You are on page 1of 12

INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION

2016, VOL. 10, NO. 4, 354–364


http://dx.doi.org/10.1080/15568318.2014.945672

An examination of recent trends in multimodal travel behavior among American


motorists
Ralph Buehler and Andrea Hamre
Department of Urban Affairs and Planning, Virginia Tech, Alexandria, VA, USA

ABSTRACT ARTICLE HISTORY


According to national statistics, 87% of all trips in the United States are by automobile and 90% of Received 2 December 2013
commuters typically get to work by car. Statistics for individual trips or the main mode of commuting do Revised 6 July 2014
not capture variability in individual travel behavior over time. This article uses the 2001 and 2009 National Accepted 14 July 2014
Household Travel Surveys to analyze recent trends in the share of multimodal motorists who use a car and KEYWORDS
also walk, bicycle, or ride public transport during a day or week. This article identifies trends of multimodal Multimodal car user;
behavior among car users in the United States and provides profiles of these multimodal motorists. sustainable transport; travel
During a typical day about 14% of American car users make at least two trips by foot, bicycle, or public behavior; trends 2001–2009;
transport, while during a typical week about 25% of motorists make at least seven trips by means of USA
transport other than the car. Results from a bivariate analysis and logistic regressions suggest significant
shifts toward more multimodal behavior among motorists between 2001 and 2009. Multimodal motorists
tend to be younger, educated beyond high school, in households without cars, and live in high-density
neighborhoods with access to a rail system. Results suggest that planning for walking, cycling, and public
transport benefits a larger proportion of the U.S. population than suggested by traditional trip-based
analysis.

1. Introduction socioeconomic profiles of monomodal and multimodal moto-


rists in the United States. A better understanding of multimodal
Reliance on the automobile is high in the United States.
car use in the United States will help identify target groups for
According to national statistics, 87% of all trips in the United
policies aimed at increasing walking, cycling, and public trans-
States are by automobile and 90% of commuters typically get to
port use (Chlond, 2012; Diana & Mokhtarian, 2009; Kuhnim-
work by car (U.S. Census Bureau, 2010; U.S. DOT, 2010b). To
hof, Chlond, & von der Ruhren, 2006; Oram & Stark, 1996).
address negative externalities from reliance on the car for most
trips, policymakers are seeking effective ways to increase use of
public transport, walking, and cycling (Forsyth, Krizek, & 2. Literature: Overview of multimodality concept,
Rodriguez, 2009; Lucas, 2011; Small and Verhoef, 2007; U.S. data sources, and key findings
DOT, 2006, 2009, 2010a; Weiner, 2013).
2.1 Concept and terminology of multimodality
Because car ownership and use is high in the United States,
it is likely that Americans who switch to walking, cycling, or Multimodality is an emerging field of study, but it is an exten-
public transport for some trips will continue to drive for other sion of a larger body of research on intrapersonal variability of
trips. Therefore, multimodality—the use of multiple modes of travel behavior. There are four main dimensions of intraper-
transport during a day, week, or month—is gaining recognition sonal variability that have been studied: (a) temporal, (b) spa-
as a potential strategy for reducing automobile reliance by tial, (c) purpose, and (d) modal. The “temporal” dimension
switching some trips to walking, cycling, or public transport refers to variability in timing and frequency of travel. This
(Chlond, 2012; Kuhnimhof et al., 2012a; Nobis, 2007). To date, includes, for example, studies about intrapersonal variability in
little research has focused on the group of American motorists trip duration, trip start times, or frequency of trip-making dur-
who walk, cycle, or ride public transport for some trips—the ing a specific time frame (e.g., a week). The “spatial” dimension
so-called “multimodal car users.” captures variability in geographic location and extent of travel.
Based on two National Household Travel Surveys, the This includes, for example, studies on variability in route choice
NHTS 2001 and 2009, this article analyzes travel behavior of for different trips from the same origin to the same destination.
multimodal car users (motorists) in the United States. The The “purpose” dimension analyzes variability in travel pur-
analysis focuses on multimodal motorists during a travel day or poses or activities. This includes, for example, variability in trip
week. The goals of this article are to provide an overview of the purposes during a week. Last, the “modal” dimension focuses
recent multimodality literature, identify recent trends of multi- on variability in the use of means of transport over time. This
modality in the United States, and develop demographic and dimension captures the use of different modes of transport

CONTACT Ralph Buehler ralphbu@vt.edu Department of Urban Affairs and Planning, Virginia Tech, 1021 Prince Street, Alexandria, VA 22314.
© 2016 Taylor & Francis, Group LLC
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 355

during a month, week, day, or tour. Studies of intrapersonal Travel Survey, Mobilit€at in Deutschland (“MiD”) (Clifton &
variability of travel behavior suggest that travel is more var- Muhs, 2012; Nobis, 2007). MiD is based on a single-day travel
ied—temporally, spatially, modally, and by purpose—than oth- diary. The survey includes questions about how often different
erwise captured by the more common analysis focusing on modes of transport are used during a typical week. A limitation
individual trips, such as the commute (Schonfelder & Axhau- of this type of data is that reporting on travel behavior during
sen, 2010). the prior week is more prone to recall error, particularly for
Terminology and definitions of multimodality vary across short trips, than reporting on travel behavior via travel diaries.
studies (Block-Schachter, 2009). In general, multimodality is
defined as the use of at least two modes of transport during a
trip chain, day, week, or longer time period. By contrast, mono- 2.3 Recent findings on multimodal motorists
modality is defined as the use of a single mode during a speci- Empirical study of intrapersonal variability of travel behavior
fied time period. Some studies incorporate intensity measures has been limited by the scarcity of information on multiday
to further distinguish multi- and monomodality. These studies travel (Schonfelder & Axhausen, 2010). However, increasing
identify thresholds based on the percentage of trips made by a recognition of variability of travel over time has led to greater
single mode. For example, two studies defined individuals with support for multiday data collection (Kuhnimhof, 2009; Schon-
“monomodal tendencies” as using a single mode for over 70% felder & Axhausen, 2010). An overview of recent findings on
or over 90% of all trips (Nobis, 2007; Vij, Carrel, & Walker, multimodal car use since 2005 is provided in Table 1.
2011). Most national level studies focus on Germany or the Nether-
Studies further distinguish groups of multi- and monomodal lands, while U.S.-based studies only capture specific urban
individuals based on the modes of transport used. For example, areas or regions, such as the San Francisco Bay Area (Diana &
multimodality is subdivided into groups based on the combina- Mokhtarian, 2009) or the Massachusetts Institute of Technol-
tion of modes of transport used (e.g., car and bicycle, car and ogy university community (Block-Schachter, 2009).
public transport, bicycle and public transport, etc.). Monomo- Comparison of findings about multimodality across studies
dality is based on exclusive travel either by car, foot, bicycle, or is difficult, because of diverging geographic scales, data sources,
public transport. When intensity of use is incorporated, mono- timing, and definitions of multimodality. However, most stud-
modality may be based on a heavy tendency toward travel by a ies agree that multimodal car use is correlated with not owning
single mode. Some studies of multimodality focus on the car, a car, not having a driver’s license, younger age, living in a
public transport, and bicycling, and exclude walking as a mode smaller household, and having good access to public transport.
of transport (Kuhnimhof et al., 2006; Nobis, 2007). Having a driver’s license and easy access to a car makes car use
The term intermodality is typically used when referring to for all trips more feasible and is correlated with monomodal
the use of multiple modes of transport within one trip or a trip- car use. Multimodal car use among younger travelers is likely
chain. For example, public transport use is often intermodal, related to better access to steeply discounted public transport
because individuals typically access public transport stops or tickets and school environments conducive to walking, cycling,
stations by foot, bicycle, or automobile (Givoni & Rietveld, and riding public transport. Authors hypothesize that smaller
2007). Intermodality can also be used to describe the use of households without children may have less complicated travel
multiple modes of transport during a trip-chain—defined as patterns that create greater flexibility regarding travel sched-
stopping and changing a mode of transport between the origin ules, making use of walking, cycling, and public transport more
and main destination of a trip (Clifton & Muhs, 2012; Nobis, feasible. Moreover, the literature suggests that convenient
2007). access to public transport makes it more likely that a traveler
can substitute away from car use for at least some trips. The
2.2 Typical data sources findings regarding gender have been more variable and seem to
relate to the specific modal combinations used in conjunction
Studies of multimodality typically rely upon the following types with driving. For example, one study found men more likely to
of data: (a) multi-week travel surveys, (b) weeklong travel sur- combine driving with cycling, but women more likely to com-
veys, and (c) one-day travel surveys with questions about travel bine driving with other modes in general (Nobis, 2007). A
during longer time periods. As the time period of observation more detailed discussion of the literature and studies referenced
increases, the share of the population exhibiting multimodal here can be found elsewhere (Buehler & Hamre, 2013).
behavior increases (Clifton & Muhs, 2012; Nobis, 2007). How-
ever, most studies suggest that survey periods of one week tend
to capture most of the variability in everyday habitual travel 3. Data sources and methods
behavior (Kuhnimhof et al., 2006). In those studies, a week is
3.1 Data sources: NHTS 2001 and 2009
described as a “natural increment for the cyclical recurrence of
many day-to-day activities” (Block-Schachter, 2009; Nobis, Data for this analysis originate from the 2001 and 2009
2007). Longer multi-week survey periods additionally capture National Household Travel Surveys (NHTS). NHTS 2001
more occasional travel behavior (Schlich & Axhausen, 2003). and 2009 have only minor differences and many similarities
Multimodality can also be studied using data derived from in methods and variable definitions that render an analysis
single-day surveys that additionally contain questions about of trends over time meaningful (U.S. DOT, 2010b). For
habitual or occasional mode use. For example, two studies of both surveys, households were contacted via random digit
multimodality rely in part on the German National Household dialing of landline telephone numbers. Both surveys
356 R. BUEHLER AND A. HAMRE

