Citation
Alleviating metropolitan roadway congestion

Material Information

Title:
Alleviating metropolitan roadway congestion the efficacy of alternate urban performance measures at the land use and transportation interface in analysis and policy development
Creator:
Tsai, Te-I
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
158 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Traffic congestion ( lcsh )
City planning ( lcsh )
Transportation and state ( lcsh )
City planning ( fast )
Traffic congestion ( fast )
Transportation and state ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 148-158).
Thesis:
Design and planning
General Note:
College of Architecture and Planning
Statement of Responsibility:
by Te-I Tsai.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
53873579 ( OCLC )
ocm53873579
Classification:
LD1190.A74 2003d T72 ( lcc )

Full Text
ALLEVIATING METROPOLITAN ROADWAY CONGESTION:
THE EFFICACY OF ALTERNATE URBAN PERFORMANCE MEASURES AT
THE LAND USE AND TRANSPORTATION INTERFACE IN ANALYSIS AND
POLICY DEVELOPMENT
by
Te-I Albert Tsai
B.L.A. Chinese Culture University, 1989
M.L.A. University of Pennsylvania, 1993
M.A.U.D. University of Colorado at Denver, 1994
A thesis submitted to the
University of Colorado
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Design and Planning


2003 by Te-I Albert Tsai
All rights reserved.


This thesis for the Doctor of Philosophy
Degree by
Te-I Albert Tsai
has been approved
by
Thomas A. Clark
Deborah Thomas
J^bc, I 1,2002.
Date


Tsai, Te-I Albert (Ph.D., Design and Planning)
Alleviating Metropolitan Roadway Congestion: The Efficacy of Alternate Urban
Performance Measures at the Land Use and Transportation Interface in Analysis and Policy
Development
Thesis directed by Professor Thomas A. Clark
ABSTRACT
With economic and population growth and technological advances, travel congestion has
emerged as a universal urban problem. Air pollution and rising per capita energy
consumption are among the first consequences, and these in turn compromise our health,
economic productivity, and quality of life. Researchers have tried to use many urban
measurements such as population density and vehicle miles of travels (VMT) to monitor
roadway congestion, and some believe that encouraging compactness would enhance transit
usage, and result in VMT reduction, which in turn alleviates traffic congestion. However,
the reduction of VMT or trip length cannot directly ease roadway congestion, if the
roadways are insufficient and the internal developed area is too dense. Today, they are still
unable to completely understand the interaction between land use and transportation, and
unable to employ the appropriate approach to link indicator applications with urban
development policies. Therefore, the main purpose of this dissertation is to unveil the
underlying relationships among land use, transportation, and urban form, and to extract
from this means for alleviating work related traffic congestion.
Statistical examination of the congestion variation in 46 metropolitan areas during the AM-
peak hours yields these conclusions: population densification coupled with insufficient
roadway capacity tends to worsen traffic congestion; higher transit usage probably does not
help to relieve congestion effectively, if the transit system could not capture enough
commuters; and increasing jobs-housing balance does not necessarily mean less congestion.
IV


Indeed, such balance may result from higher roadway congestion, which leads commuters
to choose alternative modes for avoiding traffic.
To ease work related traffic congestion, two supply-side strategies expanding the existing
infrastructure and adopting transportation system management are recommended. Other
demand-side strategies include distributing population more uniformly (to urban villages) to
encourage non-single-occupant vehicle usage, promoting transit options and jobs-housing
balance/imbalance based on the economic characters of an MSA, educating commuters to
shift their modal choices, and encouraging the usage of telecommunications for work.
Finally, an approach of ensuring enough transit users based on an MSAs employment
density is proposed to further facilitate congestion relief.
This abstract accurately represents the content of the candidates thesis. I recommend its
publication.
Thomas A. Clark
v


DEDICATION
I dedicate this thesis to my parents and my wife for their complete support, while I was
preparing and writing this dissertation for pursuing my doctoral degree in Design and
Planning.


ACKNOWLEDGMENT
I would like to first thank my advisor, Professor Thomas A. Clark, for his support and
patience during the past five years. Without his guidance, I would have not been able to
finish this dissertation. Secondly, I would also like to thank my committee members.
Professor Yuk Lee and Professor Sarosh I. Khan have helped me with transportation issues
and other related issues. Professor Brian Muller and Professor Deborah Thomas have
provided many useful suggestions on urban form issues. Moreover, I would like to show
my appreciation to Professor Willem van Vliet, Dr. Peter Newman, and Professor Ernesto
Arias, who have supported me and provided general advice during the research process.
Finally, I would like to express my thanks to Richard Sandoval, Maiyann O'Brien, and
Cindy Higgins-Mulhem, who have spent their valuable time in helping me with editing this
dissertation.


CONTENTS
Figures.............................................................................xi
Tables............................................................................xiii
Chapter
1. Introduction....................................................................1
1.1 Problems and Purposes...........................................................1
1.2 Findings and Other Related Issues...............................................4
2. Literature Review...............................................................5
2.1 Historical Antecedents..........................................................5
2.2 The Early Land Use and Transportation Planning Strategies Relating to Urban
Form............................................................................7
2.3 The Use of Urban Indicators in Monitoring Urban Performance...................10
2.3.1 The Institutional Indicator Usage.............................................11
2.3.2 The Academic Indicator Usages................................................14
2.3.3 The Important Difference between Environmental Conditions and Quality........14
2.4 Possible Alternatives for Alleviation of Traffic Congestion in Urban Areas....16
2.4.1 Traffic Congestion Measurements...............................................16
2.4.2 Alternative One: Adding More Capacity........................................21
2.4.3 Alternative Two: Transportation Demand Management............................21
2.4.4 Alternative Three: New Urbanism Development..................................23
2.4.5 Alternative Four: Land Use and Transportation Interaction....................23
2.5 The Discussions on Land Use and Transportation Interaction..................24
2.5.1 The Jobs-Housing Balance......................................................25
2.5.2 Mixing Land Use..............................................................29
2.5.3 Density......................................................................31
2.5.4 Air Pollution................................................................32
viii


2.5.5 The Constraints............................................................34
3. Methodology for Examining Traffic Congestion based on Land-use and
Transportation Interaction...................................................37
3.1 Methodology.................................................................37
3.1.1 Method.....................................................................40
3.1.2 Data Collection............................................................43
3.1.3 Assumptions and Hypotheses.................................................44
3.2 Selected Metropolitan Statistical Areas.....................................46
3.3 Selected Variables for Analyses and Their Definitions.......................48
3.4 Summary.....................................................................54
4. The Examination of Forty-Six U.S. Metropolitan Areas.........................56
4.1 Characteristics of the Selected Metropolitan Areas.:.........................56
4.1.1 The Congestion in 1990.....................................................57
4.1.2 Population and Employment Density in 1990..................................59
4.1.3 The Economic Condition of MS As............................................62
4.1.4 Median Household Income in 1989............................................65
4.1.5 Car Ownership..............................................................66
4.1.6 Transit Ridership..........................................................68
4.1.7 Urban Form Index...........................................................69
4.2 The Results of Cluster and Factor Analyses..................................71
4.3 The Results of Correlation Analysis.........................................75
4.3.1 Congestion vs. Non-jobs-housing Balance Variables..........................75
4.3.2 Congestion vs. Jobs-housing Balance Variables..............................77
4.3.3 Industrial vs. Occupational Composition....................................79
4.4 The Results of Fixed Nonlinear and Stepwise Multiple Regressions............81
4.4.1 The Results of Fixed Nonlinear Regressions.................................81
4.4.2 The Results of the Stepwise Multiple Regression...........................102
4.5 Summary....................................................................117
5. Conclusions and Policy Implications.........................................122
5.1 Supply-side Strategies.....................................................124
IX


5.1.1 Supply-side Strategy 1 Expanding the Existing Infrastructure............124
5.1.2 Supply-side Strategy 2 Adopting Transportation System Management (TSM).... 126
5.2 Demand-side Strategies....................................................127
5.2.1 Demand-side Strategy 1 Distribute Population More Uniformly (to urban
villages) to Encourage Non-single-occupant Vehicle Usage..................127
5.2.2 Demand-side Strategy 2 Promote Transit Options..........................130
5.2.3 Demand-side Strategy 3 Promote the Jobs-housing Balance/Imbalance Based
on the Economic Characters of an MSA or a Municipality....................133
5.3 Summary and Additional Foreseeable Studies................................136
Appendix
A. Other Results of the Fixed Nonlinear Regression............................139
B. A Letter from a Professional Editor........................................146
References......................................................................148

x


FIGURES
Figure 2.1 Traffic Congestion Reduction Alternatives.......................r...20
Figure 2.2 Number of Vehicle Trips during Weekdays by Trip Purpose in 1990......35
Figure 2.3 Number of Vehicle Trips during Weekdays by Trip Purpose in 1995......36
Figure 3.1 Research Approach Diagram...........................................39
Figure 3.2 The Selected 47 Metropolitan Statistical Areas.......................47
Figure 3.3 The Shape of Tucson, AZ MSA..........................................52
Figure 3.4 The Shape of Colorado Springs, CO MSA................................52
Figure 3.5 The Shape of New York-Northern New Jersey-Long Island, NY-NJ.........53
Figure 4.1 The Result of Cluster Analysis for 46 Selected MSAs by the SIC.......73
Figure 4.2 Mean Median Speed vs. Macro-scale Roadway Density....................86
Figure 4.3 Mean Median Speed vs. Employment Density.............................87
Figure 4.4 Mean Median Speed vs. Population Density.............................88
Figure 4.5 Transit Ridership vs. Employment Density.............................89
Figure 4.6 Population Density vs. Transit Ridership.............................90
Figure 4.7 Median Household Income vs. Transit Ridership........................91
Figure 4.8 Median Household Income vs. Worker Car Ownership.....................92
Figure 4.9 Coefficient of Variation of Financial, Insurance, and Real Estate vs.
Calibrated Mean Median Speed.........................................93
Figure 4.10 Coefficient of Variation of Armed Forces vs. Calibrated Mean Median
Speed...............................................................94
Figure 4.11 Calibrated Mean Median Speed vs. Observed Coefficient of Variation of
Finance, Insurance, and Real Estate Industrial Sector Based on 30
Congested MSAs..................................................................95
xi


Figure 4.12 Calibrated Mean Median Speed vs. Coefficient of Variation of Armed
Forces Industrial Sector Based on 30 Congested MSAs.................96
Figure 4.13 Mean Median Speed Calibrated by Macro-scale Roadway Density vs.
Transit Ridership....................................................97
Figure 4.14 Transit Ridership vs. Worker Car Ownership...........................98
Figure 4.15 The Diagram for Balancing Transit Ridership and Mean Median Speed
to Reduce Traffic Congestion through Controlling Employment Density.... 116
Figure A. 1 Mean Median Speed vs. Median Household Income......................140
Figure A.2 Mean Median Speed vs. Worker Car Ownership.........................141
Figure A.3 Median Household Income vs. Car Ownership (car per capita)...........142
Figure A.4 Mean Median Speed vs. Transit Ridership............................143
Figure A.5 Mean Median Speed vs. Urban Form Index.............................144
Figure A.6 Coefficient of Variation of Manufacturing Durable Goods vs. Mean
Median Speed........................................................145
xii


TABLES
Table 2.1 The Comparison between the TTTs Roadway Index with Mean Median
Travel Speed...........................................................19
Table 2.2 Number of Vehicle Trips during Weekdays by Trip Purpose in 1990.........35
Table 2.3 Number of Vehicle Trips during Weekdays by Trip Purpose in 1995.........36
Table 3.1 The Applications of Statistical Analyses................................42
Table 3.2 The Applications of Geographic Information Systems......................43
Table 3.3 The Relationships between Explanatory Variables and the Dependent
Variable................................................................45
Table 3.4 The Selected 47 MSAs................................... ..............48
Table 3.5 The Definition of the Dependent Variable and Explanatory Variables.....50
Table 3.6 Three Examples of the Urban Form Index Application.....................51
Table 3.7 The Relationships between Explanatory Variables and the Dependent
Variable (same as Table 3.3)............................................54
Table 4.1 The Congestion Index (Mean Median Travel Speed).........................58
Table 4.2 Population Density......................................................60
Table 4.3 Employment Density......................................................61
Table 4.4 The Characteristics of 46 Selected MSAs.................................63
Table 4.5 Median Household Income in 1989.........................................65
Table 4.6 Car Ownership Rate for 46 Selected MSAs.................................66
Table 4.7 Worker Car Ownership Rate for 46 Selected MSAs..........................67
Table 4.8 Transit Ridership Rate for 46 Selected MSAs.............................68
Table 4.9 The Urban Form Index for 46 Selected MSAs...............................70
Table 4.10 The Result of Factor Analysis by the SOC...............................72
Table 4.11 The Result of Factor Analysis by the SIC...............................72
xiii


Table 4.12 The Correspondent Codes for Figure 4.1.................................74
Table 4.13 The Result of the Correlation Based on Congestion vs. Non-jobs-housing
Balance Variables........................................................76
Table 4.14 The Result of Correlation Based on Congestion vs. Jobs-housing Balance
Variables.............................................................. 78
Table 4.15 The Result of Correlation between the SOC vs. the SIC..................80
Table 4.16 The Results of the Fixed Nonlinear Regression..........................84
Table 4.17 The Result of Stepwise Regression by Forcing All Explanatory Variables.... 104
Table 4.18 The Result of the Forward Stepwise Regression for Selected Explanatory
Variables...............................................................110
Table 4.19 The Result of Forward Stepwise Regression between Transit Ridership
with Selected Explanatory Variables....................................114
Table 4.20 The Regression Result between Employment Density with Transit
Ridership and Mean Median Speed.........................................115
Table 4.21 A Summary of Regression Analyses.......................................121
Table 5.1 A Summary of Policy Implication..........................................138
xiv


1. Introduction
1.1 Problems and Purposes
According to the 2000 World Banks report, more than 2.7 billion people (almost half the
worlds population) live in urbanized areas, and this number is expected to reach 5.1 billion
by 2030 (The World Bank 2000). With population and economic growth and technological
advances since the Industrial Revolution, air pollution and traffic congestion have become
universal urban problems affecting our health, economic productivity, and quality of life.
To reduce air pollution and traffic congestion, preserve open space, and enhance the quality
of life, planners and researchers have been trying to discover new and more effective
growth strategies, such as the Compact Development, Urban Growth Management, New
Urbanism, Smart Growth, Balancing Growth, and Sustainable Development. They also
have tried to employ urban performance indicators to monitor the urban environment.
Some urban performance indicators such as population density and vehicle miles of travels
(VMT) have been used to monitor traffic congestion by many researchers, who believe that
encouraging compactness will enhance transit usage and result in the reduction of VMT
which in turn will alleviate traffic congestion. However, these applications may not be
appropriate, since the reduction of VMT or trip length cannot directly ease roadway
congestion, if the roadways are insufficient and the internal developed area is too dense.
Another indicator car ownership was used by the World Bank to monitor traffic
congestion, and this was based on the statement that:
ownership of passenger cars has increased in recent years, and the expansion of
economic activity has led to the transport by road of more goods and services over
greater distances. These developments increased demand for roads and vehicles,
adding to urban congestion, air pollution, health hazards, traffic accidents, and
injuries (The World Bank 2000).
However, a higher car ownership does not necessarily result in serious roadway congestion,
if more people use the public transit instead of driving or if more roadways are built.
1


Around the world, many researchers have tried to use various urban measurements such as
population density, car ownership, vehicle miles of travels (VMT), and energy consumption
to monitor the urban setting and traffic congestion within it. However, they are unable to
completely understand the interaction between land use and transportation, and to employ
the appropriate approach to link indicator applications with urban development policies.
Today, there is perhaps an emerging consensus about the proper questions to ask, the
appropriate data to examine, and the limited potential to solve transportation problems
through the manipulation of activities and land users in urban space.
This research seeks to expose the factors that affect traffic congestion, especially work
related trips at AM-peak hours. One justification for this focus is that traffic congestion
seems to be the root cause of many urban problems (air pollution and general environmental
degradation). The National Air Quality and Emissions Trends Report of 1991 from the U.S.
Environmental Protection Agency indicates that transportation sources remain a primary
contributor to the air quality problems, particularly in metropolitan areas, in spite of the
overall drop in emissions between 1970 and 1991. Reduction of traffic congestion would
surely result in enhancement of air quality and perhaps the quality of life in cities as well.
Therefore, based on the fact that work related trips have contributed more than 50 percent
of the morning peak of roadway congestion since the 1990s (but not the PM-peak-hour
congestion), this dissertation focuses upon 46 metropolitan statistical areas (MSAs) and
uses three types of urban measurements (land use, transportation, and urban form) to
explain the underlying determinants of AM-peak-hour work related traffic congestion.
Based on the hypothesis that the more dispersed the municipalities in an MSA, the more
the AM-peak-hour work trips generated (more people commute from one municipality to
the central city), and the trips in turn result in congestion, this dissertation aims to probe
the relationship between several metropolitan form elements and overall regional
transportation congestion. Attributes of areal shape and land use factors, and of the
transportation system per se are applied to explain metropolitan roadway congestion. Since
congestion (dependent variable) arises from the interaction between travel demand and
aggregate transportation system capacity (supply), it is statistically regressed against several
system attributes. Several geographic information system (GIS) programs such as
TransCAD, AutoCAD Map 2000, and ArcView are used to obtain the minimum joumey-to-
2


work distances and other explanatory attributes, since each program has its own specific
application functions. Moreover, all of the census and transportation data employed here
are from the 1990 Census of Transportation Planning Package (CTPP), since the package
provides the most detailed demographic and transportation information for both origins
(residential locations) and destinations (workplaces). By using the CTPPS and matching
the distance of the shortest path with the average travel time for each selected route from
the home to the workplace, hopefully, this dissertation will be able to define the most
reliable relationships between land use, transportation, and urban form.
u
3


1.2 Findings and Other Related Issues
Evidence of causality based on a cross-sectional examination (46 MSAs) of the interaction
among the three categories of explanatory attributes and between these three and congestion
itself yields the following conclusion: population densification and insufficient roadway
supply tend to worsen traffic congestion, higher transit usage does not contribute to less
congestion, and a greater jobs-housing balance does not necessarily result in less congested
roadways either. In fact, the relationship between the jobs-housing balance and the
roadway congestion may vary depending on the characteristics of the industrial sector in
each MSA. Some evidence also indicates that greater balance probably results from serious
traffic congestion because households are more prone to shorten work trips in more highly
congested settings.
Roadway expansion is surely able to alleviate congestion in the short term, but securing
right-of-way for road construction within the built-up area is costly. Another remedy such
as compact development is likely to raise the population density within the developed
interior area, and may result in even worse roadway congestion. This research has
recommended two types of strategies to reduce work related AM-peak-hour traffic
congestion. On the supply-side, we may expand the existing infrastructure and adopt the
transportation system management. Demand-side strategies include distributing population
more uniformly to several urban villages to encourage non-single-occupant vehicle usage,
promoting transit options and the jobs-housing balance/imbalance based on the economic
characters of the MSA, educating commuters to shift mode choice, and encouraging the
usage of telecommunications for work. Also, a solution based on the multiple regression is
proposed to accommodate transit usage enhancement combined with congestion reduction
through controlling employment density, since the higher the employment density, the
higher the transit usage, but the higher the employment density, the more serious the
roadway congestion.
4


2. Literature Review
Most implemented plans have been based on what we already had, and it is veiy rare for
planning to start from scratch (Catanese and Snyder 1988). Through understanding the
evolutionary process of land use and transportation planning and reviewing contemporary
responses to urban problems, we may begin to understand what planning has attempted to
achieve in the past and what should not be repeated. Therefore, this literature review starts
with the historical evolution of land use and transportation planning to understand their
interaction. After that, several basic concepts of indicators applications are introduced.
Finally, alternatives to ease roadway congestion are discussed, which become the basic
foundation of my methodology. The subjects discussed in the following are: (1) Historical
Antecedent; (2) The Early Land Use and Transportation Planning Strategies; (3) The Use of
Urban Indicators in Monitoring Urban Performance; (4) The Alternatives to Alleviate
Traffic Congestion; and (5) The Interaction between Land Use and Transportation.
2.1 Historical Antecedents
The earliest urbanization began around 4000 B.C. (Catanese and Snyder 1988, 4). Later, a
city became a marketplace for the agricultural products of the surrounding lands; it was a
place where trade and commerce occurred (ibid., 4), and the center where ceremonies,
myths, and power were accumulated (Lynch 1987). Walking, water transportation, and
horse-drawn vehicles during this period of time were efficient enough to move around their
residents and the various goods produced and consumed. Until 1769, with the invention of
the steam engine, the Industrial Revolution (beginning in England in the late eighteenth
century) generated more job opportunities, increased productivity, opened up mass markets
for goods (Catanese and Snyder 1988, 13; Hartshorn 1992, 33), and extended the urban
spatial configuration from central city to suburb. Through transportation innovations (a key
to Industrialization) paced by railroad and steamship advances, factories were made more
productive and then were able to generate additional capital to support urban growth (ibid.).
5


