UC Berkeley

Earlier Faculty Research

Title
Correlation or causality between the built environment and travel behavior?
Evidence from Northern California

Permalink
https://escholarship.org/uc/item/5b76c5kg

Authors
Handy, Susan
Cao, Xinyu
Mokhtarian, Patricia L.

Publication Date
2005-06-01

eScholarship.org Powered by the California Digital Library

University of California
 Transportation Research Part D 10 (2005) 427–444
www.elsevier.com/locate/trd

Correlation or causality between the built environment

and travel behavior? Evidence from Northern California
a,*
Susan Handy , Xinyu Cao b, Patricia Mokhtarian b

a
Department of Environmental Science and Policy, University of California, One Shields Avenue,
Davis, CA 95616-8762, USA
b
Department of Civil and Environmental Engineering, University of California,
One Shields Avenue, Davis, CA 95616-8762, USA

Abstract

The sprawling patterns of land development common to metropolitan areas of the US have been blamed
for high levels of automobile travel, and thus for air quality problems. In response, smart growth pro-
grams—designed to counter sprawl—have gained popularity in the US. Studies show that, all else equal,
residents of neighborhoods with higher levels of density, land-use mix, transit accessibility, and pedestrian
friendliness drive less than residents of neighborhoods with lower levels of these characteristics. These stud-
ies have shed little light, however, on the underlying direction of causality—in particular, whether neigh-
borhood design influences travel behavior or whether travel preferences influence the choice of
neighborhood. The evidence thus leaves a key question largely unanswered: if cities use land use policies
to bring residents closer to destinations and provide viable alternatives to driving, will people drive less
and thereby reduce emissions? Here a quasi-longitudinal design is used to investigate the relationship
between neighborhood characteristics and travel behavior while taking into account the role of travel pref-
erences and neighborhood preferences in explaining this relationship. A multivariate analysis of cross-
sectional data shows that differences in travel behavior between suburban and traditional neighborhoods
are largely explained by attitudes. However, a quasi-longitudinal analysis of changes in travel behavior
and changes in the built environment shows significant associations, even when attitudes have been
accounted for, providing support for a causal relationship.
Ó 2005 Elsevier Ltd. All rights reserved.

*
Corresponding author. Tel.: +1 530 752 5878; fax: +1 530 752 3350.
E-mail address: [email protected] (S. Handy).

1361-9209/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.trd.2005.05.002
428 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

Keywords: Built environment; Travel behavior; Smart growth; Land-use

1. Introduction

Sprawling patterns of land development have been blamed for high levels of automobile travel,
and thus for air quality problems. The defining characteristics of sprawl include: low-density
development, unlimited outward expansion, and leap-frog development (Burchell et al., 2002).
This low-density pattern of growth has two important effects on travel: longer trip distances
and greater reliance on the car. In response to these travel effects and other impacts, smart growth
programs designed to counter sprawl have gained popularity in the US. The American Planning
Association (2002) defines smart growth as ‘‘the planning, design, development and revitalization
of cities, towns, suburbs and rural areas to create and promote social equity, a sense of place and
community, and to preserve natural as well as cultural resources’’. Specific land use policies used
in smart growth programs include mixed-use zoning, infill development, brownfield development,
and transit-oriented development, as well as bicycle and pedestrian infrastructure. The hope is
that these strategies will bring residents closer to destinations and provide viable alternatives to
driving, and thus help to reduce automobile use.
Recognizing this potential, the US Environmental Protection Agency (EPA) now accepts land-
use policies as an effective tool for improving air quality and allows state and local communities to
account for the air quality benefits of smart growth strategies in state air quality plans (EPA,
2001). However, the estimation of the emissions effects of land use policies is based on limited
empirical evidence, and little is known about the sensitivity of air quality to changes in land
use. Studies in the US show that, all else equal, residents of neighborhoods with higher levels
of urban density, land-use mix, transit accessibility, and pedestrian friendliness (among other
characteristics) drive less than residents of neighborhoods with lower levels of these characteris-
tics. These studies have not established the underlying direction of causality, in particular,
whether neighborhood design influences travel behavior or whether travel preferences influence
the choice of neighborhood. The available evidence thus leaves a key question largely unanswered:
If cities use land use policies to bring residents closer to destinations and provide viable alterna-
tives to driving, will at least some people drive less, thereby reducing emissions?
This paper provides new evidence that helps to answer this question; it aims to provide a stron-
ger basis for assessing the potential for land-use policies to reduce automobile travel, and thus
vehicle emissions. The study uses a quasi-longitudinal design to investigate the relationship be-
tween neighborhood characteristics and travel behavior while taking into account the role of tra-
vel attitudes and neighborhood preferences in explaining this relationship.