Table 1. Recent findings on determinants of multimodal car use published since 2005.

Author(s), year Level of analysis Main data source(s) Main method(s) Multimodal car use findings
(study area)

Kuhnimhof et al. (2012b) National (Germany) Kontiv 1976; German Mobility Panel Descriptive quantitative analysis, Multimodal car use tends to be
(MOP) for 1995–2009; German multivariable regression analysis associated with younger age, less
Income and Expenditure Survey for car availability, not having a
1998 and 2008 driver’s license, higher
educational attainment, being
female, living in an urban area,
auto restraint measures, and
awareness of environmental
impacts.
Chlond (2012) Subnational City of Karlsruhe Office of Statistics for Case study analysis Multimodal car use tends to be
(Karlsruhe, 1980–2005; other City of Karlsruhe associated with younger age, less
Germany) secondary sources car availability, living in an urban
area, auto restraint measures, and
awareness of environmental
impacts.
Vij et al. (2011) Subnational Survey of Karlsruhe and Halle/Salle Descriptive quantitative analysis, Multimodal car use tends to be
(Karlsruhe and residents based on six-week data multivariable regression analysis associated with more bicycle
Halle/Salle, collection in 1999 (MobiDrive) availability, less car availability,
Germany) being female, and small
household size.
Brons and Rietveld (2010) National (The Rail use data from Dutch Railways (NS); Descriptive quantitative analysis, Multimodal car use tends to be
Netherlands) rail performance data from ProRail; multivariable regression analysis negatively associated with travel
car use data from CBS-Statline; time unreliability (regarding
additional station, postcode area public transport).
and timetable data from various
sources, for 2004–2005
Diana and Mokhtarian (2009) Subnational (San Survey of San Francisco Bay Area Cluster analysis Multimodal car use tends to be
Francisco Bay residents in 1998; survey of associated with less car
Area, CA, and employees of the French National availability, higher educational
France) Institute for Transport and Safety attainment.
Research in 2004
Block-Schachter (2009) Subnational Surveys of MIT students, faculty, and Descriptive quantitative analysis Multimodal car use tends to be
(Massachussetts staff based on weeklong data associated with less car
Institute collection in 2004, 2006, and 2008 availability, longer tenure in
of Technology current residence, neighbors with
university multimodal patterns, proximity to
community) public transport, and residing in
an urban area.
Brons et al. (2009) National (The Survey by Dutch Railways (NS) on Descriptive quantitative analysis, Multimodal car use tends to be
Netherlands) customer satisfaction principal component analysis, positively associated with the
multivariable regression analysis quality and level of station
accessibility.
Nobis (2007) National (Germany) German Mobility Panel (MOP) for Descriptive quantitative analysis, Multimodal car use tends to be
1999–2004; Mobilitat in multivariable regression analysis positively assoicated with younger
Deutschland (MiD) for 2002 and older age groups, less car
availability, not having a driver’s
license, no employment, small
household size, and residing in an
urban area.
Givoni and Rietveld (2007) National (The Survey by Dutch Railways (NS) on Descriptive quantitative analysis, Multimodal car use tends to be
Netherlands) customer satisfaction in 2005 multivariable regression analysis associated with station
accessibility improvements.
Kuhnimhof et al. (2006) National (Germany) German Mobility Panel (MOP) for Descriptive quantitative analysis, Multimodal car use tends to be
1996–2003; Mobilitat in multivariable regression analysis associated with younger age, less
Deutschland (MiD) for 2002 car availability, using public
transport as a commute mode,
not having a driver’s license,
higher educational attainment,
small household size, and living in
an urban area.

Sources. Block-Schachter (2009), Brons et al. (2009), Brons and Rietveld (2010), Chlond (2012), Diana and Mokhtarian (2009), Givoni and Rietveld (2007), Kuhnimhof et al.
(2012b), Kuhnimhof et al. (2006), Kuhnimhof (2009), Nobis (2007), Vij et al. (2011a).

combined telephone interviews with single-day travel diaries memory jogger. Both surveys asked respondents about the
to record trips of all household members during a randomly number of trips by walking and cycling during the previous
assigned travel day. Once the travel day passed, households week, as well as public transport use during the last two
were called to report their travel using the diary as a months (2001) or month (2009).
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 357

