CALIFORNIA TRAVEL TRENDS AND
DEMOGRAPHICS STUDY
Final Report
prepared by
Randall Crane, Abel Valenzuela,
Dan Chatman, Lisa Schweitzer, and
Peter J. Wong
Institute of Transportation Studies
University of California, Los Angeles
with
Chris Williamson and Erik Kancler
Solimar Research Group, Inc.
prepared for
California Department of Transportation
Division of Transportation Planning
Office of State Planning
December 2002
CALIFORNIA TRAVEL TRENDS AND
DEMOGRAPHICS STUDY
Final Report
Prepared by
Randall Crane ( Co- Principal Investigator)
Abel Valenzuela ( Co- Principal Investigator)
Dan Chatman
Lisa Schweitzer
Peter Wong
Institute of Transportation Studies
School of Public Policy and Social Research
University of California, Los Angeles
Los Angeles, CA 90095
with
Chris Williamson
Erik Kancler
Solimar Research Group, Inc.
973 East Main Street
Ventura, CA 93001
December 2002
Report For Contract # 74A0034
This project was funded by the California Department of Transportation. This report is the
independent product of university research and does not necessarily reflect the views of the
Department.
California Travel Trends and Demographics December 2002
Final Report
i
EXECUTIVE SUMMARY
The purpose of the Transportation Trend Analysis and Demographic Projection Study was to
analyze past population and travel trends, and project future trends, in order to support the state
infrastructure and development planning process. Tasks included:
♦ Projecting population to 2025 for the state of California at the tract level, including socio-demographic
variables likely to influence travel choice and opportunity;
♦ Developing a spatial database so that the Department of Transportation and its planning
partners can access and manipulate the projections;
♦ Implementing and testing an empirical model of travel demand using data from urban areas
in California;
♦ Combining the results of the empirical model and population projections to forecast
statewide travel trends at the Census tract level in 2015 and 2025; and
♦ Explaining how the projected population changes and travel demand trends can be used to
inform the planning of the state transportation system.
Demographic Changes and Challenges for Policymaking
We project that the population of California will increase from 33.9 million residents in 2000 to
about 48.6 million in 2025, a 44 percent increase. The share of elderly is expected to increase
significantly over this period, as is the share of non- White residents, particularly Hispanics.
How will changes in the service population affect travel needs from a policy perspective, and
what are some policy options in addressing these needs? What are the policy options to address
road congestion and continued expected preferences for automobile travel? The research
reported here provides an important input to the State’s planning to address these questions.
Travel Demand Trends
Aggregate travel by all modes will increase substantially in California. For example, auto trips
are estimated to rise nearly 40 percent from 2000 to 2025. Since most population growth will be
in urban centers, traffic congestion will worsen. The following are key findings of our study:
♦ The number of car trips per capita will decline slightly, and some travel will shift to transit
and non- motorized travel. In response to higher congestion, jobs and residences will
suburbanize.
♦ The travel impacts of an aging population will vary by area depending on the projected age
distribution. While the oldest drivers drive less often and travel shorter distances, take transit
more, and make fewer passenger- serving trips, the middle of the age distribution makes a
larger number of auto trips.
♦ Transit demand is projected to rise as a share of all trips— substantially so in parts of some
metropolitan areas. However, the net share is expected to be less than 10 percent in most
California Travel Trends and Demographics December 2002
Final Report
ii
Census tracts. The share of walk/ bike trips is expected to increase at the same rate, but from a
substantially higher base statewide.
♦ The largest percentage increases in population and travel are projected to occur in the Central
Valley and peripheral exurbs/ edge cities at the fringe of the state’s traditional metropolitan
areas, and in the highway corridors linking these areas. The degree to which these will
translate into additional road infrastructure demand depends on current and future capacity
utilization.
♦ “ Smart growth” land use and governance strategies play a limited though potentially
important role in managing transportation demand.
♦ The evolving ethnic mix of the state has numerous impacts on the transportation system. To
the extent that non- Whites and recent immigrants are more likely to have low incomes,
access to employment and transit dependence will continue to have both economic growth
and equity consequences.
The travel demand projections are based on a number of assumptions, two of which are
particularly important. First, we assume that transportation infrastructure will be provided
statewide at levels similar to the Bay Area counties in places where land use density and
population accessibility are similar. Second, we assume that measured influences of age and
race/ ethnicity on travel will stay consistent over time. These assumptions are the most reasonable
ones available given the inherent uncertainty of forecasting.
Recommendations to the State and the Department of Transportation
♦ Use the travel projections at the Census tract level statewide to compare expected future
impacts on transportation infrastructure given Department of Transportation information on
current and future state road capacity by region.
♦ Acknowledge and plan for inevitable large increases in traffic congestion. Road maintenance
and building programs are important, but large scale road infrastructure is extremely costly,
even in areas where additional right- of- way is available. Given likely constraints in funding,
focus on strategies that manage congestion wisely, such as congestion pricing.
♦ Be sensitive to the needs of the carless and transit- dependent, particularly in areas that will
experience high amounts of auto demand. Such areas may be the appropriate recipients of
any funds for paratransit, auto ownership assistance, and van programs.
♦ Provide state support for walking and biking infrastructure, since these modes have
substantially higher shares of travel than transit, and will experience greater increases in
demand.
♦ Target “ smart growth” and transit development planning or funding in areas that anticipate
high demand for walk/ bike and transit modes. Carefully identify areas that will exceed
population accessibility thresholds ( for example, areas with more than 200,000 population
within a five mile radius— see Sections 4 and 7) as the best candidates.
California Travel Trends and Demographics December 2002
Final Report
iii
ACKNOWLEDGEMENTS
In addition to the authors listed on the cover page, other individuals contributed to this report.
Most of the GIS maps were prepared by Kimiko Shiki. Sheryll Del Rosario and Melissa Hatcher
carried out research supporting the demographic projections. We particularly thank Kimiko
Shiki, Doug Miller, and Doug Houston for their assistance in creating ArcView scripts to
calculate accessibility indices.
In order to complete the work, we solicited data and information from numerous agencies.
Planners at the local and regional level throughout the state provided us with their agencies’
demographic projections. Regional agency staff graciously shared their travel diary data. In
particular, Charles Purvis and Kenneth Vaughn of the San Francisco Bay Area Metropolitan
Transportation Commission made the Bay Area Travel Survey available, explained the data set,
and provided other data that were key inputs to the travel modeling process. In addition, Pablo
Guttierez from the Southern California Association of Governments and Gillian van Oosten
Biedler from the Sacramento Council of Governments made their regional travel diary
information available.
We conducted our research in a very supportive research environment. We want to thank D.
Gregg Doyle and Brian Taylor for making their US Department of Transportation- sponsored
literature review on women and transportation available to us. We received the benefit of Dowell
Myers’s considerable expertise in demography and housing issues. Discussions with colleagues
including Brian Taylor, Evy Blumenberg, and Hiro Iseki provided useful feedback. In our
business office, Gertrude Lewis, Janet Peltier, Robert Duncan, and Anna Diep managed the
administrative side of the project, handled contracts, kept the bills paid, and responded to the
research team’s questions and requests. At the Center for the Study of Urban Poverty, Gretchen
Baumhover provided assistance in arranging travel, meetings, and report preparation.
The upper- left photo on the cover depicts a pedestrian crosswalk at the 12th Street BART station
in Oakland. The lower- right photo was taken on Hollywood Boulevard near the Hollywood and
Highland development in Los Angeles. The photos are provided courtesy of Terry Parker in the
Division of Mass Transportation, California Department of Transportation.
California Travel Trends and Demographics December 2002
Final Report
iv
CONTENTS
Executive Summary....................................................................................................................... i
Demographic Changes and Challenges for Policymaking................................................... i
Travel Demand Trends ........................................................................................................ i
Recommendations to the State and the Department of Transportation .............................. ii
Acknowledgements ...................................................................................................................... iii
Contents ............................................................................................................................... ........ iv
List of Figures........................................................................................................................ ..... vii
List of Maps........................................................................................................................... .... viii
List of Tables ............................................................................................................................... xi
1. Introduction................................................................................................................... ........... 1
1.1 Policy Context............................................................................................................... 1
1.2 Research Objectives...................................................................................................... 2
1.3 Data ............................................................................................................................... 3
1.4 Research Approach ....................................................................................................... 4
1.5 Organization of the Report............................................................................................ 4
1.6 Principal Findings and Recommendations.................................................................... 4
2. Demographics, Land Use, and Travel..................................................................................... 7
2. 1 Race/ Ethnicity, Sex, and Mobility ............................................................................... 7
2.2. Travel and the Elderly.................................................................................................. 9
Increasing per capita travel ....................................................................................... 10
The impact of health concerns .................................................................................. 10
Driving safety............................................................................................................ 11
Location and auto dependence.................................................................................. 12
2.3 Land Use Influences on Travel ................................................................................... 12
Development density ................................................................................................ 12
Mixed land uses ........................................................................................................ 16
Accessibility.............................................................................................................. 17
Summary ................................................................................................................... 18
2.4 Lessons for the California Demographics and Trend Study....................................... 18
3. California Travel Today & Yesterday .................................................................................. 21
3.1 Trip Purpose................................................................................................................ 21
California Travel Trends and Demographics December 2002
Final Report
v
3.2 Temporal Distribution................................................................................................. 23
3.3 Modal Distribution...................................................................................................... 24
3.4 Trip Length ................................................................................................................. 25
3.5 Individual and Household Travel Behavior................................................................ 28
Variations by sex....................................................................................................... 28
Variations by income ................................................................................................ 28
Car licensing ............................................................................................................. 29
3.6 Car Ownership, Household Size and Income ............................................................. 30
4. Empirical Travel Modeling.................................................................................................... 33
4.1 Notes on Empirical Models ........................................................................................ 33
Random utility theory ............................................................................................... 33
Activity- based models .............................................................................................. 34
Aggregate versus disaggregate models..................................................................... 34
4.2 The Bay Area Travel Survey ...................................................................................... 35
Trips ......................................................................................................................... 36
Trips by purpose ....................................................................................................... 37
Trips by mode ........................................................................................................... 39
Travel duration.......................................................................................................... 42
4.3 Demographic Characteristics ...................................................................................... 43
Race and ethnicity..................................................................................................... 44
Age ......................................................................................................................... 45
Sex ......................................................................................................................... 47
Discussion of demographic variables ....................................................................... 48
4.4 Land Use Variables..................................................................................................... 49
Gross residential density ........................................................................................... 50
Population accessibility ............................................................................................ 52
Discussion of land use variables ............................................................................... 53
4.5 Basic Travel Models ................................................................................................... 53
Independent variables ............................................................................................... 53
Dependent variables for basic travel models ............................................................ 54
Results from the basic trip count models.................................................................. 55
Results from the basic travel duration models .......................................................... 67
4.6 Complex Travel Models ............................................................................................. 69
Enriched demographic models.................................................................................. 69
Enriched land use models ......................................................................................... 70
5. Population Projections............................................................................................................ 73
5.1 Projection Methodology.............................................................................................. 73
5.2. Revised Methodology ................................................................................................ 74
5.3 Confidence in Results ................................................................................................. 77
