THE EFFECTS OF GENDER ON COMMUTER BEHAVIOR CHANGES
IN THE CONTEXT OF A MAJOR FREEWAY RECONSTRUCTION
by
Patricia L. Mokhtarian
Department of Civil and Environmental Engineering
and Institute of Transportation Studies
University of California, Davis
One Shields Avenue
Davis, CA, 95616, USA
Phone: 1 530 752 7062
e- mail: plmokhtarian@ ucdavis. edu
Liang Ye
PhD Candidate in Transportation System Planning
School of Transportation Engineering, Tongji University
4800 Cao’an road, Shanghai, 201804, P. R. CHINA
e- mail: yel1231@ gmail. com
Visiting Researcher at the Institute of Transportation Studies
University of California, Davis
One Shields Avenue
Davis, CA, 95616, USA
Phone: 1 530 400 1952
Fax: 1 530 752 6572
e- mail: lye@ ucdavis. edu
and
Meiping Yun
School of Transportation Engineering, Tongji University
4800 Cao’an Road, Shanghai, 201804, P. R. CHINA
email: yunmp@ tongji. edu. cn
Visiting Researcher at the Institute of Transportation Studies
University of California, Davis
One Shields Avenue
Davis, CA, 95616, USA
Phone: 1 530 554 6600
e- mail: myun@ ucdavis. edu
April 2010
Presented at the 4th International Conference on
Research on Women’s Issues in Transportation, October 2009
Posted with permission of the Transportation Research Board of the National Academies
2
ABSTRACT
To study the commuter travel behavior impacts of a nine- week reconstruction of Interstate 5 ( I- 5)
in downtown Sacramento, California, a series of three internet- based surveys was conducted.
This paper offers a preliminary analysis of the first two of those surveys, focusing on the role of
gender in commuters’ responses. Avoiding rush hour and changing route were the most common
responses, and women were more likely than men to employ them. Among the changes that
reduce vehicle- miles traveled, increasing transit use and increasing telecommuting were the most
common. Overall, women were 21% more likely to make at least one change than men were. A
binary logit model of the choice to increase transit use suggests that persuading current transit
users to increase their transit use was easier than convincing nonusers to switch. Respondents
who heard about the increased level of transit service were more likely to increase transit use.
Employer transit subsidies supported increases in transit use ( but only for women), while
variable work hours ( for women) discouraged them. Men in managerial/ administration and
women in larger households were also more likely to increase their transit use.
Keywords: Gender differences, Behavior change, Mode choice, Transit use, Freeway
reconstruction, Transportation Demand Management
1 INTRODUCTION
In the nine weeks between May 30 and July 31, 2008, a one- mile stretch of Interstate 5 ( I- 5)
in downtown Sacramento, California was intermittently closed for a reconstruction project (“ the
Fix I- 5 project”, or “ the Fix”). This portion of I- 5 is part of a major north- south conduit for
interregional traffic, as well as a key commute route serving downtown Sacramento ( the state
capital) and other job locations in the region. To mitigate the impact of the Fix, a number of
strategies were implemented by Caltrans and other public agencies, including providing extensive
information on the Fix and commute alternatives, increasing transit service, and offering
reduced- rate parking at some facilities. Two weeks before the Fix ( May 16), Gov. Schwarzenegger
issued Executive Order S- 04- 08 ( http:// gov. ca. gov/ index. php?/ press- release/ 9631/, accessed
August 31, 2009) directing State executive agencies to promote commute alternatives for their
employees to the fullest extent possible, and encouraging other public agencies and private
companies to do the same. The Executive Order also authorized a study of the impacts of the Fix,
with a view to evaluating the effectiveness of the commute modification strategies being
promoted, and the extent to which more sustainable commute patterns would be the longer- term
result. As the interstate highway system and other elements of the transportation infrastructure
continue to age, reconstruction projects like Fix I- 5 will occur quite frequently. Therefore, it is
important to learn as much as possible about commuter reactions to such projects, to enable
future projects to be implemented in the most effective way. At the same time, such projects
offer valuable opportunities to disrupt habitual behavior in a natural way, and to use that
disruption to motivate shifts toward more sustainable commute choices ( Gardner, 2009).
For both of these purposes, it is of interest to examine whether there are any significant
differences by gender. Do women, with generally more complex activity schedules ( Bianco
and Lawson, 1996; Hjorthol, 2004), experience adverse impacts of such reconstruction projects
more heavily than men? Are they less likely than men to make changes, and if so is it because
3
they have more constraints, or because they have already built in more flexible options such as
flextime or telecommuting that allow them to absorb disruptions more robustly? Or are women
more likely than men to make changes, and if so, is it because of experiencing more adverse
impacts, a generally greater receptivity to change ( Dobbs, 2007), or a greater environmental
sensitivity ( Matthies et al., 2002) giving them a stronger internal motivation to change? The
present study can provide some preliminary insights into these questions.
A series of three internet- based surveys was conducted to evaluate the impacts of the Fix
on commuter behavior. The present paper offers an initial analysis of the first two of those
surveys. Specifically, we address the active choices made by commuters to cope with the closure
( e. g. taking vacation, telecommuting, or changing mode), and how those choices may differ by
gender. In addition to presenting descriptive statistics, we develop a binary logit model of the
most frequently adopted active choice that reduces vehicle- miles traveled ( VMT), namely the
choice to increase the use of transit during the Fix.
The remainder of the paper is organized as follows. The next section briefly reviews
previous related research, and the following section describes the data collection. Subsequently,
we discuss descriptive statistics for baseline work schedule/ commute characteristics, and the
active choices made by commuters. The binary logit model is presented in the penultimate
section. The final section summarizes the study and suggests future research directions.
2 LITERATURE REVIEW
Numerous studies have addressed gender differences in various indicators of travel behavior, and
likely reasons for those differences. Some key findings ( especially in, though not limited to, the
U. S. context) include the following:
 Women continue to exhibit shorter commute times and distances than men ( Crane, 2007).