The surveys were conducted between March 2001 and April relies on number of trips (and not trip distance or travel time),
2002 as well as March 2008 and April 2009. Responses were because NHTS does not include travel distance or travel time
collected for all days of the year, including weekdays, weekends, for the weekly data. Moreover, this analysis includes walk trips,
and holidays. The final samples included 160,758 individuals because the NHTS survey protocol features specific prompts
living in 69,817 households making 642,292 daily trips for the reminding respondents to report walk trips. Following Nobis
2001 survey, and 324,184 individuals living in 150,147 house- (2007) and Vij et al. (2011) the “momomodal category” for this
holds making 1,167,321 daily trips for the 2009 survey. study comprises both monomodal car users, who drove for all
The surveys used stratified sampling of all states, Census trips, and car users with monomodal tendencies, who made
regions, and metropolitan areas, such that each survey is repre- few trips by modes other than the car.
sentative of the United States. A complex weighting procedure At the day level the cutoff value between the monomodal
accounts for nonresponse as well as the exclusion of households and multimodal categories was set at drivers with two or more
without landline telephone service. This analysis applies the additional trips by other modes of transport. Drivers who only
2001 and 2009 weights to ensure statistically representative esti- made one trip by foot, bicycle, or public transport were consid-
mates for the United States as a whole. ered individuals with monomodal tendencies and grouped with
respondents who drove for all trips. This definition provides a
conservative threshold for multimodality, by treating drivers
3.2 Methods: Analyzing monomodal and multimodal car
who only made one additional trip during the travel day as
use with the NHTS
monomodal drivers. Empirically, about 22% of multimodal
To evaluate population shares of multimodal motorists in the drivers reported only one additional non-car trip. Drivers with
United States, this analysis relied on data from the daily trip two additional non-car trips constituted the largest group
and person data files of the 2001 and 2009 NHTS surveys. Data among multimodal respondents (52%). Driving plus three or
about variability in mode choice of driving, public transport, more non-car trips accounted for 26% of daily multimodal
walking, and cycling during the day originate from the trip file drivers, with no subsequent categories (motorist plus four, plus
based on the single-day travel diary. Individuals recorded all five, etc.) representing more than 9% of multimodal drivers.
trips made during a randomly assigned travel day. Trips were At the week level, the cutoff value between the monomodal
defined as travel between two addresses. Additionally, the anal- and multimodal categories was set at driving and seven or more
ysis includes separate modes of transport reported for trips to trips by other modes of transport during the week—one trip by
and from public transport. The mode of transport used for an alternative mode per day. A statistical average of one addi-
individual trips and for public transport access and egress was tional trip by walking, cycling, or public transport per weekday
aggregated and matched to the trip maker and then added to can be considered habitual daily use of modes other than driv-
the person data set. ing. Empirically, the weekly data had more gradual breaks than
In addition to the 24-hour trip data, the analysis also the daily data presented above. The percentage shares of multi-
includes weekly data available in the NHTS person file with modal motorists were: 12% for one non-car trip per week; 16%
responses to specific questions about the number of trips made for two non-car trips; 16% for three non-car trips; 11% for four
by walking and cycling during the previous week. Moreover, non-car trips; 11% for five non-car trips; and 6% for six non-car
the authors measured the use of public transport during the trips. Slightly more than 14% reported seven non-car trips per
last week based on a variable capturing public transport use in week, representing a local maximum in the distribution; subse-
the previous two months provided in the 2001 survey and pub- quent categories (motorist plus eight, motorist plus nine, etc.
lic transport use in the previous month recorded in the 2009 non-car trips per week) only accounted for small shares (<4.5%
survey. Neither survey included questions about car use in the each) of the distribution of multimodals.
previous week or month. Thus, the data for weekly car use orig- The analysis first compares population shares of multimodal
inate from the single-day travel diary and a question asking motorists in 2001 and 2009 at the aggregate level and for popu-
about the usual mode of transport for the commute to work in lation subgroups using the “day C 2” and “week C 7” thresh-
the previous week. olds. This part of the analysis identifies levels of multimodality
The literature review above showed that there is no standard in population subgroups without controlling for other varia-
definition of multimodality. Some studies distinguish between bles. The second part of the analysis compares statistical signifi-
monomodals who exclusively use one mode of transport, and cance of differences in multimodal driving for population
multimodals who use multiple modes. Other studies combine subgroups in 2001 and 2009 using logistic regressions based on
those who exclusively use one mode together with individuals the 2001 and 2009 NHTS. The regressions calculate the likeli-
with monomodal tendencies who make the majority of their hood of multimodal car use in 2001 and 2009, after controlling
trips by one mode. Studies define multimodality groups based for the impacts of all the other variables analyzed. Wald tests
on travel distance, travel time, or number of trips. Some studies compare coefficients of the 2001 and 2009 regressions to assess
exclude walking, either because walk trips are often underre- statistical significance of changes over time. The first and sec-
ported or because walking is a poor indicator in studies that ond parts of the analysis complement each other. The bivariate
derive multimodality from trip distance—a measure dominated analysis shows levels (population shares) of multimodality not
by faster motorized modes. shown in the regression analysis, but cannot control for the
Results from the literature review and an empirical analysis influence of other variables. The regression produces estimates
of the 2001 and 2009 NHTS informed the definition of mono- of coefficients for individual variables while controlling for
modal and multimodal groups for this analysis. This analysis other variables in the analysis.
358 R. BUEHLER AND A. HAMRE

3.3 Limitations or more trips by modes other than the car rose from 22.3% to
25.1%.1 The table also lists population shares for week and day
There are important limitations to this analysis of multimodal
levels across both years for the independent variables. Control
car use based on the NHTS data sets:
variables were chosen based on data availability in NHTS and
First, the data sets provide detailed information about all
variables identified in the literature review including gender,
modes of transport used during one travel day. Weekly data
race/ethnicity, age, household life cycle, education, employment
rely on self-reported trips during the past week for walking and
status, income, car access, land use, rail access, and weekend
cycling and during the past 1–2 months for public transport,
versus weekday travel (day level only). Population shares for
and are likely not as reliable as information from travel diaries.
independent variables only changed minimally between the
Second, there was no information about car use during the
two years (between C/¡0.0% and C/¡2.0%), with the excep-
week other than the mode for the usual commute and driving
tion of two age groups. Between 2001 and 2009, the share of
for trips reported for the travel day. Thus, the weekly data do
individuals between 25 and 34 decreased significantly from
not capture the intensity of driving (e.g., number of car trips
20.6% to 14.5%, while the share of 50- to 64-year-olds increased
per week). Moreover, the analysis may underestimate the share
significantly from 20.1% to 26.3%. In both years the weekly
of multimodal drivers, because of the lack of information about
data has slightly more cases than the daily data. Some trip mak-
the usual car use during the week.
ers who did not drive during the travel day reported commut-
Third, this analysis excluded individuals who did not report
ing by car during the week.
any trips for the assigned travel day. Even if they reported walk,
The vast majority of multimodal motorists reported
bike, or public transport trips during the week, missing infor-
walking as their only non-automobile mode of transport at
mation about their car use during the travel day inhibited accu-
the “day C 2” (78.5% in 2001 and 81.8% in 2009) and
rately categorizing them into one of the modality groups.
“week C 7” (65.7% in 2001 and 68.2% in 2009) levels. The
Fourth, the 2009 survey only asked respondents older than
share of multimodal motorists reporting the use of two dif-
15 about travel during the last week. Thus, this analysis
ferent modes of transport in addition to the car was higher
excludes children 15 and younger. Many analyses of multimo-
at the week than at the day level. In 2009, 14.7% of multi-
dality exclude children who cannot drive, because children
modal motorists in the “week C 7” category combined
legally have to rely on others (often their parents) if they wish
walking and cycling, while the share was 18.7% for the
to travel by automobile (Kuhnimhof et al., 2006; Vij et al.,
combination of walking and public transport. Using all four
2011). Only including individuals 16 and older (driving age in
modes of transport was rare, with 2.6% of multimodal
most U.S. states) focuses the analysis on those who can obtain a
motorists in the “week C 7” category reporting walking,
driver’s license (or provisional driving permit) and legally drive
cycling, and riding public transport. In 2009, 18.5% of daily
a car.
monomodal car users were weekly multimodal motorists at
Last, our analysis compares changes over time using two
the “week C 7”0 level. The NHTS trip file allows for a more
independent cross-sectional data sets. Therefore, the analysis
detailed analysis of the number of daily trips, average daily
cannot claim to observe changes in travel behavior for the same
trip distance, and daily trip purpose. In 2009, individuals in
individual over time. An analysis to track changes in travel
the “day C 2” multimodal car use category made more trips
behavior of individuals would rely on a panel data set with the
per person per day than monomodal car users (6.4 vs. 4.4).
same individuals reporting on their daily and weekly travel
Daily monomodal car users made longer trips with an aver-
over time. Moreover, a quasi-experimental study that would
age trip distance of 10.8 miles compared to 6.4 miles. Aver-
allow comparison between a “treatment group” and a “control
age total daily travel distance was greater for monomodal
group” would allow researchers to isolate causal factors for
car users (47.0 miles) than multimodal car users (40.7
multimodality. This analysis relies on data of independently
miles). Monomodal motorists made a higher share of trips
sampled observations from 2001 and 2009. In recognition of
for shopping, personal errands, and schooling (31.8% of all
potential changes in the distributions of variables between sur-
trips) than multimodal motorists (24.2%). Work and work-
vey years, this study employs Wald tests of our logistic regres-
related trips accounted for a higher share of trips among
sion coefficients to test for significant changes over time. We
monomodal than multimodal motorists (22.7% vs. 17.3%).
also present descriptive statistics of the independent variables
By contrast, compared to monomodal car users, multimodal
and discuss the potential influence of changes in the composi-
motorists made a greater share of their trips for recreational
tion of the population between the two surveys.
purposes (33.8% vs. 24.2%).