6. Travel Forecasts...................................................................................................................... 81
6.1 Methodology............................................................................................................... 81
California Travel Trends and Demographics December 2002
Final Report
vi
Interpreting Forecast Results .................................................................................... 82
6.2 Statewide Forecast Results ......................................................................................... 84
6.3 Interpreting Travel Demand Trends............................................................................ 92
6.4 Summary..................................................................................................................... 93
7. Conclusions.................................................................................................................... ......... 95
7.1 Racial/ Ethnic Diversity............................................................................................... 95
7.2 Transporting Seniors................................................................................................... 97
7.3 Managing a Changing Population Distribution .......................................................... 97
7.4 Recommendations....................................................................................................... 99
Works Cited.......................................................................................................................... .... 101
APPENDICES
A. Demographic GIS File Documentation.............................................................................. 107
State Elevation and Landform Images............................................................................ 107
Reference Features.......................................................................................................... 107
Population Projections Files ........................................................................................... 108
B. Functional Forms for Travel Demand Models.................................................................. 111
C. Complex Models: Demographics........................................................................................ 113
D. Complex Models: Land Use ................................................................................................ 129
E. Demographic Projection Maps ........................................................................................... 145
F. Travel Trend Maps: Statewide............................................................................................ 158
G. Travel Trend Maps: Bay Area and Sacramento Region.................................................. 175
H. Travel Trend Maps: Southern California ......................................................................... 184
California Travel Trends and Demographics December 2002
Final Report
vii
FIGURES
Figure 1. Vehicle Miles Traveled in California, 1960 to 2000....................................................... 2
Figure 2. Trip Duration in Rural Regions..................................................................................... 27
Figure 3. Trip Duration in Los Angeles Region .......................................................................... 27
Figure 4. Trips per Person ( Simple Definition) ............................................................................ 36
Figure 5. Trips Per Person ( Refined Definition)........................................................................... 37
Figure 6. Auto Trips Per Person, Two- Day Period....................................................................... 41
Figure 7. Transit Trips Per Person for Transit Riders................................................................... 42
Figure 8. Predicted Trips using Age, Age Squared, and Age Cubed ( Negative Binomial
Regression) ........................................................................................................................... 46
Figure 9. Predicted Trips Using Age Categories .......................................................................... 46
Figure 10. Licensing Rates by Age Category ............................................................................... 49
Figure 11. Gross Residential Density for Respondent Transportation Analysis Zones ............... 51
Figure 12. Population Access Index for Respondents by Transportation Analysis Zones.......... 52
California Travel Trends and Demographics December 2002
Final Report
viii
MAPS
Map 1. Areas Newly Exceeding Population Accessibility of 200,000, Southern California Area,
2000 to 2025 ......................................................................................................................... 83
Map 2. Absolute Per Capita Trip Growth, 2000 to 2025 ( Trips per Person per Day).................. 86
Map 3. Changes from 2000 to 2025 in Southern California- Trips per Capita per Day................ 87
Map 4. Changes from 2000 to 2025 in Southern California- ....................................................... 88
Map 5. Changes from 2000 to 2025 in Bay Area: Trips per Capita per Day ............................... 89
Map 6. Changes in Daily Trips per Square Mile, Bay Area/ Sacramento, 2000 to 2025.............. 90
Map 7. Changes in Daily Per Capita Auto/ POV Trips, Bay Area/ Sacramento, 2000 to 2025..... 91
Map 8. Changes in Car/ POV Trips per Capita, Southern California, 2000 to 2025..................... 92
Map 9. Projected Change in Percentage Share of Non- Whites, 2000 to 2025 ............................. 96
Map 10: Census 2000 Population Density, State of California .................................................. 146
Map 11: 2015 Projected Population Density, State of California............................................... 147
Map 12: 2025 Projected Population Density, State of California............................................... 148
Map 13: Census 2000 Population Density, Bay Area................................................................. 149
Map 14: 2015 Projected Population Density, Bay Area ............................................................. 150
Map 15: 2025 Projected Population Density, Bay Area ............................................................. 151
Map 16: Census 2000 Population Density, Southern California ................................................ 152
Map 17: 2015 Projected Population Density, Southern California............................................. 153
Map 18: 2025 Projected Population Density, Southern California............................................. 154
Map 19: Census 2000 Population Density, Southern Central Valley......................................... 155
Map 20: 2015 Projected Population Density, Southern Central Valley ..................................... 156
Map 21: 2025 Projected Population Density, Southern Central Valley ..................................... 157
Map 22: Changes in Daily Total Trips, 2000 to 2025 ................................................................ 159
Map 23: Changes in Daily Car Trips, 2000 to 2025 ................................................................... 160
California Travel Trends and Demographics December 2002
Final Report
ix
Map 24: Changes in Daily Transit Trips, 2000 to 2025 ............................................................. 161
Map 25: Changes in Daily Walk/ Bike Trips, 2000 to 2025 ....................................................... 162
Map 26: Changes in Daily Trips Per Capita, 2000 to 2025 ........................................................ 163
Map 27: Changes in Daily Car Trips Per Capita, 2000 to 2025 ................................................. 164
Map 28: Changes in Daily Transit Trips Per Capita, 2000 to 2025............................................ 165
Map 29: Changes in Daily Walk/ Bike Trips Per Capita, 2000 to 2025...................................... 166
Map 30: Changes in Daily Trip Density, 2000 to 2025.............................................................. 167
Map 31: Changes in Daily Auto Trip Density, 2000 to 2025..................................................... 168
Map 32: Changes in Daily Transit Trip Density, 2000 to 2025 ................................................. 169
Map 33: Changes in Daily Walk/ Bike Trip Density, 2000 to 2025............................................ 170
Map 34: Changes in Daily Trip Duration Density, 2000 to 2025............................................... 171
Map 35: Changes in Daily Auto Trip Duration Density, 2000 to 2025...................................... 172
Map 36: Changes in Daily Transit Trip Duration Density, 2000 to 2025 .................................. 173
Map 37: Changes in Daily Walk/ Bike Trip Duration Density, 2000 to 2025 ............................ 174
Map 38: Changes in Daily Trip Density, Bay Area/ Sacramento, 2000 to 2025......................... 176
Map 39: Changes in Daily Auto Trip Density, Bay Area/ Sacramento, 2000 to 2025................ 177
Map 40: Changes in Daily Transit Trip Density, Bay Area/ Sacramento, 2000 to 2025 ............ 178
Map 41: Changes in Daily Walk/ Bike Trip Density, Bay Area/ Sacramento, 2000 to 2025 ...... 179
Map 42: Changes in Daily Trips Per Capita, Bay Area/ Sacramento, 2000 to 2025................... 180
Map 43: Changes in Daily Auto Trips Per Capita, Bay Area/ Sacramento, 2000 to 2025.......... 181
Map 44: Changes in Daily Transit Trips Per Capita, Bay Area/ Sacramento, 2000 to 2025 ...... 182
Map 45: Changes in Daily Walk/ Bike Trips Per Capita, Bay Area/ Sacramento, 2000 to 2025 183
Map 46: Changes in Southern California Daily Trip Density, 2000 to 2025 ............................. 185
Map 47: Changes in Southern California Daily Auto Trip Density, 2000 to 2025 .................... 186
Map 48: Changes in Southern California Daily Transit Trip Density, 2000 to 2025................. 187
California Travel Trends and Demographics December 2002
Final Report
x
Map 49: Changes in Southern California Daily Walk/ Bike Trip Density, 2000 to 2025........... 188
Map 50: Changes in Southern California Daily Trips Per Capita, 2000 to 2025 ....................... 189
Map 51: Changes in Southern California Daily Auto Trips Per Capita, 2000 to 2025 .............. 190
Map 52: Changes in Southern California Daily Transit Trips Per Capita, 2000 to 2025 ........... 191
Map 53: Changes in Southern California Daily Walk/ Bike Trips Per Capita, 2000 to 2025 ..... 192
California Travel Trends and Demographics December 2002
Final Report
xi
TABLES
Table 1: Built Environment and Land Use Characteristics Thought to Affect Individual and
Household Travel Behavior .................................................................................................. 13
Table 2. Percent of Weekday Trips by Purpose in California ..................................................... 22
Table 3. Trip Purpose by Metropolitan Region ............................................................................ 23
Table 4. Trip Timing..................................................................................................................... 24
Table 5. Travel Mode by Travel Time from SCAG and SACOG................................................ 25
Table 6. Trip Length in Minutes................................................................................................... 26
Table 7. Percentage of Licensed Drivers by Age and Sex........................................................... 30
Table 8. Vehicles per Household in California Regions............................................................... 31
Table 9. Vehicles per Household by Income: SCAG ................................................................... 31
Table 10. Vehicles per Household: SACOG ................................................................................ 32
Table 11. Vehicles per Household: BATS.................................................................................... 32
Table 12. Activities by Type......................................................................................................... 38
Table 13. Trips Away from Home, By Purpose ........................................................................... 39
Table 14. Trip Purposes Included in Nonwork Category ............................................................. 39
Table 15. Travel Mode for All Trips and Trip Segments ............................................................. 40
Table 16. Travel Mode for Trips Away from Home .................................................................... 40
Table 17. Average Trip Duration by Mode .................................................................................. 43
Table 18. Total Travel Duration by Mode, 2- Day Period............................................................. 43
Table 19. BATS Survey Respondents by Race/ Ethnicity............................................................. 44
Table 20. Average Trips by Racial/ Ethnic Group By Purpose..................................................... 45
Table 21. Average Trips by Racial/ Ethnic Group by Mode ......................................................... 45
Table 22. Persons by Age Category.............................................................................................. 47
Table 23. Average Trips by Sex by Purpose, All Ages ................................................................ 47
California Travel Trends and Demographics December 2002
Final Report
xii
Table 24. Average Trips by Sex by Mode .................................................................................... 48
Table 25. Licensing Rates by Age and by Sex ............................................................................. 50
Table 26. Basic Trip Count Model, Part 1: Demographic Variables............................................ 57
Table 27. Average Number of Work and School/ Daycare Trips by Age Category ..................... 60
Table 28. Average Nonwork Trips, by Age and by Mode............................................................ 61
Table 29. Basic Trip Count Model, Part 2: Land Use Variables .................................................. 62
Table 30. Basic Travel Duration Model, Part 1: Demographic Variables.................................... 65
Table 31: Basic Travel Duration Model, Part 2: Land Use Variables .......................................... 68
Table 32. Variables Available from AGS..................................................................................... 75
Table 33. Total Trip and Trip Duration Trends, 2000 to 2025 ..................................................... 84
Table 34. Daily Per Capita Trip and Travel Duration Trends, 2000 to 2025 ............................... 85
Table 35. Fields for County_ projections_ final ........................................................................... 108
Table 36. Fields for Tract_ projections_ final .............................................................................. 109
California Travel Trends and Demographics December 2002
Final Report
1
1
INTRODUCTION
In collaboration with its regional and citizen planning partners, the California Department of
Transportation is currently developing a long- term, multimodal transportation plan for the state
of California. The California Travel Trends and Demographics study was designed to support the
data requirements of the statewide plan. The purpose of this project is to enable the State to
develop overall policy to accommodate future statewide trends. The research did not include
identifying transportation infrastructure needs for specific geographical areas or transportation
corridors.
Phase I of the project, completed by UC Berkeley, is a comprehensive overview of the major
social and economic forces that will affect transportation in California over the next 25 years.
Phase II, conducted by UCLA and its research partner Solimar Research Group, developed
population projections by Census tract for years 2015 and 2025, and integrated those projections
with Census 2000 geography in a GIS database for the state. Phase III of the research, completed
by UCLA, projects travel demand trends to 2015 and 2025, applying several empirical travel
models to the projections developed in Phase II.
1.1 Policy Context
California’s total population is projected to grow by about 15 million residents over the next 25
years. Many newcomers to the state will be recent immigrants, many of whom are young and
whose children will grow up and remain in California. But a substantial part of California’s
growth is expected to come from natural increase, that is, from the state’s existing residents
having and raising children in California. As the population grows over time, so does the demand
for travel in the state ( see Figure 1, below).
As metropolitan areas grow and disperse outward, existing communities in the inner cities and in
older suburbs contend with spatial isolation from jobs and procedural inequities in growth
management decisions. Affordable access to opportunities assumes great importance in the light
of growth pressures. Welfare reform and the transition from state dependency to work likewise
hinges, in part, on understanding how transportation services can either open up or deny
opportunities to vulnerable groups in California.
In addition to concerns about social equity, all Californians have a stake in future land and
infrastructure development. Countering residents’ need for mobility and housing is the equally
compelling need to protect California’s unique natural resources from the ravages that have
accompanied previous development. Wildfire destruction, utility crises, air quality well below
federal standards, and water quality issues loom as the possible consequences of poor planning
California Travel Trends and Demographics December 2002
Final Report
2
and foresight. These needs will become more pressing as California’s near- capacity
transportation system prepares to take on the demands of future growth.