It was not until the late 18th century that transportation (streetcars and rails) had become a
major factor changing urban settings from compact to decentralized. Before the Industrial
Revolution, work was carried out in homes or in shops close to homes. With mechanization,
industrial production became more centralized in large factories (Catanese and Snyder 1988,
14). Workers homes were frequently distant from their workplaces. Raw materials with
the facilitation of the transportation system were brought to the factories and finished
products distributed to market areas (ibid., 14). From the 1890s until World War II,
streetcars and transit systems were fully deployed to transport people. On the other hand,
the automobile was used largely for recreational purposes due to relative high prices and a
restricted market. .In 1920, nearly 90 percent of urban commutes occurred either by
streetcar or by rail transit (Hartshorn 1992, 164). A phenomenon of city life appearing at
the time of Industrialization was the journey to work through taking public transit (Catanese
and Snyder 1988, 14). It is clear that the old walking cities had been swept aside as electric
streetcars and trains led to more dispersed development patterns (Newman and Kenworthy
1999, 40).
After World War II, transportation modes shifted from transit to automobile. The housing
boom together with a massive federal highway facilitated automobile usage, and this in turn
encouraged urban sprawl. Since then, suburban residential development has accelerated,
supported by federal mortgage insurance and the mobility conferred by car ownership
(Barnett 1995, 2). Urban problems such as air pollution and traffic congestion have arisen
with population growth due to the urban spatial configuration changes from a centralized
city to leapfrog, low-density, and dispersed suburbs. Moreover, this urban form change was
caused not only by the federal governments investments of transportation projects and the
interaction between land use and transportation, but also by the American dream of living in
a suburb with a single lot at a residential site (Barnett 1995).
6


2.2 The Early Land Use and Transportation Planning
Strategies Relating to Urban Form
Urban form not only is an expression of land use patterns in urban areas, but also addresses
the shape of the city in terms of radial extent, population density, and the ratios of housing
capital to land (Thrall 1987). Two major factors shaping the urban form are land use and
transportation; they interact mutually and affect each other. After the Industrial
Revolution, with the emergence of streetcars and the rail transit, the spatial relationship
between work and residence districts became more widely separated and more specialized;
administrative, commercial, and industrial zones emerged in the cities (Hartshorn 1992,
219). Workers often lived far away from their residences with the facilitation of public
transit. On the contrary, before Industrialization, work and residence frequently were co-
located on the same street mostly within walking distance or horse-drawn carrying range.
Today, in many areas, urban form has evolved from the concentric zone (E. W. Burgesss)
to the multiple nuclei pattern (Harris and Ullman 1945; Hartshorn 1992, 233),
emphasizing the linkages (transportation networks) among different activities having
different spatial affinities.
Regarding transportation planning strategies, in 1969, the National Environmental Policy
Act (NEPA) included a transportation element and specified that major transportation
proposals need to prepare an environmental impact statement to express the purpose of the
project and to help the decision-making process. In 1970, the Clean Air Act established
national ambient air quality standards and required States to develop plans to meet the
standards, since motor vehicle emissions were identified as major contributors to air
pollution. In 1982, the Federal Highway Administration (FHWA) proposed a 3Cs
(continuing, comprehensive, and coordination) transportation planning process to address
the social, economic, and environmental impacts, and to ensure community participation.
In 1991, the Intermodal Surface Transportation Efficiency Act required metropolitan
planning organizations (MPOs) to incorporate transportation improvement programs in
metropolitan plans to tie land use and transportation into policy development. In 1998, the
Transportation Equity Act for the Twenty-First Century (TEA-21) included a continued
emphasis on transportation enhancement projects and encouraged joint public-private
initiatives for transportation development. It is clear that the federal government monitors
7


and guides the transportation planning process through the funding incentives and
disincentives. In addition, transportations spatial imprint is regional, often crossing
different jurisdictions. On the other hand, land use planning is predominantly a local
governments responsibility. Therefore, to solve transportation related problems (e.g.
traffic congestion and air pollution), the coordination between different governments is a
key to the success. Without full cooperation within the different hierarchy of governments,
these urban problems are unlikely to be resolved completely.
Regarding land use planning strategies, the Garden City movement probably was the
earliest concept in dealing with crowding and congested issues resulting from
Industrialization. In 1898, Ebenezer Howard proposed the Garden City concept, in which
cities were built outside developed areas and provided with good transportation networks,
so that the crowding and congestion of industrial cities could be relieved and people could
return to nature. He used four criteria to define garden cities and they were the following: (a)
all land, about 1,000 acres, would be owned singly in the public trust; (b) population and
development would be staged until a maximum of 30,000 people were housed; (c) a
greenbelt of 5,000 acres of agricultural land would surround the city; and (d) there would be
a mixture of land uses to ensure social and economic self-sufficiency (Catanese and Snyder
1988, 16). Generally, this Garden City concept is similar to the Urban Villages concept.
Newman and Kenworthy (1999) characterized the Urban Villages as follows:
1. High-density land use at the center and everything in the village is within
walking and cycling distance;
2. Mixed land use with offices, shops, businesses, and community facilities
to encourage more local activity;
3. A mixing of public, private, and cooperative housing;
4. A heavy rail or light rail station near the core;
5. A pedestrian circulation system linked with parking facilities placed underground
to create a traffic-free and people-oriented environment;
6. Public spaces with strong design features (water, street furniture, playgrounds);
and
7. A high degree of self-sufficiency in the community to meet local needs, but with
good rail and bus links to the wider city for employment, higher education, and
so on. (Newman and Kenworthy 1999, 166)
8


Although self-sufficient cities/villages/satellites seems to be an alternative approach to
minimize travel then to help to relieve congestion, people living in suburbs inevitably still
need to commute from place to place or travel to major central cities to obtain the main
public services such as courts, or visit museums, zoos, and so on. Therefore, on the
municipality scale, even if traffic congestion within the villages or suburban cities is
reduced by avoiding longer commutes, the congestion from village to central city trips on
the metropolitan scale might be increased, unless the suburban cities are fully self-sufficient.
Based on the evidence that suburb housing was common at the periphery of cities, while
lower-paid workers lived in the congested central area since the late nineteenth or the early
twentieth centuries, the Garden City concept aims "to promote rural protection, to create
compact and efficient smaller urban forms within defined boundaries as an alternative to
suburban sprawl, to minimize commuting and even to promote gardening [which] remain
impressively green goals" (Ward ed. 1992, 205). However, the Garden City concept clearly
inevitably caused suburban sprawl and the location of traffic congestion has shifted from
inner village to outer village. Moreover, it is also clear that planners have been trying to
encourage a high density, mixed land use, and the usage of transit, as well as to emphasize
the walking and cycling environment to build so-called self-sufficient cities.
As our metropolitan areas become more and more fragmented while suburban areas
continue to sprawl and some satellite cities have the transit system to commute to central
cities, the transit system alone is unable to take care of the congestion, unless highway
systems and other land use and transportation policies are integrated. Moreover, it is very
rare to have a real self-sufficient municipality or village, since a satellite city or village
inevitably has to interact with other suburban villages or central cities to obtain full public
service. A non-congested road network within an inner village does not mean that people
can move easily from that village to other places. The concept of Garden City or urban
village or new urbanism actually might be able to reduce inner-city traffic congestion
and pollution, but it probably could not solve the congestion problems completely on a
metropolitan scale. To address congestion properly, an appropriate way to measure traffic
congestion for an urban setting should be based on either a metropolitan or urbanized scale
and should consider the interaction between land use and transportation systems. Moreover,
to control traffic congestion and air pollution, a well-coordinated government has to be in
place.
9


2.3 The Use of Urban Indicators in Monitoring Urban
Performance
From the late 1800s to the beginning of the twentieth centuiy, the rapid growth of cities had
led to the development of many urban problems. Since then (1968), the British passed their
Public Health Act in England to mainly deal with housing standards, and the Americans
established the Housing and Urban Development Act and the New Community
Development Act (Catanese 1988, 16). The intensive study of indicators applications
concerning the quality of life were probably first started during this period of time.
However, the policy implications resulting from such indicators were often unrelated to the
precise requirements and linked to unknown causalities, since the methods of selecting a set
of indicators for monitoring the built environment were mostly based on discretionary
decision-making without any planning theory supported (Port 1996). Hart and Farrell (1998,
7) found that:
the choice of a particular indicator is guided by two considerations: what one
wishes to know and how the information will be used. Generally, scientists and
analysts are interested in seeing the raw data and interpreting it themselves.
Policymakers are more interested in summary information that is clearly related to
policy objectives, evaluation criteria, and targets, and they usually do not want to
perform much analysis themselves, although they may be interested in how it is
done. (Hart and Farrell 1998, 7)
Although scientists and analysts tend to employ raw data for interpretations, these are often
based on the scientists and analysts own intuition.
To further understand how indicators have been employed to monitor built environment,
Hart and Farrell (1999) found that:
an indicator as something that provides useful information about a physical, social,
or economic system (usually in numerical terms). It can be used to describe the
state of the system, to detect changes in it, and to show cause-and-effect
relationships. For instance, the level of water in a reservoir is a state, drawdowns
represent change in that state, and comparisons of these variables over time can
10


reveal cause-and-effect relationships such as the impact of conservation policies on
water usage. (Hart and Farrell 1999, 7)
Indicators have also been developed and used by many municipal and national authorities
and some international organizations (e.g. the Organization for Economic Cooperation and
Development [OECD], the World Bank, and the United Nations) to identify and study
specific urban conditions and problems. Furthermore, they have substantially been used to
gauge environmental, urban, social, and health conditions (OECD 1997). They have also
been used to address economic performance, energy consumption, urban growth, and
sustainability issues in general. Unfortunately, most of these indicators do not explain the
underlying causes: consequently, they are not appropriately applicable to guide ameliorative
actions due to insufficiently supported evidence. Even though indicators can help to
identify problems, which ought to be analyzed through using quantitative and qualitative
data and information (OECD 1997), the underlying explanations linking an indicator and a
specific action need to be built and based on the foundation of planning theories. Therefore,
researchers (Cervero 1989 and 1996; Crane 1999 and 2000; Dunphy 1996; Frank 1995 and
2000; Handy 1996 and 1997) have studied different indicator usages such as jobs-housing
ratio, density, mixing land use, travel time, vehicle-miles of travels, and travel duration, to
describe and explain travel behavior and try to find solutions to reduce congestion and air
pollution. Unfortunately, these works are still inconclusive and have no complete
consensus (Boamet and Samiento 1998). However, we could classify two types of indicator
usages based on their applications. One is the institutional usage and the other is the
academic usage. These two usages are discussed below.
2.3.1 The Institutional Indicator Usage
In the 1997 OECDs report, an indicator is defined as an empirical interpretation of reality
but not as the reality itself (OECD 1997, 14). Indicators are commonly used to present a
quantitative account of a complex situation or process; they can be used to point out or
identify something that is not immediately visible, audible or perceived in a precise
situation; they can also translate data and statistics into succinct information that can be
readily understood and used by several groups of people including scientists, administrators,
11


politicians and citizens with a wide range of interests (ibid.). Moreover, indicators are not
criteria or standards, but they need a criterion or standard to interpret their meanings (ibid.).
Many of the current indicator applications do not either imply the underlying causalities or
link to policy implications; indicators have been only treated as statistical information in
general. Take the World Banks 2000 World Development Indicators for example.
Indicators such as vehicle ownership, numbers of motor vehicles, road traffic
(vehicle-kilometers of travel), traffic accidents, travel time to work, income
differential, and housing price are selected to gauge the traffic condition for different
nations (The World Bank 2000). Later, a conclusion was made that congestion is the most
visible cost of expanding vehicle ownership (ibid., 161). It is unfortunate that vehicle
ownership may not directly relate to congestion, since congestion happens when the
roadway traffic volume exceeds or already has exceeded the roadway design capacity. A
higher vehicle ownership does not necessarily imply a higher number of vehicles driven on
road networks, and therefore a higher vehicle ownership may not result in serious roadway
congestion. In this research, I have found that vehicle ownership has no relationship with
congested roadways, which is discussed in Chapter 4.
Another popular institutional indicator usage is in association with performance
standards. This usage aims to identify the correspondence between ambient conditions and
a stated goal, threshold, or policy, and is intended to show achievements with respect to
redefined goals or targets (OECD 1997, 23). Porter (1996) showed that there were at least
nine different indicator applications. These are planned unit developments, industrial
standards, flexible zoning, point systems, environmental standards, adequate public
facilities ordinances, thresholds, benchmarks, and project performance. A good example of
this type of usage is the point system employed by the city of Boulder, Colorado, which is
also used to evaluate their growth management policies. In this system, there are criteria
and points (scores) defined for each indicator, and the indicators include, for example,
distance to transit route, joint parking, and site next to historic preservation locations.
If a development project does not make any impact upon the surroundings based on the
above measurements, it will gain the highest point and will be allowed to be developed. On
the contrary, if another project could possibly make impacts to the area, each of the adverse
impacts would result in a lower point total. If the aggregated points are less than the
minimal requirement, that project will not be allowed to proceed. In general, this type of
12


indicator usage is a variation of the traditional planning and zoning approach. Although the
institutional indicator usage at the municipality level has been popularly utilized for a long
time, it has at least two shortcomings:
First, it is very difficult to define a set of right numbers for performance indicators. Port
(1996) pointed out that:
in many communities, the performance standards have been passed down through
previous ordinances or borrowed from the community next door. Although
technical studies and comparative analyses can shed light on the pros and cons of
alternative levels or targets, decisions on quantitative measures often rest more on
common understandings and perceptions than factual data. (P. 4)
Then, Port (1996) also used the Level of Service (LOS) in transportation planning as an
example:
The LOS indices were established by a committee of the Institute of
Transportation Engineers as a guide to describe the degree of traffic congestion at
intersections and on highway segments. Traffic counts at the intersections during
peak commuting hours establish the LOS levels from A (free-flowing traffic) to
F (traffic flows often interrupted), thus indicating capacities for additional traffic
generated by future development or, alternatively, needs for road improvements or
traffic demand measures to provide capacity for new development, (ibid.)
The comparison is the key approach while employing the LOS. However, using a traffic
count at AM-peak hours (usually from 6:30 AM to 8:30 AM) for example, the LOS is
probably unable to tell the whole traffic story, since the intersection at a particular location
or in an urbanized area may have different starting time and duration. The LOS standard
that was established based on weekdays might be inappropriate for applying to weekends.
In my research, I have found that work trips are not the main contributor to AM-peak-hour
traffic congestion, when a metropolitan area is identified by entertainment or agriculture as
its economic character.
Second, the employed performance standard often depends on discretionary decision-
making due to the difficulty of defining the right number or the difficulty of employing a
sophisticated evaluation system. If the number needed to be adjusted according to a
different situation or project site, this number of the performance standard may not be able
13


to be updated, since public officials probably do not have the full capability to calibrate
these numbers; they often have relied on their own intuition to establish the measurement or
to adjust the number.
2.3.2 The Academic Indicator Usages
Indicators or performance measures such as vehicle-miles traveled (VMT), travel time to
work, vehicle speed, car ownership, population density, income, employment,
jobs-housing ratio, and degree of mixing land use have been used to monitor air
pollution and traffic congestion, and to document transportation mobility. However, the
implemented policies based on these indicator applications are sometimes problematic,
since they do not solve the urban problems as they are expected to. For example, the
reduction of VMT, trip length, and car ownership were expected to ease roadway
congestion. But, the reduction of VMT or trip length may be unable to ease roadway
congestion, if the roadways are insufficient and the internal developed area is too dense.
Also, higher car ownership does not necessarily result in more serious roadway congestion,
if more people use public transit instead of driving or if more roadways are built. The
further discussion of the academic indicator usage is in section 2.5.
2.3.3 The Important Difference between Environmental
Conditions and Quality
To appropriately use an indicator, one must understand the difference between
environmental conditions and environmental qualities, since each requires a different
method for monitoring the built environment. Environmental conditions can be measured
objectively by using indicators, since they are the physical facts, which can be described
and measured with reasonable precision based on an explanation of what is happening to
our environment (Milbrath 1978). The examples of this type of measure are many, such as
levels of cleanliness of air and water, numbers of hospital beds per residents, miles of
developed roads for a region, gross national product per capita, average level of education
for a region, numbers of automobiles per household, and so on (ibid., 38). On the other
14


hand, environmental quality can only be measured subjectively by employing indicators,
since quality related measurement is more associated with an individuals perceptions
(ibid.). Take the quality of life for example. If a person is satisfied with his living
environment and believes that his home is of high quality (subjective scoring), the
environmental quality of his home will be a good standard to him (ibid.). Similarly, if a
person believes that his water is clean and it tastes good to him, the water quality will be
ranked as a high quality, even if the water might contain impurities and may harm him
(Milbrath 1978). In some cases, high scores on the objective measures of environmental
conditions do not always lead to a higher environmental quality (ibid., 38). Therefore, the
study of traffic congestion and air quality is about the environmental conditions type of
research. Both roadway congestion and air quality can be measured objectively by
employing scientific method differently. On the other hand, the study of the quality of life
is the environmental qualities type of research. Researchers have used the public
participation process for studying the quality of life by interviewing people who live in the
community. However, sometimes, both types of studies might eventually have a mutual
affect. Take congestion and pollution reduction for example. Since congested highways
are a symptom of the deteriorating quality of life in a community and since congested road
conditions can have a detrimental effect on air quality (Meyer 1997), reducing traffic
congestion not only improves an environmental condition (air quality), but it also enhances
the quality of life (The World Bank 2000). The details of roadway congestion
measurements are discussed in section 2.4.1.
To evaluate quality of life, Hartshorn (1992) provides two approaches and both are
perception studies. One is the structural approach and the other is the evaluative approach.
The structural approach focuses on the identity and form of geographical space perception
and how information about the environment is obtained (Hartshorn 1992, 202); Lynchs
Image of the City is an example of this approach (ibid.). On the other hand, the evaluative
approach goes one step further in the organization of space by placing emphasis on the
manner in which individuals respond to information from their perceived environment
(Hartshorn 1992, 202). In other words,
the evaluative approach is concerned with the behavioral-spatial impact of the
environment on individual activity. . Fundamentally, the idea underlying the
evaluative approach is that people have the capability of perceiving the most
15


important elements in their environment and the recognition of these elements
affects their decision-making process, (ibid., 202)
2.4 Possible Alternatives for Alleviation of Traffic Congestion
in Urban Areas
Based on literature review, to alleviate traffic congestion, there are at least four approaches.
Before introducing these four possible alternatives, I would like to discuss the possible
methods to monitor roadway congestion.
2.4.1 Traffic Congestion Measurements
People always believe that they know exactly what congestion is. When they are asked
for a definition of congestion, they are likely to give different definitions based on their
perception. Therefore, congestion is a term which is difficult to define, as people believe
they know exactly what it means. Generally, congestion arises when demand levels
approach the capacity of a facility at the time required for its use (travel through it), and
when demand increases well above the average under low demand conditions (Ortuzar and
Willumsen 1994). In other words,
traffic congestion means that there are more people trying to use a given
transportation facility during a specific period of time than the facility can
accommodate with what are considered to be acceptable levels of delay or
inconvenience. In a broader sense, a congested facility is just one element of
transportation systems ability to provide mobility and accessibility. Delays at
particular locations in a transportation network are certainly aggravating to those
using the system, but these delays are part of a much larger picture of how a
transportation system allows people and goods to move around in a metropolitan
area. (Meyer 1997, 2)
Traditionally, congestion has been forecasted by using the standard four-step transportation
demand modeling approach (from trip generation, trip distribution, mode choice, to traffic
assignment). The result of traffic assignment usually includes traffic volume and speed for
each link of the road networks, and congestion is determined by comparing volume and
speed with the level of service (LOS) standard. In addition, traffic engineers employ
16


delay time at intersections and use the weighted mean method to aggregate the volume
and travel time for each directional traffic flow to gauge congestion. Another approach of
determining congestion is based on the Urban Mobility Study published by the Texas
Transportation Institute (TTI), in which the TTI employs a roadway congestion index
(RCI) to document traffic congestion for all-trip purposes on roadway networks in the
United States every year. This congestion index is calculated by using the aggregation of
demanded VMT divided by the aggregation of designed VMT, and the equation (2.1) is
as follows:
RCI = [(Freeways VMT/Lane-Mile x Freeways VMT) + (Principal Arterial
Streets VMT/Lane-Mile x Principal Arterial Streets VMT)] [(13,000 x
Freeways VMT) + (5,000 x Principal Arterial Streets VMT)] (2.1)
Basically, this equation is similar to the concept of real traffic volume over design
capacity (V/C). Whenever the index exceeds 1.0, it implies a degree of seriously
congested roadways. Moreover, no matter what method is used to describe traffic
congestion, the LOS developed by the Institute of Transportation Engineers remains the
basic standard and method that researchers and planners have used to measure roadway
condition. Based on the LOS theory, there are various ways to gauge traffic congestion,
such as speed, volume, density, and real traffic volume divided by design capacity (V/C).
The roadway congestion index is just a different format of the V/C through using the
VMT. Although many traffic congestion studies employed either VMT, travel time, or
travel duration in length to gauge roadway congestion (these are the most popular examples
of indicators misuse), these studies fail to include the capacity concept in the research.
In this dissertation, because of data availability, I use mean median travel speed (a
variation of measure similar to the V/C concept) to monitor work-related AM-peak-hour
traffic congestion. The comparison between the TTIs Roadway Index with my Mean
Median Travel Speed is shown in Table 2.1.
To study traffic congestion, one needs to identify what type/purpose of trip that is addressed.
By reviewing the 1990s and the 1995s Nationwide Personal Transportation Survey reports,
we know that AM-peak-hour congestion is closely associated with work related trips, but
not for PM-peak-hour congestion (50 % of work trips contributing to the moming-peak-
17


hour traffic congestion). Therefore, this research focuses only on work related AM-peak-
hour traffic congestion.
Moreover, generally, there are two types of approaches (supply side and demand side)
to ease roadway congestion (Ferguson, 1998). The supply-side approaches include
increasing supply, increasing efficiency of existing supply by shifting supply schedule, and
decreasing supply by forcing demand management (ibid.). The demand-side approaches are
similar with the supply-side and include increasing demand, increasing efficiency of
existing demand by shifting demand schedule, and decreasing demand by educating
commuters shifting mode choices (adopting transit systems as an attraction) (ibid.). If the
supply is the infrastructure for example, the demand will be the travel behavior. In this
research, I have categorized at least four different alternatives for alleviating roadway
congestion (see Figure 2.1). They are to do the following: (1) redesign or add more
infrastructure to increase capacity; (2) employ transportation system management or
transportation demand management by encouraging mode shift or changing individual
travel behavior; (3) encourage new urbanism development by minimizing unnecessary
travel; and (4) rearrange land use patterns and promote public transit usage.
18