4. Mode choice and multimodality


1
4.1 Trends in multimodal driving at the day and week Our analysis included mode choice for access and egress trips to public transport.
levels and independent variables 2001–2009 Excluding public transport access and egress trips from our analysis would
decrease the share of individuals considered multimodal. However, because the
Table 2 presents descriptive statistics and definitions for the share of all trips by public transport is relatively low in the United States, the dif-
ference in results is not large. We prefer to include public transport access and
dependent and independent variables used in this analysis— egress trips to more accurately capture modal variability at the person level.
both for the day and week level and for 2001 and 2009. Between However, we recognize that travel demand forecasting typically excludes these
2001 and 2009, the share of drivers making two or more daily trips. Excluding public transport access and egress would mean a change in the
population share of the “day C 2” group from 11.0% to 10.2% in 2001 and
trips by foot, bicycle, or public transport increased from 11.0% 13.5% to 12.5% in 2009, and a change in the share of the “week C 7” group
to 13.5%. At the week level, the share of drivers making seven from 22.3% to 22.0% in 2001 and 25.1% to 25.0% in 2009.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 359

Table 2. Descriptive statistics and measurement of variables in the analysis.

Day Week

2001 2009 2001 2009 Description & Measurement


   
Variable Categories Mean SD Mean SD Mean SD Mean SD Share of motorists who…

Day modal groups Driver with maximum of 1 0.890 0.313 0.865 0.341 NA NA NA NA … make 0 or 1 walk, bike, and/or
non-car trip per day public transport trips per day
Driver with at least 2 0.110 0.313 0.135 0.341 NA NA NA NA … who make 2 or more walk, bike,
non-car trips per day and/or public transport trips per
day
Week modal groups Driver with maximum of 6 NA NA NA NA 0.777 0.416 0.749 0.434 … make 0 or 1–6 walk, bike, and/or
non-car trips per week public transport trips per week
Driver with at least 7 NA NA NA NA 0.223 0.416 0.251 0.434 … make 7 or more walk, bike, and/or
non-car trips per week public transport trips per week
Gender Female 0.504 0.500 0.504 0.500 0.503 0.500 0.503 0.500 … are female
Male 0.496 0.500 0.496 0.500 0.497 0.500 0.497 0.500 … are male
Race/ethnicity Non-White 0.247 0.431 0.231 0.421 0.248 0.432 0.230 0.421 … identified as Non-White
White 0.753 0.431 0.769 0.421 0.752 0.432 0.770 0.421 … identified as White
Age Age 16–24 0.148 0.355 0.129 0.335 0.148 0.355 0.129 0.335 … are 16 to 24 years old
Age 25–34 0.206 0.405 0.146 0.353 0.206 0.405 0.146 0.353 … are 25 to 34 years old
Age 35–49 0.319 0.466 0.326 0.469 0.320 0.466 0.327 0.469 … are 35 to 49 years old
Age 50–64 0.201 0.401 0.263 0.440 0.201 0.401 0.263 0.440 … are 50 to 64 years old
Age 65C 0.125 0.331 0.136 0.342 0.125 0.331 0.135 0.342 … are 65 years old or above
Household life cycle Singles 0.101 0.301 0.117 0.321 0.101 0.301 0.117 0.321 … live in households with a single
adult and no children
2C adults only 0.387 0.487 0.384 0.486 0.387 0.487 0.384 0.486 … live in households with two adults
and no children
With children 0.512 0.500 0.499 0.500 0.512 0.500 0.500 0.500 … live in households with children
Education Status Education < D high school 0.379 0.485 0.354 0.478 0.378 0.485 0.354 0.478 … have education up to high school
Education >high school 0.621 0.485 0.646 0.478 0.622 0.485 0.646 0.478 … have education beyond high
school
Employment status Not employed 0.286 0.452 0.281 0.450 0.285 0.452 0.279 0.449 … are not employed
employed 0.714 0.452 0.719 0.450 0.715 0.452 0.721 0.449 … are employed
Income quartile Lower income quartile 0.321 0.467 0.337 0.473 0.321 0.467 0.338 0.473 … are in the first income quartile
Middle income quartiles 0.461 0.498 0.443 0.497 0.461 0.498 0.442 0.497 … are in the second and third
income quartiles
Upper income quartile 0.218 0.413 0.220 0.414 0.218 0.413 0.220 0.414 … are in the fourth income quartile
Household car availability 0 Household cars 0.018 0.132 0.019 0.137 0.018 0.133 0.019 0.137 … live in households with zero
household automobiles
1 Household car 0.200 0.400 0.212 0.409 0.200 0.400 0.212 0.409 … live in households with one
household automobile
2 Household cars 0.429 0.495 0.418 0.493 0.429 0.495 0.418 0.493 … live in households with two
household automobiles
3C Household cars 0.353 0.478 0.351 0.477 0.353 0.478 0.350 0.477 … live in households with three
household automobiles
Residential density Density <10,000 0.903 0.296 0.910 0.286 0.902 0.297 0.909 0.287 … live in locations with population
density per square mile below
10,000
Density > D 10,000 0.097 0.296 0.090 0.286 0.098 0.297 0.091 0.287 … live in locations with population
density per square mile at or
above 10,000
Rail access Not in MSA with rail 0.741 0.438 0.730 0.444 0.741 0.438 0.729 0.444 … live outside a Metropolitan
Statistical Area with a rail system
MSA with rail 0.259 0.438 0.270 0.444 0.259 0.438 0.271 0.444 … live in a Metropolitan Statistical
Area with a rail system
Travel day Weekday 0.716 0.451 0.723 0.448 NA NA NA NA … traveled between Monday and
Friday
Weekend 0.284 0.451 0.277 0.448 NA NA NA NA … traveled Saturday or Sunday
# of cases 93,270 192,575 93,731 193,654

Sources. NHTS 2001 and 2009.