Figure 1. Vehicle Miles Traveled in California, 1960 to 2000
VMT
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
160,000
180,000
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
Year
Total VMT
Urban
Rural
SOURCES: 1960- 1972 data from Table VM- 2C of “ Historical State Highway, County Road and City Street
Statistics 1960 – 1972” provided by Division of Highways Traffic Branch; 1973- 1977 data from “ California
Table TA- 1, Statewide Mileage, Travel and Non- Fatal Accidents” by Highway Planning and Research Branch;
1978- 1995 data from the yearly tabulation “ Statewide TA- 1 Data”, Department of Transportation, Traffic
Operations Program. Program. 1997- 1999 data provided by Traffic Operations publication: 1999 Accident
Data on California State Highways, Statewide Travel and Accidents Rates ( page 7).
In order to address these complex issues, planning agencies in the state need information on the
interactions among socioeconomic, activity, land use, and travel behavior in California over a
long planning horizon.
1.2 Research Objectives
We analyzed past transportation and population trends in order to look at the possible
consequences of future infrastructure and development policies. The purpose of this project was
to provide high- quality population forecasts with substantial geographic and demographic detail,
and to understand how demographic and land use changes in the state will affect future travel
demand.
California Travel Trends and Demographics December 2002
Final Report
3
Our specific research objectives were to:
♦ Project population for the state of California at the Census tract level, including socio-demographic
variables likely to influence travel choice and opportunity;
♦ Develop a spatial database and GIS files so that the Department of Transportation and its
planning partners can access and manipulate the projections;
♦ Estimate and test an empirical model of travel in California, based on socio- demographic and
policy variables;
♦ Use the results of the empirical model and population projections to forecast travel in 2015
and 2025; and
♦ Recommend ways that the research can used to inform planning and policy making.
Given the extent of the work required on this project, in this report we summarize the results of
the demographic projections, empirical modeling effort, and travel demand forecasts rather than
describing all the results in detail. The appendix to this report contains further information. The
detailed demographic projections, travel demand trends, and maps will be made available in
electronic form.
1.3 Data
Our study draws on a wide array of local, regional, and national data sources:
♦ Micro- data and block- group level data from the 1970, 1980, and 1990 US Census that
include demographic, employment, and transportation characteristics;
♦ Data from the Nationwide Personal Transportation Survey ( NPTS) of the US Department of
Transportation;
♦ Population projections created and maintained by the California Department of Finance;
♦ Population projections prepared by local, county, and regional agencies throughout the state
of California;
♦ Population projections to 2011 prepared by the Applied Geographic Solutions, a private
company;
♦ Tract- level data from the 2000 Census that include demographic, employment, and
transportation characteristics;
♦ Travel survey data from the Southern California Association of Governments, the
Sacramento Council of Governments, and the Metropolitan Transportation Commission; and
♦ Data from the 2000- 2001 California statewide travel survey.
More detail on the data and methodology for this study is included in each section of the report.
California Travel Trends and Demographics December 2002
Final Report
4
1.4 Research Approach
The research effort for this projection consisted of six stages: a review of relevant literature
background information, data collection, population projections, GIS mapping and development,
empirical demand modeling, and travel forecasts. Each of these stages is described in a separate
section in the remainder of the report.
We used national, state, regional and local data for information about existing population and
travel behavior characteristics. The population data were used to construct demographic
projections to 2015 and 2025. In order to forecast travel, travel diary data were used to develop
and test an empirical model of current travel choices ( trips and travel duration), focusing on
readily available measures of demographics and land use. Using the coefficients from this model
and the demographic projections, we developed travel demand projections for the state of
California.
1.5 Organization of the Report
This report is divided into eight sections, including this introduction and appendices:
Section 1 summarizes the report goals, data, methods, and findings.
Section 2 presents and interprets key findings of research on travel behavior,
demographics, and urban form.
Section 3 describes the existing population and current travel patterns in California.
Section 4 develops an empirical travel model that quantifies relationships between
individual travel behavior and demographic and land use variables.
Section 5 describes population projection modeling methods and describes statewide
results.
Section 6 applies the results of the empirical travel model to the population projections in
order to project travel demand trends.
Section 7 presents conclusions and recommendations.
1.6 Principal Findings and Recommendations
We carried out more than a hundred empirical travel models using the Bay Area survey data,
varying by:
• travel measure ( trips, and time spent traveling),
• mode ( personally operated vehicle, transit, or walk/ bike),
• trip purpose ( work/ school/ daycare, non- work, and passenger- serving),
• a set of independent variables used to explain the travel behavior measure ( e. g.,
race/ ethnicity, sex, age, household income, household structure, household vehicle
ownership, employment status, licensing status, and various measures of land use in
and around the household residence zone).
California Travel Trends and Demographics December 2002
Final Report
5
Despite the complexity, some common themes emerge from the basic empirical model that is
used for the travel demand forecasts. Age, sex and race/ ethnicity are correlated with trip making
by mode by purpose, and to travel duration by mode, in the following ways, when controlling for
all three factors simultaneously ( as well as for gross population density and a five- mile radius
population accessibility index, explained below).
First, increasing age up to the 40 to 50 age category is associated with an increasing number of
trips for all purposes ( work, non- work, and passenger- serving). After that time, increasing age
implies a decrease in trip making and travel duration. For example, overall time spent traveling
on all modes decreases about five minutes per day for every five- year increase in age over age
50. This difference accelerates rapidly in older cohorts, so that those aged 80 and above travel
about 45 minutes to an hour less on average than those in the peak 40 to 50 age range.
Second, those in non- white race/ ethnicity categories make fewer trips than Whites, but travel
about the same amount of time per day. The difference is apparently because these groups make
a higher share of their trips via transit. African- Americans make almost a half- trip less than
Whites per day, while Asian Americans / Pacific Islanders and Hispanics make about a third of a
trip less per day. Most of these differences are not due to work trips. Controlling for the other
variables included in the basic model, Asian Americans / Pacific Islanders and Hispanics make
about the same number of auto trips for work/ school/ daycare purposes, while African Americans
make just slightly fewer ( about one- tenth of a work trip by car less). Hispanics and African
Americans make just slightly more work trips by transit, and Asian Americans / Pacific Islanders
and Hispanics make just slightly fewer work trips on foot or bike. In the non- work trip category,
the non- White groups make fewer auto trips, averaging about a quarter trip less per day than
Whites, and fewer walk/ bike trips. African Americans make slightly more non- work transit trips
than the other groups ( about a tenth of a trip per day).
Third, women currently make fewer work trips than men across age categories, but consistently
make more passenger- serving and non- work trips. These differences are primarily due to
differences in trips by auto; by mode, women's share of all trips by walk/ bike and by transit is
higher than men's, to the extent that their number of work trips by transit and walk/ bike is very
close to that of men for all three trip purposes.
These relationships decline somewhat in importance when household income is added to the
models. Higher household income increases trip making by auto and decreases it by transit, with
an ambiguous effect on walk/ bike trips.
Despite the statistically significant relationships in the Bay Area survey data, the magnitude of
the relationships is relatively small, accounting for ten percent or less of individual variation in
trip making. Since unobserved factors are clearly more important than observed factors in
influencing travel behavior, forecasts based on observed factors must be interpreted with caution.
The empirical models are used to forecast travel demand by Census tract statewide. These
projections are mapped for the state, for Department of Transportation districts, and for selected
regions in Appendices F through H.
California Travel Trends and Demographics December 2002
Final Report
6
As explained below, the travel projections require careful interpretation and should be thought of
as broadly indicative rather than precise. Of course, they primarily show that we can expect
travel to be concentrated where the population is most concentrated. Beyond this, some
interesting results emerge. For example, under the assumption that transit options are available
everywhere, the projections show that the highest per capita demand for transit would be
predicted to increase slightly over time in areas that exceed particular density thresholds. In other
words, if transit were provided in such places, it would be used at a slightly higher rate over
time. These results are discussed in more detail in Sections 6 and 7.
California Travel Trends and Demographics December 2002
Final Report
7
2
DEMOGRAPHICS, LAND USE, AND TRAVEL
In this section we review empirical research in two main areas: the variance of travel behavior
and demographics, with special attention to travel of the elderly; and the influence of land uses
on travel behavior. The intent of the review is not to describe issues in California, though many
of these studies were conducted using California data. Instead, it is to motivate the empirical and
forecast models, as well as to assist in interpreting and supplementing the results of those
models.
2. 1 Race/ Ethnicity, Sex, and Mobility
Research on travel behavior has often concerned itself with urban inequality and economic
isolation. Two categories of research stand out: work that has quantified differences in travel by
population subgroup ( e. g., ethnicity, age, and sex), and “ spatial mismatch” research, which has
examined the effects of changing urban labor and spatial structures on inner city residents. In this
section we focus primarily on representative literature in the first category.
Rosenbloom ( 1995) finds that women make more person trips per day than do men in the US.
However, women make shorter trips, whereas men travel 27 percent more person- miles than
comparable women in urban areas and 16 percent more in rural areas. Low income people of
both sexes in urban areas and low income women in rural areas work farther from home than
comparable people from households making more money. At the very lowest income levels,
women workers traveled farther than comparable male workers.
Ethnicity is also thought to influence travel. In general, travel data suggest that white men travel
more than all other men, and white women traveled more than all other women. Hispanic women
and those from other races make fewer trips than comparable men. In a study of 1995 data from
the Nationwide Personal Transportation Survey, the difference between Hispanic men and
women on all indicators of travel were two to three times greater than the differences between
the sexes in any other grouping ( Rosenbloom 1995).
Doyle and Taylor ( 1999) study variation in metropolitan travel behavior by sex and ethnicity.
They find that ethnicity appears to be a more important influence than sex on mode choice and
commuting behavior, although sex differences persist, especially by household type. They find
that ethnicity plays a major role in commuting distance and duration. For example, African
American women have the longest commute times of any group. In addition, women of color,
especially those living in central cities, have disproportionately longer commute times, which
can be largely explained by their lower incomes, their greater tendency to use transit and walk,
their greater household responsibilities, and their lower levels of education. Finally, the authors
find that women make more trips per day on average because they make more stops for shopping
California Travel Trends and Demographics December 2002
Final Report
8
and household- serving purposes. Working women are likely to chain these errands into their
commute trips.
Giuliano ( 2000) documents racial differences in four travel categories: daily travel distances,
time spent traveling, number of person trips, and trip mode. She finds significant differences in
the distance and time traveled by different racial groups. Whites travel the farthest and make the
most trips, while African Americans have the longest travel durations. Trips made by personal
vehicle are the overwhelming majority of all person trips regardless of race/ ethnicity. Significant
differences exist among racial groups for other modes such as transit and walking. Using
multivariate analysis, Giuliano finds that racial and ethnic differences are not only limited to
effects explained by different location patterns, but rather by fundamental differences in what
motivates travel and location choices. She argues that spatial location patterns seem to provide
the best explanation of differences among whites, African Americans, Hispanics, while for Asian
Americans, differences reflect different travel choice processes.
Papers by Chu, Polzin, Rey, and Hill ( 1999) and Polzin, Chu, and Rey ( 1999) analyze both the
amount of travel and mode choice for non- work travel by people of color. Chu et al. ( 1999)
provide rich descriptive data on trip making in 1995 and an analysis of how the rate of travel
changed from 1983 to 1995, using the Nationwide Personal Transportation Survey. They find
that whites made about two percent more trips than the national average, while trip making for
people of color was lower. Among people of color, Hispanics had the highest trip rate ( about two
percent below the national average) while Asians made the fewest trips ( about 15 percent below
the national average). They also find that average non- work trip making for non- work travel
among the racial and ethnic groups changes little with personal, household, and geographic
characteristics. For all racial/ ethnic groups, non- work travel increased over time for several
different measures of mobility ( e. g., person trips, person miles, vehicle trips, vehicle miles, and
person hours). Mobility grew at a much faster rate for people of color than for the white
population during 1983- 1995. Among people of color, Hispanic mobility grew at the highest
rate, followed by African- Americans and other groups.