 Nationwide, women are ( still) more than men likely to commute by transit ( Pisarski, 2006),
though their transit use has declined over the past 20 years and the picture is more complex
when controlling for race ( Crane, 2007). In a 1994 stated preference survey of suburban
commuters in Montreal, Canada, Patterson et al. ( 2005) found that women were less likely
to choose public transit than men, suggesting that their higher revealed preference shares
often represent constrained choices.
 Women do more child- chauffeuring and make more household- serving trips ( such as
grocery shopping) than men ( Mauch and Taylor, 1997, using trip diary data from a 1990
survey of San Francisco Bay Area residents, with similar results found in Norway by
Hjorthol, 2004).
 Women are more concerned about safety ( Liss et al., 2005), and more likely to take safety
considerations into account in making travel choices ( Church et al., 2000; Kenyon et al.,
2002).
Although women’s travel patterns have changed dramatically over the past few decades, Crane
( 2007) found persistent gender differences in a nationwide ( US) panel study covering the period
1985- 2005. One Norwegian study ( Hjorthol, 2004) similarly found that gender differences did not
diminish during the 1990s, noting that women’s travel distances were ( still) more local than men’s,
men had more work trips, and women had more non- work trips.
In view of these and other differences, decision- makers have been advised to consider the
distinctive travel needs of women and men during transportation planning and operations ( Kerkin,
4
1995; Lu and Pas, 1999). Tranter ( 1994) pointed out that gender- based role differences influence
not only daily travel patterns but economic and social status which in turn affects the choices
available for women which consequently affect mobility. The Women's Planning Network Inc.
( 1995) also indicated that there is an absence of transportation policies directed at women's
transport needs. Law ( 1999) suggested the need for further research on “ gender and daily
mobility”, incorporating it within a larger theoretical project.
Interestingly, we were unable to find any studies specifically focusing on the subject of this
paper: gender- based effects associated with the execution of a planned freeway reconstruction.
However, several ( linked) literatures are generally relevant. Many or most strategies associated
with mitigating the disruption of a freeway reconstruction fall into the category of transportation
demand management ( TDM) policies, and a sizable literature exists on the adoption and
effectiveness of TDM measures. Although TDM policy seldom addresses gender differences in
travel behavior, within that field some gender- based observations have been made. For example,
Bianco and Lawson ( 1996) suggested that TDM strategies may be more burdensome on women
than on men, because women are more likely to be in the lower income strata. Other analysts also
argue that the numerous policies aimed at discouraging car use may run counter to the
transportation needs of women, particularly where complex, multi- destination trip making is
required ( Women's Planning Network Inc., 1995).
The disruption occasioned by a freeway reconstruction may be severe enough to trigger the
reconsideration of habitual behavior, and thus another literature relevant to this study deals with
voluntary behavior change, and/ or habit disruption. Research on voluntary behavior change is
often directed to reducing the use of private motor vehicles in urban areas ( Taylor et al., 2003; Fujii
and Taniguchi, 2005; Rose and Marfurt, 2007). In one study that investigated gender differences in
travel mode changes, Rose and Marfurt ( 2007) found ( using 2003/ 2004 Australia “ Ride- to Work”
annual event travel survey data) that female first- time transit riders were more likely ( 30%) than
men ( 22%) to continue riding transit five months after the event. Finally, a third relevant literature
deals with environmental concern and related behaviors. For example, Matthies et al. ( 2002),
using a survey of 187 inhabitants of a German city, found that women were more likely than men
to report an intention to reduce their car use, due to their stronger ecological norms and weaker car
habits.
3 DATA COLLECTION
Since the study was conducted just a few weeks before the closure, there was not time to
draw a rigorous, geographically- based random sample and recruit participants by mail. Instead,
respondents were recruited via e- mail invitations, disseminated through numerous State agencies,
the Fix I- 5 listserv, Transportation Management Agencies, and a press release from the University
of California, Davis. Given the ad hoc nature of the recruitment process, we can neither compute a
response rate, nor expect a completely representative sample. It is likely that State workers are
overrepresented, as well as internet- literate workers, workers with burdensome commutes, those
heavily affected by the Fix, and those who made environmentally- beneficial behavior changes.
However, self- employed and small- business workers are probably underrepresented, some of
whom ( together with the less computer- literate) may also have been heavily affected by the Fix, so
biases in the sample are counteracting each other to an unknown extent.
Accordingly, the subsequent descriptive statistics should be viewed with caution.
Nevertheless, we suggest that some data are better than no data, and we also believe that
5
comparisons in the data should be largely robust ( particularly the comparison of men and women).
Thus, we encourage the reader to focus on trends, conditional relationships, and the model ( which
is a specific type of conditional relationship): in these cases, it is less critical to have a sample that
is representative of the total population ( Babbie, 1998).
Two internet- based surveys were administered during the period of the Fix: the first,
hereafter referred to as the Wave 1 survey, sought information on behavior during the first closure
of the freeway, of all lanes in the northbound direction ( Monday, June 2 – Sunday, June 8, 2008, for
our purposes). The second survey ( Wave 2) addressed the first closure of the freeway in the
southbound direction ( June 16- 22). Another paper ( Ye et al., 2009) includes selected comparisons
between the two waves, but the present paper will focus on comparisons by gender, without
regard to wave. To avoid the complication of having a sizable fraction ( about a third) of the
sample not being independent across waves, in this analysis we have retained all of the Wave 1
cases ( 4520 after screening for completeness), and among the Wave 2 cases, only those who did
not complete Wave 1 ( 2414), for a pooled sample of 6934 cases. The pooled sample contains
2247 ( 32%) men, 4160 ( 60%) women, and 527 ( 8%) respondents who declined to state their
gender. The sizable proportion of women in the sample is not as unbalanced as it may seem, in
view of the preponderance of State employees in our study population, and the fact that in
general some 60% of state and local government workers are women ( Caldwell, 2009).