2. Literature review

If cities use land use policies to bring residents closer to destinations and provide viable alter-
natives to driving, will people drive less, thereby reducing emissions? Existing research does not
provide a clear answer. The idea that land use and design policies could be used to influence travel
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 429

behavior was not widely explored until the 1980s. Early interest focused on the connection be-
tween density and transit use. The study by Pushkarev and Zupan (1977) suggests that transit
use can be increased through polices that increase densities. A heated debate ensued in the early
1990s, over analysis by Newman and Kenworthy (1999) on the correlation between densities and
gasoline consumption for a sample of international cities (Mindali et al., 2004). In response to the
emergence of the smart growth movement and the concept of new urbanism, a number of recent
studies have examined the effect of specific characteristics of the built environment on travel
behavior at a disaggregate level in an effort to test the hypothesis that policies that shape the built
environment can be used to reduce automobile travel.
Since the early 1990s, such studies have appeared with increasing frequency. Recent reviews
document over 70 studies during the 1990s that explored and quantified these relationships
(Handy, 1996; Boarnet and Crane, 2001; Ewing and Cervero, 2001). Sorting out the extent to
which socio-economic characteristics and characteristics of the built environment impact travel
behavior is a common challenge in these studies. Ewing and Cervero (2001), after one of the most
thorough reviews of these studies, come to the conclusion that the built environment has a greater
impact on trip lengths than trip frequencies and that mode choice depends as much on socio-
economic characteristics as the built environment. Their analysis of the existing studies shows
small but statistically significant effects of the built environment on vehicle miles of travel.
In the US, the debate over the link between neighborhood design and travel behavior now cen-
ters on the issue of causality (Transportation Research Board and Institute of Medicine, 2005).
Scientific practice dictates four criteria for establishing causality between an independent variable
(the cause) and a dependent variable (the effect): the cause and effect are statistically associated
(association), the cause precedes the effect in time (time order), no third factor creates an acciden-
tal or spurious relationship between the variables (non-spuriousness), and the mechanism by
which the cause influences the effect is known (causal mechanism) (Singleton and Straits, 1999).
Most studies have met the first criterion but have not met the other three.
Almost all of the studies have used cross-sectional designs to establish a statistical association
between the built environment and travel behavior. Cross-sectional designs do not establish
whether the cause precedes the effect, however. Most studies have controlled for socio-demo-
graphic characteristics, thereby minimizing the possibility that income, for example, creates a spu-
rious relationship between the built environment and travel behavior. But few have accounted for
the effects of attitudes towards travel. By falling short on the criteria of time-order and non-
spuriousness, these studies leave open the possibility of self-selection, in which individuals who
would rather not drive choose to live in neighborhoods conducive to driving less. In this case,
the characteristics of the built environment do not cause them to drive less; rather, their desire
to drive less causes them to select a neighborhood with those characteristics. Understanding
the role of self-selection is the key to understanding the causal relationship between the built envi-
ronment and travel behavior.
A few studies have made efforts to address the self-selection issue by accounting for preferences
and attitudes, although with cross-sectional data as well, thereby addressing the non-spurious cri-
terion but not the time order criterion. Studies by Handy and Clifton (2001) and Bagley and
Mokhtarian (2002) in fact suggest that the observed associations between travel behavior and
neighborhood characteristics are largely explained by the self-selection of residents with certain
attitudes into certain kinds of neighborhoods. On the other hand, Schwanen and Mokhtarian
430 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

(2005), also using cross-sectional data but a different methodology, found that neighborhood type
does impact on travel behavior, even after attitudes are accounted for. Unfortunately for efforts to
reduce car use, they found the effect to be more extreme in the suburbs (where residents who pre-
ferred higher-density neighborhoods drove to work about as often as those who liked their lower-
density environment) than in the city (where residents who preferred lower-density neighborhoods
drove to work more often than those who liked higher densities although not as much as those
who lived in the suburbs). Cao et al. (2005) recently found that characteristics of the built envi-
ronment influence walking behavior after accounting for a preference for neighborhoods condu-
cive to walking.
The causal mechanism that might link the built environment to travel behavior has been given
limited attention by researchers. Boarnet and Crane (2001) offer an economic explanation: the
built environment influences the price of travel, through its impact on travel time and other qual-
ities of travel, which then influences the consumption of travel. A similar idea is implicit in discrete
choice models of travel behavior: the utility of a particular travel choice—what mode to take or
which destination to choose—is influenced by travel time and other characteristics of the possible
choices, that in turn are influenced by the built environment. But these explanations focus on the
immediate and direct connection between the built environment and travel behavior and do not
account for longer term decisions about residential location. It is possible that different causal
mechanisms apply in different situations, depending on the combination of the preferences of
the individual and the type of environment in which she chooses to live (Handy, 2005). It is also
possible that the built environment has a less immediate and more indirect effect on travel behav-
ior through its impact on attitudes over time. Not only has the existence of a causal relationship
not yet been established, the nature of potential causal relationships is poorly understood.

3. Methodology

The objectives of the study presented here were, first, to confirm the role of attitudes and pref-
erences in explaining the observed link between the built environment and travel behavior, and,
second, to more directly test the general hypothesis that the built environment has a causal effect
on travel behavior. Causal relationships are most validly established through experimental de-
signs, in which individuals are randomized to treatment and control groups (thereby addressing
non-spuriousness) and behavior is measured for both groups before and after the treatment of
interest (thereby addressing time order) (Singleton and Straits, 1999). Neither randomization
nor the application of a treatment is practical for studying the link between the built environment
and travel behavior. Instead, in this study, the treatment is defined as a move from one neighbor-
hood to another and the lack of randomization is addressed by accounting for preferences and
attitudes that might influence the choice of neighborhood. The specific hypotheses addressed here
are thus as follows:

1. Differences in the built environment are associated with differences in travel behavior, after
accounting for socio-demographic characteristics and for attitudes and preferences. More
specifically, environments where residents are closer to destinations and have viable alterna-
tives to driving are in fact associated with less driving.
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 431

2. Changes in the built environment are associated with changes in travel behavior, after
accounting for socio-demographic characteristics and for attitudes and preferences. More
specifically, moves to environments where residents are closer to destinations and have viable
alternatives to driving are associated with a decrease in driving.