As measured all variables have a range from 0 to 1 and the mean indicates the population share falling in the specific subgroup.

4.2 Bivariate analysis of multimodal car use in 2001 and over time is determined by calculating differences in weighted
2009 proportions or means between two independent samples (p <
0.01). The analysis also compares differences in multimodal
Table 3 presents shares of multimodal motorists at the “day C
driving for population subgroups—examining the statistical
2” and “week C 7” levels disaggregated by demographic, socio-
significance of bivariate differences using t tests and analysis of
economic, and land-use variables for the years 2001 and 2009.
variance (ANOVA) with Scheffe posttests to assess differences
The goal of this bivariate analysis is to present levels of multi-
between groups (p < 0.01). Section 4.4 will go beyond this two-
modality for different population subgroups—looking at one
variable analysis and present a logistic regression analysis con-
independent variable at a time. Statistical signifıcance in trends
trolling for the impact of other variables.
360 R. BUEHLER AND A. HAMRE

Table 3. Trends in shares of multimodal motorists at day and week levels between 2001 and 2009.

Day level Dependent variable Week level Dependent variable


(1 D 2 or more non-car trips; 0 D 0 or 1 non-car trip) (1 D 7 or more non-car trips; 0 D 0 to 6 non-car trips)

Variable 2001 2009 Change 01–09 (Ho: 01 D 09) 2001 2009 Change 01–09 (Ho: 01 D 09)

All 10.9 13.4 p < 0.01 22.1 25.1 p < 0.01


Female 11.3 13.3 p < 0.01 20.5 23.1 p < 0.01
Male 10.5 13.4 p < 0.01 23.8 27.1 p < 0.01
Non-White 12.2 14.3 p < 0.01 22.1 24.8 p < 0.01
White 10.5 13.1 p < 0.01 22.2 25.2 p < 0.01
Age 16–24 10.8 12.1 24.4 25.0
Age 25–34 12.0 15.9 p < 0.01 20.0 24.3 p < 0.01
Age 35–49 10.8 14.4 p < 0.01 21.6 25.0 p < 0.01
Age 50–64 10.6 13.3 p < 0.01 22.6 26.6 p < 0.01
Age 65C 10.3 10.0 23.5 23.5
Singles 13.4 15.8 24.5 26.4
2C Adults only 11.1 12.5 p < 0.01 23.1 26.3 p < 0.01
With kids 10.3 13.4 p < 0.01 20.9 23.9 p < 0.01
Education < D H.S. 9.4 10.2 21.1 23.7 p < 0.01
Education >H.S. 11.9 15.1 p < 0.01 22.7 25.7 p < 0.01
Not employed 11.0 12.5 p < 0.01 23.4 27.0 p < 0.01
Employed 10.9 13.7 p < 0.01 21.6 24.3 p < 0.01
Lower income 11.0 11.8 23.8 26.0 p < 0.01
Middle income 10.0 13.1 p < 0.01 21.3 24.6 p < 0.01
Upper income 13.0 16.7 p < 0.01 22.3 25.5 p < 0.01
0 HH Cars 31.9 28.0 48.1 43.5
1 HH Car 14.4 15.9 25.3 28.6 p < 0.01
2 HH Cars 10.5 13.8 p < 0.01 20.3 23.5 p < 0.01
3C HH Cars 8.3 10.5 p < 0.01 21.1 23.9 p < 0.01
Density <10,000 9.8 12.3 p < 0.01 21.0 24.5 p < 0.01
Density > D 10,000 21.5 23.8 32.5 31.5
Not in MSA with rail 9.3 11.5 p < 0.01 20.7 24.3 p < 0.01
MSA with rail 15.5 18.2 p < 0.01 26.2 27.2
Weekday 11.5 14.1 p < 0.01 NA NA
Weekend 9.4 11.3 p < 0.01 NA NA

Source. Authors’ analysis of NHTS 2001 and 2009.


Note. T tests and ANOVA with Sheffe posttests (p < 0.01) showed that most variable categories were statistically significantly different from each other. However, there
are some exceptions as listed below. In 2001 at the day level, only the age group between 25 and 34 was different from other age groups and there was no statistical
difference by employment level. In 2009 at the day level, there were no differences by gender and by race/ethnicity. In 2001 at the week level, there was no statistically
significant difference between the 16–24 and 65C, 35–49 and 65C, and 50–64 and 65C age groups, as well as between the middle and higher income groups. In 2009
at the week level, there were no significant differences for 16–24 and 25–34, 16–24 and 35–49, 25–34 and 35–49, 25–34 and 65C, and 50–64 and 65C age groups, sin-
gle and two adult households, lower and upper income quartiles, and two- and three-car households.

For all population subgroups, multimodality levels were multimodal car users and the share of multimodal motorists
greater at the week than at the day level (p < 0.01). In declined with increasing age between the age groups of 25–34
both years the week- and day-level analysis show that and 65C. The “week C 7” level showed more homogenous rates
households without cars and in the high population density of multimodal car use for all age groups ranging only from
categories display the highest shares of multimodal drivers. 23.5% to 26.6%. Between the surveys, the shares of multimodal
Similarly, in 2009 at both thresholds the individuals with a motorists increased significantly for the age groups between 25
high school degree or less, more cars, 65 and older, and and 64 at both thresholds. Multimodal car use did not change
without rail access had among the lowest shares of multi- significantly for older adults (65C) and those 16–24.
modal drivers. Moreover, the majority of population sub- At both multimodality thresholds singles were more likely multi-
groups showed increasing shares of daily and weekly modal motorists than other household life-cycle groups. In contrast
multimodal motorists between 2001 and 2009—exceptions to the other two life-cycle groups, singles did not show a significant
include the 16–24 and 65C-year-olds, singles, the carless, change in multimodal car use between 2001 and 2009 (p < 0.01).
and those at lower population densities. At both thresholds individuals with education beyond high
In 2009, there were higher shares of multimodal motorists school were more likely multimodal motorists than those with
among men than women (27.1% vs. 23.1%) and among Whites a high school degree or less. Moreover, the share of multimodal
than non-Whites (25.2% vs. 24.8%) at the “week C 7” level, but car users among those with education beyond high school
there were no differences by gender or race/ethnicity at the increased significantly over the last decade. By contrast, for
“day C 2” level (p < 0.01). Between the two surveys, the share those with a high school degree or less the share of multimodal
of multimodal motorists increased among Whites, non-Whites, motorists increased only in the “week C 7” category. Individu-
men, and women (p < 0.01). als who are not employed (or not in the workforce) were more
At both thresholds, the age group 65C had the lowest share likely multimodal motorists at both day and week levels in
of multimodal motorists in 2009 (p < 0.01). At the “day C 2” 2009. The share of multimodal motorists increased among
level the 25- to 34-year-old age group had the highest share of both employed individuals and those not employed.
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 361

Table 4. Relative likelihood of being a multimodal motorist for population subgroups at week and day levels in 2001 and 2009.