Table 2.1 The Comparison between the TTIs Roadway Index with Mean Median
Travel Speed
The Criteria for Comparison The TTIs Roadway Congestion Index Mean Median Travel Speed Employed in My Dissertation
Measurement Roadway Congestion Index (RCI) = [(Freeways VMT/Lane-Mile x Freeways VMT) + (Principal Arterial Streets VMT/Lane-Mile x Principal Arterial Streets VMT)] [(13,000 x Freeways VMT) + (5,000 x Principal Arterial Streets VMT)] Where 13,000 vehicles per lane per day is interpreted as the beginning of level-of- service D operation for highway in an urbanized area, and 5,000 is for principal arterial. Mean Median Travel Speed for a metropolitan statistical area = Average of every [Shortest travel distance of a route from place to place based on highway networks (only one route) Median travel time at AM-peak hours of the same route]. Assuming that people will use the shortest distance of the route to travel from place to place, if there is no congestion. Therefore, Mean Median Travel Speed is a least congested route in distance divided by congested travel time. The slower the speed, the higher the degree of congestion.
Congestion Period AM and PM Peak Hours AM Peak hour (6:30 a.m. to 8:29 a.m.)
Type of Trip All types of trip purpose. Work trip only. Different types of trip purpose are caused by different land use factor and human behavior. Therefore, without specifying the trip purpose, the research is likely unable to suggest any appropriate policy implication.
Study Units Urbanized Area only population density. Metropolitan Area commuting patterns and population density.
Factor Consideration Roadway characteristics including capacity. Land use, urban form, and roadway capacity.
Implication TTI could only suggest road improvement strategies such as high-occupancy lane, or encourage enhancement of transportation demand policies such as transit usage. My dissertation employs three types of factors (land use, transportation, and urban form), and therefore I could suggest future land use policies and the effectiveness of using transit systems to reduce traffic congestion.
Source: Authors editing.
19


Reduction of Air
Pollution and
Enhancement of Quality
of Life
Reduction ofTraffic
Congestion
Mode Choice Reduction of Trip Redesigning/Adding
Gene ration more Infra structures
Application of Transportation Demand Promotion of Public Rearrangement of Land Use Structure/ 5] 1 Designing Better Living Environment
Management/ Transportation Systems Management Transit H Urban Spatial Configuration I Automobile Commute
tsaiaE jr;,
U ' * 1 r- V
Through
Transportation
Planning Approaches
(Ridership, Vehicle
Ownership, Mode
Choice Survey,
v Policies Subsidy) J
/Through Urban and\
/ Transportation
/ Planning Approaches
\ (Land Use, Density,
\ Population, and
\ Operation cost)
/ Through Urban/
/ Regional Planning
/ Approaches
\ (Land Use, Population,
\pensity, Income,
\ and Urban form)

Note; The shadowed boxes are the selected approaches for this dissertation.
Figure 2.1 Traffic Congestion Reduction Alternatives
20


2.4.2 Alternative One: Adding More Capacity
The first alternative, adding more capacity to the existing infrastructure, is the approach that
transportation and traffic engineers are always asked to do. It is the strategy that has been
mostly adopted today, but the consequence is that new construction does not solve growing
congestion (Schrank and Lomax 2001). The cycle of creating road supply, which in turn
eats up the capacity, is unable to solve traffic congestion, since rezoning the future
undeveloped land facilitated by infrastructure improvement is hard to predict (at least not
until it is developed). In some cases, even if the land use plan is able to include the future
development, it does not mean that the site will be developed as the developer originally
requested, since the developer might change his proposal and ask for a higher density of
development after the accessibility is improved. Then, after the final rezoning requests
approved by the planning official, the new proposed development is likely to result in
demanding more road capacity to accommodate the final plan in the end. However, for a
dense and well-developed city, it might be very difficult to build additional new roads due
to a lack of right-of-way. On the other hand, even if there is an available right-of-way, the
government probably has to spend a high development cost to purchase the land, and
therefore expansion of existing infrastructure might be costly (not cost effective) and
infeasible.
2.4.3 Alternative Two: Transportation Demand Management
Regarding the second alternative, transportation system management (TSM) and
transportation demand management (TDM) are very similar. However, TSM focuses on
supply-side in identifying potential solutions to transportation problems. It requires projects
to conserve energy, to promote transit usage, and to provide a greater priority for high
occupancy vehicles (HOVs) (Ferguson 1998). It generally seeks for low-cost ways to
expedite traffic flows. Some management strategies suggested by Hartshorn (1992) under
TSM simply involve more effective traffic control and engineering, which include the
following:
1. Parking restrictions;
21


2. Reversible lanes;
3. Turn controls at intersections;
4. One-way streets;
5. Signing, signaling, and computerization;
6. Stricter traffic enforcement;
7. Banning deliveries and/or trucks during rush hours;
8. Contraflow lanes (traffic flowing in reverse direction of normal flow during
peak hours);
9. Turn lane flyovers into major employment centers; and
10. Goods-delivery restrictions in curb lanes. (Hartshorn 1992, 187)
On the other hand, Transportation Demand Management (TDM) is more demand oriented
than supply-side. It aims to make more efficient use of transportation resources (already in
place) by shifting demand (travel behavior) such as carpooling, or eliminating trips such as
by encouraging the usage of telecommunications for work. Some management strategies
under TDM for example are rider-share promotion and parking management. From a land-
use planning perspective, TDM is more market driven than either TSM or traditional
transportation planning (Ferguson 1998). It addresses more of a land-use issue than a
transportation issue by gaining support from the private sectors, such as large employers,
developers, and business. Thus, a collective community participation is vital to the success
of TDM programs, whether it is in the format of purely voluntary, market driven, or a
regulatory response to social issues and environmental problems (Ferguson 1998). In
general, TDM works best when the public and private sector work together on a voluntary
basis, and this approach should be developed on an individual basis to suit the particular
needs and requirements of the communities on a long term basis (ibid.).
To reduce congestion, TSM seems to work best on a municipal scale, and TDM is more
appropriate for a metropolitan scale. However, the effectiveness of employing TDM to
alleviate congestion needs to be evaluated. Since fewer people commute from suburban
cities to downtown and more people commute from one suburban city to other suburban
cities within a metropolitan area, the attempt to encourage minivan routes (most routes are
from suburban cities to a central city) or car pools has relatively little success (Barnett
1995). Moreover, many new cities are too spread out and fragmented to be served
22


effectively by the minimal bus systems (ibid., 6). Therefore, as suburbs continue to sprawl,
the strategy of using TDM to reduce traffic needs to be modified as well. It is likely that
educating people to shift their mode choice is the most important and fundamental
prescription, even though it needs a little longer before it takes effect.
2.4.4 Alternative Three: New Urbanism Development
By providing pedestrian-friendly communities based on compact and mixed-land use
development, new urbanism development approach tries to reduce automobile dependence
by making the community a more pleasant place to walk and cycle and to encourage more
face-to-face activities. However, alleviating congestion on the neighborhood scale of a new
urbanism community is only part of a much larger picture of a transportation problem, and
reducing congestion at a local level probably does not help too much on the metropolitan
scale especially for joumey-to-work commutes, since people still have to commute to other
places (outside of community) for work. Nevertheless, the New Urbanism concept is still
able to reduce the congestion at local or municipal levels by encouraging more walking and
cycling trips to compensate for vehicle trips.
2.4.5 Alternative Four: Land Use and Transportation
Interaction
The approach of land use and transportation interaction is discussed in section 2.5.
Although I have categorized three other alternatives to deal with traffic congestion, I mainly
consider this altemative/approach (land use and transportation interaction) for the following
discussion.
23


2.5 The Discussions on Land Use and Transportation
Interaction
Travel patterns, volumes, and modal distributions are largely a function of the spatial
distribution of land use (Meyer 1997, 13). On the other hand, land use pattern is influenced
by the degree of accessibility provided by the transportation systems from one activity area
to another (Meyer and Miller 1984). A developed land for a particular use results either in
the generation of new trips originating from that area or new trips attracted to that area, or
both (ibid.). This piece of developed land in an urban area creates new travel demands, and
requires a need for transportation facilities, either in the form of a new infrastructure or a
more efficient operation of existing facilities (ibid.). Then, such improvements to the
transportation systems make the land more accessible to existing activity centers, thereby
making it more desirable and affecting its monetary value (ibid., 62). Through increasing
accessibility and improving land values, these actions in turn influence the locational
decisions of individuals and firms to relocate their office to the new area. Later, this new
development in the new area for the individuals and firms again requires expansion of
transportation facilities to accommodate the need of accessibility. Generally, the cycle of
this interaction between land use and transportation will keep on going, until an equilibrium
state is reached or until some other external factor intervenes (Meyer and Miller 1984).
Recently, more and more land use and transportation interaction studies have been
conducted due to the little success of using TDM and of investing more new constructions
to relieve congestion. Williams (2000) and others stated that although it is now widely
accepted that a relationship exists between the shape, size, density, and uses of a city and its
sustainability, the consensus of the relationship is lacking about the exact nature (Williams
2000, 1). Also, the relative sustainability [in terms of urban form] of high and low urban
densities, or centralized and decentralized settlement is still disputed. . Some forms may
be sustainable locally, but may not be beneficial city wide or regionally (ibid., 1). Lynch
(1987) treated transportation (also communication) as the important asset/component for an
urban area, and activities were located according to the relative cost of reaching materials,
customers, services, jobs, or labor (Lynch 1987, 187). He also found that some types of air
pollution are positively correlated with city size as well as the travel time to work (ibid.,
241). Although planners increasingly view land-use policy as a way to manage transport
24


demand and employ land-use policies as potential solutions to solve transportation problems,
the evidence of the link between land use and travel behavior is inconclusive and there is
little evidence existing to support the belief that such policies can be effective in reducing
automobile dependence (Boamet 1998; Handy, 1997).
Crane (2000) emphasized the influence of urban form on travel, providing me many good
references on land use and transportation interaction issues. Newman and Kenworthy (1999)
also published a book on the discussion of sustainability with automobile dependence,
displaying rich empirical examples on the interaction between land use, transportation, and
urban form. Although there are also plenty of similar studies proposed and the results are
improving in several aspects, there is no consensus reached yet due to an inaccurate
interpretation of data and methodology misuses (e.g. inappropriate usage of regressions and
unsuitable variables for regressions) (Boamet 1998). However, Frank (2000) wrote an
article on land use and transportation interaction, where he pointed out in particular three
frameworks for his discussion densification, intermixing of land use, and the jobs-
housing balance. Therefore, the following discussion is based on these frameworks and will
start with the Jobs-Housing Balance, then mixing land use, then density, then air pollution,
and finally the constraint of land use and transportation interaction study.
2.5.1 The Jobs-Housing Balance
In 1989, Robert Cervero used statistics and a gravity model to analyze the jobs-housing
balance. He used joumey-to-work volume (number or work trips) to correlate with number
of employed residents, number of employed residents working in service occupations (e.g.
clerical, operations, labor, farming, sales, etc.), number of employed residents working in
professional and technical occupations, employees in service occupations (e.g. clerical,
operations, labors, farming, sales, etc.), median cost of single family, and travel time. The
empirical results based on a goodness-of-fit test by Cervero (1989) showed that:
1. All variables are statistically significant; and
2. Cost and availability of housing are among the most important factors that
shaped the residential locational choices of suburban workers. (P. 136-148)
25


Later, stepwise regression was used by Cervero to define the relationship between
congestion (based on the LOS standard) and the Jobs-housing ratio. He found that suburban
centers that are big, dense, and housing-free in character tend to suffer the worst nearby
freeway congestion (Cervero 1989, 145). Cervero then concluded that workers were forced
out of suburban employment areas where single-family homes were costly and residentially
zoned land was in short supply (ibid.). He also argued that inclusionary housing programs
and a number of incentive-based programs could reduce jobs-housing mismatches for those
suburban places with severe jobs-housing imbalances, and these policies in turn should be
able to reduce congestion on connecting freeways (ibid.).
Unfortunately, in 1996, Cervero revisited his 1989s Jobs-Housing Balance policy and
found that several communities were nearly perfectly balanced, yet fewer than a third of
their workers resided locally, and even smaller shares of residents work locally (Cervero
1996a, 492), based on the 23 large San Francisco Bay Area cities that adopted this strategy.
Since the Jobs-Housing Balance is unable to reduce work-based travel, Cervero then
concluded that more relevant to the reduction of commuting duration and VMT are adding
more housing in or near jobs-surplus cities and reducing rates of external commuting, in
part through the production of housing appropriate to the earnings and taste preferences of
workers (ibid., 516). Clearly, he tried to emphasize that providing more housing near the
job centers is more related to reducing commuting distance and vehicle miles traveled,
which therefore probably could result in congestion reduction. However, there are several
issues responding to the Jobs-Housing Balance as follows.
1. Two workers in same household:
Giuliano and others (1991) believed that workers in two-earner households usually work in
different locations and frequent job turnover reduces the ability to locate ones home close
to his/her workplace. He also pointed out that there were other factors that could affect
residential locational choice, such as the quality of schools. Therefore, there are other
possible external factors related to trip and congestion reduction, and the jobs-housing
balance policy seems to have less effect on this issue compared with other factors (Giuliano
et al. 1991). However, it is likely that people would tend to move to any residential location
near their work, if the government could provide adequate funding for good education
(through policy implication) and other public services. Moreover, according to the 1990s
26


Nationwide Personal Transportation Survey (NPTS), 50% of work trips occur during AM-
peak hours, and only 27% of church and school trips happen at the same period. In other
words, it seems that there were 50% (in 1990) and 55% (in 1995) of workers suffering AM-
peak-hour congestion, and therefore they will probably choose the residential locations near
workplaces regardless of other external factors such as the quality of schools, if the
roadways were seriously congested. In this research, in order to include the quality of
schools issue, I have employed median income as an indicator to index it through
correlating it with mean median speed, since a place having a good school is often the place
where wealthy people live.
2. Work-trip versus non-work-trip:
Richardson and Gordon (1989) argued that the jobs-housing balance has little effect on the
fastest growing travel region, since joumey-to-work trip is not the major contributor during
peak hours (P. 2). Although they used the Nationwide Personal Transportation Survey
(NPTS) and reached this conclusion, there were some debates relating to their research.
First, work trip (50% in 1990, 55% in 1995) is still the main contributor of AM-peak-hour
travel based on the 1990s and the 1995s NPTS, but not for PM-peak-hour travel. Second,
the Nationwide Personal Transportation Survey could only provide residency data, for
which there is no detailed census by workplace; also the census numbers are too general,
in that they are not based on any specific geographic location such as the metropolitan of
Los Angeles or the city of Denver. Thus, the NPTS data are not specific enough for them to
make this conclusion. Third, the indicators (commuting time and distance) used in
Richardson and Gordons (1989) study cannot provide enough details to support either the
spatial mismatch hypothesis or traffic congestion studies, since commuting time also
includes delay time (congestion time). Longer commuting time does not necessarily result
in longer commuting distance. Moreover, travel distance also includes other alternative
routes, and therefore commuting distance does not equal the direct distance between home
and workplace; if the shortest route (from home to workplace) is congested, workers are
likely to choose other routes, which inevitably will increase travel distance. Therefore, the
census distance probably cannot represent the real commuting distance either.
27


3. Problematic Spatial mismatch hypothesis or problematic travel time measurement:
Based on the Chi-square test, Gordon and others (1989) found that neither minorities nor
low-income workers have longer commutes, and their commuting patterns are very similar
to those of workers in general (Gordon at al. 1989, 315). In fact, minorities are not
constrained by income or housing to live closer to work, and women consistently have
shorter work trips (ibid., 315). They then concluded that the spatial mismatch hypothesis
received no support based on commuting data (ibid.). Although this finding somewhat
shows that people tend to achieve the jobs-housing balance themselves automatically (even
though there is no jobs-housing balance policy), the indicators (commuting time and
distance) that they used from the NPTS might be unable to provide them enough details to
reach this conclusion. Since commuting time includes both real travel time and delay
(congestion) time, longer commuting time does not necessarily mean longer travel distance
from home to workplace. Longer commuting time might result from congestion, and the
real travel distance might be very short. In a similar debate of commuting time, DeRango
(2001) argued that:
the relationship between spatial mismatch and commuting times is indeterminate, if
employment probabilities decline as the distance from job site to residence
increases. Specifically, if employment probabilities decline faster (slower) than a
threshold rate, then spatial mismatch will decrease (increase) the commuting times
of central-city minorities. Thus, commuting-based tests of the spatial mismatch
hypothesis are not just biased but mis-specified because spatial mismatch is
theoretically consistent with both the null and alternative hypotheses. Evidence
that this concern is empirically important is taken from the contradictor findings
of recent studies that use commutes to test the spatial mismatch hypothesis. (P.
1521)
Regarding the other travel time discussion, in Sustainability and Cities, Newman and
Kenworthy (1999) concluded that whatever the infrastructures in all cities are, people take
about 30 minutes for the joumey-to-work. Levinson and Kumar (1994) found that the
average automobile work-to-home time was 32.5 minutes in both 1968 and 1988. They
then supported the hypothesis of jobs and housing mutually co-located to optimize travel
times (Levinson and Kumar 1994,319).
28


2.5.2 Mixing Land Use
Cervero (1996b) found that residential densities exerted a stronger influence on
commuting mode choices than levels of land-use mixture, except for walking and bicycle
commutes (P. 319), since:
having grocery stores and other consumer services within 300 feet of ones
residence is found to encourage commuting by mass transit, walking and bicycling, -
through controlling for such factors as residential densities and vehicle ownership
levels. When retail shops are beyond 300 feet yet within 1 mile of residences, this
range of distance tends to encourage auto-commuting, because of the ability to
efficiently link work and shop trips by car (Cervero 1996b, 361).
It is clear that distance together with mixed land use is a main determinant in affecting
mode choices, and we may expect that roadway congestion will be reduced through
encouraging non-single occupant vehicle usage. Moreover, two issues relating to mixing
land use are discussed as follows.
1. Is there a direct relationship between work-trip and mixed land uses coupled with
walking distance?
The relationship (if any) between the strategy of mixing land uses and walking distance is
determined by the type of trip. Take shopping trips for example. People might be willing
to walk to a neighborhood store for groceiy shopping, if the distance is within 300 feet or a
10-minute walking distance and the neighborhood is mixed with residential houses and
retail stores. On the other hand, work trips probably are less related to mixing land use and
walking distance, since people are likely to have diverse jobs, and the distance from home
to workplace could vary. Some people might live in the suburbs but work downtown or in
suburban cities. Others might live and work downtown. Therefore, to promote walking and
bicycling commutes for work-related trips, the mixing-land-use strategy probably will be
less effective. However, for a shopping trip, it will probably reduce traffic congestion
within a community.
29