Here an innovative approach is used to examine the causal relationship between the built envi-
ronment and travel behavior while taking into account the role of attitudes, including attitudes
with respect to travel and preferences with respect to neighborhood characteristics. The research
design enables both cross-sectional and quasi-longitudinal analyses.

3.1. Survey and sampling

The data came from a self-administered 12 page survey mailed in two rounds in late 2003 to
households in eight neighborhoods in Northern California. The neighborhoods were selected to
vary systematically on three dimensions: neighborhood type, size of the metropolitan area, and
region of the state. Neighborhood type was differentiated as traditional for areas built mostly
in the pre-World II era, and suburban for areas built more recently. Although this design was in-
tended to provide ample variation across neighborhood types, and these discrete indicators of
neighborhood type are useful for descriptive comparisons, they are too simplistic for more de-
tailed analyses. For the multivariate models presented in Section 4, we used a rich set of variables
describing the neighborhoods along a variety of dimensions.
Using data from the US Census, we screened potential neighborhoods to ensure that average
income and other characteristics were near the average for the region. Four neighborhoods in
the San Francisco Bay Area—two in the Silicon Valley area and two in Santa Rosa—had been
previously studied (Handy, 1992). Two neighborhoods from Sacramento and two from Modesto
were selected to contrast with Bay Area neighborhoods (Fig. 1). The four traditional neighbor-
hoods differ in visible ways from the four suburban neighborhoods—the layout of the street net-
work, the age and style of the houses, and the location and design of commercial centers, as shown
in Fig. 2 for Sacramento as an example.
For each neighborhood, two databases of residents were purchased from a commercial pro-
vider, New Neighbors Contact Service (www.nncs.com; this service maintains an overall database
of names and addresses for residences throughout the US constructed from a variety of public
records. The database is largely used for commercial advertisement mailings): a database of mov-
ers and a database of non-movers. The movers included all current residents of the neighborhood
who had moved within the previous year. From this database, we drew a random sample of 500
residents for each neighborhood. The database of non-movers consisted of a random sample of
500 residents not included in the movers list for each neighborhood.
Survey questions were developed from surveys used in previous research projects by the authors
and other researchers. The survey was pre-tested on UC Davis students and staff, then on a con-
venience sample of Davis residents. Participants were asked to first complete the survey, then to
discuss the survey questions with the researchers, either in a group meeting or in one-on-one inter-
views. Based on these pretests, survey questions were modified and refined.
The survey was administered using a mail-out, mail-back approach. Surveys were mailed to
households in the selected neighborhoods, and the cover letter asked for ‘‘any adult household
432 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

Fig. 1. Location of neighborhoods.

member who shares in the decision making . . . and who participated in selecting your current res-
idence’’ to complete the survey. The initial survey was mailed out at the end of September 2003.
Two weeks later, a reminder postcard was mailed to the entire sample using first-class mail. At the
beginning of November, a second copy of the survey with a revised cover letter was sent to a
shorter list that excluded incorrect addresses and individuals who had already responded to the
survey. Two weeks later, a second reminder postcard was mailed to this list of residents. As an
incentive to complete the survey, respondents were told they would be entered into a drawing
to receive one of five $100 cash prizes; the winners were selected in December.
The original database consisted of 8000 addresses but only 6746 valid addresses. The number of
responses totaled 1682, for a response rate of 24.5%. This is considered quite good for a survey of
this length; the typical response rates for a survey administered to the general population are 10–
40% (Sommer and Sommer, 1997). A comparison of sample characteristics to population charac-
teristics (based on the 2000 US Census) shows that survey respondents tend to be older on average
than residents of their neighborhood as a whole, and that households with children are underrep-
resented for most neighborhoods while home owners are overrepresented for all neighborhoods
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 433

Fig. 2. Comparison of traditional and suburban neighborhoods-Sacramento.

(Table 1). However, since the focus is on explaining the relationships of other variables to travel
behavior rather than on describing travel behavior per se, these differences are not expected to
materially affect the results (Babbie, 1998).
 434
Table 1
Sample characteristics vs. population characteristics
Traditional Suburban Traditional Suburban
Silicon Santa Modesto— Sacramento— Silicon Santa Modesto— Sacramento—
Valley— Rosa— Central Midtown Valley— Rosa— Suburban Natomas
Mountain Junior Sunnyvale Rincon
View College Valley