Day level Dependent variable Week level Dependent variable


(1 D 2 or more non-car trips; 0 D 0 or 1 non-car trip) (1 D 7 or more non-car trips; 0 D 0 to 6 non-car trips)

2001 2009 2001 2009


Variable AOR AOR Change 01-09 (Ho: 01 D 09) AOR AOR Change 01-09 (Ho: 01 D 09)

Gender Female 1.000 1.000 1.000 1.000


Male 0.969 1.015 1.269 1.343 p < 0.01
Race/ethnicity Non-White 1.000 1.000 1.000 1.000
White 1.158 1.146 1.436 1.245 p < 0.01
Age Age 16–24 1.000 1.000 1.000 1.000
Age 25–34 0.935 1.135 p < 0.01 0.761 1.010 p < 0.01
Age 35–49 0.958 1.129 p < 0.01 0.907 1.134 p < 0.01
Age 50–64 0.956 1.109 p < 0.05 0.902 1.137 p < 0.01
Age 65C 0.788 0.803 0.839 0.863
Household life cycle Singles 1.000 1.000 1.000 1.000
2C Adults only 1.071 0.940 p < 0.01 1.445 1.003 p < 0.01
With kids 0.972 0.923 1.268 0.864 p < 0.01
Education status Education < D H.S. 1.000 1.000 1.000 1.000
Education >H.S. 1.485 1.672 p < 0.01 1.147 1.119
Employment status Not employed 1.000 1.000 1.000 1.000
Employed 0.963 0.943 0.961 0.817 p < 0.01
Income quartile Lower income 1.000 1.000 1.000 1.000
Middle income 1.055 1.200 p < 0.01 0.882 0.959 p < 0.01
Upper income 1.325 1.591 p < 0.01 0.891 1.036 p < 0.01
Household car availability 0 HH Cars 1.000 1.000 1.000 1.000
1 HH Car 0.321 0.485 p < 0.01 0.373 0.542 p < 0.01
2 HH Cars 0.217 0.374 p < 0.01 0.274 0.503 p < 0.01
3C HH Cars 0.168 0.299 p < 0.01 0.288 0.529 p < 0.01
Residential density Density <10,000 1.000 1.000 1.000 1.000
Density > D 10,000 2.098 1.565 p < 0.01 1.418 1.214 p < 0.01
Rail access Not in MSA with rail 1.000 1.000 1.000 1.000
MSA with rail 1.364 1.353 1.121 1.063
Travel day Weekday 1.000 1.000 NA NA
Weekend 0.814 0.805 NA NA

p < 0.01 Number of cases 93,270 192,575 93,731 193,654
Likelihood-ratio p < 0.01 p < 0.01 p < 0.01 p < 0.01

Source. Authors’ analysis of NHTS 2001 and 2009.

For the “day C 2” category, higher income was associated while controlling for other variables. Explanatory variables in
with a greater share of multimodal motorists. This was not the multiple regression analysis include all of the demographic,
found for the “week C 7” category. For both thresholds, the socioeconomic, and land-use variables introduced in Tables 2
share of multimodal car users in the mid-income groups and 3. Within each regression, adjusted odds ratios (AORs) are
increased significantly between 2001 and 2009. interpreted as the population subgroup’s likelihood of being
The share of multimodal motorists was greatest for those liv- multimodal relative to a specific reference group assigned the
ing in households without cars and the share of multimodal base value 1.000.
motorists declined as car ownership increased. Over time, the Within each column of results, AORs of independent
shares of multimodal motorists in households with cars variables are evaluated according to three criteria: (a) the
increased significantly, while the changes for individuals in sign of the coefficient (AOR), (b) its magnitude in relation
households without cars were not statistically significant (p < to the reference category, and (c) its statistical significance.
0.01). An additional column presents results from a Wald test of
At both levels the share of multimodal motorists is statistical difference comparing AORs for the 2001 and
greater at higher population densities and for those with 2009 models. For all four models, likelihood ratio tests (LR
rail access. Over time, the “day C 2” category showed sig- tests) indicate that independent variables have joint signifi-
nificant increases in multimodal car use for high density cance. Variance Inflation Factor and Tolerance tests for
and both rail access categories. However, there was no sig- multicollinearity do not indicate any serious problem.
nificant change of multimodal car use for the high-density
category at the weekly level. 4.3.1 Day level
Models 1a and 1b show no difference in the likelihood of multi-
modal car travel between men and women for 2001 and 2009.
4.3 Logistic regression analysis of multimodal car user at
In both years Whites were about 1.16 times as likely multi-
day and week levels in 2001 and 2009
modal as non-Whites (AORs 1.158 and 1.146). In 2001 and
Table 4 presents results from four logistic regression models— 2009, those 65 and older were less likely to be multimodal
one model for each year (2001 and 2009) at the day and week motorists than those 16–24 (AORs 0.788 and 0.803). Although
levels—estimating the likelihood of being a multimodal driver there was no difference in the likelihood to be a multimodal
362 R. BUEHLER AND A. HAMRE

motorist between the age groups 25–64 and those 16–24 years U.S. DOT, 2011). However, this analysis shows that longer
old in 2001, by 2009 the mid-age groups were more likely mul- time periods capture greater variability in travel behavior
timodal than the 16- to 24-year-olds (AORs 1.135, 1.129, (Schonfelder & Axhausen, 2010): 25.1% of motorists make at
1.109). Similarly, in 2001 there was no statistical difference in least seven trips by another mode of transport during a week
the likelihood to be multimodal between single and other and 13.5% of American motorists walk, cycle, or ride public
household life-cycle categories, but in 2009, households with transport for at least two trips per day.
children were associated with a lower likelihood to be multi- Between 2001 and 2009, the share of multimodal motorists
modal than single-person households (AOR 0.923). increased at the aggregate level and for most population sub-
In both years individuals with more than a high school edu- groups analyzed in this study. Increasing multimodality of
cation were more likely to be multimodal than those with a motorists may help explain overall declining car travel demand
high school degree or less (AOR 1.485 and 1.672). This differ- in the United States (ITF/OECD, 2012; Millard-Ball & Schip-
ence increased significantly between 2001 and 2009 (p < 0.01). per, 2011). The NHTS survey shows that in spite of a drop in
Similarly, in 2009 those in the middle and highest income daily vehicle miles traveled (62.3 to 56.1 per person) and num-
groups were significantly more likely multimodal than those in ber of daily trips by car (4.31 to 3.92 per person), vehicle own-
the lowest income quartile—with significant increases in AORs ership rates among car users did not change between 2001 and
between the two survey years for the higher income groups. 2009 (2.32 and 2.34 cars per household member at driving
In both years, more cars per household were associated with age). Increasing multimodality—more daily trips by foot (0.53
a decreased likelihood of being multimodal. However, the dif- to 0.63 per person), bicycle (0.02 to 0.03), and public transport
ference between individuals in carless households and those in (0.10 to 0.11)—may help explain why in 2009 Americans
households with cars diminished between 2001 and 2009—as owned automobiles at the same rate as in 2001, but drove fewer
evidenced by larger AORs for the year 2009 than for 2001 miles and made fewer trips by car. Finally, the increase in mul-
(0.485, 0.374, 0.299 vs. 0.321, 0.217, 0.168). In both years, timodality may possibly be related to the economic downturn;
higher population density and access to rail were associated higher gasoline prices; slower increases in roadway supply, traf-
with a greater likelihood of being multimodal. Lastly, weekend fic congestion and constrained individual time budgets; the
travelers were only about 0.800 times as likely multimodal as impact of new technology; or more and better infrastructure
respondents who traveled during weekdays (AOR 0.814 and for walking, cycling, and public transport (Kuhnimhof, 2012b;
0.805). Millard-Ball & Schipper, 2011; Newmann & Kenworthy, 2011).
The vast majority of multimodal motorists reported walking
4.3.2 Week level as their only non-automobile mode of transport at the day
Most variables and categories show statistically significant (»80%) and week (»67%) levels. Use of two modes of trans-
changes between AORs for the 2001 and 2009 week-level mod- port in addition to the car was rare on the day level (<10% of
els (p < 0.01)—with the exceptions of the oldest age and educa- multimodal car users), but accounted for about 30% of multi-
tion groups. In contrast to the day-level analysis that found no modal motorists at the week levels. In addition, roughly 3% of
difference between genders, men were 1.300 times as likely multimodal motorists in the “week C 7” category reported
weekly multimodals as women (AORs 1.269 and 1.343). Similar using three other modes—walking, cycling, and riding public
to the day-level analysis, Whites were more likely multimodal transport—in addition to driving.
than non-Whites (AORs 1.436 and 1.245). As for the day-level Overall, multimodal car users made more trips per day than
analysis, in both years the age group 65C was less likely multi- monomodal car users. However, average total daily travel dis-
modal than the 16- to 24-year-olds (AORs 0.839 and 0.863). In tance was shorter for multimodal than monomodal motorists
addition, like results for the day level, in 2009 those in house- (»41 vs. 47 miles). This is in line with at least one prior study
holds with children were less likely multimodal than singles from Germany, suggesting fewer miles traveled for multimodal
(AOR 0.864). Similar to the day-level analysis, those with more than for monomodal car users (Nobis, 2007). Shorter travel dis-
than a high school education had a greater likelihood to be tances among multimodal car users suggest that a higher share
multimodal than those with a high school degree or less (AORs of multimodal individuals in the population may help explain
1.147 and 1.119). On the week and day level, employed individ- reduced overall vehicle miles of travel (Chlond, 2012).
uals were less likely multimodal than those not employed or Disaggregate analysis showed diverging trends and levels
not in the workforce. Like in the day-level analysis, more cars of multimodal car use across population subgroups. Results
per household were associated with lower likelihood of multi- at the national level for the United States do not necessarily
modality and the difference between those in households with conform to findings from prior studies in Europe and at
and without cars diminished between the years (as evidenced the regional level in the United States. For example,
by larger AORs for 2009). Population density and rail access although prior studies consistently showed a lower share of
were related to a greater likelihood to be a multimodal motorist. multimodal car users among men than women in Europe
(Kuhnimhof et al., 2006; Nobis, 2007), the only significant
difference by gender in this study was a higher share of
5. Discussion
multimodal motorists among men than women at the
The analysis reveals that travel behavior of U.S. motorists is “week C 7” level. More ubiquitous car ownership in U.S.
more varied than trip-based analysis suggests. American moto- households—compared to Europe—may help explain similar
rists make roughly 93% of their daily trips by automobile and levels of multimodal car use among American women and
drive for almost 95% of regular commutes (U.S. DOC, 2011; men. The lower car ownership levels in Europe may result
INTERNATIONAL JOURNAL OF SUSTAINABLE TRANSPORTATION 363