Using descriptive statistics and multivariate analysis, Polzin et al. ( 1999) find that non- Whites
are several times as likely as whites to use public transit for non- work travel and about twice as
likely as Whites to walk for non- work travel. African Americans are nine times as likely and
other peoples of color are two to three times as likely, as whites to use public transit for non-work
travel.
One final factor that may be as important as ethnicity is immigration ( Myers and Park 1996).
Spain ( 1997) pointed out that immigrants now make up approximately 10 percent of the elderly
population, with the highest proportions of elderly foreign- born living in California, New York,
and Florida. Forty- one percent of immigrants who entered the US during the 1980s speak no
English. Economically, nearly one- quarter of the older immigrants live in poverty. Immigrants
who are poor and are not part of the workforce when they arrive in this country are likely to be
permanently limited in their travel options as they age. On the other hand, immigrants who
become part of the workforce and have rising incomes may be more likely to have gained
automobile access and continue such mobility into old age.
California Travel Trends and Demographics December 2002
Final Report
9
In summary, there are numerous differences among racial/ ethnic groups in the frequency, length,
duration and mode of travel. As a result, differences exist by income level, because non- white
ethnic groups tend to have lower incomes. Second, because these papers are national in scope,
they fail to address differences in regional or city/ urban contexts. As a result, caution should be
taken when analyzing national data, especially when it points to differential outcomes by
ethnicity. National figures on most measures of inequality often mask significant differences in
social economic indicators regarding the effect of ethnicity.
2.2. Travel and the Elderly
By the year 2030, up to 20 percent of the population of the United States— over 50 million
people— will be aged 65 years or more. While this reflects the progression of the “ baby- boom”
generation into their golden years, it also reflects the fact that health care and medical
developments have extended life expectancy for Americans. Those over 80 years of age are in
the fastest growing cohort, meaning that there will be a larger- than- ever group of people who are
particularly dependent on family, friends, or public transportation services for mobility, and
who— in the absence of these— may have seriously limited mobility and life activities.
The increasing numbers of older residents will also be more diverse, in terms of both ethnicity
and lifestyle. Spain ( 1997) found that 87 percent of the elderly were white in 1990, and estimated
that if current fertility differentials persist and immigration remains the same, 65 percent will be
white, 11 percent African American, 15 percent Hispanic, and 8 percent Asian American in
2050. In a study of Los Angeles 25 years ago, Wachs et al. ( 1976) observed that the elderly may
be as heterogeneous as younger population groups, and a variety of lifestyle groups may exist
among older populations of metropolitan communities. Thus, it may be important to identify
subgroups of elderly persons on the basis of their past travel behavior. The implication for
transportation planning is that as the population ages, the differences among the elderly will
become as important as the differences between the elderly and the non- elderly ( Spain 1997).
Wachs et al. ( 1976) noted that one important demographic effect of aging was the creation of
single- adult households, most often widows. Spain ( 1997) found that older women are more
likely than older men to be widowed and live alone. She found that the percentage of women
aged 75 and over who live alone rose from 37 to 53 percent between 1970 and 1996 ( Spain
1997). However, this tendency also varies by ethnicity. Elderly white women are more likely to
be living alone than elderly women of color. In addition, elderly white women are more likely to
reside in less dense suburban areas and as a result may require different transportation services
than needed by the elderly living in extended- family households in inner- city areas.
Critical to the analysis of elderly transportation needs in the future are demographic and
geographic trends among senior citizens. If longevity and immigration cause a larger proportion
of the elderly to live in the inner city or the suburbs, this will have implications for the types of
service likely to be needed. Spain ( 1997) argues that non- Whites lead more geographically
constricted lives than non- Hispanic Whites. Since the older population is predicted to be more
racially and ethnically diverse in the future than it is now, the increases in travel associated with
baby- boom women’s increased independence could possibly be tempered by larger proportions
of minorities who are more geographically constricted.
California Travel Trends and Demographics December 2002
Final Report
10
But we cannot be sure that the elderly living in cities tomorrow will have travel patterns similar
to the elderly living in cities today, nor that future elderly persons of color will have similar
travel patterns. If younger people of color continue to have lower rates of automobile ownership
and driver’s licensing, and tend to locate in denser central cities with good transit and walking
access, as these individuals age they may be continue to rely on public transportation or walking.
But if people of color ( particularly, immigrants) increase their ownership and use of automobiles
at the same rates that women have historically done, this may not be the case.
Increasing per capita travel
Due to both increased licensing rates— particularly among women— and to more active lifestyles
later in life, the amount of daily travel per elderly person is expected to increase, independent of
the overall size of the elderly population. According to Coughlin and Lacombe ( 1997), trends
indicate that today’s seniors are more active than previous generations. The lifestyle of what
might be called the ‘ new elderly’ includes many activities that, in years past, may have been
considered unusual pursuits for those over 65 ( Wachs et al. 1976).
Spain ( 1997) noted that for today’s older married woman, the husband is more likely to be the
driver and the wife to travel as a passenger. However, if baby boom women keep their licenses
and continue to drive into an advanced age, it would cause an increase in the number of vehicles,
number of trips, and miles traveled as compared to the elderly women generation today. In
general, as the health of the elderly improves, they are likely to travel similarly to how they
traveled when working, but without the commute trip ( Coughlin and Lacombe 1997). This
similarity has its greatest consequences with respect to women, because elderly women who do
not drive now are likely never to have been licensed. In contrast, middle- aged women driving
today are much more likely than their foremothers to drive well into old age ( Spain 1997).
The impact of health concerns
Health concerns such as the increased need for medical- related urban travel among the elderly
make it more difficult for them to travel on their own ( Spain 1997). However, frailty does not
mean that these seniors no longer wish to participate in out- of- home activities. Alternative
transportation services could be made available so that the eldest elderly may maintain as much
dignity, independence and choice as possible, for as long as possible ( Coughlin and Lacombe
1997). Strategies to accommodate the mobility needs of the elderly should incorporate many
modes. In order to facilitate mobility and access for seniors, transportation planning should
incorporate elderly residents in all possible roles— as drivers, passengers, transit riders, delivery-recipients,
cyclists, and pedestrians.
While the elderly rely primarily on their cars for mobility, there are some trips which do not
require automobile access. In 1976, Wachs et al. found that for urban residents in Los Angeles
County, a high proportion of trips were made on public transit. However, as overall transit
ridership has declined and has also shifted toward commute trips, it is likely that the proportion
of trips by the elderly on public transportation has also declined. In a more recent study,
Coughlin and Lacombe ( 1997) suggested that the elderly still walk and even ride bicycles for
some trips. The mode choice that the elderly use may largely depend on the quality of options
available and the perceived risk involved with each. For example, alternatives to driving,
California Travel Trends and Demographics December 2002
Final Report
11
including walking, cycling, and riding transit, may not be appealing if the traveler is physically
frail or feels vulnerable in more public travel settings.
Households or individuals without cars or driver's licenses are the most likely to use alternative
modes. Spain ( 1997) found that even when licensed to drive, older women now are more likely
than older licensed men to live in a household without a vehicle, 25 percent for women versus 5
percent for men. Even with equalization of licensing rates, given income constraints and longer
life expectancy women are still more likely than men to lack access to cars.
Driving safety
Gebers et al. ( 1993) noted that a substantial number of accidents involving elderly drivers are at
least partially attributable to worsening vision, poor physical coordination, cognitive confusion,
or other age- related physical and mental impairments. Howe et al. ( 1994) concurred that older
people are more likely to have deficits in visual acuity and peripheral vision, greater
susceptibility to glare, and poorer night vision and ability to focus. However, Gebers et al. ( 1993)
cautioned that chronological age per se is not a very good measure of accident risk for
individuals, because elders vary considerably in driving skills, physical/ mental abilities, point of
onset of decline, and rate of decline. Coughlin and Lacombe ( 1997) noted that although most
elderly drivers know their limits and are safe drivers, age- related physical and cognitive
deterioration, coupled with the increased likelihood of drug interaction from medical treatment,
contributes to some seniors being impaired drivers.
Some drivers may lose the ability to drive safely in their 60s, while others may drive safely well
into their 80s. Of course, while individuals vary greatly in the timing of their loss of driving
ability, there is an observable higher level of impairment in each successive cohort. Gebers et al.
( 1993) found that on a per- mile- of- travel basis, drivers over 70 years of age are as likely as
teenagers to be involved in automobile accidents. Yet licensing and re- examination procedures
do not always reflect what research has shown are the most important factors associated with this
increased risk. Further, Spain ( 1997) noted that developments in health care reforms, medical
advances, safer workplaces, and healthier lifestyles may reduce the incidence of chronic
disabilities for the elderly in the future. The most likely scenario is that people will stay healthy
longer, but will still succumb to functional limitations in later ages ( Spain 1997).
Older drivers are often well aware of the tradeoffs between their own mobility and road safety.
Gebers, et al. ( 1993) noted that due to some form of vision impairment, older drivers commonly
voluntarily limit or give up night driving and driving under conditions of reduced visibility. They
also noted that the elders who had recently given up driving reported more visual problems than
the elderly who continued to drive. As a result, when seniors decide to stop driving, it may be
due to an awareness of one’s own physical limitations. However, the lack of alternatives to
driving may lead some drivers to hold onto their license. For the elderly who have relied on
driving throughout their working lives, giving up driving is a serious sacrifice unless various
alternative transportation options exist.
Coughlin and Lacombe ( 1997) also noted that license examiners and officials and physicians are
hesitant in recommending suspension of elderly drivers’ licenses because such action may
sentence the driver to isolation and dependency. In a 1995 survey of state licensing examiners
and supervisors throughout the nation, more than half of the respondents indicated that the lack
California Travel Trends and Demographics December 2002
Final Report
12
of readily alternative transportation was an important consideration in revoking an elder’s
driving privileges. Consequently, state officials should take care to balance safety- related license
revocation policies with the availability of alternatives.
Location and auto dependence
Coughlin and Lacombe ( 1997) argue that the combination of low- density developments and
single- family housing patterns, once thought ideal for child rearing, now presents considerable
obstacles to meeting the mobility needs of elders who attempt to stay in their suburban homes.
Spain ( 1997) points out that as suburbanites age and worry less about the quality of schools and
more about their ability to drive, the high density of cities may become more appealing if there
are adequate options that reduce the need to drive. But contrary to these countervailing factors to
elderly suburbanization, many retirement communities are often still built on the suburban model
where the use of an automobile to meet the majority of a resident’s mobility needs remains an
underlying assumption of these developments ( Coughlin and Lacombe 1997).
2.3 Land Use Influences on Travel
The characteristics of the built environment at different spatial scales are thought to have distinct
effects on the travel behavior of households and individuals. Changes in the built environment
may influence travel by changing the relative attractiveness of travel modes, altering the time or
money costs of travel, or affecting the provision of transportation services ( such as transit). Table
1 contains a list of the various urban design and land use aspects that have been theorized to
change travel behavior. These questions have been addressed in the empirical literature, as
described below. The sections are organized into empirical results relating to four categories of
built environment characteristics: development density, accessibility, mixed uses, and street
pattern.
Development density
The correlation of density with higher alternative mode use and lower amounts of travel has been
widely documented in aggregate, area- based descriptive analysis. Much of the analysis of metro-wide
density effects does not deal with many complications inherent in attributing causality, such
as controlling for correlates of density ( such as transit infrastructure and city size) and the
interrelationship of residential location choice and travel decision making. However, this
literature provides a useful overview of the observed correlations between metro- area density
and travel.