2. What is mixing land use?
Grant (2002) used his Canadian experience based on mixed land-use communities for this
study and categorized five benefits for mixing land use, which also suggest several
strategies of mixing land use as follows:
(1) Providing a greater range of housing options for smaller, post-baby-boom households
(rather than just detached homes) to increase affordability and equity and to reduce the
premium;
(2) Providing a variety of land uses in a community (e.g. offices, housing, shops,
entertainment facilities, and so on) to enable residents to live near places where they
can shop, work, or play, and
(3) Providing housing near commercial and civic activities for residents to reduce the
dependence of cars, to make optimum uses of infrastructure which creates an active
urban environment at all hours, and to increase pedestrian, cycling, and transit use,
which would help to alleviate the environmental consequences associated with
automobile usage.
It is clear that mixing land use aims to enhance community vitality and environmental
quality, and to promote social equity. The reduction of roadway congestion is not the direct
consequence resulting from mixing land use. However, Frank and Pivo (1995) found that
transit usage and walking increase as density and land-use mix increase, whereas single-
occupant vehicle (SOV) usage declines (Frank and Pivo 1995, 44). They then stated that
increasing the level of land-use mix at the trip origins and destinations is related to a
reduction in SOV travel and an increase in transit and walking (ibid.). The increasing
usages of transit and walking will definitely result in less traffic congestion at the trip
origins and destinations (along with stations or activity nodes) due to modal shift. However,
mixing land use might not directly link with the traffic congestion in a metropolitan area. If
a metropolitan area has insufficient roadways and the design roadway capacity is unable to
absorb the traffic volume, the reduction of SOV resulting from mixing land use in a
community might not be able to reduce the congested roadways in a metropolitan area.
Researchers always assume that higher mixing land use could promote alternative
transportation modes. Then, by shifting automobile usage to other modes such as mass
30


transit, researchers expect that the roadway congestion could be reduced due to fewer
automobile trips. This assumption is likely to be true on a neighborhood or community
scale, but not at the metropolitan level. Take Asian cities for example. Many metropolitan
areas of Asian cities are dense and have mixed land uses. However, they are not free from
congestion. Roadways in most Asian cities are congested, and people are not willing to
change to other modes due to a higher usage of motorcycles for convenience. Therefore,
the potential of employing mixing land use strategy to reduce roadway congestion is limited,
since it is not applicable to all geographic scales (Frank 2000).
2.5.3 Density
There are numerous studies demonstrating that residential density is highly correlated with
travel behavior (Newman and Kenworthy 1989; Cervero 1989). Levinson and Kumar
(1997) concluded that a threshold of the residential density between 7,500 and 10,000
persons per square mile tends to have the shortest duration of automobile commutes
(Levinson and Kumar 1997, 147), and this density will result in less congestion on the road
networks based on a cross-sectional examination of many metropolitan statistical areas in
the United States. They also suggested that metropolitan residential density serves
principally as a surrogate for city size, and local residential density measures relative
location (accessibility) within the metropolitan region and indexes the level of congestion
(ibid.). Newman and Kenworthy (1989) found that density is highly correlated with fuel
consumption for both central city and peripheral regions, and that the higher the density,
the lower the gasoline consumption. And the appropriate city density regarding efficient
fuel consumption is about 30 persons per hectare or 7800 persons per square mile.
Newman and Kenworthy (1999) further found that density patterns are obviously closely
linked to transportation and energy usage, based on their study of international cities. They
stated that there appears to be a critical point (about 20 to 30 persons per ha) below which
automobile-dependent land use patterns appear to be an inherent characteristic of the city
(Newman and Kenworthy 1999, 100).
Although there were many studies showing that density is significantly related to
automobile ownership, VMT reductions, and less auto travel per capita (Dunphy and Fisher
31


1994), little effort has been made to identify the trade-offs between densification and other
environmental factors such as air pollution. In fact, some evidence shows that the
concentrations of air pollution along congested transportation corridors could result from
the intensification of land uses. Wendell Cox (1999) from the Heritage Foundation found
that higher concentrations of urban residential and employment density are likely to
produce higher concentrations of automobile traffic and air pollution. Asian cities for
example are suffering serious air pollution due to dense population. On the other hand,
other research has shown that higher density could promote transit usage, and therefore it
could reduce traffic and then reduce congestion. Whether densification will reduce
congestion and air pollution, it is important to understand that higher density might not
directly have an influence on travel behavior, which results in less traffic congestion. In
fact, a higher population density is able to support a higher transit usage (to cover the
operation cost), but a higher transit usage does not necessarily contribute to less congested
roadways, if the transit systems within a metropolitan area cannot capture enough
commuters or the transit systems do not have convenient networks to serve commuters
effectively. Therefore, encouraging people to shift their mode choices through education
probably is more important than densification per se.
2.5.4 Air Pollution
Air pollution comes from many different sources: (1) stationary sources such as factories,
power plants, and smelters, and smaller sources (e.g. dry cleaners and degreasing
operations); (2) mobile sources such as cars, buses, planes, trucks, and trains; and (3)
naturally occurring sources such as windblown dust, and volcanic eruptions. However,
mobile sources seem to be a primary contributor to air quality problems, particularly in
metropolitan areas, according to the National Air Quality and Emissions Trends Report of
1991. The major components of smog from cars are Carbon monoxide (CO), Nitrogen
oxides (NOx), Ozone, Volatile organic compounds (VOCs), Hazardous air pollutants
(HAPs), Choloro-fluoro-carbons (CFCs), and Particulate matter (PM-10). Although these
emissions from an individual car are generally low (relative to the smokestack pollution),
the accumulative effect of all individual cars is the single largest contributor to ground-level
ozone. Thus, it seems that the prescription to reduce air pollution (mostly caused by
32


automobile usage) is more associated with transportation planning; the remedy based on
land use and transportation interaction may be more effective than land use planning per se.
However, Bae (1993) stated the following:
the measures aimed at reducing vehicle miles traveled have only a modest
impact on reducing air pollution, and technological solutions to the automobile
emissions problem are much more important. Thus, more transit use, ridesharing,
and telecommuting are not needed to achieve clean air objectives (Bae 1993,
65).
Moreover, Bae (2001) further stated that land use options will not work, since the
settlement pattern can only be changed at the margin, and land use changes take a very long
time to implement in any event, and even if implemented, they would have negligible
effects on vehicle miles traveled (VMT) (Bae 2001, 2).
On the other hand, Newman and Kenworthy (1999) used a German project to demonstrate
the changes in the different types of driving (aggressive and calm) by comparing the
emissions with a vehicles speed reduction from fifty kilometers per hour to thirty
kilometers per hour (Newman and Kenworthy 1999, 150). They found that in both
aggressive and calm driving, emissions are reduced at the thirty kilometers per hour; the
calm driving has a generally greater reduction and fuel use is lower (ibid.). Similarly,
according to the EPAs Fact Sheet, it also shows that higher car speeds tend to produce
more emissions. Moreover, Cox (1999) found that higher levels of air pollution are
associated with higher densities, not lower densities; the higher the population density, the
greater the intensity of air pollution. He further stated that as transit-oriented development
increases traffic, it could reduce speeds and increase pollution, because higher pollution is
associated with slower, and more congested traffic. Up to now, there is still a great debate
on the relationship between air pollution, car usages, and population density. Whether the
strategies based on land use and transportation interaction will help to reduce air pollution,
researchers are continuously looking for these relationships.
33


2.5.5 The Constraints
Crane (2000) stated that much of the research on urban form and travel were more
exploratory than definitive; the literature has made substantial progress in identifying the
key questions at hand and how to ask them (Crane 2000, 4). Handy (1992) further pointed
out that the problem of most research studying the linkage between urban form and travel
behavior relied on relatively coarse data on both urban form and travel patterns. For
example, Gordon and others (1989) used the Nationwide Personal Transportation Survey
(NPTS) and found that the jobs-housing balance has little effect on the fastest growing
travel region, since the joumey-to-work trip is not the major contributor during peak hours
(Gordon et al. 1989, 315). In fact, the NPTS data are not specific enough to make this
conclusion, since it only provides the data by residency and there is no detailed census by
workplaces. Therefore, there are two basic issues in association with the studies on land-
use and transportation interaction. First is how to choose the appropriate and reliable data
for the analyses; and second is how to select suitable analytical methods (e.g. regression) to
examine the research hypotheses. These two issues seem to determine whether a study will
succeed or not. In my research, I have used the Census of Transportation Planning
Packages (CTPP) for analyzing the relationship between land use, transportation, and urban
form, since the packages provide the census data by area of residence, by area of work, and
by area of residence by area of work. Moreover, to choose the appropriate analytical
approaches, I have chosen cluster and factor analysis to facilitate explaining the regressions
results. The details are discussed in Chapter 3.
Since this dissertation is not intended to find a comprehensive remedy to solve the traffic
congestion completely, I focus only on work related trips at AM-peak hours. Based on the
1990s and the 1995s NPTSs (see Figure 2.2 and 2.3), 50% of the AM-peak-hour traffic
came from work trips (earning a living) in 1990, and 55% in 1995. Thus, it is clear that
work related trips are the major contributor to AM-peak-hour congestion. On the other
hand, other types of trips (e.g. family and personal business, church and school, social and
recreational, and other purposes) all together contribute to PM-peak hours traffic congestion,
but none of these trips alone is able to contribute more than 50% of trips to the congestion.
34


% of Vehicle Trips Taken During Weekdays by Time of Day and Trip Purpose (by Start Time) --1990


^^EARNINGA LIVING ""FAMILY AND PERSONAL BUSINESS CHURCH AND SCHOOL SOCIAL AND RECREATIONAL OTHER
\




1:00 AM-5:59 AM 6:00 AM-8:59 AM 9:00 AM-12:59 PM 1:00 PM-3:59 PM Time 4:00 PM-6:69 PM 7:00 PM-9:59 PM 10:00 PM-1:00 AM
Figure 2.2 Number of Vehicle Trips during Weekdays by Trip Purpose in 1990
Table 2.2 Number of Vehicle Trips during Weekdays by Trip Purpose in 1990
NUMBER OF VEHICLE TRIPS TAKEN DURING WEEKDAYS 1/ BY TIME OF DAY AND TRIP PURPOSE 1990 NPTS (Millions)
OCTOBER 1997 % of start time TABLE NPTS-9
EARNING A LIVING FAMILY AND PERSONAL BUSINESS CHURCH ANC SCHOOL SOCIAL AND RECREATIONAL OTHER TOTAL
1:00 AM-5:59 AM 67.46% 13.28% 1.03% 17.84% 0.40/ 100%
6:00 AM-8:59 AM 50.06% 18.75% 26.59% 4.05% 0.53/ 100%
9:00 AM-12:59 PM 14.86% 60.88% 6.18% 17.16% 0.92/ 100%
1:00 PM-3:59PM 17.83% 48.71% 17.87% 14.80% 0.77 A 100%
4:00 PM-6:59 PM 30.75% 40.79% 5.58% 22.15% 0.73 % 100%
7:00 PM-9:59 PM 14.50% 37.56% 8.41% 38.70% 0.83/ 100%
10:00 PM-1:00 AM 29.90% 24.65% 3.98% 40.65% 0.8491 100%
TOTAL 21 32.50% 49.80% 3.40% 14.20% 0.10 1/ Weekdays are defined as the time from 12:01 AM on Monday until 6:00 PM on Friday. 2/ Includes trips where time of day, trip purpose or both were unreported.
Source: The 1990 Nationwide Personal Transportation Survey, U.S. DOT.
Note: 6:00 AM 8:59 AM is morning peak hour and 4:00 PM 6:59 PM is PM peak hour.
35


% of Vehicle Trips Taken During Weekdays by Time of Day and Trip Purpose
(by Start Time) -1995
Time
Figure 2.3 Number of Vehicle Trips during Weekdays by Trip Purpose in 1995
Table 2.3 Number of Vehicle Trips during Weekdays by Trip Purpose in 1995
NUMBER OF VEHICLE TRIPS TAKEN DURING WEEKDAYS 1/ BY TIME OF DAY AND TRIP PURPOSE 1995 NPTS (THOUSANDS)
OCTOBER 1997 % of start time TABLE NPTS-9
EARNING A LIVING FAMILY AND PERSONAL BUSINESS CHURCH AND SCHOOL SOCIAL AND RECREATIONAL OTHER TOTAL
12:00 AM-fi:69 AM 66.18% 17.46% 0.50% 15.80% 0.06% 100%
6:00 AM-8:69 AM 55.38% 34.08% 4.84% 5.64% 0.06% 100%
9:00 AM-12:G9PM 21.26% 62.71% 2.95% 12.96% 0.09% 100%
1:00 PM-3:59 PM 25.01% 60.23% 2.80% 11.85% 0.09% 100%
4:00 PM-6:59 PM 35.17% 45.60% 2.67% 16.44% 0.10% 100%
7:00 PM-9:59 PM 19.42% 45.35% 5.47% 29.57% 0.17% 100%
10:00 PM-11:59 PM 34.38% 29.06% 4.41% 32.07% 0.07% 100%
TOTAL 21 32.50% 49.80% 3.40% 14.20% 0.10% 100%
1/ Weekdays are defined as the time from 12:01 AM on Monday until 6:00 PM on Friday. 2/ Includes trips where time of day, trip purpose or both were unreported.
Source: The 1995 Nationwide Personal Transportation Survey, U.S. DOT.
Note: 6:00 AM 8:59 AM is morning peak hour and 4:00 PM 6:59 PM is PM peak hour.
36


3. Methodology for Examining Traffic Congestion based on
Land-use and Transportation Interaction
Crane (2000) found that a common approach to study the interaction between urban form
and travel is to regress commute length on a measure of residential density and the
demographic characteristics of travelers and then examine the significance, sign, and
magnitude of the estimated coefficient on density (Frank and Pivo 1995; Levinson and
Lumar 1997; Crane 2000, 5). He also found that much recent literature employed land-use
factors (e.g. population density, employment location, mixed land uses) to correlate with
outcome measures (e.g. vehicle miles traveled, car ownership rate, and mode choice) to
understand the interaction between different fields (e.g. land use, travel behavior, urban
form, and so on) (Crane 2000, 4). Clearly, planners and researchers have attempted to
explore how travel behavior might be affected by manipulating urban form; they are now
more interested in using the interaction between land use and travel behavior to solve
transportation issues (ibid.). In this research, attributes of areal shape (urban spatial index),
land-use factors (jobs-housing ratio, density, median income), and the transportation system
(car ownership, transit ridership, roadway supply) are applied to explain metropolitan traffic
congestion by using several regression analyses. This chapter will start with the discussion
on the methodology, data collection, the selected study areas, and the selected variables for
analyzing. Then, I will try to show the difference of my approach with previous studies.
3.1 Methodology
Researchers (e.g. Cervero, Gordon, and Richardson) have provided many useful
demonstrations to study the interaction between land use and transportation. A most
common assumption that previous researchers have made is that the degree of traffic
congestion is strongly positively correlated with VMT. Unfortunately, VMT alone could
not index roadway congestion, unless roadway capacity is included. Therefore, their results
were inevitably debatable, especially when their studies were related to traffic congestion.
37


This dissertation starts from understanding the possible methods to measure traffic
congestion through reviewing literature, and then the attributes (land use, transportation,
and urban form) of metropolitan form relating to congestion are selected to explain their
underlying relationships. Moreover, I intend not to employ any complicated model to avoid
making unnecessary assumptions to keep the research approach as simple as possible, since
the more complicated the model used, the more assumptions have to be made, and it will
result in an unrealistic study. The analytical approach is demonstrated by a diagram (see
Figure 3.1).
Instead of using urbanized area (UA) for this research, I chose metropolitan statistical area
(MSA), since an MSA is defined by both population density and commuting patterns. On
the other hand, a UA is only based on population density. Moreover, this research only
takes commutes within an MSA into account, since the CTPP has shown that those non-
inner-MSA commutes are very complicated and therefore the commutes from one
metropolitan area to another area or to another country are not considered. Take working
overseas for example. Based on the 1990s census in the CTPP, those people who flew to
other nations for work only counted the commuting time from home to the airport and they
did not include the airtime, when they filled out the questionnaires for the 1990s Census
Survey. Therefore, the numbers (inaccurate travel time) of the non-inner-MSA commutes
are not considered. Also, median travel time ~ the midpoint (central value) of the sample
array is used for calculating travel speed (mean median speed) to avoid adopting any
inappropriate commuting time (too small or too big).
38


Figure 3.1 Research Approach Diagram
39


3.1.1 Method
To understand the congestion measurement the dependent variable in this research, 1 first
referred to the Roadway Congestion Index produced by Texas Transportation Institute
(TTI), and then I consulted the theory of the Level of Services (LOS) developed by the
Institute of Transportation Engineers. After reviewing both methods, based on the
assumption that the faster the speed a vehicle could travel, the less congested the
roadways according to the LOSs standard, mean median travel speed (the detailed
discussion of this variable is in section 3.3) is selected as an indicator to gauge congestion
due to data availability. Moreover, mean median travel speed used in this research is not
equal to real travel speed. It is the average of the shortest route distance used for work
based on least congested highway networks divided by the congested median travel
time. It is similar to the reverse concept of Existing-congested-volume divided by
Designed-uncongested-Capacity (V/C). Thus, the higher the mean median speed, the less
congested the highway networks.
There are two major tools employed in this dissertation, which are statistics and Geographic
Information Systems (GIS). Statistics is used to explain and define the underlying
relationship between land use, transportation, and urban form, and the STAT1STICA
package is used for analyzing the data statistically. Several geographic information systems
(GIS) such as ArcView, AutoCAD Map 2000, TransCAD, and TransVu, are used for data
management and extraction, and shortest-path calculation and selection. TransVu is
employed to extract census data from the 1990s Census of Transportation Planning
Packages. ArcView is for matching the selected census data with the geographic digital
entities. TransCAD is for calculating the shortest distance of the route from home to
workplace, and finally, AutoCAD Map 2000 is for obtaining the urban form index. The
detailed discussion is shown in Table 3.1 and 3.2.
In general, descriptive statistics, cluster analysis, factor analysis, and correlation analysis
are used to characterize the selected MSAs and to facilitate explaining the results of the
regression analyses. Two regressions are taken in this research. One is the fixed nonlinear
regression and the other is the multiple regression. Frank (1995) found that the relationship
40


between population and employment density and mode choice for Single-Occupant Vehicle
(SOV), transit, and walking is nonlinear for both work and shopping trips (Frank 1995, 44).
Therefore, the fixed nonlinear regression is used to make sure that the nonlinear model is
covered in the analyses. To define the relationship between explanatory variable and the
dependent variable (mean median speed), the fixed nonlinear regression and the forward
stepwise regression are employed. The statistical significance for defining the relationship
is determined by comparing the adjusted i?-square and the /7-value. Then, those explanatory
variables with most statistical significance based on the fixed nonlinear regression are
transformed to the linear model by employing either logarithm or reciprocal or square root,
so that they can be used for the multiple regression (Linear model). Finally, the forward
stepwise regression is taken to determine not only the relationship between the dependent
variable and explanatory variables, but also the statistically significant sequence/order of
explanatory variables. Through setting F value (one of the options for running the forward
stepwise regression) to minimum (0.0001), I was able to force almost all explanatory
variables into the equation to understand their initial order, although there is a multi-
collinearity by doing this. However, after understanding the statistically significant
sequence of explanatory variables, I only chose those explanatory variables with higher
statistical significance for the final stepwise multiple regression, including one variable
from each SIC/SOC group categorized by factor analysis to avoid the multi-collinearity.
The results of the forward stepwise regression (multiple regression) are shown in Chapter 4.
Cervero (1989) regress the ratio of residence jobs over residence workers based on the
Standard Occupational Classification to the LOS for gauging the roadway congestion.
He assumed that whenever the ratio is greater or less than 1, it implied more traffic on the
roadways. However, the jobs-housing ratio cannot show where the traffic goes if the ratio
is less than 1, and the ratio cannot identify where the traffic comes from if the ratio is
more than 1Some traffic might either come from or go to other metropolitan areas/states
(non-inner-MSA commutes). It is clear that the measurement of the jobs-housing ratio is
not precise enough to simulate the traffic flow. In this research, I have used the coefficient
of variation to aggregate the jobs-housing ratio based on census places into a
metropolitan area. Since the coefficient of variation is aimed to measure the dispersion of
a sample, a higher coefficient of variation of the jobs-housing ratio means more traffic
traveling from homes to workplaces on the roadways in an MSA. Moreover, the coefficient
41


of variation is also used to calculate the urban form index to describe metropolitan form.
The detailed discussion is in section 3.3.
Table 3.1 The Applications of Statistical Analyses
Statistics Purpose Specific Method
Descriptive Statistics 1. To identify the characteristics of a metropolitan area. 2. To test statistical significance. 3. Use to calculate the Coefficient of Variation for the Jobs-Housing ratio and the Urban form index. Mean, Median, Standard Deviation, Coefficient of Variation
Cluster Analysis 1. To identify the economic characteristics for a metropolitan area based on the SIC and the SOC. 2. To facilitate explaining the regression results. Complete Linkage (based on Tree-clustering and Euclidean distance)
Factor Analysis 1. To minimize the multi-collinearity problem. 2. To categorize the SIC and the SOC into fewer groups. 3. To facilitate explaining the regression results. Principal components (based on Varimax normalized factor rotation, Eigenvalues, Scree plot)
Correlation Analysis 1. To determine initial relationship between explanatory variables and the dependent variable. 2. To facilitate explaining the regression results. Two lists partial correlation (correlation coefficient and p- levels)
Nonlinear Regression and 2D Scatterplots 1. To find the relationship between an explanatory variable and a dependent variable. 2. To obtain a regression equation for further analysis. Piecewise linear regression, Fixed nonlinear regression (based on X2, X3, X4, X5, X5, SQRT(X), LN(X), LOG(X), ex, 10x, 1/X.
Multiple Regression 1. To find the relationship between explanatory variables and the dependent variable. 2. To determine the sequence of statistical significance between the selected explanatory variables. Stepwise regression (based on partial correlation, F-Test, r-Test,/?-value (p-Level), R- square, Adjusted F-square)
Note: All of the above statistical analyses are proceeded by using STATISTICA.
42