S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

Sample characteristics
Number 228 215 184 271 217 165 220 182 898 784
Percent female 47.3 54.3 56.3 58.2 46.9 50.9 50.9 54.9 54.1 50.7
Average auto 1.80 1.63 1.59 1.50 1.79 1.66 1.88 1.68 1.62 1.76*
ownership
Average age 43.3 47.0 51.3 43.4 47.1 54.7 53.2 45.6 45.1 48.9*
Average HH size 2.08 2.03 2.13 1.78 2.58 2.19 2.41 2.35 1.99 2.40*
Percent of HHs 21.1 18.6 21.7 8.9 42.4 24.8 25.5 31.9 16.9 31.5*
w/kids
Average number 1.60 0.16 1.83 1.58 1.65 1.59 1.98 1.64 1.65 1.71
of kids
Percent home 51.1 57.8 75.6 47.0 61.1 68.7 81.0 82.4 56.4 73.3*
owner
Median HH 98.7 55.5 45.5 64.2 95.0 49.5 55.5 55.3 68.5 55.5
income (k$)
Population characteristicsa
Population 5,493 9,886 13,295 7,259 14,973 13,617 19,045 13,295
Average age 36.1 36.3 36.5 42.7 35.9 38.3 38.1 31.7
Average HH size 2.08 2.21 2.46 1.79 2.66 2.48 2.51 2.57
Percent of 19.3 20.3 32.9 12.4 35.3 35.4 34.2 41.7
HHs w/kids
Percent home 34.3 31.2 58.8 34.3 53.2 63.5 61.4 55.2
owner
Median HH 74.3 40.2 42.5 43.8 88.4 49.6 40.2 46.2
income (k$)
Percent of units 54.3 37.2 21.4 22.7 79.9 90.3 94.6 97.6
built after 1960
*
Differences between traditional and suburban significant at 1% level.
a
Source: 2000 U.S. Census of Population and Housing.
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 435

3.2. Variables

Travel behavior was variously measured through a series of questions on commute trips, non-
work trips, and walking trips. In addition, respondents were asked to list vehicles currently avail-
able to the household, and to estimate how many miles they drive in a typical week. Although it is
unlikely that respondents know exactly how many miles they drive in a week, these estimates
should provide a reasonably accurate indicator of their level of driving. Reported vehicle miles
driven (VMD) per week is used as the dependent variable in cross-sectional models. Change in
travel behavior from either just before the move (for the movers) or from 1 year ago (for the non-
movers) was measured using 5-point scales; for example, respondents were asked to indicate
whether they drive ‘‘a lot less now’’, ‘‘a little less now’’, ‘‘about the same’’, ‘‘a little more
now’’, or ‘‘a lot more now’’.

Table 2
Factors for neighborhood characteristics
Factor Statement Loadinga
Accessibility Easy access to a regional shopping mall 0.854
Easy access to downtown 0.830
Other amenities such as a community center available nearby 0.667
Shopping areas within walking distance 0.652
Easy access to the freeway 0.528
Good public transit service (bus or rail) 0.437
Physical activity options Bike routes beyond the neighborhood 0.882
Sidewalks throughout the neighborhood 0.707
Parks and open spaces nearby 0.637
Good public transit service (bus or rail) 0.353
Safety Quiet neighborhood 0.780
Low crime rate within neighborhood 0.759
Low level of car traffic on neighborhood streets 0.752
Safe neighborhood for walking 0.741
Safe neighborhood for kids to play outdoors 0.634
Good street lighting 0.571
Socializing Diverse neighbors in terms of ethnicity, race, and age 0.789
Lots of people out and about within the neighborhood 0.785
Lots of interaction among neighbors 0.614
Economic level of neighbors similar to my level 0.476
Outdoor spaciousness Large back yards 0.876
Large front yards 0.858
Lots of off-street parking (garages or driveways) 0.562
Attractiveness Attractive appearance of neighborhood 0.780
High level of upkeep in neighborhood 0.723
Variety in housing styles 0.680
Big street trees 0.451
a
Represents the degree of association between the statement and the factor.
436 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

Table 3
Factors for travel preferences
Factor Statement Loadinga
Pro-bike/walk I like riding a bike 0.880
I prefer to bike rather than drive whenever possible 0.865
Biking can sometimes be easier for me than driving 0.818
I prefer to walk rather than drive whenever possible 0.461
I like walking 0.400
Walking can sometimes be easier for me than driving 0.339
Pro-travel The trip to/from work is a useful transition 0.683
between home and work
Travel time is generally wasted time 0.681
I use my trip to/from work productively 0.616
The only good thing about traveling is 0.563
arriving at your destination
I like driving 0.479
Travel minimizing Fuel efficiency is an important factor 0.679
for me in choosing a vehicle
I prefer to organize my errands so that I 0.617
make as few trips as possible
I often use the telephone or the Internet to 0.514
avoid having to travel somewhere
The price of gasoline affects the choices 0.513
I make about my daily travel
I try to limit my driving to help improve air quality 0.458
Vehicles should be taxed on the basis of 0.426
the amount of pollution they produce
When I need to buy something, I usually 0.332
prefer to get it at the closest store possible
Pro-transit I like taking transit 0.778
I prefer to take transit rather than drive whenever possible 0.771
Public transit can sometimes be easier for me than driving 0.757
Walking can sometimes be easier for me than driving 0.344
I prefer to walk rather than drive whenever possible 0.363
Safety of car Traveling by car is safer overall than walking 0.753
Traveling by car is safer overall than taking transit 0.633
Traveling by car is safer overall than riding a bicycle 0.489
The region needs to build more highways 0.444
to reduce traffic congestion
Car dependent I need a car to do many of the things I like to do 0.612
Getting to work without a car is a hassle 0.524
We could manage pretty well with one 0.418
fewer car than we have (or with no car)
Traveling by car is safer overall than riding a bicycle 0.402
I like driving 0.356
a
Represents the degree of association between the statement and the factor.
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 437

The explanatory variables are classified into:

Neighborhood characteristics and neighborhood preferences: Respondents were asked to indicate
how true 34 characteristics are for their current and previous (only for movers) neighborhood, on
a four-point scale from 1 (‘‘not at all true’’) to 4 (‘‘entirely true’’). The characteristics of these
neighborhoods as perceived by survey respondents reflect fundamental differences in neighbor-
hood design. The importance of these items to respondents when/if they were looking for a
new place to live was also measured on a four-point scale from 1 (‘‘not at all important’’) to 4
(‘‘extremely important’’). The comparison of individualsÕ perceived neighborhood characteristics
for their current residence and their preferences for neighborhood characteristics indicates how
well their current neighborhoods meet their preferences. Since some of these characteristics mea-
sure similar dimensions of the built environment and are highly correlated, we conducted a factor

Table 4
Vehicle miles driven and explanatory variables by neighborhood type
Average Average p-valuec p-valuec p-valuec
for for traditional traditional suburban
traditional suburban vs. suburban only only
Vehicle miles driven per weeka 148 175 0.00 0.66 0.64
Perceived neighborhood characteristicsb
Accessibility 0.15 0.18 0.00 0.00 0.00
Physical activity options 0.01 0.01 0.45 0.00 0.05
Safety 0.14 0.16 0.00 0.00 0.00
Socializing 0.09 0.12 0.00 0.00 0.00
Outdoor spaciousness 0.00 0.01 0.82 0.00 0.00
Attractiveness 0.28 0.33 0.00 0.00 0.00
Selected accessibility measures
# business types within 400 m 2.6 0.8 0.00 0.00 0.09
# eat-out places within 400 m 0.7 0.2 0.00 0.00 0.44
Meters to nearest eat-out place 526 789 0.00 0.00 0.00
Preferences for neighborhood characteristicsb
Accessibility 0.03 0.04 0.14 0.00 0.00
Physical activity options 0.01 0.02 0.60 0.00 0.00
Safety 0.18 0.21 0.00 0.00 0.48
Socializing 0.05 0.05 0.05 0.00 0.00
Outdoor spaciousness 0.05 0.06 0.02 0.00 0.03
Attractiveness 0.04 0.05 0.04 0.00 0.00
Travel attitudesb
Pro-bike/walk 0.20 0.23 0.00 0.00 0.00
Pro-travel 0.03 0.03 0.27 0.03 0.01
Travel-minimizing 0.01 0.01 0.69 0.23 0.02
Pro-transit 0.15 0.17 0.00 0.00 0.00
Safety of car 0.27 0.31 0.00 0.00 0.00
Car dependent 0.06 0.07 0.01 0.03 0.00
a
Six respondents reported over 1000 miles per week; these values were treated as outliers and recoded to 1000.
b
Scores normalized to a mean value of 0 and variance of 1.
c
p-values for F-statistics from analysis of variance (ANOVA).
438 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

analysis to identify underlying constructs of perceived and preferred neighborhood characteristics.

Through this analysis, these items were reduced to six factors (some items were dropped due to
their poor conceptual interpretability): accessibility, physical activity options, safety, socializing,
outdoor spaciousness, and attractiveness (Table 2).
Objective measures of accessibility were estimated for each respondent, based on distance along
the street network from home to a variety of destinations classified as institutional (bank, church,
library, and post office), maintenance (grocery store and pharmacy), eating-out (bakery, pizza, ice
cream, and take-out), and leisure (health club, bookstore, bar, theater, and video rental). Acces-
sibility measures included the number of different types of businesses within specified distances,
the distance to the nearest establishment of each type, and the number of establishments of each
business type within specified distances. Commercial establishments were identified using on-line
yellow pages, and ArcGIS was used to calculate network distances between addresses for survey
respondents and commercial establishments.
Travel attitude: To measure attitudes regarding travel, the survey asked respondents whether
they agreed or disagreed with a series of 32 statements on a 5-point scale from 1 (‘‘strongly dis-
agree’’) to 5 (‘‘strongly agree’’). Factor analysis was then used to extract the fundamental dimen-
sions spanned by these 32 items, for reasons similar to those for neighborhood characteristics.
As shown in Table 3, six underlying dimensions were identified: pro-bike/walk, pro-travel, travel
minimizing, pro-transit, safety of car, and car dependent.
Socio-demographics: The survey also contained a list of socio-demographic variables that may
help to explain travel behavior. These variables include gender, age, employment status, educa-
tional background, household income, household size, the number of children in the household,
mobility constraints, residential tenure, and so on. Some changeable socio-demographics such as
household structure and income were measured currently as well as before moving for movers and
from 1 year ago for non-movers.

Table 5
Regression model for ln(VMD)
Variable Coefficient Standardized t-statistic p-value
coefficient
Constant 3.646 11.317 0.000
Female 0.282 0.140 5.650 0.000
Working 0.298 0.112 4.034 0.000
Age 0.006 0.094 3.296 0.001
DriverÕs license 1.050 0.086 3.519 0.000
Cars per adult 0.170 0.069 2.852 0.004
Pro-bike/walk attitude 0.055 0.054 1.973 0.049
Pro-transit attitude 0.048 0.046 1.784 0.075
Safety of car attitude 0.060 0.058 2.255 0.024
Car dependent attitude 0.271 0.260 10.566 0.000
Outdoor spaciousness preference 0.054 0.052 2.110 0.035
N 1466
R-square 0.160
Adjusted R-square 0.154
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 439