Dunphy and Fisher ( 1996) investigate relationships between driving, transit use, and density at
two geographic scales: cities and zip codes using 1990 data from the Nationwide Personal
Transportation Survey ( NPTS). City- based aggregate comparisons show an inverse relationship
between density and vehicle miles traveled, and a positive relationship between density and
transit use. The authors suggest that the road and transit networks also play a large role.
Kockelman ( 1995) investigated commute mode choice as a function of density and income in the
San Francisco Bay Area. In an aggregate analysis at the city level, population density was much
more strongly correlated with the percentage of workers driving alone to work ( correlation of -
0.524) than was income ( 0.213).
California Travel Trends and Demographics December 2002
Final Report
13
Table 1: Built Environment and Land Use Characteristics Thought to Affect Individual
and Household Travel Behavior
Site design:
Building setbacks
Placement of garages and parking
Architectural attractiveness
Presence/ absence of front porches and picket fences
Design of transit stops
Neighborhood built environment / land use characteristics:
Development density
Availability of commercial, residential, industrial, office, and recreational land uses
Cost, availability, and placement of on- street and off- street vehicle parking
Spatial relationship to regional transportation network and activity centers
Metropolitan/ regional built environment / land use characteristics:
Development density
Land use segregation
Development clustering ( e. g., share of employment in high- density nodes such as central
business district, pattern and size of activity centers)
Transportation network design characteristics:
Percentage of land devoted to roads and parking
Number of street intersections
Curb radius length
Number of curb cuts ( driveways)
Rear location of parking and building services
Lineal amount of street and sidewalk
Sidewalk connectivity
Average block size
Loops and cul- de- sacs per mile of road
Average street width
Extent of vehicle/ pedestrian network separation
Presence/ extent of “ traffic calming” devices
Presence/ extent street and sidewalk amenities ( e. g., trees, benches, lamps)
Number and proximity of transit stops
Some work has investigated the correlations between density and transit service. Pushkarev and
Zupan ( 1977) found that residential density of seven units per acre was needed to make provision
of transit services financially feasible in the New York metropolitan region. In a more recent
study of Dade County, Florida, Messenger and Ewing ( 1996) find that residential density of 19.4
dwellings per acre is necessary to support 25- minute headways at the transit agency’s average
productivity level ( 8.4 dwelling units per acre for the “ minimum” productivity level).
Studies using aggregate data for Census tracts or municipalities tend to find that higher
development density reduces auto use, in some cases dramatically. Holtzclaw ( 1994) examined
the relationship between land use patterns and areawide average household car ownership and
VMT in 27 sub- municipalities ranging from 11,000 to 724,000 in population in San Francisco,
Los Angeles, San Diego and Sacramento. Holtzclaw regressed average household vehicle
ownership and odometer readings on population, household, and residential unit density, as well
as the availability of transit, access to commercial establishments, and an index of pedestrian
California Travel Trends and Demographics December 2002
Final Report
14
accessibility. He found that higher average residential density was associated with lower auto
ownership and less driving. Transit accessibility was also a statistically significant predictor of
household VMT. Frank and Pivo ( 1994) used data from the 1989 Puget Sound Transportation
Study to investigate how Census tract average mode choice for shopping and work trips was
related to gross population and employment density at both the trip origin and destination, as
well as a measure of mixed use. An average of gross population and employment density at the
residence and workplace zones was the most consistently significant variable in the six
correlations presented.
Dunphy and Fisher ( 1996) investigated the effect of zip code level population density on
household travel. They found that people averaged 3.5 trips per day in lower density zip codes of
up to 4,500 residents per square mile, reaching an average of 1.9 personal vehicle trips in areas
of 30,000 residents per square mile. In higher density areas the total number of trips per capita by
all modes does not decrease very much, but a greater share of bus, rail, and walking trips results
in substantially fewer vehicle miles traveled per capita. Dunphy and Fisher also found that
density was highly correlated with lower income, lower auto ownership, and shorter distances to
the nearest transit stop. In turn, these characteristics are associated with higher transit mode share
and lower per capita vehicle miles traveled, possibly explaining much of the correlation of
density with travel behavior.
Messenger and Ewing ( 1996) included the log of combined employment and population density
as an explanatory variable in regressions of bus mode share for traffic analysis zones in the
urbanized portion of Dade County, Florida. Density was negatively related to bus mode share
when auto ownership was included in the model, but was positively related to the proportion of
households with no cars or only one car, implying that “ as density rises, automobile ownership
falls; as automobile ownership falls, density rises” ( 150). Thus, automobile ownership was a
primary influence on travel behavior, as were local jobs- housing balance and transit service. In
turn, auto ownership was affected by development density, income, and transit access.
Studies using disaggregate data are more reliable, because aggregate zonal travel conceals
important variations and masks relationships between demographics and travel. Some of these
disaggregate studies continue to find strong relationships between land use and travel.
Kockelman ( 1995) carried out a disaggregate, trip- based binomial logit regression model for the
decision to drive to work, with population density of the residential and workplace Census tract,
income, and an accessibility index as independent variables. The accessibility index for origin
and destination was the most significant variable in this model, accounting for most of the
probability of choosing to drive alone, with income a distant second and density coming in last.
However, development density and accessibility were strongly correlated and are conceptually
interrelated. The accessibility and density measures were likely both highly correlated with
parking costs, congestion, and other factors affecting the analysis.
Many of these authors emphasize the importance of correlates with development density that are
not controlled for in their analysis, particularly better transit service, shorter distances to transit
stops, and road congestion.
Ewing ( 1995) regressed household vehicle hours traveled on demographic characteristics and
land use variables at both the residential location and the employment location of households in
California Travel Trends and Demographics December 2002
Final Report
15
Palm Beach County. Unexpectedly, he found that higher employment density in the zone of
employment location increased the vehicle hours traveled per household. Ewing interpreted this
result to mean that “ when workplaces are accessible to other activities, so many additional trips
are generated as to overwhelm the favorable effect of accessibility on trip lengths” ( 20).
However, the results could be due to slower travel speeds in dense employment areas.
Using a disaggregate data set of households, Sun, Wilmot and Kasturi ( 1998) found that
employment density had a small statistically significant negative impact on total trip- making, but
no significant impact on VMT. The authors also found that the correlation of income with
population density in Portland was not very significant, but that both auto ownership and
household life cycle were significantly correlated with population density. They used a measure
of employment density in linear regressions investigating the effect of demographic
characteristics and land use on vehicle miles traveled and total trips by households in Portland,
Oregon in 1994. Accessibility indices were also included in their analysis and found to be
statistically significant in reducing total trips and decreasing VMT. The inclusion of accessibility
indices probably accounted for the negligible impact of density, since the index essentially
accounts for density simultaneously with mixed use. This is a common finding ( e. g., Kockelman
1997).
Schimek ( 1996) investigated the impact on travel behavior of the gross population density for the
residential zip code area, using data from the 1990 Nationwide Personal Transportation Survey.
Schimek employed a sequential equations model to first predict vehicle ownership and then
vehicle use, and controlled for the endogeneity of residential location, auto ownership, and auto
use by using predicted gross population density from an instrumental variables regression in the
auto ownership and use equations, instead of observed density. In Schimek’s models, income,
household size, and the number of workers were more strongly correlated than population
density with the number of vehicles in the household and the household vehicle distance
traveled. However, a one percent increase in gross density was associated with one- tenth of a car
less per household. As for usage, the direct and indirect effects of density combined accounted
for a statistically significant reduction of 2,185 personal VMT per percentage increase in density,
and a daily reduction of 0.37 household vehicle trips.
In studies using 1990 and 1995 NPTS data, Pickrell and Schimek carried out an analysis of
household auto travel using a modeling structure that controlled for income, household size,
race/ ethnicity, and size of the urban area. The analysis used both gross population density, and
density squared, as well as a specification using the residual of density that was not explained by
household income, household size, employment status of household members, racial and ethnic
characteristics, the size of the urban area, and geographic region. The authors found that
population density of residential Census blocks and zip codes reduced household auto trips and
the proportion of trips made by auto, but only at levels above 4,000 people per square mile; the
most significant reductions were for households in areas above 7,500 persons per square mile,
densities “ typically found only in central city neighborhoods of the nation’s largest urban areas”
( Pickrell 1999: 427).
Boarnet and Greenwald ( 2000) carry out three sets of regression models using 1994 Portland
activity diary data. ( This work is similar to that of Crane and Crepeau ( 1998) and Boarnet and
Sarmiento ( 1998); for brevity, these earlier works are not described here.) The authors include a
California Travel Trends and Demographics December 2002
Final Report
16
number of variables as built environment measures: gross population density and gross retail
density for the residential Census tract, percentage of the quarter- mile- radius area covered by a
gridlike street pattern, a pedestrian accessibility index, a dummy variable indicating whether the
home is within a half- mile of light rail, and the proportions of multifamily, single family
attached, and single family detached housing in the Census tract. In their initial one- stage
ordered probability models, population density is associated with an increased number of
nonwork auto trips when speeds are not included among the explanatory variables, while retail
employment density is negatively related when speeds are included.
In the second model, the authors first regress median trip speed and median trip distance on the
built environment measures listed above. Predicted trip speeds and distances from that model are
then used as instruments in a second ordered probit model, which does not include any of the
built environment measures. The predicted distance from the Census tract level model is
statistically significant with the expected sign, while the zip code- level model’s predicted
distance and the two variables for predicted speed are not statistically significant. This result
implies that Census tract level land use characteristics affect the number of car trips by reducing
trip distances, but not through average speeds, while zip code- level land use characteristics do
not affect the number of car trips.
Finally, in their third set of models, the authors carry out a number of regressions in which
predicted land use characteristics ( in an instrumental variables procedure) are used to account for
the possibility that individuals simultaneously choose their residential locations and make travel
decisions based on built environment characteristics. In these regressions, the ( predicted)
proportion of single family homes and the ( predicted) proportion of multifamily housing are both
positively correlated with the number of auto trips, while ( predicted) retail employment density
is negatively correlated. Other land use characteristics are not significant in these regressions.
Mixed land uses
A number of other studies focus in particular on how mixed land uses at the sub- metropolitan
level affect travel behavior. Cervero ( 1988) studied the impact of mixed uses in employment
centers on commute mode choice using data on 57 suburban employment centers with at least
one million square feet of office space in the 26 largest US metropolitan areas. Cervero
hypothesized that increased car commuting to such locations is caused by the fact that “ those
who work in many campus- style office parks are almost stranded in the midday if they don’t
drive their car to work” and that single use centers are pedestrian- unfriendly because they are
dominated by parking. The study employed a stepwise OLS regression process, with the
percentage share of commuting by solo auto, carpool, and walk/ bike as dependent variables, and
selected measures of land use mix and transportation supply as independent variables. Land use
measures found to be significant in one or more of these models included the percentage of floor
space in office use, retail square footage within a 3- mile radius, jobs- housing balance within a 5-
mile radius, and size of the center ( number of full- time employees), all with the expected signs.
Transportation supply variables found to be significant in one or more models included the
number of company vans in operation, density of nearby freeway interchanges, and whether
there was a ride share coordinator at the location. Most of the relationships were of moderate or
modest magnitude.
California Travel Trends and Demographics December 2002
Final Report
17
Using land use and commuting data from the American Housing Survey ( AHS), Cervero ( 1996)
studied the impact of the availability of commercial uses on commute trip mode choice for
residents of eleven metropolitan areas. He found that the presence of commercial establishments
within 300 feet of the home significantly increased the probability of an individual walking or
biking to work and slightly increased the probability of using transit. He also found that the
presence of a grocery or drug store farther than 300 feet away but less than 1,000 feet away
decreased the use of these alternative modes. However, residential density ( as proxied for by
characteristics of nearby housing), commute distance and household car ownership were
substantially more important predictors of individual mode choice.