Table 3.2 The Applications of Geographic Information Systems
G.I.S. Programs Purposes
TransCAD 1. To select the shortest route from one census place to the other census place based on the highway networks within an MSA. 2. To obtain the distance of shortest route from one census place to the other census place within an MSA. 3. To obtain an Origins and a Destinations IDs for census data extraction. 4. To obtain the distance directly linking from one census place to the other census place within an MSA.
ArcView 1. To obtain the total distance of the selected highway networks for calculating the macro-scale roadway density. 2. To match geographic entities with census data. 3. To obtain area and perimeter for an MSA. 4. To facilitate proceeding TransCADs analysis.
AutoCAD MAP 2000 1. To establish the 8-reference lines for an MSA. 2. To obtain the distances of the 8-fixed-reference lines, and then use them for calculating the coefficient of variation (urban form index) based on these 8-reference lines.
Trans VU 1. To extract data from the Census of Transportation Planning Package.
Note: *.DBF (database file format) is the connective format to link different GIS programs.
3.1.2 Data Collection
Instead of using the NPTSs data, this research employs the data in the 1990s Census of
Transportation Planning Package (CTPP) statewide for defining the statistical relationships,
since it provides more details than the NPTS. Three types of data are provided in the CTPP;
they are census and transportation data by area of residence, census and transportation
data by area of work, and transportation data of area of residence by area of work.
Besides, the CTPP also provides both origin and destination commuting data such as travel
time, aggregate number of vehicles, and workers per carpool. Other census data such as the
1990s metropolitan population and area are found at the Census Bureau web site and are
used for demographic analysis. Generally, two types of data are collected. One is census
data (numbers), such as population, income, employee by occupations (SOC) and
industrial sectors (SIC). The other is geographic digital entities, such as highway
43


networks in an MSA, an MSAs boundary, and census-designated places in an MSA
(centroid format).
There is an inconvenience of using the CTPPs data, and that is the lack of connection
between its census data and geographic entities in several States. Fortunately, the data and
the geographic entities of these States can still be connected through the census place
codes and the geographic IDs by using ArcView and TransCAD. Also, some CD-ROMs
of the CTPP were problematic. Sometimes, census data could not be extracted by using
TransVU, which was provided by the Census Bureau for data extraction; some census data
are incorrect due to origin data input error. Although I contacted several members of the
Census Bureau technical support staff in order to solve the problems, they could not resolve
the data input error. Thus, originally, there were 47 MSAs selected for this dissertation, but
the metropolitan area of St. Louis, MO-IL was dropped due to inaccurate data. I only
include 46 MSAs for my analyses.
3.1.3 Assumptions and Hypotheses
Two assumptions and several hypotheses are made in this research. The first assumption is
that people tend to choose the shortest route to travel from one census place to the other
census place within an MSA, if there is no traffic congestion on the highway networks. The
second assumption is that the faster the speed a vehicle could travel, the less the highway
congestion.
The hypotheses are as follows, and Table 3.3 shows the assumed relationships between the
dependent variable and explanatory variables.
HI. Higher coefficient of variation of the jobs-housing ratio implies higher vehicle trips,
and therefore it might result in serious highway congestion. However, this relationship
might also vary, if the jobs-housing ratio is based on different occupations or industrial
sectors;
H2. The more dispersed the metropolitan form, the more congested the highway
networks, since a higher irregular urban form probably relates to inaccessibility and
people have fewer alternative routes to choose;
44


H3. An MSA with higher population density is likely to suffer serious highway congestion,
since the highway might be unable to absorb the demanded volume;
H4. An MSA with lower median household income is likely to have less congested
highways, since people with lower income are les likely to own a vehicle for traveling
and they tend to take public transit;
H5. An MSA with higher car ownership might also relate to higher traffic congestion, since
people having more cars may tend to drive;
H6. An MSA with higher transit ridership is likely to have less congested highways, since
commuters are willing to take public transit but not drive; and
H7. An MSA with inadequate roadway supply is likely to have even more serious roadway
congestion.
Table 3.3 The Relationships between Explanatory Variables and the Dependent
Variable
Depen- dent Variable =/ Independent Variables
Land Use Factors Transportation Factors Urban Form Factor
Mean Median Speed The COV of the Jobs- Housing ratio based on the SIC/SOC Metro- politan Popu- lation /Employ -ment density Median Income based on an MSA Car Owner -ship based on an MSA Transit Rider- ship based on an MSA Macro- Scale Roadway Density based on an MSA The COV of the Distance of the 8- reference lines for each MSA
Relationship N/A - - + - + + -
Hypothesis N/A HI H3 H4 H5 H6 H7 H2
Note: 1.
+ means positive relationship, and means negative rei
ationship.
2. COV is the coefficient of variation.
To reduce work related traffic congestion at AM-peak hours, the main difference between
my research and any previous study is that I include the concept of roadway supply and
urban form as part of the explanatory variables, and only focus on the highway networks.
In many MSAs, during the peak hours, their local arterial networks are often free from
congestion, since there are many alternative routes that drivers could choose. On the other
hand, their highway networks are always congested since there are fewer alternative routes
to be chosen by the drivers, especially when the drivers are regional commuters. Because
45


of this reason, I only choose the highway networks in this research. Moreover, I include the
concept of urban spatial distribution pattern and establish an assumption that the more
dispersed the metropolitan form, the more seriously congested the highway networks.
3.2 Selected Metropolitan Statistical Areas
In order to make sure that the selected metropolitan areas have the following characteristics,
a congested highway network with and without transit systems, and a non-congested
highway network with and without transit systems, the TTIs roadway congestion index
(RCI) of 1990 (there are 63 urbanized areas selected in the 1990s Urban Mobility Study)
and the 1988s transit system condition in Transportation Planning Handbook (Edward,
1992, 58) were used for selecting MSAs. Although the RCI is based on UA, I have
transformed the selected UAs into MSAs after the initial UAs were chosen. Four criteria
are used for selecting MSAs as follows.
a) Those urbanized areas have the RCI larger than or equal to 1. The RCI >= 1 means that
the roadway is congested. There are 19 MSAs selected.
b) Those urbanized areas have the RCI less than 1 (less congested), and have transit systems
in place. There are 11 MSAs selected.
c) Those urbanized areas have the RCI less than 1 and without transit systems. I randomly
choose 10 MSAs from this category.
d) In order to substitute some MSAs having input data error, there are 7 MSAs randomly
selected.
The selected 47 MSAs are shown on the next page, and the metropolitan area of St. Louis,
MO-IL was dropped due to inaccurate data input.
46


I
'-J
Eugene*
San Francisco, C. __
San Francisco, Cfl
San Jose, C!
FresnoVf
Bakersfield,^
Los Angeles-Long Beach,
Los Angeles-Long Beach, (
Los Angeles-Long Beach, C/^
Honolulu,HI
ston, MA PMSA
*NewYork, NY CMSA
i York, NJ CMSA
\Hous*on-BjaJfla-Galveston,
Tampa-St Petersburg-Clearwater,
Brownsvllle-rfarllngton, TX
Mlaml-f-Msuderdale, FL MSA
Figure 3.2 The Selected 47 Metropolitan Statistical Areas


Table 3.4 The Selected 47 MSAs
MSA MSA
1 Los Angeles-Long Beach, CA MSA Columbus, OH MSA
2 San Francisco, CA MSA Cleveland, OH MSA
3 San Jose, CA PMSA PortlandVancouver, ORWA CMSA
4 BoulderLongmont, CO CMSA Pittsburgh, PA PMSA
5 Denver, CO PMSA Philadelphia, PA-NJ PMSA
6 Colorado Springs, CO MSA San Antonio, TX MSA
7 HartfordNew BritainMiddletown, CT CMSA SeattleTacoma, WA CMSA
8 Washington, DCMDVA CMSA Oakland, CA MSA
9 Jacksonville, FL MSA Phoenix, AZ MSA
10 MiamiFort Lauderdale, FL CMSA Tucson, AZ MSA
11 Orlando, FL MSA Bakersfield, CA MSA
12 Atlanta, GA MSA Fresno, CA MSA
13 Honolulu, HI MSA Sacramento, CA MSA
14 ChicagoGary, 1LIN CMSA ** San Diego, CA MSA
15 New Orleans, LA MSA TampaSt. PetersburgClearwater, FL CMSA
16 Boston, MA PMSA Rochester, NY MSA
17 Baltimore, MD MSA Oklahoma City, OK MSA
18 Detroit-Ann Arbor, MI CMSA Eugene-Springfield, OR MSA
19 Omaha, NEIA MSA Austin, TX MSA
20 Albuquerque, NM MSA BrownsvilleHarlingen, TX MSA
21 Las Vegas, NV MSA DallasFort Worth, TX CMSA
22 BuffaloNiagara Falls, NY CMSA HoustonGalvestonBrazoria, TX CMSA
23 New York-Northern New Jersey-Long Island, NY-NJ CMSA Milwaukee, WI PMSA
24 St. Louis, MO--IL CMSA
3.3 Selected Variables for Analyses and Their Definitions
Crane (2000) stated that land use factors such as population density, employment location,
and mixing land uses in the neighborhood and region, are associated with the outcome
measures that include VMT, car ownership rates, and mode choice. In this dissertation,
instead of using VMT, car ownership rates, or modal choices as the dependent variable, the
traffic congestion index (mean median travel speed) is used for regressing it to other
explanatory variables, which are land use factors, transportation factors, and urban form
48


factors. Here, land use factors are the coefficient of variation of the jobs-housing ratio
based on the SIC/SOC, population/employment density, and median household income.
Transportation factors are car ownership, transit ridership, and macro-scale roadway density,
and urban form factor is the coefficient of variation of the distances based on the 8-fixed-
reference lines in each MSA.
Regarding mean median travel speed, the first step is to obtain median travel speed; it
is calculated by using the shortest distance from home to work divided by the median
travel time at AM-peak hours (6:30 a.m. to 8:29 a.m.) on the selected shortest route (O-D
pair) within the MSA. Then, the mean median travel speed is the average of the
overall median travel speeds based on all the selected shortest routes used for work at AM-
peak hours. In other words, mean median travel speed is the distance of the shortest
route used for work based on uncongested highway networks divided by the congested
travel time. This is similar to the reverse concept of Existing-Congested Volume divided
by Designed-Uncongested Capacity (V/C) for gauging traffic congestion. Therefore, a
higher mean median speed means that the highway networks are less congested, and a lower
mean median speed means that the highway networks are seriously congested.
Regarding the coefficient of variation of the jobs-housing ratio, the coefficient of
variation is used to aggregate the jobs-housing ratio (residence jobs over residence worker)
based on a census place to a metropolitan area. A higher jobs-housing ratio (>1) means that
the number of residence jobs is larger than the number of residence workers, and therefore
more workers living in other census places travel to a census place for work. On the
contrary, a lower jobs-housing ratio (<1) means that the number of residence jobs is less
than the residence workers, and therefore people living in a census place have to go to other
census places for work. In general, both jobs-housing ratios (>1 and <1) mean more
travel. In this dissertation, the coefficient of variation is used to measure the
dispersion/change of the ratios, and a higher coefficient implies serious roadway congestion
(low mean median travel speed). The definitions of explanatory variables and the
dependent variable are shown as follows.
49


Table 3.5 The Definition of the Dependent Variable and Explanatory Variables
Variables Expression Definitions
Mean Median Travel Speed (mph) MnMdnSpd Si" (Shortest Distance from home to work place divided by Median Travel Time at AM-peak hours) divided by Total number of the selected shortest routes used for work at AM-peak hours, where 1. i represents each shortest route from home to work place within each MSA at AM-peak hours (6:30a.m. to 8:29 a.m.). 2. The shortest distance of each route based on highway networks is obtained from TransCAD. 3. Median travel time is direct from the CTPP.
Coefficient of variation of the Jobs-Housing ratio (%) B02 0101-0115 B03 0101-0119 Coefficient of variation of the (Residence jobs divided by Residence Workers) by SOC/SIC for each MSA
Macro-scale Roadway Density (Car/mile) Rden (Aggregate numbers of vehicles used in travel to work at AM-peak hours within each MSA) divided by (Total distance of highway networks within each MSA)
Median Income (dollars) A15_01 Median Household Income in 1989 by area of residence based on an MSA
Population Density (Population/mile2) PopDen Total Population by area of residence based on an MSA divided by an MSAs area
Employment density (Worker/mile2) WrkDen Number of workers by area of residence based on an MSA divided by an MSAs area
Car Ownership (Car per capita) COShip (Aggregate vehicles available by area of residence within an MSA) divided by (an MSAs Population)
Workers Car Ownership (Car per worker) COshipW (Aggregate # of vehicles used in travel to work at AM- peak hours by area of residence within each MSA) divided by (Total Workers within each MSA)
Transit Ridership / Transit Usage (% Workers by transit/Total workers) (Transit includes bus/trolley, streetcar/trolley, subway/elevated, and railroad) TRShip (Means of transportation to work by area of residence within each MSA Streetcar/Trolley, Subway/Elevated, and Railroad) divided by (Total Workers within each MSA)
Coefficient of Variation for Urban Form (%) UFCOV Coefficient of Variation of the distances based on the 8- fixed-reference lines within each MSA
Vehicle Miles of Travel (Car-miles) VMT (Aggregate Number of Vehicle Occurred on Shortest Path) times (Distance of the Selected Highway Used for Work)
Note: The above variables are based on an MSA scale.
50


Regarding the coefficient of variation based on the distances of the 8-fixed-reference lines
for each MSA, these lines are first placed to each MSA in the same configuration (from
center to the edges of east, west, north, south, northeast, northwest, southeast, and
southwest) to obtain the distance. Then, the coefficient of variation is calculated based on
the distances of these eight lines from center to the MSAs edge following the 8 different
directions. A lower coefficient (urban form index) means that MSA has uniform
metropolitan shape, and it implies higher accessibility (shorter travel distance) and more
choices of alternative routes for connecting to other places, which therefore result in less
congested highways. A higher coefficient of variation (urban form index) means that MSA
has a very irregular metropolitan shape, and it implies inaccessibility (longer travel distance)
and an MSA has fewer alternative routes to be chosen to connect to other places. Therefore
fewer alternative routes result in serious highway congestion. In the following, I use
three MSAs (Tucson in AZ, Colorado Springs in CO, and New York-Northern New
Jersey-Long Island in NY and NJ), which have very different urban shape, to demonstrate
the concept of the urban form index.
Table 3.6 Three Examples of the Urban Form Index Application
MSA Tucson, AZ Colorado Springs, CO New York-Northern New JerseyLong Island, NY--NJ
Area (IVf) 63,396,409,257.84 13,938,835,293.34 253270000000
Perimeter (M) 1,203,665.34 497,728.19 3488483.6376
Standard Deviation 61441.51 7,948.33 165553.62
Mean 135182.56 63,997.35 313471.67
Coefficient of Variation 0.45 0.12 0.53
Shape Trapezoid Rectangle near Square Y-Shape Island
Line 1 (M) 83,115.23 56,291.57 564108.5341
Line 2 (M) 129,611.20 79,364.94 589146.4664
Line 3 (M) 185,623.01 58,987.69 225821.3656
Line 4 (M) 111,845.17 56,173.39 201522.5083
Line 5 (M) 92,875.14 67,384.16 228419.0889
Line 6 (M) 100,705.82 58,987.69 287077.5855
Line 7 (M) 265,839.72 66,791.56 176469.1443
Line 8 (M) 111,845.17 67,997.80 235208.6491
Note: The shape of each MSA is shown in Figure 3.3, 3.4, and 3.5.
51


Figure 3.4 The Shape of Colorado Springs, CO MSA
52


53


3.4 Summary
In sum, this research focusing on AM-peak-hour work trip is based on the comparison of a
cross-sectional examination through 46 selected MSAs by using the CTPPs (statewide).
There are two main methods used for analyses, which are statistics and GIS. GIS is used
mainly for shortest route selection, distance calculation, and data management. Statistics
such as descriptive statistics, factor analysis, cluster analysis, correlation analysis, and
regression are employed to explain the underlying relationship between traffic congestion
with three types of explanatory factors to alleviate work related traffic congestion at AM-
peak hours. Moreover, the whole research can be expressed by a mathematical equation:
Traffic Congestion Index /(Land Use Factors) //(Transportation Factors) +/(Urban Form Factors)
that is: Traffic congestion index (mean median speed) is the function of Land-use factors
(Coefficient of variation of the jobs-housing ratios based on the SIC/SOC,
Population/Employment density, Median income), Transportation factors (Car ownership,
Transit ridership, Total roadway capacity), and Urban form factor (Coefficient of variation
of the distance of the 8-fixed-reference lines).
The assumed relationships between dependent variable with explanatory variables are as
follows (same as Table 3.3).
Table 3.7 The Relationships between Explanatory Variables and the Dependent
Variable (same as Table 3.3)
Depen- dent Variable =/ Independent Variables
Land Use Factors Transportation Factors Urban Form Factor
Mean Median Speed The COV of the Jobs- Housing ratio based on the SIC/SOC Metro- politan Popu- lation /Employ -ment density Median Income based on an MSA Car Owner -ship based on an MSA Transit Rider- ship based on an MSA Macro- Scale Roadway Density based on an MSA The COV of the Distance of the 8- reference lines for each MSA
Relationship N/A - - + - + + -
Hypothesis N/A HI H3 H4 H5 H6 H7 H2
Note: 1. + means positive relationship, and means negative relationship.
2. COV is the coefficient of variation.
54


Where the hypotheses are as follows.
HI. Higher coefficient of variation of the jobs-housing ratio implies higher vehicle trips,
and therefore it might result in serious highway congestion. However, this relationship
might also vary, if the jobs-housing ratio is based on different occupations or industrial
sectors;
H2. The more dispersed the metropolitan form, the more seriously congested the highway
networks, since a higher irregular urban form probably relates to inaccessibility and
people have no alternative route to choose;
H3. An MSA with higher population density is likely to suffer serious highway congestion,
since the highway might be unable to absorb the demanded volume;
H4. An MSA with lower median household income is likely to have less congested
highways, since people with lower income are less likely to own a vehicle for traveling
and they tend to take public transit;
H5. An MSA with higher car ownership might also relate to higher traffic congestion, since
people having more cars may tend to drive;
H6. An MSA with higher transit ridership is likely to have less congested highways, since
commuters are willing to take public transit but not drive; and
H7. An MSA with inadequate roadway supply is likely to have even more serious roadway
congestion.
55


4. The Examination of Forty-Six U.S. Metropolitan Areas
According to the 1999 Transportation Statistics Annual Report, the average commute speed
for work-related trips changed from 29 miles per hour (mph) in 1983 to 34 mph in 1995
(20% increase) (Schmitt 1999, 5). On the other hand, the average travel speed for work-
related trips in the St. Louis Metropolitan area was only 7.3 miles per hour in 1990, which
was calculated based on the 1990 CTPP. Since this speed is much lower than the speed in
the annual report, there must certainly be a mistake in the Missouri State CTPP. Originally,
47 metropolitan statistical areas were selected for examination. Then, the St. Louis, MO-
IL MSA was dropped due to inaccurate data input after the mean median speed was
calculated. There are only 46 MSAs used for the following analyses. In this Chapter,
characteristics of selected metropolitan areas, the results of cluster and factor analyses, the
results of correlation analysis, and the results of the fixed nonlinear and the stepwise
multiple regressions are discussed as follows.
4.1 Characteristics of the Selected Metropolitan Areas
The characteristics of an MSA could be categorized in many ways. In this dissertation, a
congestion index (mean median speed), population and employment density, economic
condition, income, car ownership, transit ridership, and an urban form index are used to
characterize the 46 selected MSAs. The details are below.
56