4. Findings

Vehicle miles driven by the respondent per week is 18% higher for residents of suburban neigh-
borhoods than for residents of traditional neighborhoods (Table 4). This pattern holds true across
individual neighborhoods: the highest level of driving for traditional neighborhoods (161 miles
per week in Modesto Central) is still lower than the lowest level of driving for suburban neighbor-
hoods (166 miles in Sunnyvale). The difference in total vehicle miles driven appears to come from
differences in both work travel and non-work travel.
What explains these differences? According to an analysis of variance (ANOVA), traditional
neighborhoods score significantly higher than suburban neighborhoods on factors for perceived
accessibility, socializing, and attractiveness, but lower on safety; residents in these neighborhoods
are also closer to more businesses. But traditional neighborhoods also score higher on factors for
pro-bike/walk and pro-transit attitudes and lower on the safety of car attitude. To complicate
matters, differences are often significant between neighborhoods of each type on both neighbor-
hood characteristics and attitudinal factors; traditional neighborhoods are not all alike, nor are all
suburban neighborhoods alike. To sort out the relative importance of neighborhood characteris-
tics and attitudes and preferences in explaining levels of driving, multivariate models are
estimated.

4.1. Cross-sectional analysis of vehicle miles driven

Most studies have used cross-sectional models of driving behavior to test the significance of
built environment characteristics as explanatory variables. For comparison purposes, we also esti-
mated a cross-sectional model, but unlike most previous studies we incorporated preferences for
neighborhood characteristics and travel attitudes into the model to account for the possibility of
self-selection. This tests the hypothesis that environments where residents are closer to destina-
tions and have viable alternatives to driving are in fact associated with less driving, after account-
ing for attitudes and preferences as well as socio-demographic characteristics; lower levels of
driving might result from shorter and/or fewer driving trips. Because of the skewed distribution
of VMD, the natural log of VMD was used as the dependent variable and the model was esti-
mated using ordinary least squares regression. Potential explanatory variables were entered into
the model in groups, starting with socio-demographic factors, followed by travel attitudes and
preferences for neighborhood characteristics, then perceived neighborhood characteristics and
accessibility measures. At each step, insignificant variables were dropped and the model re-esti-
mated before the next set of variables was entered.
The factor for car dependent attitude had the highest standardized coefficient (Table 5). This
factor reflects a perceived need for a car, which may or may not reflect the actual availability
of alternatives to driving. Other attitudes were also significant: pro-bike/walk and pro-transit atti-
tudes were negatively associated with driving, and the safety of car attitude and a preference for
outdoor spaciousness were positively associated with driving. With these attitudes accounted for,
no measures of the actual built environment—neither accessibility measures nor perceived char-
acteristics—were significant. As a result, it appears that observed correlations between neighbor-
hood type and levels of driving are better explained by attitudes towards transportation than by
the built environment. The model does not support the hypothesis that the built environment has
440 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

a causal relationship with travel behavior and suggests that self-selection plays a significant role in
explaining the observed correlations between the built environment and travel behavior. This find-
ing differs from previous studies that found a significant relationship between the built environ-
ment and driving and demonstrates the importance of accounting for attitudes and preferences.

4.2. Quasi-longitudinal analysis of travel behavior

A stronger test of a causal relationship between the built environment and travel behavior in-
volves an examination of the association between a change in the built environment and a change
in driving. This approach addresses the time-order criterion for establishing causal validity: if the
change in the built environment precedes the change in driving, then a causal relationship is more
certain (Singleton and Straits, 1999). In the quasi-longitudinal approach used here, changes are
measured for residents who have recently moved from before to after their move, and for non-
movers from 1 year earlier to the present time. The model estimated from these data tests the
hypothesis that moves to environments where residents are closer to destinations and have viable
alternatives to driving are associated with a decrease in driving after accounting for neighborhood
preferences and travel attitudes; decreases in driving might result from a decrease in driving dis-
tances and/or a decrease in driving trips.
The change in driving and the use of other modes is measured using a 5-point scale, from ‘‘a lot
less now’’ to ‘‘a lot more now’’. For the sample of movers, changes in the built environment could
be measured by taking the difference between perceived characteristics of the current and previous
neighborhoods; the built environment was assumed unchanged for the sample of non-movers. A
simple bivariate analysis of these variables for movers (Table 6) shows several significant associ-

Table 6
Percent of respondents by change in driving or walking vs. change in neighborhood characteristicsa
Decrease in characteristic Increase in characteristic p-value
Increase in driving Decrease in driving Increase in driving Decrease in driving
or walking or walking or walking or walking
For driving
Accessibility 31.0 28.3 23.9 47.6 0.00
PA options 28.3 38.3 24.7 44.2 0.34
Safety 30.3 41.0 23.2 42.6 0.11
Socializing 28.7 38.6 24.2 44.1 0.31
Spaciousness 24.4 44.6 27.4 39.6 0.42
Attractiveness 26.3 40.3 25.8 43.0 0.77
For walking
Accessibility 37.4 27.8 55.9 16.7 0.00
PA options 35.4 28.8 58.9 15.1 0.00
Safety 44.8 28.0 54.2 14.9 0.00
Socializing 38.9 27.8 58.0 14.9 0.00
Spaciousness 50.6 22.4 50.4 17.7 0.21
Attractiveness 35.7 31.9 59.0 13.1 0.00
a
Movers only; n = 688.
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 441