TABLE 1 presents the sample statistics for some selected characteristics. The “ average”
characteristics are: age 47, female, college graduate, in a household with 2.7 members, 2.1 cars,
and having an annual income of $ 75,000 - $ 99,999. Women tend to have lower ( household)
incomes than men, at least in part because on average they have less education, are slightly
younger, are more likely to hold clerical positions, and are more likely to work part- time. They
also tend to have smaller households ( hence, generally fewer wage- earners). However, number
of vehicles, and commute distances and times are relatively similar across gender.
We applied three filters to the data at various points in the analysis. ( 1) We excluded
from the remainder of the analysis 344 respondents ( 118 men, 210 women, 16 unknown) who
were out of the region for the entire week of the closure. ( 2) From the analysis of commuters’
active choices ( Section 5), we additionally excluded 28 respondents ( 9 men, 19 women) who did
not commute to work during the closure week. ( 3) Finally, from the model of the choice to
increase transit use ( Section 6) we further excluded 1123 respondents ( 406 men, 646 women)
who already used transit for all or almost all of their commuting trips. The remaining sample for
the “ increase transit use” model is 5439 cases ( 1714 men, 3285 women).
6
TABLE 1 Selected Characteristics of the Sample, by Gendera
Pooled data Men Women
Characteristic ( sample sizes)
N (%) N (%) N (%)
Number of cases 6934 ( 100.0) 2247 ( 32.4) 4160 ( 60.0)
* Average age ( years) ( 6053, 2067, 3965) 46.5 47.3 46.1
* Average educational level b ( 6291, 2192, 4038) 4.06 4.3 3.9
* Annual household income ( 5977, 2112, 3817)
Less than $ 15,000 29 ( 0.5) 14 ( 0.7) 15 ( 0.4)
$ 15,000-$ 29,999 106 ( 1.8) 25 ( 1.2) 80 ( 2.1)
$ 30,000-$ 44,999 493 ( 8.2) 112 ( 5.3) 374 ( 9.8)
$ 45,000-$ 59,999 715 ( 12.0) 176 ( 8.3) 535 ( 14.0)
$ 60,000-$ 74,999 1070 ( 17.9) 312 ( 14.8) 752 ( 19.7)
$ 75,000-$ 99,999 1243 ( 20.8) 490 ( 23.2) 740 ( 19.4)
$ 100,000 or more 2321 ( 38.8) 983 ( 46.5) 1321 ( 34.6)
* Average household size ( 6328, 2184, 4092) 2.72 2.83 2.67
Driver’s license possession ( 6387, 2215, 4109) 6333 ( 99.2) 2200 ( 99.3) 4071 ( 99.1)
* Average number of household operational vehicles
( 5722, 1991, 3683) 2.08 2.17 2.03
* Distance from home to the nearest bus stop or light- rail
station ( 6356, 2210, 4081)
Less than a 5- minute walk 1680 ( 26.4) 676 ( 30.6) 992 ( 24.3)
5 - 10- minute walk 1729 ( 27.2) 592 ( 26.8) 1123 ( 27.5)
10 – 20- minute walk 1042 ( 16.4) 352 ( 15.9) 674 ( 16.5)
More than a 20- minute walk 1501 ( 23.6) 482 ( 21.8) 1002 ( 24.6)
Don't know 404 ( 6.4) 108 ( 4.9) 290 ( 7.1)
* Average commute minutes ( 6878, 2231, 4134) 31.9 32.8 31.4
* Average commute miles ( 6872, 2231, 4130) 17.8 18.4 17.4
* Primary job work schedules ( 6920, 2245, 4152)
Part time c 388 ( 5.6) 84 ( 3.7) 274 ( 6.6)
Conventional d 3221 ( 46.5) 1065 ( 47.4) 1899 ( 45.7)
Variable e 1461 ( 21.1) 509 ( 22.7) 858 ( 20.7)
Compressed 9- 80 work week f 1504 ( 21.7) 470 ( 20.9) 921 ( 22.2)
Compressed 4- 40 work week g 261 ( 3.8) 82 ( 3.7) 158 ( 3.8)
Other 85 ( 1.2) 35 ( 1.6) 42 ( 1.0)
* Occupation ( 6914, 2242, 4150)
Manager/ administration 1591 ( 23.0) 541 ( 24.1) 929 ( 22.4)
Professional/ technical 3929 ( 56.8) 1432 ( 63.9) 2211 ( 53.3)
Services/ repair 47 ( 0.7) 29 ( 1.3) 13 ( 0.3)
Clerical/ administrative support 1184 ( 17.1) 175 ( 7.8) 916 ( 22.1)
Sales/ marketing 71 ( 1.0) 20 ( 0.9) 43 ( 1.0)
Production / construction/ crafts 42 ( 0.6) 27 ( 1.2) 12 ( 0.3)
Other 50 ( 0.7) 18 ( 0.8) 26 ( 0.6)
* Number using transit as primary commute mode
( 6595, 2160, 3945) 1183 ( 17.9) 484 ( 21.5) 609 ( 14.6)
Number currently using transit but not primary commute
mode ( 6595, 2160, 3954) 1098 ( 16.6) 368 ( 16.4) 654 ( 15.7)
Awareness of Fix strategy that increase number of buses
( 6544, 2228, 4116) 3798 ( 58.0) 1291 ( 57.9) 2397 ( 58.2)
7
Employer- provided reduced- rate transit passes
( 6575, 2236, 4139) 4122 ( 62.7) 1383 ( 61.9) 2636 ( 63.7)
Employer- provided variable start/ end times
( 6575, 2236, 4139) 4109 ( 62.5) 1376 ( 61.5) 2617 ( 63.2)
a Traits that differ significantly by gender ( at the 0.05 level or better) are marked with a “*” in front.
b 1= Some grade or high school; 2= High school graduate; 3= Some college or technical school; 4= Four- year college,
university, or technical school graduate; 5= Some graduate school; 6= Completed graduate degree( s).
c Less than 35 hours per week.
d 7 ½ - 8 hours per day, with a start time between 8: 00 and 9: 00 a. m.
e 7 ½ - 8 hours per day, with a variable start time.
f 9- hour days, for nine days in two weeks.
g 10- hour days, for four days each week.