4.1.1 The Congestion in 1990
Unlike the Roadway Congestion Index (RCI) by the TTI, which documents the congestion
condition for both AM and PM peak hours for all types of trip purposes through measuring
VMT for both freeways and principal arterial networks, this research focuses only on work
related commutes at AM-peak hours and is only based on highway networks by assuming
that the higher the mean median speed, the less congested the highway networks.
Moreover, the mean median speed, calculated in this research, is just a congestion index and
it does not represent the real travel speed, since a real travel speed should include other
types of trip purposes. Therefore, a real average vehicle speed under congested roadway
networks is different from the mean median travel speed used in this research, and this
congestion index is different from the RCI calculated by the TTIs.
The result from the mean median travel speed calculation (see Table 4.1) shows that Las
Vegas (NV) is the most congested MSA based on work related trips at AM-peak hours.
Also, Bakersfield (CA) and Fresno (CA) are the two least congested MSAs within all 46
MSAs. Again, because mean median travel speed is based on work related trips (50%
based on the 1990s NPTS; 55% in 1995) and there are other types of trips contributing to
the congestion at AM-peak hours, the result shown below is only part of the traffic
congestion condition (work related trips) during AM-peak hours.
57


Table 4.1 The Congestion Index (Mean Median Travel Speed)
MSA Mean Median Travel Speed (mph) MSA Mean Median Travel Speed (mph)
1 Las Vegas, NV CMSA 29.20 24 Boston, MA PMSA 43.52
2 Albuquerque, NM MSA 32.42 25 Jacksonville, FL MSA 43.74
3 Washington, DC--MD-VA MSA 33.15 26 Atlanta, GA MSA 44.59
4 Denver, CO PMSA 34.67 27 Baltimore, MD MSA 45.01
5 New Orleans, LA MSA 35.33 28 Seattle-Tacoma, WA CMSA 45.43
6 San Jose, CA PMSA 35.87 29 PortlandVancouver, OR-WA CMSA 46.65
7 Honolulu, HI MSA 36.37 30 DetroitAnn Arbor, Ml CMSA 46.91
8 Pittsburgh, PA PMSA 36.71 31 San Francisco, CA MSA 47.88
9 Boulder-Longmont, CO CMSA 37.18 32 Milwaukee, WI PMSA 49.15
10 Miami-Fort Lauderdale, FL CMSA 37.35 33 Rochester, NY MSA 49.33
11 Orlando, FL MSA 37.46 34 TampaSt. PetersburgClearwater, FL MSA 49.60
12 San Antonio, TX MSA 37.78 35 Oklahoma City, OK MSA 50.64
13 Hartford--New Britain- Middletown, CT CMSA 39.30 36 Sacramento, CA MSA 54.92
14 Columbus, OH CMSA 40.41 37 Austin, TX MSA 55.96
15 Buffalo-Niagara Falls, NY CMSA 40.48 38 Brownsville-Harlingen, TX MSA 56.53
16 Colorado Springs, CO CMSA 41.91 39 Tucson, AZ MSA 58.04
17 New York-Northern New Jersey- -Long Island, NY-NJ 42.25 40 Phoenix, AZ MSA 58.67
18 Omaha, NEIA MSA 42.46 41 Houston-Galveston-Brazoria, TX CMSA 61.72
19 Oakland, CA MSA 42.59 42 DallasFort Worth, TX CMSA 62.27
20 Cleveland, OH CMSA 42.71 43 San Diego, CA MSA 68.99
21 Chicago-Gary, ILIN CMSA 43.06 44 EugeneSpringfield, OR MSA 72.33
22 Philadelphia, PA--NJ PMSA 43.34 45 Fresno, CA MSA 80.82
23 Los AngelesLong Beach, CA MSA 43.51 46 Bakersfield, CA MSA 174.07
58


4.1.2 Population and Employment Density in 1990
Based on the CTPPs data and MSAs Census, it shows that New York--Northern New
Jersey-Long Island (NY-NJ), Los Angeles-Long Beach (CA), Boston (MA), and San
Francisco (CA) are the four densest MSAs for both population density and employment
density. EugeneSpringfield (OR), Bakersfield (CA), Tucson (AZ), and Las Vegas (NV)
are the least dense MSAs for both population and worker densities. In general, the
sequences (from low to high) of MSAs population density and employment density are
very similar (see Table 4.2 and 4.3).
59


Table 4.2 Population Density
MSA Population Density (Pop. Per Sq. Mi) MSA Population Density (Pop. Per Sq. Mi)
1 EugeneSpringfield, OR MSA 62.10 24 New Orleans, LA MSA 536.60
2 Bakersfield, CA MSA 66.80 25 Atlanta, GA MSA 553.30
3 Tucson, AZ MSA 72.60 26 DallasFort Worth, TX CMSA 557.60
4 Las Vegas, NV CMS A 93.70 27 San Diego, CA MSA 594.10
5 Fresno, CA MSA 111.90 28 Pittsburgh, PA PMSA 604.90
6 Colorado Springs, CO CMSA 186.70 29 Buffalo-Niagara Falls, NY CMSA 758.70
7 Oklahoma City, OK MSA 225.70 30 HartfordNew Britain Middletown, CT CMSA 759.10
8 Phoenix, AZ MSA 230.60 31 Tampa~St. Petersburg- Clearwater, FL MSA 809.50
9 Austin, TX MSA 280.00 32 Detroit-Ann Arbor, MI CMSA 901.40
10 Brownsville-Harlingen, TX MSA 287.20 33 Baltimore, MD MSA 913.00
11 Sacramento, CA MSA 290.80 34 Milwaukee, W1 PMSA 980.90
12 Boulder-Longmont, CO CMSA 303.50 35 Washington, DCMDVA MSA 989.10
13 Omaha, NE--IA MSA 322.60 36 Miami-Fort Lauderdale, FL CMSA 1012.40
14 Portland-Vancouver, OR- WA CMSA 338.10 37 San Jose, CA PMSA 1159.80
15 Rochester, NY MSA 341.90 38 Cleveland, OH CMSA 1210.90
16 Jacksonville, FL MSA 344.00 39 Philadelphia, PA-NJ PMSA 1380.50
17 Columbus, OH CMSA 384.90 40 Honolulu, HI MSA 1393.30
18 Albuquerque, NM MSA 412.10 41 Oakland, CA MSA 1428.80
19 Orlando, FL MSA 422.70 42 Chicago-Gary, IL--IN CMSA 1435.50
20 Denver, CO PMSA 431.50 43 San Francisco, CA MSA ** 1579.10
21 Seattle-Tacoma, WA CMSA 434.40 44 Boston, MA PMSA 1630.90
22 San Antonio, TX MSA 516.80 45 Los AngelesLong Beach, CA MSA 2183.10
23 Houston-Galveston- Brazoria, TX CMSA 522.10 46 New York-Northern New JerseyLong Island, NYNJ 2320.10
60


Table 4.3 Employment Density
MSA Workers Density (Employee Per Sq. Mi) MSA Workers Density (Employee Per Sq. Mi)
1 Bakersfield, CA MSA 26.23 24 HoustonGalvestonBrazoria, TX CMSA 247.60
2 EugeneSpringfield, OR MSA 27.79 25 Pittsburgh, PA PMSA 259.29
3 Tucson, AZ MSA 31.74 26 DallasFort Worth, TX CMSA 283.69
4 Fresno, CA MSA 44.51 27 Atlanta, GA MSA 289.33
5 Las Vegas, NV CMSA 46.91 28 San Diego, CA MSA 292.65
6 Colorado Springs, CO CMSA 92.84 29 BufFalo-Niagara Falls, NY CMSA 338.81
7 Brownsville-Harlingen, TX MSA 93.47 30 Tampa-St. Petersburg- Clearwater, FL MSA 358.08
8 Oklahoma City, OK MSA 105.98 31 Hartford-New Britain- Middletown, CT CMSA 392.88
9 Phoenix, AZ MSA 108.27 32 DetroitAnn Arbor, Ml CMSA 401.85
10 Sacramento, CA MSA 134.66 33 Baltimore, MD MSA 456.76
11 Austin, TX MSA 144.72 34 MiamiFort Lauderdale, FL CMSA 468.06
12 Omaha, NE--IA MSA 163.73 35 Milwaukee, WI PMSA 472.60
13 Boulder-Longmont, CO CMSA 164.10 36 Washington, DCMDVA MSA 558.23
14 Rochester, NY MSA 164.24 37 San Jose, CA PMSA 616.95
15 Portland-Vancouver, ORWA CMSA 165.76 38 Philadelphia, PA--NJ PMSA 648.24
16 Jacksonville, FL MSA 168.41 39 Chicago-Gary, 1L-IN CMSA 673.15
17 Columbus, OH CMSA 189.41 40 Oakland, CA MSA 709.54
18 Albuquerque, NM MSA 196.33 41 Honolulu, HI MSA 728.95
19 Orlando, FL MSA 219.65 42 Cleveland, OH CMSA 821.39
20 Seattle-Tacoma, WA CMSA 222.06 43 San Francisco, CA MSA 840.83
21 New Orleans, LA MSA 222.94 44 Boston, MA PMSA 845.64
22 Denver, CO PMSA 224.17 45 Los AngelesLong Beach, CA MSA 1013.61
23 San Antonio, TX MSA 225.89 46 New York-Northern New Jersey-Long Island, NY-NJ 1033.52
61


4.1.3 The Economic Condition of MSAs
The characteristics of an MSA is determined mainly by the result of cluster analysis and
facilitated by comparing the percentage of an economic/industrial sector (SIC) in an MSA
to the 46 MSAs ratio in the same sector for explanation. Whenever the percentage of an
industrial sector in an MSA is larger than the 46 MSAs ratio, that MSA is characterized by
such an economic/industrial sector. Also, an MSA may have several economic/industrial
characteristics. Here, the calculation of the percentage for an industrial sector is based on
the number of residence jobs (employees by area of work) of an economic/industrial sector
(SIC) in an MSA divided by the total number of residence jobs (employees of all
industries based on area of work) of all industrial sectors in that MSA. The way to
calculate the 46 MSAs ratio is based on the total number of residence jobs of an
economic/industrial sector for the 46 selected MSAs divided by the overall number of
residence jobs of all industrial sectors for the 46 selected MSAs (the total number of
regional employees).
Moreover, if the above comparison cannot identify and explain the character of an MSA
because none of the percentage of an industrial sector in that MSA is larger than the 46
MSAs ratio, the character of that MSA is then determined by comparing the percentage of
an industrial sector in that MSA with the ratio of an MSAs all industries sector. Here,
the method to calculate the ratio of an MSAs all industries sector is based on the total
number of residence jobs (employees of all industries based on area of work) of all
economic/industrial sectors in an MSA divided by the overall residence jobs of the 46
selected MSAs of all industrial sectors (the total number of employees of all industries
based on area of work).
The characteristics of 46 MSAs are below. It is important to know that an MSA with
different economic characteristics (e.g. agricultural and entertainment) is associated with
different travel patterns and congestion conditions (due to the different period of peak
hours). Take agricultural industry for example. It is likely the people working for the
agricultural industry (as for mining) tend to live close to the farm, and therefore they car-
pool for work, thereby resulting in less roadway congestion.
62


Table 4.4 The Characteristics of 46 Selected MSAs
MSA Feature
1 Phoenix, AZ MSA Communication + Finance + Health service
2 Tucson, AZ MSA Education + Administration
3 Bakersfield, CA MSA Agriculture + Manufacturing durable goods + Entertainment
4 Fresno, CA MSA Agriculture
5 Los Angeles-Long Beach, CA MSA Construction + Manufacturing nondurable goods
6 Sacramento, CA MSA Agriculture
7 San Diego, CA MSA Agriculture
8 San Francisco, CA MSA No specific character
9 San Jose, CA PMSA No specific character
10 Boulder-Longmont, CO CMSA Education + Manufacturing durable goods + Business
11 Denver, CO PMSA Communication + Administration
12 Colorado Springs, CO CMSA Military
13 Hartford-New Britain-Middletown, CT CMSA Finance
14 Washington, DC-MD-VA MSA Administration + Service + Business
15 Jacksonville, FL MSA Military + Transportation + Administration
16 Miami-Fort Lauderdale, FL CMSA Construction + Wholesale
17 Orlando, FL MSA Construction + Wholesale
18 Tampa-St. Petersburg-Clearwater, FL MSA Communication + Finance + Health service
19 Atlanta, GA MSA Communication + Finance + Business + Wholesale
20 Honolulu, HI MSA Military + Transportation + Administration
21 Chicago-Gary, IL-IN CMSA Manufacturing nondurable + durable goods
22 New Orleans, LA MSA Communication + Administration
23 Boston, MA PMSA Finance + Health service
24 Baltimore, MD MSA Finance + Health service
25 Detroit-Ann Arbor, Ml CMSA Manufacturing durable goods
26 Omaha, NE--IA MSA Communication + Finance + Business + Wholesale
27 Albuquerque, NM MSA Education + Administration
28 Las Vegas, NV CMSA Personal service + Construction + Entertainment
29 Buffalo-Niagara Falls, NY CMSA Manufacturing nondurable + durable goods
30 Rochester, NY MSA Manufacturing durable goods
31 New York-Northern New Jersey-Long Island, NY-NJ Transportation + Finance
32 Columbus, OH CMSA Finance + Health service
33 Cleveland, OH CMSA Manufacturing nondurable + durable goods
34 Oklahoma City, OK MSA Communication + Administration
35 Eugene-Springfield, OR MSA Education + Manufacturing durable goods
36 Portland-Vancouver, OR--WA CMSA Wholesale + Manufacturing durable goods + Wholesale
63


Table 4.4 The Characteristics of 46 Selected MSAs (Cont.)
MSA Feature
37 Pittsburgh, PA PMSA Finance + Health service
38 Philadelphia, PA--NJ PMSA Manufacturing nondurable goods
39 Austin, TX MSA Education + Administration
40 Brownsville-Harlingen, TX MSA Education + Manufacturing nondurable goods
41 Dallas-Fort Worth, TX CMSA Wholesale + Manufacturing durable goods + Business
42 Houston-Galveston-Brazoria, TX CMSA Construction + Wholesale
43 San Antonio, TX MSA Military + Transportation + Administration
44 Seattle-Tacoma, WA CMSA Wholesale + Manufacturing durable goods
45 Milwaukee, Wl PMSA Manufacturing nondurable + durable goods
46 Oakland, CA MSA ** No specific character
64


4.1.4 Median Household Income in 1989
Based on the CTPPs data, San Jose in CA is the richest MSA, and Brownsville-Harlingen
in TX is the poorest MSA.
Tab e 4.5 Median Household ! ncome in 1989
MSA Median Household Income in 1989 (dollars) MSA Median Household Income in 1989 (dollars)
l Brownsville-Harlingen, TX MSA 17,336 24 Houston-Galveston-Brazoria, TX CMSA 31,488
2 New Orleans, LA MSA 24,442 25 Milwaukee, WI PMSA 32,316
3 Eugene-Springfield, OR MSA 25,268 26 Sacramento, CA MSA 32,734
4 Tucson, AZ MSA 25,401 27 DallasFort Worth, TX CMSA 32,825
5 Tampa-St. Petersburg- Clearwater, FL MSA 26,036 28 Denver, CO PMSA 32,852
6 San Antonio, TX MSA 26,092 29 Rochester, NY MSA 34,234
7 Fresno, CA MSA 26,377 30 Detroit-Ann Arbor, MI CMSA 34,729
8 Pittsburgh, PA PMSA 26,700 31 Los Angeles-Long Beach, CA MSA 34,965
9 Oklahoma City, OK MSA 26,883 32 San Diego, CA MSA 35,022
10 Albuquerque, NM MSA 27,382 33 Seattle-Tacoma, WA CMSA 35,047
11 Buffalo-Niagara Falls, NY CMSA 28,084 34 Philadelphia, PA--NJ PMSA 35,321
12 Austin, TX MSA 28,474 35 Boulder-Longmont, CO CMSA 35,322
13 MiamiFort Lauderdale, FL CMSA 28,503 36 Chicago-Gary, 1LIN CMSA 36,005
14 Bakersfield, CA MSA 28,634 37 Atlanta, GA MSA 36,051
15 Jacksonville, FL MSA 29,514 38 Baltimore, MD MSA 36,550
16 Colorado Springs, CO CMSA 29,604 39 New YorkNorthern New Jersey-Long Island, NY-NJ 37,853
17 Cleveland, OH CMSA 30,332 40 Boston, MA PMSA 40,491
18 Omaha, NEIA MSA 30,368 41 San Francisco, CA MSA 40,494
19 Columbus, OH CMSA 30,668 42 Honolulu, HI MSA 40,581
20 Las Vegas, NV CMSA 30,746 43 Oakland, CA MSA 40,621
21 Phoenix, AZ MSA 30,797 44 HartfordNew Britain Middletown, CT CMSA 41,440
22 Portland-Vancouver, OR-WA CMSA 31,064 45 Washington, DC-MD-VA MSA 46,590
23 Orlando, FL MSA 31,230 46 San Jose, CA PMSA 48,115
65


4.1.5 Car Ownership
There are two types of car ownership rate calculated in this research. One is car ownership
rate, and the other is worker car ownership rate. According to the CTPPs data,
Brownsville-Harl ingen in TX is the poorest MSA and has the lowest car ownership rate.
The MSA of Cleveland in OH has the highest car ownership rate.
Table 4.6 Car Ownership Rate for 46 Selected MSAs
MSA Vehicle Ownership MSA Vehicle Ownership
1 Brownsville-Harlingen, TX MSA 40.80% 24 Phoenix, AZ MSA 63.03%
2 New York-Northern New Jersey- -Long Island, NY--NJ 41.83% 25 San Diego, CA MSA 63.08%
3 New Orleans, LA MSA 51.43% 26 Las Vegas, NV CMSA 63.62%
4 Chicago-Gary, IL--IN CMSA 52.91% 27 Tampa-St. Petersburg- Clearwater, FL MSA 64.05%
5 Honolulu, HI MSA 52.91% 28 Orlando, FL MSA 64.28%
6 Philadelphia, PA--NJ PMSA 53.29% 29 Austin, TX MSA 64.52%
7 Boston, MA PMSA 55.40% 30 Dallas-Fort Worth, TX CMSA 65.27%
8 Fresno, CA MSA 55.57% 31 HartfordNew Britain Middletown, CT CMSA 65.72%
9 Los Angeles-Long Beach, CA MSA 56.67% 32 Omaha, NE-IA MSA 65.75%
10 San Antonio, TX MSA 56.69% 33 Columbus, OH CMSA 65.82%
11 Miami-Fort Lauderdale, FL CMSA 57.14% 34 Oakland, CA MSA** 66.05%
12 Buffalo-Niagara Falls, NY CMSA 57.19% 35 Oklahoma City, OK MSA 66.94%
13 Pittsburgh, PA PMSA 57.32% 36 Colorado Springs, CO CMSA 67.47%
14 Bakersfield, CA MSA 57.92% 37 Sacramento, CA MSA 67.73%
15 Baltimore, MD MSA 58.21% 38 Atlanta, GA MSA 67.97%
16 Houston-Galveston-Brazoria, TX CMSA 59.24% 39 Albuquerque, NM MSA 68.93%
17 San Francisco, CA MSA 59.45% 40 PortlandVancouver, ORWA CMSA 69.03%
18 Milwaukee, W1 PMSA 59.49% 41 San Jose, CA PMSA 69.19%
19 Rochester, NY MSA 61.72% 42 EugeneSpringfield, OR MSA 70.04%
20 Detroit-Ann Arbor, MI CMSA 61.84% 43 Denver, CO PMSA 71.06%
21 Jacksonville, FL MSA 62.32% 44 Seattle-Tacoma, WA CMSA 72.00%
22 Washington, DC--MDVA MSA 62.62% 45 BoulderLongmont, CO CMSA 73.56%
23 Tucson, AZ MSA 63.03% 46 Cleveland, OH CMSA 94.73%
Vfote: The unit is % of Car per capita.
66


