess, D. B. 2001. “Effect of Free Parking on Commuter Mode Choice: Evidence
Disclosure statement from Travel Diary Data.” Transportation Research Record 1753 (1): 35–42.
doi:10.3141/1753-05.
No potential conflict of interest was reported by the author. Kamargianni, M., S. Dubey, A. Polydoropoulou, and C. Bhat. 2015.
“Investigating the Subjective and Objective Factors Influencing
Teenagers’ School Travel Mode Choice–An Integrated Choice and
References Latent Variable Model.” Transportation Research Part A: Policy and
Practice 78: 473–488.
Adams, R. J., M. Wilson, and W. C. Wang. 1997. “The Multidimensional Klöckner, C. A., and T. Friedrichsmeier. 2011. “A Multi-level Approach to
Random Coefficients Multinomial Logit Model.” Applied Psychological Travel Mode choice–How Person Characteristics and Situation Specific
Measurement 21 (1): 1–23. doi:10.1177/0146621697211001. Aspects Determine Car Use in A Student Sample.” Transportation Research
Angueira, J., K. C. Konduri, V. Chakour, and N. Eluru. 2019. “Exploring the Part F: Traffic Psychology and Behaviour 14 (4): 261–277. doi:10.1016/j.
Relationship between Vehicle Type Choice and Distance Traveled: A Latent trf.2011.01.006.
Segmentation Approach.” Transportation Letters 11 (3): 146–157. Kroesen, M., and S. Handy. 2014. “The Relation between Bicycle Commuting
doi:10.1080/19427867.2017.1299346. and Non-work Cycling: Results from a Mobility Panel.” Transportation 41
Bento, A., M. Cropper, A. Mobarak, and K. Vinha. 2003. The Impact of Urban (3): 507–527. doi:10.1007/s11116-013-9491-4.
Spatial Structure on Travel Demand in the United States. World Bank. LaMondia, J., and C. Bhat. 2011. “A Study of Visitors’ Leisure Travel Behavior in
Bhat, C. R. 1998. “Analysis of Travel Mode and Departure Time Choice for the Northwest Territories of Canada.” Transportation Letters 3 (1): 1–19.
Urban Shopping Trips.” Transportation Research Part B: Methodological. doi:10.3328/TL.2011.03.01.1-19.
Washington, D.C. 32 (6): 361–371. doi:10.1016/S0191-2615(98)00004-6. Li, J., P. Embry, S. P. Mattingly, K. F. Sadabadi, I. Rasmidatta, and M. W. Burris.
Bhat, C. R. 2000. “Incorporating Observed and Unobserved Heterogeneity in 2007. “Who Chooses to Carpool and Why? Examination of Texas
Urban Work Travel Mode Choice Modeling.” Transportation Science 34 (2): Carpoolers.” Transportation Research Record 2021 (1): 110–117.
228–238. doi:10.1287/trsc.34.2.228.12306. doi:10.3141/2021-13.
Blumenberg, E. 2009. “Moving in and Moving Around: Immigrants, Travel Liu, C. Y., and G. Painter. 2012. “Travel Behavior among Latino Immigrants:
Behavior, and Implications for Transport Policy.” Transportation Letters 1 The Role of Ethnic Concentration and Ethnic Employment.” Journal of
(2): 169–180. doi:10.3328/TL.2009.01.02.169-180. Planning Education and Research 32 (1): 62–80. doi:10.1177/
Blumenberg, E., and G. Pierce. 2014. “Multimodal Travel and the Poor: Evidence 0739456X11422070.
from the 2009 National Household Travel Survey.” Transportation Letters 6 McFadden, D. 1978. “Modeling the Choice of Residential Location.”
(1): 36–45. doi:10.1179/1942787513Y.0000000009. Transportation Research Record (673): 72–77.
Blumenberg, E., and M. Smart. 2009. “Travel in The’hood: Ethnic Mosa, A. I., and M. E. Esawey. 2013. “Modeling Individuals’ Attitudes toward
Neighborhoods and Mode Choice.” UCTC research paper no 891: 1–21. Joint Out-of-home Activity Participation with Household Members.”