ations. In general, changes in neighborhood characteristics have the strongest association with
changes in walking: for increases in all but one of the factors for neighborhood characteristics,
a significantly higher share of respondents said that their walking levels had increased than said
they had decreased. In contrast, only changes in the accessibility factor had a significant associ-
ation with changes in driving: among respondents who reported that accessibility increased, a sig-
nificantly higher share said that driving had decreased rather than increased. This finding is
interesting given that accessibility may have two opposite effects on driving: 1. higher accessibility
reduces the cost of driving and may increase levels of driving as a result; 2. higher accessibility
reduces the cost of walking and may lead to a substitution of walking for driving. The results sug-
gest that the latter effect outweighs the former, and that changes in neighborhood characteristics
are more important in explaining changes in walking rather than driving (see also Handy et al.,
2005).
The relationship between changes in the built environment and changes in driving while con-
trolling for attitudes (and changes in socio-demographics) was estimated using an ordered probit
model. This technique is appropriate for an ordinal dependent variable, and its model structure is
parsimonious. The following sets of variables were tested: current socio-demographic character-
istics, changes in socio-demographic characteristics, travel attitudes (assumed constant over this

Table 7
Ordered probit model for change in driving
Coefficient Standardized p-value
coefficienta
Constanta 1.508 1.147 0.000
Current age 0.006 0.084 0.014
Currently working 0.155 0.059 0.065
Current # kids <18 years 0.070 0.057 0.051
Limits on driving 0.678 0.074 0.000
Change in income 0.000 0.155 0.000
# groceries within 1600 m 0.014 0.066 0.048
# pharmacies within 1600 m 0.028 0.069 0.041
# theaters within 400 m 0.703 0.057 0.055
Change in accessibility factor 0.269 0.226 0.000
Change in safety factor 0.088 0.086 0.000
Car dependent 0.115 0.111 0.000
Pro-bike/walk 0.070 0.070 0.020
Threshold parameter—1 0.543 0.543 0.000
Threshold parameter—2 2.142 2.142 0.000
Threshold parameter—3 2.589 2.589 0.000
N 1490
Log-likelihood at 0 2378.038
Log-likelihood at constant 1954.785
Log-likelihood at convergence 1869.302
Pseudo-R square 0.214
Adjusted pseudo-R square 0.209
a
All independent variables standardized and model re-estimated; constant not standardized.
442 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

period), preferences for neighborhood characteristics (also assumed constant), objective accessi-
bility measures for the current neighborhood, perceived neighborhood characteristics for the cur-
rent neighborhood, and changes in perceived neighborhood characteristics. Non-movers were also
included in the model, with changes in driving and socio-demographic characteristics measured
from 1 year ago and changes in perceived neighborhood characteristics assumed to be zero.
The resulting equation can be interpreted as representing the propensity of an individual to have
a numerically larger change—either less of a decrease or more of an increase—in driving following
the move. A statistically significant association between a change in the built environment and
change in travel behavior provides evidence of a causal relationship.
Change in the accessibility factor was the most important factor in explaining changes in driv-
ing, as indicated by the standardized coefficients, with an increase in accessibility associated with
either a smaller increase or a larger decrease in driving (Table 7). Change in the safety factor was
also significant, with an increase in safety associated with either a smaller increase or a larger de-
crease in driving. Three objective measures were also significant: number of grocery stores and
number of pharmacies within 1600 m and number of theaters within 400 m (why theaters is sig-
nificant is not entirely clear, although theaters may be an indicator of a particular type of com-
mercial district). Note that objective accessibility was measured for the current neighborhood
only, rather than as the change in accessibility; however, a high current level of accessibility is
more likely to be associated with an increase in accessibility than a decrease as a result of a move.
In all of these cases, an increase in accessibility is associated with a higher propensity to drive less.
Two travel attitudes were also significant: car dependent, with a positive effect on the propensity
to drive more, and pro-bike/walk, with a negative effect on the propensity to drive more. These
results support the hypothesis that changes in the built environment are associated with changes
in driving and point to increases in accessibility as the factor having the greatest negative effect on
driving.

5. Conclusions

One lesson that emerges here is that different types of analyses can yield different answers to the
question: does the built environment have a causal relationship with travel behavior? A simple
comparison of different neighborhood types shows significant variations in levels of driving. How-
ever, a multivariate analysis of cross-sectional data shows that these differences are largely
explained by attitudes and that the effect of the built environment mostly disappears when
attitudes and socio-demographic factors have been accounted for. But a quasi-longitudinal anal-
ysis of changes in driving and changes in the built environment shows significant associations,
even when attitudes have been accounted for, providing support for a causal relationship. These
results highlight the limitations of previous studies, that mostly rely on cross-sectional analyses
and rarely account for attitudes and preferences—but also suggest that despite these limitations
their conclusions are not entirely off the mark.
These analyses are not definitive, nor do they clarify the nature of the causal relationship. More
sophisticated analyses of these data, such as structural equations modeling, will help to establish
the strength and direction of the relationships between attitudes, changes in the built environment,
changes in travel behavior, and other factors. Even so, there are limits to what the quasi-longitu-
 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444 443