4 BASELINE WORK SCHEDULE/ COMMUTE CHARACTERISTICS
Information on respondents’ baseline work schedule/ commute patterns was collected through a
series of questions focusing on a “ typical 28- day ( 4- week) period”, “ before Fix I- 5 began”. We
asked for the number of days out of 28 on which respondents:
 worked at home as the regular location of their job;
 worked at home instead of commuting to their regular workplace; and
 physically traveled to a regular workplace outside of their home.
Those with a non- zero answer to the last question were asked how many days out of that
number they
 drove alone for most of the commute;
 carpooled or vanpooled for most of the commute;
 rode a bus for any portion of the commute;
 rode light rail for any portion of the commute;
 rode Amtrak ( commuter train) for any portion of the commute;
 walked for the entire commute; and
 rode a bicycle for any portion of the commute.
Thus, only the drive alone, car/ vanpool, and walk modes are mutually exclusive, and walking ( for
the entire commute) in principle excludes any other mode. The transit and bicycle modes can occur
in any combination with each other or with the car modes. Accordingly, the sum of the
mode- specific answers can exceed the total number of commute days for a given respondent.
We analyzed the sample’s baseline participation in each of these commute options,
together with compressed work schedules since they also reduce the number of vehicle
commute- miles traveled ( for more details, see Mokhtarian et al., in progress). Engagement ( at
least one day out of 28) in compressed work schedules ( 26%), both types of working at home ( 5%
as regular workplace; 9% telecommuting), and physical commuting ( 97%) is relatively similar
across gender; the main differences appear with respect to mode choice. Women are considerably
more likely than men to drive alone ( 75% versus 68%). Women are also slightly more likely to
car/ vanpool. But ( in contrast to Pisarski, 2006) they are substantially less likely to use transit ( 30%
versus 34%), as well as less likely to bicycle ( 6% versus 18%) or walk ( 2% versus 3%).
Computing the average share of physical commute days on which each mode is used –
the measure closest to a true commute mode split available in our sample – reveals that the drive
8
alone share is relatively small at 55%, with car/ vanpooling at 18%, transit around 25%, and
walking and biking about 7% ( combined). These shares are quite different from those in the region
as a whole, which are 82%, 10%, 3%, and 5%, respectively ( SACOG, 2008). However, the latter
set of shares is for all passenger trips during the weekday peak, whereas ours are for commuting
only ( and disproportionately downtown- oriented), but without regard to time of day.
5 COMMUTE- RELATED ACTIVE CHOICES
The literature identifies a number of possible active choices that can be made in these
circumstances ( Giuliano, 1992; Möser and Bamberg, 2007, Shiftan and Suhrbier, 2002). In the
subsections below, we respectively discuss making fewer commute trips ( via telecommuting,
alternative work schedules, and/ or vacation), changing mode, and changing departure time or
route. An analysis of teleconferencing and non- work- related trips can be found in Mokhtarian et
al. ( in progress). As discussed above, this portion of the analysis excludes respondents who did not
commute at all during the closure week.
5.1 Making Fewer Commute Trips
Respondents were asked whether, during the closure week, they made “ fewer commute trips than
you normally would”, with response options “ Yes, because of Fix I- 5”, “ Yes, for some other
reason”, and “ No”. We discuss only the changes made because of the Fix. Some 14.1% of
respondents made fewer commute trips ( because of the Fix) during the closure week ( see TABLE
2). We estimate ( see Mokhtarian et al., in progress for details) that increased telecommuting,
compressed work schedules, and vacation days were adopted by about 5.6%, 3.1%, and 3.1% of
the sample, respectively.
Women were significantly more likely than men to reduce their commute travel because
of the Fix ( 15% versus 11%); interestingly, however, they were about 16% less likely than men
to do so by compressing their work schedules ( in contrast to Mokhtarian, et al., 1997). Instead,
women were more likely than men to reduce their commute travel by taking some vacation days
( 28% versus 23%).
5.2 Changing Modes
Respondents were asked whether, during the closure week, they traveled “ to or from work using
a different means of transportation than you normally would”, with the same response options as
in Section 5.1. Again, we analyze only the changes made because of the Fix. Respondents
reporting such changes were then asked which mode( s) they used on more occasions than normal,
and what they would have ordinarily done on those days ( multiple answers were possible, for
both questions).
The vast majority ( at least 93%) of those making commute mode changes increased their
use of more sustainable modes, most often at the expense of driving alone. Increasing transit use
was the most common change: 64.9% of those who altered their commute mode patterns made
that choice ( 5.3% of the total eligible respondents), among which 70.1% would otherwise have
driven alone to work. About 2.5% of eligible respondents increased their walking or biking,
and 1.4% increased ridesharing; only 0.6% increased driving alone.
There are substantial differences in these patterns by gender, however. Taking all modes
together, women were 16% more likely to change modes ( 8.1%) than men ( 7.0%). Among
9
TABLE 2 Commuters’ Work- Trip- Related Active Choices, by Gendera
a Traits that differ significantly by gender ( at the 0.05 level or better) are marked with a “*” in front.
b The survey directly asked about changes to commute trips and to non- work trips, but not about changes to ( non- commute) work trips ( such as business meetings). The
responses reported here were obtained from write- in descriptions to the “ other ( please specify)” option on the commute reduction question described in footnote c. As such,
they certainly understate the actual level of changes made to non- commute work trips in the sample, and are included here only for completeness.
c These responses were also write- ins to the “ other ( please specify)” option on the commute reduction question.
d During the week of [ June 2- 8 or June 16- 22] ( Mon- Sun)… … , how did you avoid the commute trips that you would have otherwise made? ( Options: Telecommuted to work;
Adopted a compressed work week; Took vacation time; Other). Categories are not mutually exclusive.
e This unexpectedly high average suggests that a number of respondents failed to read the word “ avoid” in the question (“ how many days did you avoid physically travelling to
and from work by …”), and responded with the number of days they commuted to work. The same could be true of the telecommuting and vacation strategies, as well.
f Cases which reported they made fewer commute trips but did not indicate how they avoided the commute trips.