For worker car ownership rate, the MSA of Detroit-Ann Arbor in MI has the highest worker
car ownership rate, and the MSA of New YorkNorthern New JerseyLong Island in NY-
NJ has the lowest rate. Based on the fact that the MSA of New YorkNorthern New
JerseyLong Island in NY-NJ has the highest transit usage rate, New Yorkers probably use
public transit a lot to avoid high parking fees, and therefore it has lowest worker car
ownership rate.
Table 4.7 Worker Car Ownership Rate for 46 Selected MSAs
MSA Workers Vehicle Ownership MSA Workers Vehicle Ownership
1 New York-Northern New Jersey-Long Island, NY-NJ 56.80% 24 Sacramento, CA MSA 81.53%
2 San Francisco, CA MSA 61.87% 25 Phoenix, AZ MSA 81.69%
3 Honolulu, HI MSA 67.08% 26 Austin, TX MSA 81.77%
4 Washington, DC-MD-VA MSA 69.70% 27 Milwaukee, W1 PMSA 81.88%
5 Boston, MA PMSA 70.28% 28 Las Vegas, NV CMSA 81.89%
6 Chicago-Gary, ILIN CMSA 72.67% 29 Miami-Fort Lauderdale, FL CMSA 81.94%
7 Philadelphia, PA--NJ PMSA 73.20% 30 Bakersfield, CA MSA 82.09%
8 Oakland, CA MSA 74.41% 31 BuffaloNiagara Falls, NY CMSA 82.36%
9 Boulder-Longraont, CO CMSA 76.27% 32 Houston-Galveston-Brazoria, TX CMSA 82.61%
10 Pittsburgh, PA PMSA 76.68% 33 Jacksonville, FL MSA 82.84%
11 Los Angeles-Long Beach, CA MSA 77.05% 34 Rochester, NY MSA 83.22%
12 San Diego, CA MSA 77.23% 35 San Jose, CA PMSA 83.42%
13 Baltimore, MD MSA 77.28% 36 Albuquerque, NM MSA 83.50%
14 New Orleans, LA MSA 77.78% 37 Atlanta, GA MSA 83.82%
15 Eugene-Springfield, OR MSA 78.56% 38 Orlando, FL MSA 84.24%
16 Tucson, AZ MSA 78.72% 39 Cleveland, OH CMSA 84.40%
17 Brownsville-Harlingen, TX MSA 78.92% 40 HartfordNew Britain- Middletown, CT CMSA 84.45%
18 Seattle-Tacoma, WA CMSA 78.93% 41 Columbus, OH CMSA 84.92%
19 PortlandVancouver, ORWA CMSA 79.49% 42 TampaSt. Petersburg- Clearwater, FL MSA 84.99%
20 Colorado Springs, CO CMSA 80.78% 43 Dallas-Fort Worth, TX CMSA 85.01%
21 San Antonio, TX MSA 81.31% 44 Omaha, NE1A MSA 85.18%
22 Denver, CO PMSA 81.44% 45 Oklahoma City, OK MSA 86.48%
23 Fresno, CA MSA 81.44% 46 DetroitAnn Arbor, MI CMSA 87.37%
Note: The unit is % of Car per worker.
67


4.1.6 Transit Ridership
Based on the CTPPs data, the MSA of New YorkNorthern New JerseyLong Island in
NY-NJ (the MSA with the lowest worker car ownership rate) has the highest transit
ridership rate. On the contrary, the MSA of Phoenix in AZ has the lowest transit share rate.
Table 4.8 Transit Ridership Rate for 46 Selected MSAs
MSA Transit Ridership MSA Transit Ridership
1 Phoenix, AZ MSA 0.02% 24 HartfordNew Britain Middletown, CT CMSA 3.61%
2 Oklahoma City, OK MSA 0.55% 25 San Antonio, TX MSA 3.61%
3 Bakersfield, CA MSA 0.92% 26 Houston-Galveston-Brazoria, TX CMSA 3.67%
4 Colorado Springs, CO CMSA 1.00% 27 Miami-Fort Lauderdale, FL CMSA 4.21%
5 BrownsvilleHarlingen, TX MSA 1.12% 28 Denver, CO PMSA 4.30%
6 Tampa-St. Petersburg- Clearwater, FL MSA 1.33% 29 Buffalo-Niagara Falls, NY CMSA 4.45%
7 Orlando, FL MSA 1.45% 30 Cleveland, OH CMSA 4.49%
S Fresno, CA MSA 1.49% 31 Atlanta, GA MSA 4.59%
9 Albuquerque, NM MSA 1.71% 32 Milwaukee, WI PMSA 5.18%
10 Las Vegas, NV CMSA 1.87% 33 PortlandVancouver, ORWA CMSA 5.36%
11 Omaha, NE--IA MSA 1.97% 34 Seattle-Tacoma, WA CMSA 6.17%
12 Jacksonville, FL MSA 1.97% 35 Los AngelesLong Beach, CA MSA 6.44%
13 Dallas-Fort Worth, TX CMSA 2.26% 36 New Orleans, LA MSA 6.87%
14 DetroitAnn Arbor, MI CMSA 2.31% 37 Baltimore, MD MSA 7.39%
15 Eugene-Springfield, OR MSA 2.35% 38 Pittsburgh, PA PMSA 8.42%
16 Sacramento, CA MSA 2.35% 39 Honolulu, HI MSA 9.03%
17 Columbus, OH CMSA 2.66% 40 Oakland, CA MSA 9.04%
18 San Jose, CA PMSA 2.94% 41 Philadelphia, PANJ PMSA 11.54%
19 Tucson, AZ MSA 3.11% 42 Washington, DCMDVA MSA 13.34%
20 Rochester, NY MSA 3.12% 43 Chicago-Gary, 1L--IN CMSA 13.57%
21 San Diego, CA MSA 3.20% 44 Boston, MA PMSA 13.86%
22 Austin, TX MSA 3.21% 45 San Francisco, CA MSA 18.99%
23 BoulderLongmont, CO CMSA 3.35% 46 New YorkNorthern New Jersey-Long Island, NY-NJ 26.78%
Note: 1. Transit includes bus/trolley, streetcar/trolley, subway/elevated, and railroad.
2. The unit is % of Workers by transit divided by Total workers.
68


4.1.7 Urban Form Index
The approach to calculate urban form index has been discussed in section 3.3, and three
selected MSAs have been used as examples for explanation. Because of the difficulty of
using an accurate word to describe MSAs metropolitan shape, the usage of the urban form
index (coefficient of variation of the 8-fixed-reference-line for each MSA) probably is more
appropriate than using a word. Again, a lower coefficient (index) means that an MSA has
more uniform metropolitan shape, and it implies higher accessibility (shorter travel
distance), for which drivers have more choice of alternative routes for connecting to other
places. Therefore higher accessibility results in less congested highways. A higher
coefficient (urban form index) means that an MSA has a very irregular metropolitan shape,
and it implies inaccessibility (longer travel distance), for which drivers have fewer
alternative routes to choose to connect to other places. Therefore inaccessibility results in
serious highway congestion. Also, the urban form index below 0.15 implies a square or
rounded circle shape. The urban form index above 0.5 implies a highly irregular
metropolitan shape, and the index from 0.15 to 0.5 implies a linear or similar to linear
metropolitan shape. The urban form index for 46 selected MSAs is shown on the next page.
69


Table 4.9 The Urban Form Index for 46 Selected MSAs
(UTM Projection -from TransCAD to AutoCAD MAP 2000) MSA State (Sq. Meters) Area (Meters) Perimeter Stdev Averaae Coefficient Variation cov Shaoe
Phoenix. AZ MSA AZ 59.589.915.502.28 1,295,591.72 34,640.52 123,980.22 0.28 L-Rounded
Tucson. AZ MSA AZ 63,396.409.257.84 1,203,665.34 61,441.51 135,182.56 0.45 Trapezoid
Bakersfield. CA MSA CA 41,977,109.928.04 985,614.71 25,653.30 106,494.55 0.24 Regular Trapezoid
Fresno. CA MSA CA 39,573,021,395.06 1,070,261.06 45,566.00 109,137.85 0.42 Similar to Rectangle
Los Anqeles-Lonq Beach, CA MSA CA 21,130,414,477.62 707,446.41 15,399.42 81,871.28 0.19 Trapezoid Almost Square
Sacramento, CA MSA CA 19,429,871,865.08 842,747.89 39,321.29 79,737.76 0.49 Linear Island
San Dieoo, CA MSA CA 24,516,893,134.91 690,281.33 12,497.69 86,446.29 0.14 Trapezoid Near Square
San Francisco,CA MSA CA 6,006,242,363.60 431,394.19 11,967.36 41,606.01 0.29 Imperfect Olive
San Jose. CA PMSA CA . 4,844,498,304.94 565,665.16 36,952.81 33,440.11 1.11 V-Linear Shape
Boulder-Lonqmont. CO CMSA CO 4,535,674,408.69 305,313.67 7,623.98 37,498.60 0.20 Rectangle near Square
Denver, CO PMSA CO 23,691,507,007.74 754,539.17 40,649.16 78,344.22 0.52 L-Rectangle
Colorado Springs, CO CMSA CO 13,938,835,293.34 497,728.19 7,948.33 63,997.35 0.12 Rectangle near Square
Hartford-New Britain-Middletown. CT CMSA CT 38,049,452,740.44 1,072,259.53 25,815.58 109,735.40 0.24 Rounded Trapezoid
Washinqton. DC-MD-VA MSA DC-MD-VA 111,660,000,000.00 2,326,422.19 77,193.29 187,558.48 0.41 Long Island
Jacksonville. FL MSA FL 127,270,000,000.00 2,053,844.83 77,488.42 194,096.86 0.40 Similar to Long Trapezoid
Miami-Fort Lauderdale, FL CMSA FL 229,470,000,000.00 2,528,758.55 42,134.27 262,693.55 0.16 Unparallel Trapezoid
Orlando. FL MSA FL 139,280,000,000.00 2,241,997.09 95,531.32 193,502.00 0.49 Long Island
Tampa-St. Petersburq-Clearwater, FL MSA FL :. ;- -:-m. 148,860,000,000.00 1,768,656.58 51,083.01 221,995.67 0.23 Irregular Parallelogram
Atlanta. GA MSA GA 130,450,000,000.00 2,119,911.08 33,133.53 213,960.15 0.15 Similar to Circle
Honolulu. HI MSA HI 3.246,525,405.73 248,796.41 8,009.10 31,786.56 0.25 Olive Island
Chicaao-Gary. IL-IN CMSA ** IL. IN 68,185,396,864.45 1,243,436,72 35,876.51 151,715.26 0.24 Two Square (Small and Big)
New Orleans, LA MSA LA 97,394,521,086.41 2,396,523.96 100,601.35 185,610.93 0.54 V-Shape Irregular Island
Boston. MA PMSA MA 55,589,705,027.64 2,090,827.89 66,437.70 149,749.37 0.44 Long Island
Baltimore. MD MSA MD 79.904,141,741.33 1,934,514.14 56,903.02 166,527.34 0.34 Long Island
Detroit-Ann Arbor. Mi CMSA Ml 61.694,209.940.22 1,243,896.01 71,919.59 137,423.41 0.52 Irregular X-Shape
Omaha. NE-IAMSA NE-IA 14.125.239.144.16 654,180.43 20,677.84 62,774.05 0.46 T-shape
Albuaueraue. NM MSA NM 8,527,196,608.83 442,250.84 24,711.94 54,462.54 0.45 Long Trapezoid
Las Veaas. NV CMSA NV 44,468,617,413.97 1,077,024.96 47,084.58 120,*2.64 0.39 L-Shape
Buffalo-Niaaara Falls, NY CMSA NY 32,188,054,769.54 966,996.75 48,896.05 92,567.97 0.53 Long Island
Rochester. NY MSA NY 70,605,069,865.13 1,386,308.17 50,303.50 147,478.74 0.34 Fat-L-Shape
New York-Northern New Jersev-Lonq Island, NY-NJ ** NY. NJ 253,270,000,000.00 3,488,483.64 165,553.62 313,471.67 0.53 Y-Shape Island
Columbus. OH CMSA OH 54,122,086,681.99 1,206,716.59 58,224.22 131,325.25 0.44 Irregular Rhombus
Cleveland. OH CMSA OH 67,380,588,861.92 1,098,245.05 48,828.96 147,504.50 0.33 Trapozoid
Oklahoma City. OK MSA OK 44,043,042,692.36 1,266,654.05 51,712.56 126,555.54 0.41 Fat-Y-Shape
Euaene-Sprinqfield. OR MSA OR 18,609,584,735.47 794,749.15 46,874.62 70,907.93 0.66 Highly Irregular Trapezoid
Portland-Vancouver, OR-WA CMSA OR-WA 17,118,593,325.00 914,275.29 38,991.13 76,344.23 0.51 Highly Irregular Fat-C
Pittsburgh. PA PMSA PA 69,367,709,483.45 1,530,710.45 50,396.88 151,969.22 0.33 Irregular Rhombus
Philadelphia. PA-NJ PMSA PA-NJ 130,320,000,000.00 2,301,924.81 73,312.75 180,237.24 0.41 H+C Highly Irregular Shape
Austin, TX MSA TX 38,768,384,211.37 845,281.79 22,027.64 106,849.91 0.21 Similar to Trapezoid
Brownsville-Harlingen, TX MSA TX 22,289,992,239.48 934,067.04 21,606.22 79,659.96 0.27 Similar to Parallelogram
Dallas-Fort Worth, TX CMSA TX 89,086,142,268.60 1,491,517.52 39,327.65 164,557.95 0.24 Bouble"+" Shape
Houston-Galveston-Brazoria, TX CMSA TXr^y#?'-' 128,820,000,000.00 2,489,962.83 63,089.60 191,013.43 0.33 Fat-C Shape
San Antonio. TX MSA TX 35,358,866,585.67 934,021.16 33,419.75 102,141.98 0.33 "8" Shape
Seattle-Tacoma. WA CMSA WA 23,540,198,144.33 853,540.17 28,722.28 82,572.28 0.35 Similar to Rectangle
Milwaulkee. Wl PMSA Wl 27,454,892,374.80 697,401.47 21,871.62 94,455.79 0.23 Trapezoid
Oakland. CA MSA CA : 6,859,271,025.85 427,756.08 7,284.51 47,077.38 0.15 Almost Square


4.2 The Results of Cluster and Factor Analyses
Generally, both cluster analysis and factor analysis are used to facilitate explaining the
results of regressions. Factor analysis has been used to minimize a multi-collinearity
problem. Cluster analysis has been used for categorizing the economic characteristics of an
MSA, facilitated by comparing the percentage of an economic/industrial sector in an MSA
with the 46 MSAs ratio (see section 4.1.3). In this research, factor analysis is able to group
the coefficient of variation of the jobs-housing ratio based on both SIC and SOC to
eliminate the multi-collinearity, before running regression. The result of factor analysis
(see Table 4.10) shows that the 14 occupations based on the coefficient of variation of the
Jobs-housing ratio in the SOC have been grouped into three groups, and each group is
closely related to income status. The result of cluster analysis (joint tree) is shown in Table
4.11, Figure 4.1, and Table 4.12. Thel8 industrial sectors (SIC) based on the coefficient of
variation of the Jobs-housing ratio in the SIC have been grouped into 4 groups. In general,
these 4 groups are correspondent to the primary, secondary, congestion most related, and
special groups (Stanley 993, 26-27). The primary industrial group (factor 3), which is less
related to traffic congestion, includes agriculture and mining sectors. The secondary
industrial group (factor 4) manufacturing non-durable goods is associated with roadway
congestion. The congestion most related group includes manufacturing durable goods,
wholesale trade, retail trade, finance, business, health services, education services, other
professional services, and armed forces. The special group includes entertainment and
public administration, and its relationship with roadway congestion varies. For example,
the entertainment sector is less related to traffic congestion, but the public administration
sector is partially associated with congestion.
71


Table 4.10 The Result of Factor Analysis by the SOC
Factor Loadings (Varimax normalized) (Sheetl in J-W-A.stw) Extraction: Principal components (Marked loadings are > .700000) JHW-FctO.stw
Occupations (SOC) Variable Factor 1 Factor 2 Factor 3
Executive, administrative, and managerial B02 0102 0.792912 0.072370 0.183117
Professional specialty B02 0103 0.773354 0.085443 0.290597
Technicians and related support B02 0104 0.732379 0.335007 0.381270
Sales B02 0105 0.502238 0.083798 0.734614
Administrative support B02 0106 0.644934 0.171964 0.567885
Private household B02 0107 0.040116 0.118654 0.777264
Protective service B02 0108 0.073919 0.797887 0.253708
Service except protective and household B02 0109 0.286495 0.159612 0.652700
Farming, forestry, and fishing B02 0110 0.009416 0.729066 0.165512
Precision production, craft, and repair B02 0111 0.896934 0.212614 0.153526
Machine operators, assemblers, and inspectors B02 0112 0.658903 0.464239 0.283507
Transportation and material moving B02 0113 0.432164 0.687543 -0.008173
Handlers, eguipment cleaners, helpers, and laborers B02 0114 0.496731 0.631316 0.142321
Armed forces B02 0115 0.314916 0.343425 0.625256
Expl.Var 4.291912 2.619015 2.762628
69.096825% Prp.Totl 0.306565 0.187073 0.197331
Table 4.11 The Result of Factor Analysis by the SIC
Factor Loadings (Varimax normalized) (Sheetl in J-W-A.stw) Extraction: Principal components (Marked loadings are > .700000) JHW-FctO.stw
Industrial Sectors (SIC) Variable Factor 1 Factor 2 Factor 3 Factor 4
Agriculture, forestry, and fisheries B03 0102 0.119853 0.168417 0.765308 0.027256
Mining B03 0103 0.129499 0.066948 0.733812 0.189558
Construction B03 0104 -0.087070 0.629354 0.016647 0.607858
Manufacturing, nondurable goods B03 0105 0.118385 -0.003338 0.159102 0.875190
Manufacturing, durable goods B03 0106 0.849478 0.148944 0.124274 0.020986
Transportation B03 0107 0.673687 0.505519 -0.062161 0.065260
Communications and other public utilities B03 0108 0.505567 0.322017 0.246985 0.445175
Wholesale trade B03 0109 0.873047 0.309998 -0.021216 0.215680
Retail trade B03 0110 0.814617 0.315659 0.202220 -0.080320
Finance, insurance, and real estate B03 0111 0.773SS7 -0.067282 0.126847 0.036919
Business and repair services B03 0112 0.872636 0.384211 0.047021 0.123306
Personal services B03 0113 0.391367 0.596600 0.290422 -0.006932
Entertainment and recreation services B03 0114 0.311880 0.781987 0.335163 -0.056395
Health services B03 0115 0.73431S 0.146530 0.255182 0.011276
Educational services B03 0116 0.851599 0.340532 0.055710 0.022595
Other professional and related services B03 0117 0.720775 0.535573 0.157971 0.138466
Public administration B03 0118 0.367824 0.751617 0.019234 0.215429
Armed forces B03 0119 0.755336 0.200545 0.249581 0.239689
Expl.Var 7.009713 3.155110 1.642236 1.571441
74.324997% Prp.Totl 0.389428 0.175284 0.091235 0.087302
72


Figure 4.1 The Result of Cluster Analysis for 46 Selected MSAs by the SIC


Table 4.12 The Correspondent Codes for Figure 4.1
Code MSA Code MSA
C_ 1 Phoenix, AZ MSA C_ 24 Baltimore, MD MSA
c_ 2 Tucson, AZ MSA C_ 25 DetroitAnn Arbor, MI CMSA
C_ 3 Bakersfield, CA MSA C_ 26 Omaha, NELA MSA
C_ 4 Fresno, CA MSA C_ 27 Albuquerque, NM MSA
C_ 5 Los Angeles-Long Beach, CA MSA C_ 28 Las Vegas, NV CMSA
C_ 6 Sacramento, CA MSA C_ 29 BuffaloNiagara Falls, NY CMSA
C_ 7 San Diego, CA MSA C_ 30 Rochester, NY MSA
C_ 8 San Francisco, CA MSA C_ 31 New YorkNorthern New Jersey- Long Island, NY-NJ
C_ 9 San Jose, CA PMSA C_ 32 Columbus, OH CMSA
C_ 10 BoulderLongmont, CO CMSA C_ 33 Cleveland, OH CMSA
C_ 11 Denver, CO PMSA C_ 34 Oklahoma City, OK MSA
c_ 12 Colorado Springs, CO CMSA C_ 35 Eugene-Springfield, OR MSA
c_ 13 Hartford-New Britain- Middletown, CT CMSA C_ 36 PortlandVancouver, ORWA CMSA
c_ 14 Washington, DC-MD-VA MSA C_ 37 Pittsburgh, PA PMSA
c_ 15 Jacksonville, FL MSA C_ 38 Philadelphia, PA-NJ PMSA
c_ 16 MiamiFort Lauderdale, FL CMSA C_ 39 Austin, TX MSA
c_ 17 Orlando, FL MSA C_ 40 BrownsvilleHarlingen, TX MSA
c_ 18 TampaSt. Petersburg Clearwater, FL MSA C_ 41 DallasFort Worth, TX CMSA
c_ 19 Atlanta, GA MSA C_ 42 HoustonGalvestonBrazoria, TX CMSA
c_ 20 Honolulu, HI MSA c_ 43 San Antonio, TX MSA
c_ 21 ChicagoGary, ILIN CMSA c_ 44 SeattleTacoma, WA CMSA
c_ 22 New Orleans, LA MSA c_ 45 Milwaukee, WI PMSA
c_ 23 Boston, MA PMSA c_ 46 Oakland, CA MSA
74


4.3 The Results of Correlation Analysis
Since correlation analysis is based on the linear model by comparing two variables, the
evidence from previous research has shown that the relationship between land use and
transportation may not be linear. Thus, the correlation analysis here is used to find the
initial relationship between the congestion index (mean median speed) with the explanatory
variables (land use, transportation, and urban form). However, this result (correlation
coefficient) is not limited to the other possible non-linear relationships. Therefore, the 2-D
scatterplot, the fixed nonlinear regression, and the nonlinear estimation (piecewise linear
regression) are used to define the non-linear relationships, which are discussed in section
4.4. Moreover, in this research, correlation analysis is also used to show the relationship
between the SOC and the SIC. Three types of correlation analysis are discussed below,
which are congestion vs. non-jobs-housing balance variables, congestion vs. jobs-housing
balance variables, and the SOC vs. the SIC.
4.3.1 Congestion vs. Non-jobs-housing Balance Variables
Originally, based on the criterion that p-level is less than 0.05, the correlation between mean
median speed with each non-jobs-housing balance variable almost does not show any linear
relationship except for metropolitan area, population density, and residence employment
density. This is probably because their relationship is non-linear. However, after using the
calibrated mean median speed by using the reciprocal of population density or the
reciprocal of macro-scale roadway density, most of the explanatory variables are negatively
correlated with the congestion index.
75