Blumenberg, E., and M. Smart. 2010. “Getting by with a Little Help from My Transportation Letters 5 (4): 213–228. doi:10.1179/
Friends . . . and Family: Immigrants and Carpooling.” Transportation 37 (3): 1942787513Y.0000000005.
429–446. doi:10.1007/s11116-010-9262-4. Neoh, J. G., M. Chipulu, and A. Marshall. 2017. “What Encourages People to
Boarnet, M. G., and S. Sarmiento. 1998. “Can Land-use Policy Really Affect Carpool? an Evaluation of Factors with Meta-analysis.” Transportation 44
Travel Behaviour? A Study of the Link between Non-work Travel and (2): 423–447. doi:10.1007/s11116-015-9661-7.
Land-use Characteristics.” Urban Studies 35 (7): 1155–1169. doi:10.1080/ NYMTC and NJTPA. 2013. “2010–2011 Regional Household Travel Survey.”
0042098984538. New York Metropolitan Transportation Council and the North Jersey
Charles, C. Z. 2000. “Neighborhood Racial-composition Preferences: Evidence Transportation Planning Authority.
from a Multiethnic Metropolis.” Social Problems 47 (3): 379–407. Pinjari, A. R., R. M. Pendyala, C. R. Bhat, and P. A. Waddell. 2007. “Modeling
doi:10.2307/3097236. Residential Sorting Effects to Understand the Impact of the Built
Chatman, D. G. 2003. “How Density and Mixed Uses at the Workplace Affect Environment on Commute Mode Choice.” Transportation 34 (5): 557–573.
Personal Commercial Travel and Commute Mode Choice.” Transportation doi:10.1007/s11116-007-9127-7.
Research Record 1831 (1): 193–201. doi:10.3141/1831-22. Polzin, S. E., X. Chu, and J. R. Rey. 1999. Mobility and Mode Choice of People of
Cheng, S., and J. S. Long. 2007. “Testing for IIA in the Multinomial Logit Color for Non-work Travel. Center for Urban Transportation Research.
Model.” Sociological Methods & Research 35 (4): 583–600. doi:10.1177/ Tampa, Florida.
0049124106292361. Ross, R. B. 1975. “Measuring the Influence of Soft Variables on Travel Behavior.”
Chorus, C. 2012. “What about Behaviour in Travel Demand Modelling? an Traffic Quarterly 29 (3): 333–346.
Overview of Recent Progress.” Transportation Letters 4 (2): 93–104. Scheepers, E., W. Wendel-Vos, E. van Kempen, L. I. Panis, J. Maas,
doi:10.3328/TL.2012.04.02.93-104. H. Stipdonk, . . . J. Schuit. 2013. “Personal and Environmental
Dieleman, F. M., M. Dijst, and G. Burghouwt. 2002. “Urban Form and Travel Characteristics Associated with Choice of Active Transport Modes
Behaviour: Micro-level Household Attributes and Residential Context.” versus Car Use for Different Trip Purposes of Trips up to 7.5
Urban Studies 39 (3): 507–527. Kilometers in the Netherlands.” PloS One 8 (9): e73105. doi:10.1371/
Ellis, M., R. Wright, and V. Parks. 2006. “The Immigrant Household and Spatial journal.pone.0073105.
Assimilation: Partnership, Nativity, and Neighborhood Location.” Urban Schlossberg, M., J. Greene, P. P. Phillips, B. Johnson, and B. Parker. 2006.
Geography 27 (1): 1–19. doi:10.2747/0272-3638.27.1.1. “School Trips: Effects of Urban Form and Distance on Travel Mode.”
Eluru, N., V. Chakour, and A. M. El-Geneidy. 2012. “Travel Mode Choice and Journal of the American Planning Association 72 (3): 337–346. doi:10.1080/
Transit Route Choice Behavior in Montreal: Insights from McGill University 01944360608976755.
Members Commute Patterns.” Public Transport 4 (2): 129–149. doi:10.1007/ Schwanen, T., M. Dijst, and F. M. Dieleman. 2001. “Leisure Trips of Senior
s12469-012-0056-2. Citizens: Determinants of Modal Choice.” Tijdschrift Voor Economische En
Emerson, M. O., K. J. Chai, and G. Yancey. 2001. “Does Race Matter in Sociale Geografie 92 (3): 347–360. doi:10.1111/tesg.2001.92.issue-3.