Frank and Pivo ( 1994) included a measure of mixed uses in their study of Census tract average
commute mode choice and land use. Mixed use levels at trip origins and destinations were
calculated using an “ entropy index” based on Cervero ( 1988: 57) using seven land use categories
applied to building square footage from the county assessor. This index was not significant when
density, demographics, and transit service were controlled, except in one case: the commute
walking share was significantly related to mixing of uses at both workplace and residence,
although not as strongly as to densities.
Ewing ( 1995) examined a number of different characteristics of land use with respect to total
vehicle hours of travel. He separated land use measures for the workplace and the residential
location, and included one mixed use measure in his model for the residential location, which
was a measure of jobs- housing balance. Other variables for land use were accessibility indices
and employment density. The mixed use measure was not significant in his model.
Kockelman ( 1997) carried out several disaggregate multiple regression models of varying types
to investigate the relative significance and influence of a variety of measures of urban form on
household vehicle kilometers traveled, automobile ownership, and mode choice. After
demographic characteristics were controlled for, measures of accessibility, land use mixing, and
land use balance were statistically significant with respect to all measures. In some cases, land
use measures were found to be more relevant than demographic characteristics. Except for the
vehicle ownership models, the impact of density was negligible after accessibility was
controlled.
Studying residents of Austin, Handy and Clifton ( 2001) found that the availability of local
shopping opportunities in neighborhoods was correlated with a higher number of long- distance
shopping trips and a somewhat lower use of auto for local trips. The authors did not control for
the size of stores. In focus groups with respondents as well as a follow- up regression analysis,
other factors than distance appeared to also be important in mode choice of local shopping trips,
such as having to cross busy streets to get to stores and other pedestrian amenities, as well as the
person’s strolling frequency ( intended to proxy for basic attitude toward walking). Based on
interviews with respondents, the authors suggest that most walk trips to the store replace driving
trips rather than being additional trips.
Accessibility
Accessibility measures are typically based on the “ gravity model,” consisting of sums of
employment by zone ( or, less commonly, residential population) divided by an exponential
function of distance from the measurement zone. Most accessibility measures include all
California Travel Trends and Demographics December 2002
Final Report
18
possible trip destinations, limited only by the geographical coverage of the data set, and the
measures often distinguish between types of employment or residential population.
Some researchers distinguish between local and regional accessibility. Handy’s ( 1992) measure
of local accessibility is essentially a measure of retail, service, and “ other” employment within
the traffic analysis zone, divided by an exponential function of average intra- zonal travel time,
while her measure of regional accessibility is similar to other gravity model- based measures.
Handy found that for shopping trips regional accessibility was sometimes more strongly
correlated with travel behavior than neighborhood accessibility, although they are clearly
complementary and often act as consumption substitutes.
Ewing, Haliyur and Page ( 1994) found that higher employment accessibility in selected
communities in Palm Beach County, Florida was correlated with a greater tendency to chain trips
in “ multipurpose tours” by car rather than making numerous separate car trips. Multipurpose
tours were also more commonly characterized by carpooling. While transit and walking modes
were rarely used in the county, carpooling was relatively common. The authors conclude that
high residential accessibility seems to be associated with fewer vehicle hours traveled, but not
with higher transit or walk share.
In a follow- up study, Ewing ( 1995) used a travel diary data set of 548 households and regressed
vehicle hours of travel on socio- demographic characteristics and land use variables, both at the
place of work and at the place of residence. He constructed four accessibility indices for the
residential location: work, shopping, social- recreational, and other. For the workplace, he
constructed a general accessibility index for all activity types. Zonal employment density was
also included in the model. The accessibility index for home- based other trips, which measures
the proximity of all possible destinations to the residential location ( other housing, all job types,
and school enrollment), was significant, but the other accessibility indices were not. Ewing
concludes that regional accessibility to all types of land use is a more important predictor of
travel decisions than employment- only or shopping- only measures.
Summary
The literature relating built environment and land use characteristics to travel choices does show
moderate to modest relationships between reduced auto use and higher development density, a
greater presence of commercial activities in residential areas, and higher accessibility indexes.
However, in the more methodologically sophisticated studies, the relationships are often more
difficult to discern. The literature suggests that accessibility measures may be more strongly
related to lower car use than the more direct measures of development density or mixed land
uses. However, this may be because high accessibility is even more correlated with high road
congestion, better transit, and a higher quality pedestrian environment than those other measures.
Such correlations are largely unexplored empirically, though often noted and commented upon.
2.4 Lessons for the California Demographics and Trend Study
The primarily descriptive literature on travel behavior leaves many questions about mobility and
equality in travel, even if it does establish differences in travel by sex, race, disability status, and
age. One important consensus, however, has arisen out of the travel behavior literature. Although
travel differs among women according to ethnicity, women of all ethnicities tend to travel
California Travel Trends and Demographics December 2002
Final Report
19
differently than similarly situated men. This seems to result from different work types and
responsibilities between women and men, both in the home and out ( Hanson and Pratt 1995;
Handy 1996).
Ethnicity and race variables, however, are somewhat different. Unlike differences in household
and work activities that has explained differences in travel by sex, differences in travel by
ethnicity are likely attributable to class and to spatial segregation ( which overlaps with class).
Class differences between different ethnic groups pertain not just to income, but to differences in
asset wealth, social networks, stigmas attached to work and private life, and residential
segregation. Thus, aggregate measures of ethnicity— that is, treating ethnicity in isolation from
these factors, may lead to misleading conclusions. Thus, our modeling efforts will be careful to
test for differences in travel by ethnicity, but the interpretation of model results must recognize
the myriad dynamics for which ethnicity provides a proxy.
Similarly, age as an explanatory variable in urban travel models conveys a lot of information
about ability and health status, income, and license possession ( at both young and old ages). All
of these factors influence aggregate levels of demand for total travel and travel by various
modes. Perhaps more importantly, a knowledge of age enriches the policy choices and
recommendations that the modeling and forecasting support.
Perhaps more than anything else, the travel behavior literature establishes the need to consider
the interactivity of race, class, sex, immigrant status, and age on individual travelers. Although
these factors are treated separately in the preponderance of the literature, they influence
individual opportunity for travel and economic citizenship. Including socio- economic variables
will— if treated simultaneously— add many dimensions to the empirical model and the
subsequent travel forecasts, thus complicating the analysis and the computational demands. Yet,
this level of detail is exactly what is needed if the forecasts are to guide the state’s decision-making
and improve Title VI compliance.
Similar problems challenge efforts to capture the effect of land use on travel behavior. On one
hand, the literature is entirely consistent with the theory that land use affects travel in the basic
ways: by changing the relative utility of travel by mode; by changing the relative time and
money costs of travel by mode; by affecting the provision of transport service; and through
dynamic effects. On the other, it is difficult to assess the relative contributions of these different
effects to observed travel behavior patterns. Most studies assume that land use affects travel
either by changing the relative utility of modes, or by affecting the relative cost of traveling.
Some authors make it clear that they are aware of both substitution and budget effects, but they
do not always explicitly investigate both.
Our review of this literature relating land use and travel suggests two main conclusions. First,
threshold effects are likely to be important both conceptually and empirically. This suggests that
instead of modeling the effects of land use with continuous variables, it may be appropriate to
segment the variables with dummies to represent thresholds. Second, interactive effects are also
important, implied most strongly by the accessibility index results.
California Travel Trends and Demographics December 2002
Final Report
20
The next section describes the existing demographic and travel behavior in California, providing
a link between the general themes developed in this section and the empirical research discussed
in the latter sections.
California Travel Trends and Demographics December 2002
Final Report
21
3
CALIFORNIA TRAVEL TODAY & YESTERDAY
This section takes a brief look at the data that are available on current travel in California. Many
datasets that have information on ethnicity, sex, income, and travel behavior are not sufficiently
disaggregated to convey the context- sensitive data most useful for good planning. Those datasets
that are sufficiently disaggregated to provide in- depth information on personal activities often do
not contain income data or ethnicity, or they are not available for the state as whole. Our
discussion focuses on the 2000 Census, the Nationwide Personal Transportation Survey, the
2001 California Statewide Household Travel Survey, and data from travel surveys carried out in
the San Francisco Bay Area, the five- county greater Los Angeles metropolitan region, and the
Sacramento area.
Aggregate travel flow tends to be characterized in five major ways: trip purpose, temporal
distribution, modal distribution, trip length, and spatial distribution. This categorization provides
a useful way to organize our discussion of travel behavior in California.
3.1 Trip Purpose
Based on 1995 data from the Nationwide Personal Transportation Survey ( NPTS), about half of
person trips and a third of person miles in the US are attributable to family and personal
business, which includes shopping, running errands, and trips to drop off or pick up passengers.
A quarter of person trips and 31 percent of person miles are for social/ recreational purposes.
Travel to and from work accounts for 18 percent of person trips and 23 percent of person miles
( FHWA 1995: 11). Passenger- serving trips, where the main activity is to pick- up or drop off a
passenger, make up 11 percent of trips by women and seven percent of trips by men; almost all
passenger- serving trips are made in private vehicles ( FHWA 1995: 16).
In the early part of the century, most travel was attributed to trip to work and back. Since that
time the prevalence of other kinds of trips has increased greatly. According to the Federal
Highway Administration, about 80 percent of the current miles traveled by individuals in the US
are for non- work purposes.
Many non- work trips occur during the week, but a large number of these trips occurs on the
weekend. As a result, Sunday and Saturday are typically the days with the highest trip making.
But shopping trips are spread fairly evenly throughout the week, with 77 percent of shopping
trips occurring on weekdays ( FHWA 1995: 15). In fact, many shopping trips are likely often
chained with work trips.
Table 2 ( below) shows the distribution of trip purpose for travelers by region in California.
California Travel Trends and Demographics December 2002
Final Report
22
Table 2. Percent of Weekday Trips by Purpose in California
Region
Home-
Other
Other-
Other
Work-
Other
Home-
Work
Home-
Shopping
Western Slope/
Sierra Nevada 38% 26% 9% 19% 8%
AMBAG 38 21 12 21 8
MTC 39 23 12 18 8
SACOG 38 21 11 21 9
SCAG 49 20 9 21 1
Rural 37 25 11 19 8
Butte 37 27 10 16 9
Fresno 38 14 10 30 9
Kern 39 18 10 26 6
Merced 38 21 10 24 7
San Diego 41 23 11 17 7
San Joaquin 39 20 9 24 8
San Luis Obispo 42 24 9 17 9
Santa Barbara 43 21 10 18 8
Shasta 37 25 10 19 8
Stanislaus 38 17 10 29 6
Tulare 38 26 8 17 11
Statewide 43% 21% 10% 20% 5%
SOURCE: 2000- 2001 California Statewide Household Travel Survey Final Report, Table 8.11. Data are for
households living in single- family homes, though the data for households in multifamily units are similar.
These statistics are remarkably uniform across regions. Home- to- work trips account for only
about 20 percent of weekday trips statewide, with lows in San Luis Obispo and Tulare. This is
close to the national figure of 18 percent ( based on 1995 NPTS data). Some variation exists,
however. Home- to- work trips accounted for more than 26 percent of trips in Fresno, Stanislaus,
and Kern counties. Thus, nonwork trips account for about 75 to 80 percent of weekday trips
across California regions.
The findings are very similar for the major California urban regions. We examined three sets of
travel diary databases: a 1991 survey carried out in the five- county greater Los Angeles
metropolitan area by the Southern California Association of Governments ( SCAG); a 2000
survey of Sacramento area residents commissioned by the Sacramento Area Council of
Governments ( SACOG); and a 2000 activity diary survey of the nine- county San Francisco Bay
Area ( BATS) administered by the Bay Area Metropolitan Transportation Commission ( MTC).
For these data, we grouped trip destinations into four main categories: work/ school/ daycare; non-work
trips; passenger- serving trips, where the main activity is to pick- up or drop off a passenger;
and at- home activities, where the home is the final trip destination. Table 3 shows summary
statistics based on this grouping.
California Travel Trends and Demographics December 2002
Final Report
23
Table 3. Trip Purpose by Metropolitan Region
Primary activities
Bay Area LA area Sacto.