Table 4.13 The Result of the Correlation Based on Congestion vs. Non-jobs-housing
Balance Variables
Case No=46 Correlations (Sheetl in OVarsAll.stw) Marked correlations are significant at p < .05000 N=46 (Casewise deletion of missing data) Mean Median Observed Speed Mean Median Travel Speed Calibrated by Population Density Mean Median Travel Speed calibrated by Macro- scale Roadway Density
Aggregate number of vehicles used for work trip -0.13 -0.40 -0.43
Residence workers per vehicle -0.13 -0.14 -0.10
Total housing units -0.12 -0.35 -0.37
Aggregate number of vehicles available -0.13 -0.39 -0.42
Macro-scale roadway density -0.24 -0.46 -0.72
Population -0.11 -0.34 -0.35
MSA AREA 0.44 0.50 0.21
Population Density (Population/mile2) -0.29 -0.60 -0.58
Residence Employment density (Workers/mile2) -0.30 -0.60 -0.59
Median Household Income in 1989 (dollars) -0.24 -0.46 -0.52
Car ownership (% Car per capita) -0.09 0.09 -0.09
Workers car ownership (% Car per worker) 0.12 0.20 0.24
Transit ridership rate (% Workers by transit/Total workers) -0.23 -0.38 -0.35
Coefficient of variation of urban form index -0.14 0.12 0.00
VMT -0.07 -0.17 -0.17
76


4.3.2 Congestion vs. Jobs-housing Balance Variables
Originally, based on the criterion that p-level is less than 0.05, the correlation between mean
median speed with each jobs-housing balance variable for both SIC and SOC almost shows
no linear relationship (except for manufacturing durable goods). This is also because the
relationship is non-linear. However, after using the calibrated mean median speed
(calibrated by the reciprocal of population density, the reciprocal of macro-scale roadway
density, and the reciprocal of employment density), some explanatory variables based on
the SOC and SIC (especially when the mean median speed is calibrated by the reciprocal of
macro-scale roadway density as the dependent variable) are negatively correlated with the
congestion index. These occupations are sales, administrative support, and service
except protective and household. For the SIC, the industrial sectors are transportation,
finance, insurance, and real estate, health service, and armed forces. Here, the SOC
seems to be less associated with the congestion index, since the SOC is not associated with
geographic distribution, and therefore it has veiy weak relationship with the traffic
congestion. On the other hand, the SIC is more associated with roadway congestion, since
it is closely related to land use patterns and urban spatial configuration.
77


Table 4.14 The Result of Correlation Based on Congestion vs. Jobs-housing Balance
Variables
Correlations (Sheetl in Imported from C:\PhdData\Statistc\JHW-Axls) Marked correlations are significant at p < .05 N;=46 (Casewise deletion of missing data) Definition Code Mean Median Observed Speed VMT Mean Median Speed Calibrated by Population Density Mean Median Speed calibrated by Macro- Scale Roadway Density Mean Median Speed Calibrated by Employ- ment Density
1 ' "All Decimations B02..0101 01S -0 24 ijjdssi 0.29 -0 25
Executive, administrative, and managerial B02_0102 0.08 0.06 0.00 0.01 0.03
Professional specialty B02_0103 -0.11 0.02 -0.11 -0.18 -0.11
Technicians and related support B02_0104 0.00 0.17 -0.18 -0.11 -0.17
Sales B02_0105 -0.24 0.18 -0.33 -0.36 -0.33
Administrative support B02_0106 -0.25 0.15 -0.32 -0.30 -0.33
Private household B02_0107 0.01 0.16 -0.22 -0.27 -0.21
Protective service B02_0108 0.03 0.21 -0.19 -0.20 -0.18
Service except protective and household B02_0109 -0.20 0.09 -0.28 -0.37 -0.29
Farming, forestry, and fishing B02_0110 -0.12 0.15 -0.45 -0.26 -0.44
Precision production, craft, and repair B02_0111 -0.03 0.03 -0.10 -0.18 -0.10
Machine operators, assemblers, and inspectors B02_0112 0.02 0.15 -0.12 -0.11 -0.11
Transportation and material movina B02_0113 0.11 0.16 0.00 -0.09 0.01
Handlers, equipment cleaners, and laborers B02JJ114 -0.11 -0.02 -0.24 -0.28 -0.24
Armed forces B02_0115 -0.08 0.26 -0.27 -0.32 -0.27
i All ii cisi'.es B03_bl01' t01B o:Cr jf\-0.26 ' \^"-0Z7 " -0.27
Aqriculture. forestry, and fisheries B03_0102 -0.11 0.19 -0.39 -0.17 -0.38
Mining B030103 0.14 0.15 -0.06 0.01 -0.03
Construction B03_0104 -0.09 0.13 -0.18 -0.18 -0.17
Manufacturing, nondurable qoods B03_0105 -0.23 0.17 -0.29 -0.27 -0.28
Manufacturing, durable aoods B03_0106 0.37 0.01 0.00 -0.03 0.03
Transportation B03_0107 0.12 0.31 -0.27 -0.34 -0.25
Communications and other public utilities B03_0108 0.13 0.17 -0.13 -0.06 -0.10
Wholesale trade B03_0109 0.14 0.03 -0.15 -0.22 -0.12
Retail trade B03_0110 0.26 -0.04 -0.10 -0.15 -0.06
Finance, insurance, and real estate B03_0111 0.02 0.02 -0.20 -0.29 -0.18
Business and repair services B03_0112 0.19 0.01 -0.14 -0.22 -0.11
Personal services B03_0113 0.23 -0.04 -0.08 -0.08 -0.07
Entertainment and recreation services B03_0114 0.23 0.14 -0.16 -0.05 -0.12
Health services B03_0115 0.02 0.02 -0.21 -0.33 -0.20
Educational services B03_0116 0.18 -0.04 -0.06 -0.23 -0.04
Other professional and related services B03_0117 0.13 0.04 -0.15 -0.26 -0.13
Public administration B03_0118 0.20 0.08 -0.16 -0.23 -0.14
Armed forces B03_0119 0.02 0.15 -0.24 -0.35 -0.23

78


4.3.3 Industrial vs. Occupational Composition
The correlation of the coefficient of variations between the SOC and the SIC has shown that
not every industrial sector has the same occupational structure. Only the construction
industrial sector provides almost all occupational categories (except farming, forestry, and
fishing occupations), and its correlation is positive. On the other hand, mining,
manufacturing durable goods, transportation, retail trade finance, insurance and real estate,
and health service industrial sectors seem to provide fewer occupational opportunities.
79


08
i a S' s I Public administration I Other professional and related services I Educational services 1 Health services I Entertainment and recreation services! 1 Personal services! i Business and repair services! I Finance, insurance, and real estate I Retail trade! j I Communications and other public utilities! s. 9 e i 1 I Manulacturlrta. durable qoods I Manufacturinq. nondurable aoodsl I Construction! I Mininql I Aqricufture, forestry, end fisheries! i Ail Industries! Correlations (Sheetl in Imported from C:\PhdData\Statistc\JHW-A.xls) Marked correlations are significant at p < .05000 N=46 (Casewise deletion of missinq data)
IB03 0119 | |B03 0116 I IB03 0117 I IB03 0116 I IB03 0115 I IB03 0114 I IB03 0113 | IB03 0112 | |B03 0111 ! I ouo coal I BOS 0109 I [B03 0106 | |B03 0107 | IB03 0106 I IB03 0105 I |B03 0104 I I BOS 0103 | a> a u> o a to IB03 0101 1 n o o. a z u S
b 8 p u o ft p Ls B 4k Q ft o 4k n p NS B O io B la p la B B la B o k) p 03 B 1m B e> Ui p w N B 4k B S p NJ p O p B02JD102 Executive, administrative, and managerial
o K o & p Q> O Ul o io o B 4k Q) B 4k CD k a s o io a p Ns to B 4k O cn O o -o B Ik to B K O CD a fcj B kj NJ B02 0103 Professional specialty
B 'ft b Si P CM o O 03 p u> B k e B 4k B s} a s p ro B la B p B NS a y p m B la os P NS p CD B k B kj ei B02JH04 Technicians and related support
a (A Cn o y p u la B w s to p B B is B S o io a B la o> B 4k B o io Nl p <0 p CM la B e> cm a io cn o io cn p CO B02_0105 Sales
o ft o bi o k o 8 o io O io or p NJ B B la B p CD p o hi B ft o bi p on B 4k B S p ai Q K B B IO B02_0106 Administrative support
p CO O O D O CjJ p o p p a a *s| O b o I -0,04 cb b o o O Cl o b a o b B la e> B la B o ki O CD B la CO B02_0107 Private household
o in o K O o 1 -0.06 (b 03 o cn O bi 6 8 o io U) s I -0.09! O io VI a s p o p ro B la Cm o y O CO NJ b k* o KJ p p O <0 p o is O io B la p CO p M o io o a 8 p NS B p B p 03 a bt B A aj B02J109 Service except protective and household
p la ro b ft p lu U1 o cn P Ns B B ft B k a io o a io CO o io CD O io o p la In p fa La O a) O on O ki B la -O p B o io NJ B020110 Farming, forestry, and fishing
b is o s p B k CD O 4k B ft B ll o to p y B tk B B CM B is B B 8 B kj ro a o O bi p k B02_0111 Precision production, craft, and repair
p la p u> u B a N o ro a a B N> ta P la B B la B N o k> p u> p B o a o a p Vl O 03 o io B p 03 o ro B la B a y p ID O y B 4k B a 8 a 8 B 4k CM B02_0113 Transportation and material moving
p B & B is Q p o B is a io cn o y O hi p a ro o B la a y b 03 o ki o B e> e> CD g O id B CM B02JJ114 Handlers, equipment cleaners, helpers, and laborei
B B p o a rs P vl a CD O io ID 1 B la Ns p a 03 o ro p In Cm o <£> o <£} B la B 4k Cm O io B ft B Cm 00 B02_0115 Armed forces
Table 4.15 The Result of Correlation between the SOC vs. the SIC


4.4 The Results of Fixed Nonlinear and Stepwise Multiple
Regressions
Frank and Pivo found that the relationship between population and employment density
and mode choice for single-occupant vehicle (SOV), transit, and walking is nonlinear for
both work and shopping trips [emphasis added] (Frank and Pivo 1995, 44). Similar
evidence has also been found in this dissertation. To cover either linear or nonlinear models
for analyses, three types of regressions are used for this research: One is the fixed nonlinear
regression; the other is the forward stepwise multiple regression; and another is the
piecewise linear regression. The fixed nonlinear regression and the piecewise linear
regression were used to determine the relationship of pair variables, and 21 pairs were
examined. However, not all 21 pairs are statistically significant, and therefore only those
pairs with statistical significance (p<0.01 and adjusted R2>0.15) are discussed in the
following. Also, since explaining the result of the piecewise linear regression sometimes is
difficult, almost all the relationships defined in this research are based on the fixed
nonlinear regression. Later, after understanding the relationships of the 21 pairs between
the dependent and explanatory variables, the forward stepwise multiple regression is used to
define the entire relationship between the congestion index (mean median speed) with other
explanatory variables.
4.4.1 The Results of Fixed Nonlinear Regressions
The 21 pairs of the fixed nonlinear regressions are shown in the following table. Each pair
was tested though 11 possible types of relations (X2, X3, X4, X5, X0'5, SQRT(X), LN(X),
LOG(X), ex, 10x, 1/X). Only those relations with the lower p-level (< 0.01) are chosen and
listed in the following. Also, according to the STATISTICA manual (1994), the p-level of a
p value represents the following:
the probability of error that is involved in accepting the result as valid, and a p-
level of 0.05 indicates that there is a 5% probability that the relation between the
variables found in the test is a fluke. Normally, the results, which are significant, at
the p < 0.01 level are considered statistically significant, and p < 0.005 or p < 0.001
levels are highly significant (Statistica 1994, 1291-1292).
81


Therefore, the criterion of p < 0.01 is used for determining whether the following results are
statistically significant or not.
The pairs of variables with statistical significance based on 46 MSAs are the following:
mean median speed vs. the reciprocal of macro-scale roadway density (equation 4.1 and
figure 4.2), mean median speed vs. the reciprocal of residence employment density
(equation 4.2 and figure 4.3), mean median speed vs. the reciprocal of population density
(equation 4.3 and figure 4.4), transit ridership vs. employment density squared (equation 4.4
and figure 4.5), population density vs. the logarithm of transit ridership (equation 4.5 and
figure 4.6), median household income vs. the logarithm of transit ridership (equation 4.6
and figure 4.7), median household income vs. the square root of worker car ownership
(figure 4.8), calibrated mean median speed vs. the reciprocal of transit ridership (figure
4.13), and transit ridership vs. the negative of worker car ownership (equation 4.7 and
figure 4.14).
Alternative approach for testing the relationship that was statistically insignificant
Since the relationship between mean median speed with the coefficient of variation of the
jobs-housing ratio variable shows statistical insignificance based on 46 MSAs, I reran the
fixed nonlinear regressions and correlated mean median speed calibrated by macro-scale
roadway density to the coefficient of variation of the jobs-housing ratio based on 31
selected most congested MSAs. The result (figure 4.11) show that calibrated mean median
speed vs. the coefficient of variation of finance, insurance, real estate and armed forces
industrial sectors is statistically significant. Although the coefficients of variation of the
jobs-housing ratio for some industrial sectors are strongly related to calibrated mean median
speed, these relationships need to be further studied.
82


The causality between mean median speed and the coefficient of variation of the jobs-
housing ratio
To determine the causality between mean median speed (congestion index) and the
coefficient of variation of the jobs-housing ratio based on the SIC (see figure 4.9 and
figure 4.10), I compare the adjusted R2 and the p-level for each directional pair of
variables. The results of the fixed nonlinear regression (the coefficient of variation of
finance, insurance, real estate and armed forces industrial sectors vs. mean median
speed calibrated by macro-scale roadway density) seem to imply that people tend to move
closer to their workplaces if they suffer serious traffic congestion, since the value of the
adjusted R2 is higher and the p-level is lower, when the coefficient of variation of the jobs-
housing ratio is the dependent variable (compared with when the calibrated mean median
speed is the dependent variable). In other words, it seems that a greater jobs-housing
balance might result from serious roadway congestion. However, this relationship needs
to be further studied by comparing the same relationship based on two points of time
(1990s and 2000s) longitudinal study. Moreover, the reason that observed mean
median speed does not correlate with the coefficient of variation of finance, insurance,
real estate and armed forces industrial sectors probably is because of a higher cluster of
the samples (MSAs).
83


Table 4.16 The Results of the Fixed Nonlinear Regression
Dependent Variable Explanatory Variable Relationship R Adjusted R2 p-level Acceptable Page
Mean Median Speed Macro-Scale Roadway Density Reciprocal Negative 0.68 0.44 0.0000003 Yes 86
Mean Median Speed Residence Employment density Reciprocal Negative 0.67 0.44 0.0000004 Yes 87
Mean Median Speed Population Density Reciprocal Negative 0.61 0.36 0.000007 Yes 88
Transit Ridership (%) Employment density Square Positive 0.75 0.56 0.00000001 Yes 89
Population Density Transit Ridership Logarithm Positive 0.72 0.52 0.00000001 Yes 90
Mean Median Speed Median Household Income Reciprocal Negative 0.35 0.10 0.015 No 140
Median Household Income Transit Ridership Logarithm Positive 0.58 0.32 0.0006 Yes 91
Mean Median Speed Worker Car Ownership (% workers' cars per capita) Piecewise Linear 0.78 Variance explained 61% Yes, but unable to explain 141
Median Household Income Worker Car Ownership Square Root Negative 0.43 0.17 0.003 Yes 92
Median Household Income Car Ownership (% cars per capita) Piecewise Linear 0.81 Variance explained 67% Yes, but unable to explain 142
Mean Median Speed Transit Ridership Reciprocal Negative 0.39 0.13 0.0106 No 143
Mean Median Speed Urban Form Index Piecewise Linear 0.59 Variance explained 56% Yes, but unable to explain 144
Mean Median Speed Calibrated by Macro-scale Roadway Density ! Coefficient of Variation of Finance, fnsurance,;and Real ^Estate industrial Sector p' based on 46 MSAs Logarithm- Positive * 0 31 - V 0.038 .NO . V. Same,?, as 93 <
Coefficient of Variation of- Finance, Insurance, and-"' Real Estate Industrial '., Sector based on 46 MSAs -, Mean Median Speed Calibrated by Macro-scale Roadway Density > > > * y A Logarithm -Positive^ 0.39 > 0.00797 Yes Same as 93
, Mean Median Speed Calibrated by Macro-scale-, ^Roadway Density y. Coefficient of Variation of Armed ForcesTrvdustriar Sector based.on46 MSAs Logarithm Positive f '4 &A C. r - 0 37 A* ; rk*- <-0.12- ,,v 0 011 v- - Same as 94f;
Coefficient of Vanation of Armed Forces Industrial Sector based on 46 MSAs Mean Median Speed [ v. Calibrated by Macro-scale' Roadway Density >~4 Logarithm Positive life 0.001715 ,v*, t *t Yes Mean Median Speed Calibrated by Macro-scale Roadway Density Coefficient of Variation of Finance, Insurance, and Real Estate Industrial Sector based on 30 Congested MSAs Logarithm Positive 0.53 0.26 0.00195 Yes 95


Table 4.16 The Results of the Fixed Nonlinear Regression (Cont.)
Dependent Variable Explanatory Variable Relationship R Adjusted R2 p-level Acceptable Page
Mean Median Speed Calibrated by Macro-scale Roadway Density Coefficient of Variation of Armed Forces Industrial Sector based on 30 Congested MSAs Logarithm Positive 0.59 0.33 0.000457 Yes 96
Mean Median Speed Calibrated by Macro-scale Roadway Density Transit Ridership Reciprocal Negative 0.43 0.16 0.0033 Yes 97
' 1 < -Jy r* Transit Ridership Worker Car Ownership (% -. workers cars per capita) Reciprocal Negative.'- 0 92 0 81 0.00000001 Hes; '98
Transit Ridership Worker Car Ownership (% -' workers cars per capita) Linear Negative 0.90 0.84 0.00000001 > Yes Same as 98.
Mean Median Speed Coefficient of Variation of Manufacturing Durable Goods Industrial Sector Linear Positive 0.36 0.11 0.01253 No 145
Coefficient of Variation of Manufacturing Durable Goods Industrial Sector Mean Median Speed Square Positive 0.37 0.12 0.01096 No Same as 145


00
Os
Mean Median SDeed vs. Macro-scale Roadwav Density Fixed Nonlinear Regression Summary for Dependent Variable: MNAMDNSPD (Sheetl in OVarsAll.stw) R= .67560617 Ra= .45644369 Adjusted R2= .44409014 Bius4 ss | F(1,44)=36.948 (F(^O.OS)^ OB) p<.00000 SW.F-.Tor of estimate: 16.205 tylSI1 ifegjr" i&SSj p a 0.0000003; MnMdnSpd 24.8659 + 17267.8809 x (1 / RDen) :iirvii7ti/RDefl):o.67560s:o.ini47.ii7287.8ai284oeos:6.o7a5i8 jooooooo: Model is: Piecewise linear regression with breakpoint (Sheetl in OVarsAII-Stw) r iconstBO i RDEN ConstBO RDEN Breaknt Dependent variable: MNAMDNSPD Loss: Least squares Final loss: 12029.049761 R=.65888 Variance exolained: 43.412%

160.00 £ at 0) a. Bakersfielc (Std. Residua . CA 3.95)


c .2 S 0) Observed mean median Speed Predicted mean median Speed (1/Rden) A Predicted mean median Speed (Piecewise)
e n 0) El
t) '£
CL o e 0) £ 0) 0) g 40.00 isn; A 825) [land, CA a Los Angeles Long Beach, CA (Std. Residual 0.82) &
Pittsburgh, PA* (Std. Residual-1.70) Las Vegas, NV (Std. Residual -1.56| s a cfi (Std. F E3 esidual 0.65) san Jc ($9d. Res se, CA dual 0.30)0
0.00 0.
00 500 .00 1,00 R 0.00 1,50 )adway Supply (II 0.00 2,00 lacro-Roadway D 0.00 2,50 ensity (Car per leng 0.00 3,00 th) 0.00 3,500.00
Figure 4.2 Mean Median Speed vs. Macro-scale Roadway Density