Number making these choices ( sample
sizes)
Pooled data
N (%)
Men
N (%)
Women
N (%)
Average number of times ( by those who made
this choice)
Pooled
data
Men
Women
* Made fewer commute trips
( 6552, 2118, 3925)
923 ( 14.1) 240 ( 11.3) 578 ( 14.8)
Made fewer work- related trips
( 6552, 2118, 3925) b 3 ( 0.1) 1 ( 0.0) 2 ( 0.1)
Made fewer trips, trip purpose not specified
( 6552, 2118, 3925) c 36 ( 0.5) 10 ( 0.5) 24 ( 0.6)
Means of avoiding the commute trips
( 627, 180, 391) d
Means of avoiding the commute trips
( 433, 125, 271)
2.63 2.56 2.59
Telecommuted 297 ( 47.4) 85 ( 47.2) 186 ( 47.6) Days avoided commuting by telecommuting
( 290, 84, 182)
2.17 2.24 2.04
Adopted compressed work week 163 ( 26.0) 52 ( 28.9) 95 ( 24.3) Days avoided commuting by compressed work
week ( 157, 51, 93)
2.92 e 2.63 e 3.01 e
Took some vacation 165 ( 26.3) 42 ( 23.3) 109 ( 27.9) Days avoided commuting by taking vacation
( 99, 26, 63)
1.80 1.73 1.83
Other 2 ( 0.3) 1 ( 0.6) 1 ( 0.3)
Missing responses for means of avoiding
the commute trips f ( 923, 240, 578)
315 ( 34.1) 75 ( 31.3) 192 ( 33.2)
Changed mode ( 6356, 2073, 3837) 493 ( 7.8) 148 ( 7.0) 318 ( 8.1)
* Avoided rush hour ( 5506, 1745, 3428) 2642 ( 48.0) 751 ( 43.0) 1725 ( 50.3) Days avoided rush hour 3.90 3.87 3.91
* Changed route ( 5254, 1650, 3282) 2371 ( 45.1) 681 ( 41.3) 1540 ( 46.9) Days changed route 3.44 3.40 3.45
10
changers, women were considerably more likely than men to increase their use of transit ( 73%
versus 58%) and car/ vanpooling ( 20% versus 12%), and much less likely to increase bicycling or
walking ( 21% versus 49%).
5.3 Changing Departure Time or Route
The active choices discussed so far are considered to be potentially somewhat costly, in that they
may require lifestyle adjustments and/ or affect other people ( Salomon and Mokhtarian, 1997). By
contrast, the remaining changes shown in TABLE 2 are considered relatively low- cost, entailing
minimal change to established patterns. Accordingly, a number of studies have found such changes
to be the most common responses to increasing congestion or disrupted commute patterns
( Mokhtarian et al., 1997), and the present analysis is no exception. Nearly half the respondents
reported making a special effort to avoid rush hour during the closure week ( on an average of four
days out of the week), and 44% of those driving/ carpooling to work that week reported making
planned changes to their commute route ( on an average of 3.4 days). Women are considerably
more likely than men to employ each of those strategies, consistent with the findings of
Mokhtarian et al. ( 1997) for changing departure time.
Altogether, about 60% of respondents adopted at least one of the active changes to the
commute that have been analyzed in Section 5. However, women were 21% more likely to make
at least one change than men were ( 64% versus 53%). At least two explanations for this greater
activism on the part of women present themselves. The first is that women were more adversely
impacted by the Fix, and thus had a greater need to make a change. Mokhtarian et al. ( in progress)
found that women were more likely than men to report passive impacts, both good and bad, due
to the Fix. The greater reported incidence of negative impacts on women lends credence to the
idea that they had a greater external motivation to make changes. The second explanation is that
women may have a greater internal motivation to make changes, i. e. that they may be more
inclined to make socially/ environmentally beneficial changes to their commute in general. This
interpretation is consistent with the independent findings of Matthies et al. ( 2002) and Rose and
Marfurt ( 2007) discussed in Section 2.
6 THE “ INCREASE TRANSIT USE” MODEL
Although it is helpful to see descriptive statistics on how commonly various changes were made,
it is also desirable to better understand the types of people who make a given change. This is
perhaps best done in the context of a model, in which multiple covariates can be controlled for
simultaneously. Space does not permit providing models for all the changes we see in our sample,
but we present a model for the most commonly- chosen change that reduces vehicle travel,
namely the choice to increase the use of transit for commuting.
The dependent variable is created from the survey question which asks ( of those who
previously indicated “ travel[ ing] to or from work using a different means of transportation
[ during the closure week] than you normally would”, because of the Fix), “ During the week
of … which did you use on more occasions than you normally would have?” with seven possible
response options: “ Carpool or vanpool”, “ Bus”, “ Light rail”, “ Amtrak train”, “ Walking or
biking”, “ Driving alone”, and “ None of the above”. In our model, we combine “ Bus”, “ Light
rail” and “ Amtrak train” into a single “ transit” variable, equal to 1 if respondents selected any of
those modes, and 0 otherwise. Among the “ 0” respondents we include those who did not change
modes because of the Fix, based on the previous question, but exclude those who ( 1) were out of
11
the region the entire week, ( 2) did not travel to work any days that week, or ( 3) already take
transit for all or almost all of their commute trips, since in none of those cases would increasing
transit use be a feasible option. This left 5439 cases, including 249 ( 4.6%) who increased their
transit use due to the Fix. A number of potential explanatory variables are available, including
sociodemographic traits and the availability of various employer- based commute modification
instruments.