Work/ school/ daycare 18.8% 22.8% 16.2%
Non- work 32.3 34.5 29.3
Passenger serving
12.4 7.7 7.4
At- home activities 40.4 35.0 47.0
As shown above, work, school and day care trips made up about 23 percent of the trips in the
Los Angeles region, almost 19 percent of the Bay Area trips, and over 16 percent of the
Sacramento trips. Roughly a third of trips were considered non- work trips, to access activities
such as shopping, social activities, recreation, banking and personal business. Another way to
look at the data is to eliminate the at- home activities category because it largely represents the
return to home trips after the primarily purpose outbound trips. Once the at- home activities are
eliminated, the percentage of work trips increases to 40 percent for the Los Angeles region and
30 percent for both the Sacramento area and the Bay Area, while the percentage of non- work
trips jumps to 50 percent in the Los Angeles region and the Bay Area, and 60 percent for the
Sacramento region.
3.2 Temporal Distribution
In urban areas, the highest traffic flows occur during the morning and evening commutes. These
flows are typically about twice as high as flows at other times of day, and can last for up to four
hours in some congested metropolitan areas. The evening peak period is often longer and more
intense than the morning commute period.
The work trip is an important contributor to the daily peak periods during the week. Peak
commute travel is three to four times as great as non- peak commute travel. However, on average
across the United States, during the 6 to 9 a. m. peak commute period less than 40 percent of all
trips are trips to and from work, and during the 4 to 7 p. m. peak period the share falls to less than
20 percent ( FHWA 1995: 14).
Although the commute remains an important trip, it is declining as a share of all trips. This is
because it is generally not as flexible in terms of scheduling as non- work trips, and because for
the individual worker, the trip to work often dictates when, where and how his/ her other travel is
accomplished ( FHWA 1995: 12). In other words, workers often carry out non- work trips on the
way to and from work, and this contributes to the peaking patterns.
Trips for non- work purposes soften the overall peaking pattern somewhat by keeping flows high
during the rest of the day. For example, about half of the shopping trips occur between 9 a. m.
and 3 p. m., and social/ recreational trips ( including eating out) exhibit a major peak between 6
p. m. and 10 p. m. ( Barber 1995: 85). The overall peaking pattern is also muted by truck traffic
which accounts for 15 percent of all vehicle trips in urban areas. Truck trips tend to be on the
road network between the peak commute times, i. e. during typical business hours ( Barber 1995:
86).
California Travel Trends and Demographics December 2002
Final Report
24
Data on trip timing ( peak and off- peak) are difficult to come by. They were available from the
Los Angeles area and Sacramento area travel diaries, but not from the Bay Area data because a
substantial portion of that data was collected on the weekends where peak and off- peak are not
easily defined. The Sacramento and L. A. data were grouped into three categories: 1) morning
peak, between 6 am to 9 am; 2) evening peak, between 3: 30 pm and 6: 30 pm; and 3) non- peak,
all other times throughout the day.
Table 4. Trip Timing
Temporal Category LA area Sacto.
Morning peak 21.1% 16.0%
Evening peak 24.9 19.7
Non- peak 54.1 64.3
Of the 137,055 trips in the Los Angeles database, 21 percent occurred during morning peak
hours and 25 percent occurred during evening peak hours; more than half occurred during non-peak
hours. The Sacramento distribution ( 43,086 trips in the survey database) was 16 percent
during peak hours, 19.7 percent during evening hours and over 64 percent during non- peak
hours. Both the Los Angeles area and Sacramento area distributions suggest that the evening
peak commute is more intense than the morning commute.
3.3 Modal Distribution
In US urban areas, transit trips account for less than ten percent of commute trips. For all trip
purposes nationwide, the transit share is about two percent. Nationwide, school buses account
for almost as many person trips as public transit ( FHWA 1995: 17). In metropolitan areas, the
share for walking and biking combined is generally higher than the combined transit/ school bus
share, regardless of population density ( Ross and Dunning 1997: 16). About 44 percent of transit
trips take place during peak commute periods ( FHWA 1995: 17). This is a much stronger
peaking pattern than for overall travel, which is dominated by personal vehicle trips.
Transit use nationwide hit its peak after World War II, when almost 23 billion yearly trips were
made on transit. It fell off dramatically afterwards, and has steadily declined as a percentage of
all trips since leveling off in 1960 at billion yearly trips ( Barber 1995: 89). There has been a
gradual spreading of peak daily period for all travel, but for public transit the peak has remained
intense or become more intense, because for non- work off- peak trips, transit is particularly
uncompetitive with personally operated vehicles.
To examine urban modal distribution in California, we summarized modal information from our
three urban travel databases into five categories: 1) car/ van/ truck/ motorcycle, including all
private vehicle trips; 2) public transit including bus and rail; 3) walking trips; 4) bicycle trips;
and 5) school bus trips. In the Los Angeles and Sacramento regions the private vehicle category
was the mode of choice for roughly nine out of ten trips. Private vehicle use in the Bay Area was
slightly lower than the other two regions, at eight out of ten trips. Even during peak commuting
hours, private vehicles accounted for 84 to 95 percent of the trips in the Los Angeles and
Sacramento regions. Total walking trips accounted for five percent of trips in the Sacramento
region, eight percent in the Los Angeles region, and 11 percent in the Bay Area. Public transit
California Travel Trends and Demographics December 2002
Final Report
25
accounted for about one percent of the trips in the SCAG and SACOG region. Although the
percentage of trips on public transportation in the BATS region was three times the size of the
other regions at 4.5 percent, it is still less than five percent of total trips for the region. Within the
public transit trips, over 50 percent of the transit trips in the SCAG region were during the peak
commuting hours. In comparison, only 37 percent of the transit trips in the SACOG region
occurred during peak commuting hours. One plausible explanation on the difference between
these two regions in transit use during commuting hours is that SCAG is a more heavily
urbanized region than SACOG and as a result have more developed public transit corridors such
as the Metro Blue Line, Metrolink and the El Monte Busway to facilitate commuter travel during
peak hours.
Table 5. Travel Mode by Travel Time from SCAG and SACOG
SCAG SACOG
Primary mode
morn
peak
even
peak
non-peak
Total
morn
peak
even
peak
non-peak
Total
Personal vehicle
( motorized) 84.2% 91.0% 88.2% 88.1%
94.9% 90.4% 90.9%
91.0%
Public transit 1.3 0.9 0.8 1.0 0.5 1.7 1.1 1.2
Walk 9.3 5.9 8.2 7.8 3.7 4.8 5.2 5.0
Bicycle 1.2 1.2 0.9 1.0 0.8 2.0 1.6 1.6
School bus 3.1 0.5 1.4 1.5 0.1 0.9 1.1 1.0
Other/ dk 0.9 0.5 0.5 0.6 0.1 0.2 0.2 0.2
BATS Rural
Personal vehicle
( motorized) — — —
82.1%
— — — 92.2%
Public transit — — — 4.5 — — — 0.5
Walk — — — 11.0 — — — 3.9
Bicycle — — — 1.4 — — — 0.5
School bus — — — — — — — 2.7
Other/ dk — — — 0.0 — — — 0.0
Statewide
Personal vehicle
( motorized) — — —
90.0%
Public transit — — — 1.7
Walk — — — 6.0
Bicycle — — — 0.6
School bus — — — 1.4
Other/ dk — — — 0.4
SOURCE: California Statewide Household Travel Survey 2001, Table 8.9, SACOG and SCAG trip information.
3.4 Trip Length
Average trip distances are greater in larger cities, but the spatial structure ( i. e. density) of a city is
related to average trip distance, with denser cities having shorter trip distances on average,
controlling for city size. On average, work trips are longer, in both distances and time, than non-work
trips. The distance of average commute trip lengths has been rising somewhat over time,
California Travel Trends and Demographics December 2002
Final Report
26
while the average time of the work trip has remained relatively constant until recently ( Barber
1995: 94- 96).
Although we do not have distance- based information in the travel dairies, the SACOG data set
has a variable on the time duration of each trip. We grouped the trip duration data into seven
categories: 1) 5 minutes or less; 2) 6- 10 minutes; 3) 11- 15 minutes; 4) 16- 20 minutes; 5) 21- 30
minutes; 6) 31- 45 minutes; and 7) more than 45 minutes.
Table 6. Trip Length in Minutes
SACOG
Minutes Work Non- work
Passenger
serving
5 15.0% 42.5% 12.2%
6- 10 16.6 39.3 10.8
11- 15 19.9 35.9 9.3
16- 20 23.7 32.0 8.1
21- 30 27.3 29.7 6.3
31- 45 29.1 29.7 6.0
more than 45 27.6 29.8 4.6
Of the 33,954 trips in the SACOG travel, 66 percent of the trips were 15 minutes or shorter and
only ten percent of the trips were over 30 minutes. By tabulating trip duration with trip
activities, we can get a distribution of the types of destination activities and corresponding travel
duration. For trips, 5 minutes or less, over 42 percent of the trips were for non- work related
travel which is consistent with the literature that work trips are generally longer than non- work
trips. In fact, over 71 percent of non- work related trips were 15 minutes or less compared to 55
percent of the work related trips were 15 minutes or less. Furthermore, passenger- serving trips
where the driver is picking up or dropping someone else off at a destination also tend to be
shorter than work trips with 76 percent of these trips taking 15 minutes or less.
The data from the California Travel Survey demonstrates similar characteristics to the SACOG
data. These data are shown for the SCAG region and for the rural sections of the travel survey in
Figure 2 and Figure 3. The data shown are for all trips, and they show a significant skew; that is,
most trips are of comparatively short duration.
California Travel Trends and Demographics December 2002
Final Report
27
Figure 2. Trip Duration in Rural Regions
Figure 3. Trip Duration in Los Angeles Region
0%
5%
10%
15%
20%
25%
30%
5 10 15 16 25 30 35 40 45 50 55 60 65 70 75 80 80
Minutes
n= 3,878,846
mean= 19
median= 10
0%
5%
10%
15%
20%
25%
30%
5 10 15 16 25 30 35 40 45 50 55 60 65 70 75 80 80
Minutes
n= 65,149,709
mean= 21
median= 15
California Travel Trends and Demographics December 2002
Final Report
28
3.5 Individual and Household Travel Behavior
Individual/ household and travel patterns, licensing rates, and vehicle ownership vary by sex,
income and race and ethnicity. These differences begin to explain some of the aggregate travel
patterns discussed previously.
Variations by sex
As discussed in the previous section, women exhibit markedly different travel patterns from
men. According to Pucher, Evans and Wegner ( 1998: 27), “ The main differences between men
and women are the much higher incidence of carpooling by women, their greater use of buses
and taxis, and their much lower rate of bicycling. Women are also much more likely to travel at
off- peak periods, to make a lower percentage of work trips, and to make shorter trips than men.”
In 1990 women made more overall trips, but fewer vehicle trips, then men and they traveled
fewer miles because their trips were shorter. These differences were partially attributable to
differences in income, licensing and auto ownership among men and women. But they are also
largely due to differences in responsibility for household activities ( Rosenbloom 1995: 2.9).
The greater use of buses and taxis by women has been diminished over time. As women become
more likely to be employed, they continue to bear the majority of the responsibility for
household functions such as shopping, child- related activities, and elder care ( Rosenbloom
1995). Employed women often find the use of transit and non- motorized modes inconvenient,
because these modes do not easily enable chains of trips to accomplish several different
purposes, a necessary adaptation to a more constrained time budget. Women also make two
thirds of passenger serving trips, which are carried out almost exclusively in privately owned
vehicles ( FHWA 1995).
Variations by income
Transit users are much more commonly from low- income households, but peak users tend to
have higher incomes than off- peak users. ( Pucher, Evans, and Wegner 1998) There are not
significant income differences in peak and off- peak travel for personally operated vehicles.
