The Influence of Built-Form and Land Use
on Mode Choice
Evidence from the 1996 Bay Area Travel Survey
Michael
Reilly and John Landis
Final Summary Report
University of California Transportation
Center Year 12
Introduction
Interest in how the built environment influences transportation behavior has increased significantly in recent years. Concerned that traffic congestion and urban sprawl are overwhelming the human scale of American metropolitan areas, an increasing number of planners, architects, urban policy analysts and even politicians are promoting the reorganization and redesign of cities as a means of reducing the problems associated with an auto-dominated transportation system. Plans and neighborhood development forms that emphasize pedestrian comfort and convenience, these advocates argue, will promote increased walking and transit use, thereby reducing auto use and freeway congestion. And because such forms will necessarily be more compact, they will also help to reign in sprawl.
The most vocal advocates of this position are the New Urbanists and other promoters of neo-traditional design. Underlying the New Urbanist movement is a belief that designing neighborhoods, communities, and regions to be more compact and walkable will result in increased pedestrian activity, increased transit use, and decreased reliance on the private auto. Backing up this belief is the observation that people in suburbs and lower-density neighborhoods generally drive more while people in center cities and higher-density neighborhoods are more likely to take transit or walk. A closer look at the issue, however, raises deeper questions regarding cause and effect. Walking is typically easier, transit is more ubiquitous, and parking is generally dearer in older, denser cities—raising the possibility that residents of such places self-select precisely to take advantage of the greater diversity of transportation services. If indeed communities can be designed not just to be more transit- and pedestrian-friendly, but to actually get people out of their cars, then the effect of urban form on travel behavior—most notably mode choice—must be found to be consistent and robust for the fullest possible array of travelers and trip types.
This
research is an empirical investigation of the effects of land use form on
home-based non-work travel behavior among residents of the San Francisco Bay
Area. It draws on prior theoretical and empirical research into the effects of
neighborhood-scale urban form on mode choice decisions. It makes use of the
results of a detailed survey of trip-making behavior conducted by the
Metropolitan Transportation Commission in 1996 as well as geographic
information systems (GIS) measurements of
of urban form at the scale of the trip-maker rather than at the zonal
scale. It focuses on non-work trips—already the majority of daily household
trips—as opposed to more frequently studied work trips. it makes use of
The full study reviews recent research into the influences of built-form on transportation behavior, then presents a series of discrete choice models of non-work mode choice incorporating trip-maker demographic characteristics, selected trip characteristics and transit supply measures, and multiple measures of built-form. These models are tested and their results are interpreted. Finally, the research is summarized and future research directions are suggested.
research Approach
The
general form of the models tested is as follows:
Prob [mi |
M] = f [ SEi, Ti,
UFij]
where:
m denotes the non-work
travel mode chosen by traveler i
from the set of possible travel modes, M
SEi indicates the
socioeconomic characteristics of traveler i
Ti indicates selected
characteristics of the trip taken by traveler i
UFij indicates the urban
form characteristics of the home location of traveler i, measured at multiple scales, j
Following
logic first suggested by McFadden (1974), individuals are assumed to make
travel decisions to obtain the greatest amount of satisfaction (i.e., utility)
possible within the constraints imposed by their income, household role, time,
location, and transportation supply. An individual’s preferences determine how
the various characteristics of potential choices are evaluated in order to
arrive at the utility-maximizing choice. Because the interaction between these
various constraints and heterogeneous tastes is extremely complex and because
information on the relevant price signals involved is generally lacking, this
type of decision is most often modeled in a reduced form discrete choice
The BATS96 Sample
Detailed
travel behavior and household characteristic data were obtained from the
Metropolitan Transportation Commission’s 1996 Bay Area Travel Survey (BATS96),
a two-day travel diary containing travel and socio-demographic data for 14,431
individuals belonging to 5,861 households. Although broadly representative of
all Bay Area households, the BATS96 sample frame was constructed to slightly
over-sample geographic areas with extensive transit service.
Of
the 10,269 home-based non-work trips
made by BATS96 respondents over the age of sixteen, 7,915 were by car, 461 were
by bus or rail transit, and 1,893 involved walking or bicycling
Prior
to releasing the BATS96 dataset, MTC geocoded the street addresses of BATS96
respondents to latitude and longitude coordinates. This allows comparison of
the locations of BATS96 respondents with other geographic datasets, including:
(1) the Association of Bay Area Government’s 1995 hectare-scale land use
database; (2) the Census Bureau’s 1995 and 2000 TIGER street files; and, (3)
assessor’s parcel data, as obtained from Metroscan. Altogether, a total of 3,089 home locations were successfully
address-matched. Person and household socio-demographic variables are described
in the table that follows.