To identify gender- related differences in the choice to increase transit use, we first
developed best models for each gender separately, and for the pooled sample. Based on those
models, we created a fully gender- specific single model ( i. e. a single equation in which every
variable was interacted with gender, so that coefficients of each variable could differ by gender).
Then, as indicated by statistical tests and conceptual considerations, we deleted insignificant
variables and constrained coefficients to be equal across genders, resulting in a final hybrid model
in which some coefficients are gender- specific while others are based on the pooled sample. The
final sample necessarily excludes cases where gender was not reported.
Due to missing data on gender and other variables, our preferred model ( TABLE 3) has 214
( 4.6%) respondents who increased their transit use during the closure week and 4422 who did not.
Overall, as mentioned in Section 5.2, women were considerably ( 45%) more likely than men to
increase their use of transit ( 6.1% versus 4.2% in the final estimation sample), perhaps because,
using it less than men to start with ( Section 4), they had more room to do so.
The  2 goodness- of- fit measure ( Ben- Akiva and Lerman, 1985), with the equally- likely
model as base, is 0.766, which, taken at face value, is considered quite good in the context of
disaggregate discrete choice models. With shares this unbalanced, however, the market- share
model alone ( the model containing just the constant term) has a  2 value of 0.730, initially
suggesting that the true explanatory variables only add 0.036 to the goodness of fit. However, the
final model is significantly better than the market- share model (  2 = 232 with 7 d. f., p= 0.000).
Further, when we re- estimate the same model except without the constant term ( not shown), we
find a  2 value of 0.499, which indicates that most ( 65%) of the explanatory power of the full
model lies in the “ true” variables. That is, the true variables are substantively helping to explain
why the shares are so unbalanced.
Seven variables besides the constant are retained in the model: two mode usage variables,
one awareness variable, two employer strategies ( gender- specific), and two ( gender- specific)
sociodemographic variables. We discuss each of these in turn.
We hypothesized that people who ( had) already used transit to some extent would be
more likely to increase their use of it than others would be to start using it. Accordingly, we
tested several indicators of transit use. Two dummy variables, marking those who use transit as
their primary commute mode and those who currently use transit but do not have it as their
primary commute mode, are strongly significant and positive, as expected. Our interpretation is
that people who currently use transit are familiar with the schedule, stop/ station locations, and
riding experience, and would therefore find it easier to step up their use. However, those who
already use transit as their primary commute mode have less room to increase their use than
respondents who currently use transit but do not have it as their primary commute mode. By
contrast, nonusers may have already firmly concluded that transit is not suitable for their needs,
whether based on erroneous impressions ( e. g. an overestimation of travel time by transit, as
found by Fujii et al., 2001 and van Exel and Rietveld, 2010), or accurate ones.
12
TABLE 3 Binary Logit Model of Increased Transit Use, Pooled Data ( 1= Increased transit
use, 0= Did not increase transit use)
Variable Coefficient P- value
Constant - 4.302 0.000
Sociodemographics
Household size ( female) 0.120 0.019
Manager/ administration occupation ( male) 0.644 0.022
Transit usage
Use transit as primary mode 1.350 0.000
Currently use transit but not primary mode 2.229 0.000
Awareness of Fix impact mitigation strategies
Increased number of buses 0.378 0.027
Employer- provided commute strategies
Reduced- rate transit passes ( female) 0.569 0.002
Variable start/ end times ( female) - 0.434 0.013
Valid number of cases, N 4636
( yes: 214; no: 4422)
Final log- likelihood, LL(  ) - 751.21
Log- likelihood for market share model, LL( MS) - 867.168
Log- likelihood for equally- likely ( EL) model, LL( 0) - 3213.43
No. of explanatory variables, K ( including constant) 8
2 1 LL( ) / LL( 0) ELbase     0.766
Adjusted 2 1 [ LL( ) K] / LL( 0) ELbase      0.764
2 1 LL( ) / LL( MS) MSbase     0.134
2 1 LL( MS) / LL( 0) MS    0.730
 2 ( between the final model and the EL model) 4924.44
 2 ( between the final model and the MS model) 231.916
It is valuable to realize that it may be more effective to try to persuade current transit
users to increase their use than to try to convince nonusers to switch – although we recognize
that both potential markets are important. On the other hand, it could also be argued that
including user history variables are not as insightful as identifying the “ first principles” that
influence whether one does currently use transit or not, and including those in the model instead.
Thus, we also tested excluding these two variables from the model ( results not shown). In the
best alternative model, the only difference is that the female- specific household size variable also
dropped out, and the  2 value fell to 0.73. This indicates that the transit experience variables are
not dominating the explanatory power of the model.
One awareness variable is significant in our model, with the expected sign: it is
unsurprising that respondents who had heard about the increased number of buses are more
likely to increase their transit use.
Two employer strategies are significant in our preferred model, with expected signs. Not
surprisingly, the availability of reduced- rate transit passes substantially increases the probability of
using transit more, but it is interesting that this is only the case for women, while the availability of
reduced fares has no significant effect on men. It is possible that men’s higher incomes make them
13
less sensitive to transit costs. Similarly, the availability of flextime decreases the propensity to
increase transit use, but again only for women. The interpretation of the latter result is that if
people can change their departure times to avoid congestion caused by the Fix, then their current
commute mode may remain viable and they have less incentive to switch to transit. This may be
true for women but not for men for the same reasons leading fewer women than men to use transit
in normal times.
Finally, two gender- specific sociodemographic traits are also significant. Men in
managerial/ administrative occupations are substantially more likely to increase their use of
transit than are other groups. It is possible that men in high- end white- collar occupation are more
likely to live in locations better- served by commuter train, light rail, and express bus service. The
positive coefficient of household size ( for women only) was initially unexpected, but is also
saying something meaningful. We assume that people in larger households have more complex
activity- travel patterns, and therefore our original expectation was that commuters in such
households would be more likely to find the car to be most practical, and more likely to stick
with the car. However, further reflection and comments on open- ended questions of the survey
suggest another possibility. It may be that the more complex patterns of larger households are
more vulnerable to disruption, and have less ability to absorb a disruption without much change.