( Barber 1995: 87) In general, higher income people tend to make more trips of longer duration,
increasingly in personally operated vehicles ( Pucher, Evans, and Wegner 1998). A study of
transportation and minority women’s employment in New York showed that higher income
groups have consistently higher use of auto modes. ( McLafferty and Preston 1998: 363)
Income appears to have its strongest effects on travel behavior by increasing the likelihood of
owning an auto. Ethnic/ racial differences in travel behavior often appear to be insignificant when
auto ownership is taken into account. For example, Johnston- Anumonwo ( 1998) found that when
travel times of auto users are compared, ethnic/ racial differences often are reduced or disappears
completely.
Variations by race/ ethnicity
Despite making up a minority of the population, non- Anglos accounted for almost two- thirds of
transit riders in the US in 1995 ( Pucher, Evans, and Wegner 1998: 15). In urban areas, Anglos
use public transit for 1.9 percent of trips, while African Americans use it for 10.3 percent of the
trips and Hispanics for 7.5 percent of trips; the average African- American person makes six
California Travel Trends and Demographics December 2002
Final Report
29
times as many trips by transit as the average Anglo ( 95 versus 15 per year). But for all three
groups in urban areas, walking is more prevalent than transit use, at 7.2, 17.3 and 12.9 percent
respectively ( Barber 1995: 94).
Part of the reason for this greater use of transit and walking is lower car ownership. Based on
1990 NPTS data, Pisarski found the on average, more than 30 percent of African- American
households do not own vehicles, and in central cities the number is over 37 percent. Hispanics
have an overall rate of vehicle- less households of 19 percent, with the central city rate rising to
27 percent ( Pisarski 1996: xv).
Another reason that minority groups drive less than Anglos is that they are less likely to have
driver’s licenses. While 90 percent of all White women 16- 64 were licensed, only 70 percent of
African- American women and 66 percent of Hispanic women had a license ( Rosenbloom 1995:
2.6).
Johnston- Anumonwo’s literature review suggests that there are racial differences in travel
behavior that are not entirely explained by various control factors such as auto ownership,
income, occupation, and domestic role. It is not clear from her review whether other factors such
as education have been controlled for. But her review does suggest that a large share of
differences is explained by these factors, particularly auto ownership. Auto ownership, in turn,
can be largely seen as a function of income.
Car licensing
Between 1969 and 1990, the population of the United States increased 21 percent, from 197
million to 239 million people. Licensed drivers increased at a rate substantially greater than
population growth. The number of male drivers increased 38 percent, while the number of
female drivers increased 84 percent. ( Lave, 1993) In California, both the growth in population
and license drivers are even more dramatic. The California population increased by more than 50
percent from 19.7 million in 1969 to 30 million in 1990. For the same time period, licensed
drivers increased by 75 percent from 11.4 million to 19.9 million in 1990.
Our examination of the SACOG, SCAG and BATS travel dairies revealed that a very high
percent of Californians are licensed to drive. Of the 7,756 persons in the SACOG sample that are
14 years of age and above, over 89 percent of them are licensed drivers with over 91 percent of
the men and 88 percent of the women licensed to drive. Similarly, in the 1991 SCAG travel
diaries, of the 31,146 persons age 14 and above, 89 percent are licensed drivers with over 92
percent of the men and 86 percent of the women licensed to drive. The licensing rates in the
BATS region were very similar to the two other regions with over 91 percent of the 14 and over
licensed to drive. In fact, over 95 percent of people between the ages 40 to 44 surveyed in the
travel dairies have licenses, which suggests that the number of license drivers in California has
probably reach a saturation point. Further examination of licensing rates within the age
categories further revealed that licensing rates remain over 90 percent for driver up to 75 years of
age. After age 75, the number of licensed drivers began to drop.
California Travel Trends and Demographics December 2002
Final Report
30
Table 7. Percentage of Licensed Drivers by Age and Sex
SACOG SCAG BATS
Age
category Male Female Total Male Female Total Male Female Total
14 to 17 33.5 30.5 32.1 32.8 35.8 34.3 31.1 31.5 31.3
18 to 20 83.0 74.1 78.9 79.0 70.6 74.6 84.4 85.2 84.8
21 to 24 88.1 90.2 89.2 88.7 80.2 84.2 92.5 90.1 91.1
25 to 29 92.8 92.9 92.9 92.8 86.5 89.5 95.5 94.7 95.3
30 to 34 90.5 91.9 90.9 95.4 91.7 93.5 96.9 97.3 97.2
35 to 39 94.6 91.3 92.9 96.5 93.4 94.9 98.6 97.0 97.8
40 to 44 93.9 98.0 96.0 96.6 93.7 95.0 98.7 97.1 97.9
45 to 49 95.7 96.2 96.0 97.4 92.5 94.9 97.8 97.4 97.6
50 to 54 95.4 96.5 96.0 97.8 93.1 95.3 98.2 97.0 97.6
55 to 59 97.1 95.8 96.4 97.4 91.7 94.4 98.8 96.5 97.6
60 to 64 95.8 94.6 95.0 96.3 88.8 92.4 98.4 95.3 96.8
65 to 69 94.5 90.1 92.3 96.5 86.8 91.1 97.6 93.4 95.4
70 to 74 97.0 88.3 92.5 94.5 86.4 90.0 96.4 91.8 93.9
75 to 79 91.6 84.9 88.1 90.9 74.2 81.3 91.7 86.5 88.9
80 to 84 95.0 72.1 83.1 79.6 60.8 68.0 89.9 72.3 80.1
85 to 100 80.2 58.2 65.5 73.1 37.3 49.6 67.6 39.2 50.2
Total 90.7 87.8 89.1 91.9 86.1 89.2 92.2 90.2 91.3
3.6 Car Ownership, Household Size and Income
The number of household vehicles has more than doubled in the last thirty years. From 1969 to
1995, a period in which household size decreased by 17 percent, the number of cars per
household increased from one to two. ( FHWA 1995: 3) The ratio of cars per licensed driver has
also increased nationally. The number of vehicles per licensed driver has increased from 0.7 in
1969 to 1.01 in 1990. ( Lave 1993) In contrast to the national ratio, California's vehicle to
licensed driver dropped between 1969 to 1990 period. In fact, it was almost a mirrored opposite
of the national trend. California had 11.42 million licensed drivers and 11.45 million passenger
and commercial vehicles, which is virtually one vehicle per licensed driver for a ratio of 1.0. In
1990, California had a population of 30 million and 22 million vehicles for a ratio of 0.73.
However, the more meaningful of the two car ownership measurements is the household
number, because it is the availability of a car to the household that mostly determines the ability
of licensed drivers to have access to a vehicle. In fact, it is through the household measurement
that we can get information on car ownership of demographic subgroups such as African
Americans and Hispanics. Using 1990 NPTS data for the US, Pisarski found that on average
more than 30 percent of African- American and 19 percent of Hispanic households do not own
vehicles, and in central cities the number is over 37 percent for African- Americans and 27
percent for Hispanics ( Pisarski 1996: xv).
The data on the number of vehicles per household across the three regions were very similar.
Between 2.5 percent to slightly over 4 percent of the households in the three data sets do not own
vehicles which means 96 percent of the households surveyed in the SCAG, SACOG and BATS
California Travel Trends and Demographics December 2002
Final Report
31
travel diaries own at least one car. Of the households with cars, almost half of them have two
vehicles.
Table 8. Vehicles per Household in California Regions
Vehicles Per Household So. Ca. Sacto. Bay Area
None 4.2% 4.3% 2.5%
One Vehicle 27.9 30.3 21.1
Two Vehicles 45.1 42.0 49.4
Three Vehicles 14.9 15.9 19.6
Four or more 7.9 7.5 7.5
SOURCES: SCAG ( 1991), SACOG ( 2000), and MTC ( 2000) travel surveys.
Further examination of the vehicle and household size variables revealed that smaller households
are more likely to be without a car. Seventy- three percent of households without vehicles were
one- person households. In fact, households with two or less persons accounted for over 90
percent of the households not owning a car. In contrast, households with more than two persons
own cars at very high rates. The data sets show that 97 percent of SCAG and 98 percent of
SACOG households with more than two persons have at least one car.
Another way to look at vehicle ownerships is to examine the ability of a household to afford a
vehicle. For example, transit users are much more commonly from low- income households. As a
result, income appears to have its strongest effects on travel behavior by increasing the
likelihood of owning an auto.
We grouped the income data into five categories: 1) low- income ( less than $ 15,000); 2) medium-low
income ($ 15K to $ 30K); 3) medium- income ($ 30K to 50K); 4) medium- high income ($ 50K
to $ 75); and 5) high income ( above $ 75K).
Table 9. Vehicles per Household by Income: SCAG
Vehicles per Household
Income categories None One Two Three Four or more
$ 15,000 or less 71.0% 26.2% 6.1% 4.1% 5.3%
$ 15,001 to $ 30,000 16.2 35.1 18.3 12.7 11.6
$ 30,001 to $ 50,000 8.0 26.8 32.2 27.0 23.3
$ 50,001 to $ 75,000 2.3 8.1 25.0 29.2 26.1
$ 75,001 or greater 2.5 3.9 18.4 27.1 33.7
Total 100.0 100.0 100.0 100.0 100.0
The above table from the SCAG travel data shows that 71 percent of the households without cars
are in the low income category. In fact, households earning less than the $ 30,000 threshold
accounted for 87 percent of the households without cars. In comparison, only five percent of the
household earning more that $ 50K do not own a car.
California Travel Trends and Demographics December 2002
Final Report
32
Table 10. Vehicles per Household: SACOG
Vehicles per Household
Income categories None One Two Three Four or more
$ 14,999 or less 59.4% 16.9 4.5 2.5 1.5
$ 15,000 to $ 29,000 25.2 33.8 12.1 7.4 5.6
$ 30,000 to $ 49,999 10.2 28.0 25.9 22.8 21.3
$ 50,000 to $ 74,999 2.6 13.1 26.8 27.7 31.0
$ 75,000 or greater 2.6 8.1 30.7 39.6 40.7
Total 100.0 100.0 100.0 100.0 100.0
Table 11. Vehicles per Household: BATS
Vehicles per Household
Income categories None One Two Three Four or more
$ 15,000 or less 16.8% 2.7% 0.2% 0.2% 0.2%
$ 15,001 to $ 30,000 37.7 20.0 4.4 2.3 1.7
$ 30,001 to $ 50,000 23.9 30.0 14.6 10.5 8.6
$ 50,001 to $ 75,000 14.6 23.9 23.7 22.7 15.8
$ 75,001 or greater 7.1 23.4 57.1 64.3 73.9
Total 100.0 100.0 100.0 100.0 100.0
The SACOG data in Table 10 shows that the vehicle per household by income data is very
similar to the SCAG data. Nearly 60 percent of the households without cars are in the low
income category and only five percent of the households earning over $ 50K do not own a car.
Table 10 shows that the BATS data on vehicle per household by income are more varied than the
other two regions. For example, over 60 percent of the household without cars have incomes
between $ 15K to $ 50K and over 20 percent of households without cars have household earnings
of $ 50K and greater. One plausible explanation on the difference between the BATS results and
the two other regions might be that the more heavily urbanized land use patterns in the BATS
region affects the rate of vehicle ownership. As a result, in developing our empirical models, we
can control for some of the variations such as land use for better predictability.
California Travel Trends and Demographics December 2002
Final Report
33
4
EMPIRICAL TRAVEL MODELING
This section of the report describes the data, assumptions, conceptual bases, and results of the
empirical travel models. Two types of model are developed. The first model is applied directly to
the Phase II demographic forecasts and relies on the basic demographic variables provided in
those forecasts: age, sex and race/ ethnicity. In addition, two simple measures of land use are
included in the forecast models: gross residential density and a population accessibility index.
The second type of model goes beyond this basic set of variables to investigate other important
correlates of travel, such as household income, the presence of children in the household, and a
wider variety of land use characteristics.
4.1 Notes on Empirical Models
Since travel is complex, empirical investigation of travel behavior takes into account numerous
potential causal factors. The most sophisticated empirica