dinal approach can do. For example, because it is not feasible to retrospectively measure attitudes,
we have data on current attitudes only, and our interpretation of the results of the change model is
predicated on the assumption that attitudes (those unmeasured as well as measured) remained
constant over time and hence are controlled for. But we cannot rule out the competing hypothesis
that an attitude change preceded and (partly) prompted the residential location change. To the
extent that is true, the attitude change is confounded with the change in built environment and
may account for some of the apparent effect of the built environment seen here.
Future studies that adopt research designs that more closely resemble a true experimental de-
sign will provide more definitive evidence yet. Two types of studies are possible: true panel studies
of residents who move from one type of neighborhood to another (with measurements of attitudes
as well as behavior before and after, and further exploration of the reasons behind the move), and
natural experiments that examine the impact on driving in response to a change in the built envi-
ronment, such as the implementation of a traffic calming program. Only with such evidence can
we be sure that by increasing opportunities for driving less through land use policies, cities will
help to reduce driving and thus emissions.
In the meantime, the results presented here provide some encouragement that land-use policies
designed to put residents closer to destinations and provide them with viable alternatives to driv-
ing will actually lead to less driving. In particular, it appears that an increase in accessibility may
lead to a decrease in driving, all else equal. Policies that could increase accessibility in new areas
include mixed-use zoning that allows for retail and other commercial uses within close proximity
to residential areas and street connectivity ordinances that ensure more direct routes between res-
idential and commercial areas. Policies that could increase accessibility in existing areas include
Main Street programs designed to enhance and revitalize traditional neighborhood shopping
areas, incentives for infill development and redevelopment of underutilized shopping centers,
and the like. Although this study does not definitively prove that land use policies can reduce driv-
ing, it provides evidence that supports the adoption of such policies.

Acknowledgements

This research was conducted under a project funded by the California Department of Trans-
portation and was supported by grants from the Robert Wood Johnson Foundation and the Uni-
versity of California Transportation Center. Thanks to Ted Buehler, Gustavo Collantes, and Sam
Shelton for their work on the implementation of the survey.

References

American Planning Association, 2002. Planning for Smart Growth: 2002 State of the States. Chicago, IL, February.
Available at: http://www.planning.org/growingsmart/pdf/states2002.pdf. Accessed: 11/1/02.
Babbie, E., 1998. The Practice of Social Research, Eighth Ed. Wadsworth Publishing Company, Stamford, CT.
Bagley, M.N., Mokhtarian, P.L., 2002. The impact of residential neighborhood type on travel behavior: a structural
equations modeling approach. Annals of Regional Science 36, 279–297.
Boarnet, M.G., Crane, R., 2001. Travel by Design: the Influence of Urban Form on Travel. Oxford University Press,
New York.
444 S. Handy et al. / Transportation Research Part D 10 (2005) 427–444

Burchell, R.W., Lowenstein, G., Dolphin, W.R., Galley, C.C., Downs, A., Seskin, S., Gray Still, K., Moore, T., 2002.
Costs of Sprawl 2000. TCRP Report 74, Transit Cooperative Research Program, Transportation Research Board,
Washington, DC.
Cao, X., Handy, S., Mokhtarian, P.L., 2005. The Influences of the Built Environment and Residential Self-Selection on
Pedestrian Behavior. Transportation. Forthcoming.
Environmental Protection Agency (EPA), 2001. EPA Guidance: Improving Air Quality Through Land Use Activities.
EPA420-R-01-001, Transportation and Regional Programs Division, Office of Transportation and Air Quality,
Washington, DC.
Ewing, R., Cervero, R., 2001. Travel and the built environment. Transportation Research Record 1780, 87–114.
Handy, S., 1992. Regional versus Local Accessibility: Variations in Suburban Form and the Effects on Nonwork
Travel. University of California Berkeley, Unpublished dissertation.
Handy, S., 1996. Methodologies for exploring the link between urban form and travel behavior. Transportation
Research D 1, 151–165.
Handy, S., 2005. Critical Assessment of the Literature on the Relationships among Transportation, Land Use, and
Physical Activity. Prepared for the Transportation Research Board and Institute of Medicine Committee on
Physical Activity, Health, Transportation, and Land Use. Washington, DC. January. Available at http://trb.org/
downloads/sr282papers/sr282Handy.pdf.
Handy, S.L., Clifton, K.J., 2001. Local shopping as a strategy for reducing automobile travel. Transportation 28, 317–
346.
Handy, S., Cao, X., Mokhtarian, P.L., 2005. Does Self-selection Explain the Relationship Between Built Environment
and Walking Behavior? Empirical Evidence from Northern California. Journal of the American Planning
Association, forthcoming.
Mindali, O., Raveh, A., Salomon, I., 2004. Urban density and energy consumption: a new look at old statistics.
Transportation Research A 38, 143–162.
Newman, P., Kenworthy, J., 1999. Sustainability and Cities: Overcoming Automobile Dependence. Island Press,
Washington, DC.
Pushkarev, B., Zupan, J., 1977. Public Transportation and Land Use Policy. Bloomington University Press,
Bloomington, IN.
Schwanen, T., Mokhtarian, P.L., 2005. What affects commute mode choice: neighborhood physical structure or
preferences toward neighborhoods? Journal of Transport Geography 13, 83–99.
Singleton, R.A., Straits, B.C., 1999. Approaches to Social Research, Third Ed. Oxford University Press, New York and
Oxford.
Sommer, B., Sommer, R., 1997. A Practical Guide to Behavioral Research: Tools and Techniques, Fourth Ed. Oxford
University Press, Oxford and New York.
Transportation Research Board and Institute of Medicine, 2005. Does the Built Environment Influence Physical
Activity? Examining the Evidence, TRB Special Report 282. Washington, DC.