Thus, commuters in larger households may have a greater need to change than those in single- or
two- adult households, while at the same time; there might be more limitations on the ability of
such commuters to choose other actions such as changing route or departure time. The fact that
the variable is significant only for women suggests that it may also reflect an income effect, as
women in smaller households are more likely to be single mothers.
Besides the seven variables just discussed, we tested a number of other variables,
including work schedule type, commute time and distance, the distance of the nearest bus stop or
light rail station from home, income, number of vehicles per household member and per licensed
driver, and a geographically- based indicator of how strongly the respondent’s commute might
have been affected by the Fix. However, none of these variables were significant in the final
model.
7 SUMMARY AND SUGGESTIONS FOR FUTURE RESEARCH
This study offers a preliminary analysis of commuters’ responses to the Fix I- 5
reconstruction project, focusing on gender differences. The easiest options ( such as avoiding rush
hour, adopted by 48%, and changing route, 45%) were the most common responses, and women
were considerably more likely than men to employ each of those strategies. About 8% of eligible
respondents altered their commute mode choices during the closure week. The vast majority
( 93%) of those increased their use of more sustainable modes, most often at the expense of
driving alone. Increasing transit use was the most common change ( made by 5.3% of the eligible
respondents), and there were substantial differences in mode change patterns by gender ( with
women considerably more likely than men to increase their use of transit and car/ vanpooling).
Overall, women were more likely to make at least one change ( 64%) than men were ( 53%).
A binary logit model was built to better understand the choice to increase transit use ( the
most popular VMT- reducing option) during the Fix. Two mode usage variables, one awareness
variable, two employer strategies ( gender- specific), and two ( gender- specific) sociodemographic
variables were significant. Being a current transit commuter positively influenced the propensity
to increase transit use, suggesting that persuading current transit users to increase their use may
be easier than convincing nonusers to switch to transit. The availability of reduced- rate transit
14
passes and flextime decreased the propensity to increase transit use, but only for women. The
latter result in particular illustrates the conflict that sometimes arises among policy instruments:
making work hours more flexible is considered beneficial for reducing congestion and for
balancing work and family needs, but the increased flexibility in avoiding peak- period
congestion may enhance the appeal of continuing to commute by driving alone. Finally, men in
managerial/ administrative occupations, and women in larger households, were more likely to
increase their use of transit. We interpret the latter result to reflect a greater vulnerability to
disruption of the complex activity patterns in larger households, with the result that commuters
in such households have a stronger impetus to make a change when disruption occurs.
Several directions for future research are indicated. Using the same data set, we can
model not only other behavioral changes beyond the one explored here ( increased transit use),
but also the reported likelihood of continuing to use a strategy adopted ( or increased) during the
Fix. With the addition of geocoded home and work location information ( for those who reported
nearby street intersections), we can explore a number of geographic relationships with observed
outcomes. Ultimately, we will be merging the attitudinal and behavioral data collected in Wave 3
with Waves 1 and 2, to provide a rich basis from which to further investigate the persistence of
behavioral changes prompted by the Fix.
8 ACKNOWLEDGMENTS
Data collection was funded by a grant from the California Air Resources Board. Kristin Lovejoy,
Laura Poff, and Julia Silvis were essential to the design and administration of the survey, and in
initial data handling and analyses. Giovanni Circella also ably assisted with data handling and
other tasks. Ryan Huang, Calvin Iwan, Charmine Solla, and Aayush Thakur provided invaluable
assistance in data cleaning. The second author was supported by a fellowship under the State
Scholarship Fund and the Key National Natural Science Foundation of China, Urban
Transportation Planning Theory and Methods under the Information Environment ( No.
50738004/ E0807). The third author was supported by a fellowship under the China Scholarship
Council.
REFERENCES
Babbie, E. ( 1998) The Practice of Social Research. Wadsworth Pub. Co., Belmont, CA.
Ben- Akiva, M. and S. R. Lerman ( 1985) Discrete Choice Analysis: Theory and Application to
Travel Demand. The MIT Press, Cambridge, Massachusetts.
Bianco, M. and C. Lawson ( 1996) Trip- chaining, childcare, and personal safety: critical issues in
women’s travel behavior. Proceedings from the Second National Conference on Women’s
Travel Issues. Publication No. FHWA- PL- 97- 024, US Department of Transportation, Federal
Highway Administration, pp. 123- 143.
Caldwell, Christopher ( 2009) The pink recovery: men are paying a higher price in this recession
than women. Perhaps that’s fair. Time, August 24, p. 21.
Church, A., M. Frost, and K. Sullivan ( 2000) Transport and social exclusion in London. Transport
Policy 7, 195– 205.
Crane, R. ( 2007) Is there a quiet revolution in women’s travel? Revisiting the gender gap in
commuting. Journal of the American Planning Association 73( 3), 298- 316.
15
de Jong, Martijn G., Jan- Benedict E. M. Steenkamp, Jean- Paul Fox, and Hans Baumgartner ( 2008)
Using item response theory to measure extreme response style in marketing research: a global
investigation. Journal of Marketing Research 45 ( February), 104- 115.
Dobbs, L. ( 2007) Stuck in the slow lane: Reconceptualizing the links between gender, transport
and employment. Gender, Work & Organization 14( 2), 85- 108.
Fujii, S., T. Gärling, and R. Kitamura ( 2001) Changes in drivers' perceptions and use of public
transport during a freeway closure: Effects of temporary structural change on cooperation in a
real- life social dilemma. Environment and Behavior 33, 796- 808.
Fujii, S. and A. Taniguchi ( 2005) Reducing family car- use by providing travel advice or requesting
behavioral plans: An experimental analysis of travel feedback programs. Transportation
Research Part D 10( 5), 385- 393.