Residential Segregation? Exploring the Preferences of White Americans.” Smart, M. J. 2015. “A Nationwide Look at the Immigrant Neighborhood Effect
American Sociological Review 66: 922–935. doi:10.2307/3088879. on Travel Mode Choice.” Transportation 42 (1): 189–209. doi:10.1007/
Fiebig, D. G., M. P. Keane, J. Louviere, and N. Wasi. 2010. “The Generalized s11116-014-9543-4.
Multinomial Logit Model: Accounting for Scale and Coefficient Smart, M. J., and N. J. Klein. 2013. “Neighborhoods of Affinity: Social
Heterogeneity.” Marketing Science 29 (3): 393–421. doi:10.1287/ Forces and Travel in Gay and Lesbian Neighborhoods.” Journal of the
mksc.1090.0508. American Planning Association 79 (2): 110–124. doi:10.1080/
Fujii, S., and R. Kitamura. 2003. “What Does a One-month Free Bus Ticket Do 01944363.2013.883227.
to Habitual Drivers? an Experimental Analysis of Habit and Attitude Spears, S., D. Houston, and M. G. Boarnet. 2013. “Illuminating the Unseen in
Change.” Transportation 30 (1): 81–95. doi:10.1023/A:1021234607980. Transit Use: A Framework for Examining the Effect of Attitudes and
Giuliano, G. 2003. “Travel, Location and race/ethnicity.” Transportation Perceptions on Travel Behavior.” Transportation Research Part A: Policy
Research Part A: Policy and Practice 37 (4): 351–372. and Practice 58: 40–53.
12 S. YUM

Stern, E., and H. W. Richardson. 2005. “A New Research Agenda for Modelling Van Wee, B., H. Holwerda, and R. Van Baren. 2002. “Preferences for Modes,
Travel Choice and Behaviour.” Social Dimensions of Sustainable Transport: Residential Location and Travel Behaviour.” European Journal of Transport
Transatlantic Perspectives 144–163. and Infrastructure Research 2 (4): 1–12.
Sztompka, P. 2000. “The Turn toward Soft Variables in Sociological Theory.” Vredin Johansson, M., T. Heldt, and P. Johansson. 2005. “Latent Variables in
Trust: A Sociological Theory. Cambridge: Cambridge University Press, 1–17. a Travel Mode Choice Model: Attitudinal and Behavioural Indicator
Van Acker, V., and P. L Mokhtarian., & Witlox, F. 2010. Refining the lifestyle Variables.” Working Paper 2005 (5): 1–36.
concept in travel behaviour research. Spatial and social variations in travel Wang, T., and C. Chen. 2012. “Attitudes, Mode Switching Behavior, and the
behaviour. Incorporating lifestyle and attitudes into travel behaviour-land Built Environment: A Longitudinal Study in the Puget Sound Region.”
use interaction research. Zelzate: DCL Print & Sign. Leegstraat, Belgium Transportation Research Part A: Policy and Practice 46 (10): 1594–1607.
van der Waerden, P., A. Lem, and W. Schaefer. 2015. “Investigation of Factors Washbrook, K., W. Haider, and M. Jaccard. 2006. “Estimating Commuter Mode
that Stimulate Car Drivers to Change from Car to Carpooling in City Center Choice: A Discrete Choice Analysis of the Impact of Road Pricing and
Oriented Work Trips.” Transportation Research Procedia 10: 335–344. Parking Charges.” Transportation 33 (6): 621–639. doi:10.1007/s11116-005-
doi:10.1016/j.trpro.2015.09.083. 5711-x.
Van Vugt, M., P. A. Van Lange, and R. M. Meertens. 1996. “Commuting by Car Yaylali, M., A. K. Çelik, and Ö. Dilek. 2016. “Analyzing Key Socio-economic and
or Public Transportation? A Social Dilemma Analysis of Travel Mode Socio-demographic Drivers of Domestic Passengers’ Airline Choice Behavior
Judgements.” European Journal of Social Psychology 26 (3): 373–395. in Turkey Using Multinomial and Mixed Logit Models.” Transportation
doi:10.1002/(ISSN)1099-0992. Letters 8 (3): 121–130.

You might also like