in greater differences in car access across genders. For 6. Conclusion


example, in households where men work outside the home
For many transport planners, a small share of trips by foot,
and drive to work, women may be forced to make more
bicycle, and public transport in the United States suggests that
trips without the car (Giuliano & Narayan, 2004). Some
only a small minority of Americans walk, cycle, or ride public
readers may think that the difference is related to the inclu-
transport. However, statistics for individual trips or the main
sion of walk trips in this analysis, while walk trips were
mode of commuting to work do not capture variability in indi-
excluded in other studies. A separate analysis excluding
vidual travel behavior over time. The analysis of the 2001 and
walk trips, however, still shows the same results, including
2009 NHTS finds that during a typical day about 14% of Amer-
more multimodality among men than among women at the
ican car users make at least two trips by foot, bicycle, or public
week level.
transport. Moreover, during a typical week about 25% of Amer-
At all thresholds the age group 65C had among the lowest
ican motorists make at least seven trips per week by means of
share of multimodal motorists in this study. This is in line with
transport other than the car.
most prior studies that indicate less multimodality among older
The analysis of trends over time suggests a significant shift
individuals. Moreover, between the surveys, the shares of multi-
toward multimodal behavior among motorists between 2001 and
modal motorists increased significantly for the age groups
2009 at the aggregate level and for most population subgroups. In
between 25 and 64 at all multimodality thresholds. This study
spite of increases in multimodality for many population subgroups,
thus expands upon the recent findings of increasing multimo-
multimodal motorists tend to be younger, educated beyond high
dality among young adults (Kuhnimhof et al., 2012a,b) by sug-
school, live in households without cars, in high-density neighbor-
gesting increasing multimodality for a larger share of the
hoods, and in areas with a rail system. Policies that promote walk-
population.
ing, cycling, and public transport use for everyone may help
This study showed higher levels of multimodal car use
increase multimodal car use among current monomodal motorists
among those with higher education. This is consistent with
and enforce the trend toward higher shares of multimodality
prior findings of higher levels of walking and cycling among
among motorists who already use alternatives to driving but wish
individuals with higher levels of education (Buehler, Pucher,
to increase the intensity of their multimodality. Policies that facili-
Merom, & Bauman, 2011). One explanation may be that while
tate access to public transport and increase population densities
attending university, individuals learn to appreciate walking,
with shorter trip distances may help to further increase the share of
cycling, and riding public transport as viable modes of trans-
multimodal drivers in the United States. In contrast to European
port (Khattak, Wang, Son, & Agnello, 2011; Transit Coopera-
countries, multimodality in the United States is hindered by
tive Research Program [TCRP], 2004; Wang, Khattak, &
sprawled settlement patterns with low population densities and
Sanghoon, 2012). This experience may then be reflected in their
long trip distances that often render walking, cycling, and public
later travel behavior after leaving university.
transport impractical.
Individuals in single households were more likely multi-
More and better data about travel behavior during the week
modal car users than those in households with children or two
could strengthen future analysis. The weekly analysis captures
or more adults. This confirms prior studies that suggest that
more variability in travel than found by analysis of just one travel
households with more adults, and especially those with chil-
day. This suggests that household travel surveys should attempt to
dren, have more complicated trip patterns than individuals.
capture longer time periods than is currently standard practice in
Picking up or dropping off other household members likely
the United States to obtain a more holistic picture of the variability
requires a car for most trips in the United States—even if direct
of individual travel. Additionally, a panel survey, with repeat meas-
individual trips (without stops along the way) could be made
ures of the same respondents over time, would facilitate the analysis
by modes other than the car.
of changes in travel behavior of individuals. Recognizing these
Car ownership is an important factor influencing monomo-
potential improvements, the findings of this study suggest that
dal versus multimodal car use. The share of multimodal moto-
planning for walking, cycling, and public transport directly benefits
rists was greatest for those living in households without cars at
a larger proportion of the U.S. population than suggested by tradi-
all four thresholds. Moreover, the share of multimodal moto-
tional trip-based analysis.
rists declined with higher car ownership levels. These findings
are consistent with prior studies that find car ownership to
have a strong influence on multimodal car use (Kuhnimhof Funding
et al., 2012b; Kuhnimhof et al., 2006; Nobis, 2007; Vij et al.,
2011). However, over time the share of multimodal motorists This research has been supported by the Mid-Atlantic University Trans-
increased for households with cars. This means that a higher portation Center (MAUTC) of U.S. DOT’s Research and Innovative Tech-
nology Administration (RITA), grant number DTRT12-G-UTC03.
share of those in households with automobiles also make trips
by foot, bicycle, or public transport.
For both years and both the week and day level the share of References
multimodal motorists is greater at higher population densities
and in areas with rail access. These findings are consistent with Block-Schachter, D. (2009). The myth of the single mode man: How the
prior studies that connect living in urban areas and near public mobility pass better meets actual travel demand (Master’s thesis).
Retrieved from http://dspace.mit.edu/handle/1721.1/51677
transport to higher levels of multimodal car use (Kuhnimhof Brons, M., Givoni, M., & Rietvield, P. (2009). Access to railway stations
et al., 2012b; Kuhnimhof et al., 2006; Nobis, 2007; Vij et al., and its potential in increasing rail use. Transportation Research Part A:
2011). Policy and Practice, 43, 136–149.
364 R. BUEHLER AND A. HAMRE