Select Person- and Household-Level
Socio-demographic Variables
|
Trip-maker Characteristic |
Mean |
Standard Deviation |
Observations |
|
Age |
37.4 |
21.1 |
7,873 |
|
Share over 65 |
0.11 |
0.31 |
7,873 |
|
Share under 18 |
0.26 |
0.44 |
7,873 |
|
Female share |
051 |
0.50 |
7,975 |
|
African-American share |
0.06 |
0.24 |
7,803 |
|
Hispanic share |
0.12 |
0.32 |
7,803 |
|
Asian-American share |
0.09 |
0.29 |
7,803 |
|
Share disabled |
0.01 |
0.08 |
7,819 |
|
Share w/Driver’s License |
0.74 |
0.44 |
7,969 |
|
Share Employed |
0.70 |
0.46 |
6,434 |
|
Full-Time Student share |
0.20 |
0.36 |
7,990 |
|
Share in Family |
0.92 |
0.27 |
2,352 |
|
Income |
9.53 |
3.69 |
3,113 |
|
Automobiles owned |
1.8 |
0.99 |
3,618 |
|
Auto ownership share |
0.63 |
0.48 |
3,599 |
|
Share residing in detached home |
0.63 |
0.48 |
3,618 |
Built-Form and Mode Choice: Relationships
and Measurement Issues
Geographic
information systems (GIS) was then used to develop measures of urban form at
multiple scales. The grid cells (rasters) used in this analysis measure 100
meters on a side and have an area of one hectare (10,000 square meters) or
about 2.5 acres.GIS is used to identify and summarize the built-form
characteristics of a series of multiple “spatial neighborhoods” of increasing
radius measured around each grid cell. Depending on the characteristic of
interest, a spatial neighborhood can range in width from 100 meters (the width
of a single grid cell) to four miles. Neighborhood measurements may include
counts, averages, maximum values, minimum values, majority and minority values,
and diversity measures. The values of the resulting neighborhood metrics are
then returned to the grid cell at their center and are assigned to the home
locations of all trip-makers whose addresses fall within each cell. This
process makes it possible to assign spatial measurements of urban form to
trip-maker locations at a variety of scales.
The built-form
measurements used here are all based on circular “walking neighborhoods,”
ranging in radius from one-quarter mile to one mile. Access measurements are
developed for circular neighborhoods ranging in size from one-quarter mile to
four miles.
We
considered nine measures of built-form, including population density, proximity
to commercial development, the proportion of commercial land uses within each
neighborhood, land use homogeneity, residential heterogeneity, block size,
intersection density, parcel size, and visual heterogeneity. The following
table reports these metrics for the full BATS96 sample.
Built-Form and Access Variables as Measured for BATS96
Respondent Homes
|
|
Scale
(radius) |
Mean |
Standard
Deviation |
Minimum |
Maximum |
Median |
|
Density: Persons
per hectare |
¼ mi |
36.7 |
39.5 |
0 |
331.7 |
25.7 |
|
(Source: 1990 Census) |
½ mi |
34.2 |
35.0 |
0 |
256.2 |
24.4 |
|
|
1 mi |
30.5 |
28.5 |
0 |
159.8 |
22.3 |
|
|
4 mi |
17.3 |
13.1 |
0.002 |
57.1 |
13.9 |
|
Proportion of commercial land |
¼ mi |
0.095 |
0.151 |
0 |
1 |
0.020 |
|
uses within neighborhood |
½ mi |
0.101 |
0.118 |
0 |
0.751 |
0.066 |
|
(Source: ABAG) |
1 mi |
0.100 |
0.091 |
0 |
0.553 |
0.078 |
|
|
4 mi |
0.063 |
0.030 |
0 |
0.190 |
0.066 |
|
Distance to Commercial (km) |
|
0.567 |
0.658 |
0 |
11.043 |
0.361 |
|
Dissimilarity Index |
Adj |
0.176 |
0.229 |
0 |
1 |
0 |
|
(Source: ABAG) |
¼ mi |
0.276 |
0.226 |
0 |
0.980 |
0.225 |
|
|
½ mi |
0.363 |
0.221 |
0 |
0.990 |
0.320 |
|
|
1 mi |
0.434 |
0.203 |
0 |
0.998 |
0.399 |
|
Urban Dissimilarity Index |
Adj |
0.083 |
0.267 |
-1 |
1 |
0 |
|
(Source: ABAG) |
¼ mi |
0.150 |
0.250 |
-0.980 |
0.980 |
0.102 |
|
|
½ mi |
0192 |
0.252 |
-0.995 |
0.970 |
0.147 |
|
|
1 mi |
0.208 |
0.244 |
-0.999 |
0.964 |
0.173 |
|
Transit Access |
¼ mi |
1.918 |
1.283 |
0 |
5 |
1.551 |
|
|
½ mi |
1.896 |
1.243 |
0 |
5 |
1.609 |
|
|
1 mi |
1.844 |
1.182 |
0 |
4.729 |
1.606 |
|
Intersection Density |
¼ mi |
0.515 |
0.252 |
0 |
1.918 |
0.490 |
|
(Source: 1995 TIGER) |
½ mi |
0.471 |
0.230 |
0.005 |
1.635 |
0.452 |
|
|
1 mi |
0.424 |
0.215 |
0.003 |
1.177 |
0.400 |
|
Mean Block Size |
¼ mi |
366.566 |
1238.366 |
0.834 |
14723.12 |
15.647 |
|
(Source: TIGER) |
½ mi |
449.028 |
1296.179 |
1.200 |
14723.14 |
26.668 |
|
|
1 mi |
594.248 |