Gardner, B. ( 2009) Modelling motivation and habit in stable travel mode contexts. Transportation
Research Part F 12, 68- 76.
Giuliano, G. ( 1992) Transportation demand management. Journal of the American Planning
Association 58( 3), 327- 342.
Hjorthol R. J. ( 2004) Gender aspects of travel behavior development: a Norwegian example, Paper
presented at the Conference on Research on Women's Issues in Transportation. Chicago, IL,
November 18- 20.
Kenyon, S., G. Lyons, and J. Rafferty ( 2002) Transport and social exclusion: investigating the
possibility of promoting inclusion through virtual mobility. Journal of Transport Geography
10( 3), 207- 219.
Kerkin, K. ( 1995) Women and urban policy: Strategies for change. Paper presented at the Women
and Transport Conference in Adelaide, Australia.
Kwan, M. ( 1999) Gender, the home- work link, and space- time patterns of non- employment
activities. Economic Geography 75( 4), 370- 394.
Law, R. ( 1999) Beyond ‘ women and transport’: towards new geographies of gender and daily
mobility. Progress in Human Geography 23( 4), 567- 588.
Liss, Susan, Nancy McGuckin, Sheila Moore, and Timothy Reuscher ( 2005) Our Nation’s Travel:
Current Issues. Publication No. FHWA- PL- 05- 015, US Department of Transportation,
Federal Highway Administration, Washington, DC. Available at
http:// nhts. ornl. gov/ 2001/ pub/ Issues. pdf, accessed August 31, 2009.
Lu, X. and E. I. Pas ( 1999) Socio- demographics, activity participation and travel behavior.
Transportation Research Part A 33, 1- 18.
Matthies, E., S. Kuhn, and C. A. Klöckner ( 2002) Travel mode choice of women: The result of
limitation, ecological norm, or weak habit? Environment and Behavior 34( 2), 163- 177.
Mauch, M. and B. D. Taylor ( 1997) Gender, race, and travel behavior: Analysis of household-serving
travel and commuting in San Francisco Bay Area. Transportation Research Record
1607, 147- 153.
Mokhtarian, P. L., E. A. Raney, and I. Salomon ( 1997) Behavioral response to congestion:
identifying patterns and socio- economic differences in adoption. Transport Policy 4( 3),
147- 160.
Mokhtarian, P. L., G. Circella, R. Huang, K. Lovejoy, S. Pike, L. Poff, J. Silvis, A. Thakur, D. van
Herick, L. Ye, S. Yui, M. Yun ( in progress) Commuter Impacts and Travel Behavior Changes
Resulting from the Sacramento Fix I- 5 Project. Final report prepared by the Institute of
Transportation Studies, University of California, Davis, for the California Air Resources Board
under Agreement # 07- 017. Available from the first author.
16
Möser, G. and S. Bamberg ( 2007) The effectiveness of soft transport policy measures: A critical
assessment and meta- analysis of empirical evidence. Journal of Environmental Psychology 28,
10- 26.
Patterson, Z., G. Ewing, M. Haider ( 2005) Gender- based analysis of work trip mode choice of
commuters in suburban Montreal, Canada, with stated preference data. Transportation
Research Record 1924, 85– 93.
Pisarski, Alan ( 2006) Commuting in America III: The Third National Report on Commuting
Patterns and Trends. Washington, DC: Transportation Research Board of the National
Academies of Science.
Rose, G. and H. Marfurt ( 2007) Travel behaviour change impacts of a major ride to work day event.
Transportation Research Part A 41( 4), 351- 364.
Sacramento Area Council of Governments ( SACOG) ( 2008) Metropolitan Transportation Plan
2035, Chapter 11, Access and mobility.
http:// www. sacog. org/ mtp/ 2035/ finaldocs/ mtp/ 11- Access- and- Mobility. pdf, accessed
August 28, 2009.
Salomon, I. and P. L. Mokhtarian ( 1997) Coping with congestion: Understanding the gap between
policy assumptions and behavior. Transportation Research Part D 2( 2), 107- 123.
Shiftan, Y. and J. Suhrbier ( 2002) The analysis of travel and emission impacts of travel demand
management strategies using activity- based models. Transportation 29, 145- 168.
Shoup, D. C. ( 1992) Employer- paid parking: The problem and proposed solutions. Transportation
Quarterly 46( 2), 169- 192.
Shoup, D. C. ( 1997) Evaluating the effects of cashing out employer- paid parking: eight case
studies. Transport Policy 4( 4), 201- 216.
Taylor, M. A. P. and E. S. Ampt ( 2003) Travelling smarter down under: policies for voluntary
travel behavior change in Australia. Transport Policy 10( 3), 165- 177.
Tranter, P. ( 1994) Behind closed doors: women, girls and mobility. Conference proceedings of
Women on the Move, But where to from here? Debating the issues that affect women and public
transport. South Australia.
Turner, T., D. Niemeier ( 1997) Travel to work and household responsibility: New evidence.
Transportation 24( 4), 397- 419.
van Exel, Nicolaas Jacob Arnold and Piet Rietveld ( 2010) Perceptions of public transport travel
time and their effect on choice- sets among car drivers. Journal of Transport and Land Use
2( 3/ 4), 75- 86.
The Women's Planning Network Inc. ( 1995) Women’s Transport Needs. Prepared by Perrott Lyon
Mathieson Pty Ltd. and Greg Tucker and Associates Pty Ltd.
http:// www. wpn. org. au/ researchresource/ womenstransportneeds/ Womens_ Transport_ Needs_
Report_ 1995. pdf, accessed August 30, 2009.
Ye, L., P. L. Mokhtarian, and G. Circella ( 2009) Commuter impacts and behavior changes during a
temporary freeway closure: The Fix I- 5 Project in Sacramento, California. Paper No.
10- 3275 presented at the 89th Annual Meeting of the Transportation Research Board,
Washington, DC, January 2010.