Brons, M. and Rietveld, P. (2010). The impact of travel time unreliability on Lucas, K. (2011). Driving to the breadline. In K. Lucas, E. Blumenberg, &
the choice of rail, access mode, and departure station. 12th World Con- R. Weinberger (Eds.), Auto motives: Understanding car use behaviours
ference on Transport Research, July 11–15, 2010, Lisbon, Portugal. (pp. 209–224). Bingley, England: Emerald Group.
Buehler, R., & Hamre, A. (2013). Multimodal travel behavior in the United Millard-Ball, A., & Schipper, L. (2011). Are we reaching peak travel?
States. University Park, PA: Mid Atlantic University Transportation Trends in passenger transport in eight industrialized countries. Trans-
Center (MAUTC). port Reviews, 31, 357–378.
Buehler, R., Pucher, J., Merom, D., & Bauman, A. (2011). Active travel in Newmann, P., & Kenworthy, J. (2011). Peak car use: Understanding the
Germany and the U.S. American Journal of Preventive Medicine, 41, demise of automobile dependence. World Transport Policy and Prac-
241–250. tice, 17, 31–42.
Chlond, B. (2012). Making people independent from the car: Multimodal- Nobis, C. (2007). Multimodality: Facets and causes of sustainable mobility
ity as a strategic concept to reduce CO2-emissions. In T. I. Zachariadas behavior. Transportation Research Record: Journal of the Transporta-
(Ed.), Cars and carbon: Automobiles and European climate policy in a tion Research Board, 2010, 35–44.
global context (pp. 269–293). The Netherlands: Springer. Oram, R. L., & Stark, S. (1996). Infrequent riders: One key to new transit
Clifton, K., & Muhs, C. D. (2012). Capturing and representing multimodal ridership and revenue. Transportation Research Record: Journal of the
trips in travel surveys. Transportation Research Record: Journal of the Transportation Research Board, 1521, 37–41.
Transportation Research Board, 2285, 74–83. Schlich, R., & Axhausen, K. W. (2003). Habitual travel behaviour: Evidence
Diana, M., & Mokhtarian, P. L. (2009). Desire to change one’s multimodal- from a six-week travel diary. Transportation, 30, 13–36.
ity and its relationship to the use of different transport means. Trans- Schonfelder, S., & Axhausen, K. W. (2010). Urban rhythms and travel
portation Research Part F: Traffic Psychology and Behaviour, 12, 107– behaviour: Spatial and temporal phenomena of daily travel. Surrey,
119. England: Ashgate Publishing.
Forsyth, A., Krizek, K., & Rodriguez, D. (2009). Non-motorized travel Small, K. A., & Verhoef, E. T. (2007). The economics of urban transporta-
research and contemporary planning initiatives. Progress in Planning, tion. New York, NY: Routledge.
71, 170–183. TCRP (Transit Cooperative Research Program). (2004). Report 95 (Chap-
Giuliano, G., & Narayan, D. (2004). A comparison of work and non-work ter 12): Transit pricing and fares: Traveler response to transportation
travel: The U.S. and Great Britain. In P. Rietveld & R. Stough (Eds.), system changes. Retrieved from http://onlinepubs.trb.org/onlinepubs/
Barriers to sustainable transport (pp. 111–129). London, England: E.T. tcrp/tcrp_rpt_95c12.pdf
Spoon. U.S. Census Bureau. (2010). American FactFinder. Washington, DC: U.S.
Givoni, M., & Rietveld, P. (2007). The access journal to the railway station Department of Commerce.
and its role in passengers’ satisfaction with rail travel. Transport Policy, U.S. DOC (U.S. Department of Commerce). (2011). United States Census.
14, 357–365. Washington, DC: United States Department of Commerce, U.S. Census
ITF/OECD (International Transport Forum/Organisation for Economic Bureau.
Co-operation and Development). (2012). Peak travel, peak car and the U.S. DOT (U.S. Department of Transportation). (2006). Building livable com-
future of mobility: Evidence, unresolved issues, policy implications, and munities with transit. Washington, DC: Federal Transit Administration.
a research agenda. Retrieved from http://www.internationaltransportfo U.S. DOT. (U.S. Department of Transportation). (2009). Transportation
rum.org/jtrc/DiscussionPapers/DP201213.pdf and climate change clearinghouse. Washington, DC: United States
Khattak, A., Wang, X., Son, S., & Agnello, P. (2011). Travel by university Department of Transportation.
students in Virginia: Is this travel different from travel by the general U.S. DOT. (U.S. Department of Transportation). (2010a). The National
population? Transportation Research Record: Journal of the Transporta- Bicycling and Walking Study: 15-year status report.
tion Research Board, 2255, 137–145. U.S. DOT. (U.S. Department of Transportation). (2010b). National House-
Kuhnimhof, T.G. (2009, January). Measuring and modeling multimodal hold Travel Survey 2009 (Version 2.0/2010). Washington, DC: U.S.
mode use in the longitudinal section. Paper presented at the Transporta- Department of Transportation, Federal Highway Administration.
tion Research Board 88th Annual Meeting, Washington, DC. U.S. DOT. (U.S. Department of Transportation). (2011). Summary of
Kuhnimhof, T., Armoogum, J., Buehler, R., Dargay, J., Denstadli, J., & travel trends: 2009 National Household Travel Survey. Washington,
Yamamoto, T. (2012a). Men shape a downward trend in car use among DC: United States Department of Transportation, Federal Highway
young adults – Evidence from six industrialized countries. Transport Administration.
Reviews, 32, 761–779. Vij, A., Carrel, A., & Walker, J. L. (2011). Capturing modality styles using
Kuhnimhof, T., Buehler, R., Wirtz, M., & Kalinowska, D. (2012b). Travel behavioral mixture models and longitudinal data. Retrieved from
trends among young adults in Germany: Increasing multimodality and http://www.ivt.ethz.ch/vpl/research/mobidrive/vij_carrel_wal
declining car use for men. Journal of Transport Geography, 24, 443– ker_2011_capturing_modality_styles.pdf
450. Wang, X., Khattak, A., & Sanghoon, S. (2012). What can be learned from
Kuhnimhof, T., Chlond, B., & von der Ruhren, S. (2006). Users of trans- analyzing university student travel demand. Transportation Research
port modes and multimodal travel behavior: Steps toward understand- Record: Journal of the Transportation Research Board, 2322, 129–137.
ing travelers’ options and choices. Transportation Research Record: Weiner, E. (2013). Urban transportation planning in the United States: His-
Journal of the Transportation Research Board, 1985, 40–48. tory, policy, and practice (4th ed.). New York, NY: Springer.
Copyright of International Journal of Sustainable Transportation is the property of Taylor &
Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

You might also like