University of Southern California Dissertations and Theses

Urban spatial transformation and job accessibility: spatial mismatch hypothesis revisited

(USC Thesis Other)

Urban spatial transformation and job accessibility: spatial mismatch hypothesis revisited

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

URBAN SPATIAL TRANSFORMATION AND JOB ACCESSIBILITY:
SPATIAL MISMATCH HYPOTHESIS REVISITED

by
Lingqian Hu

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POLICY, PLANNING AND DEVELOPMENT)

May 2010

Copyright 2010 Lingqian Hu

ii
DEDICATION
To my parents
Jiewei Hu and Ying Zhang

iii
ACKNOWLEDGMENTS
I would like to express my sincere gratitude to my committee chair, Prof. Genevieve Giuliano.
Her hard work and perfectionism have inspired and encouraged me throughout the completion of
this dissertation. Some people have to make a detour in their lives to find out what exactly they
want to do and how much they can accomplish, and I definitely was one of them. But no matter
where I was, Prof. Giuliano has always been there ready to lend a hand and help me out.
I have to thank many of my professors at the University of Southern California. I have been
working with Prof. Christian Redfearn for more than five years. He continually offers his
constructive critiques and encouragement which benefit me greatly. I am also grateful to Prof.
James Moore, Prof. Harry Richardson, Prof. Lisa Schweitzer, Prof. Peter Gordon, Prof. Eric
Heikkila and Prof. Gary Painter for their help at various stages of my Ph.D. study.
I am always deeply moved by the friendship and selfless help from many good friends at the
University of Southern California and the Southern California Association of Governments. I
consider myself a very lucky person to have the chances to meet so many wonderful people.
My husband, Jie (Jimmy) Wang is always there for me, especially during the last year when both
of us were under a lot of pressure with our own work. We supported and put up with each other
through the difficult times. Finally, I want to thank my parents, Jiewei Hu and Ying Zhang,
whom I owe too much to ever pay back. They probably did not expect me to go this far when
they started teaching mathematics to a toddling girl. I would like to dedicate this dissertation to
them.

iv
TABLE OF CONTENTS
DEDICATION ii
ACKNOWLEDGMENTS iii
LIST OF TABLES v
LIST OF FIGURES vi
ABSTRACT vii
CHAPTER 1: INTRODUCTION 1
1.1 Research Background 1
1.2 Research Statement 6
CHAPTER 2: LITERATURE REVIEW 10
2.1 Spatial Mismatch Hypothesis 11
2.2 Low-income Residential and Employment Locations 15
2.3 Job Accessibility 20
2.4 Job Accessibility and Labor Market Outcomes 23
2.5 Chapter Summary 30
CHAPTER 3: DATA AND METHODOLOGY 32
3.1 Data 33
3.2 Methodology 36
3.3 Chapter Summary 58
CHAPTER 4: LOW-INCOME JOB SEEKERS’ JOB ACCESSIBILITY 60
4.1 Descriptive Analysis 61
4.2 Job Accessibility Results 67
4.3 Chapter Summary 83
CHAPTER 5: IMPACT OF JOB ACCESSIBILITY 86
5.1 Descriptive Analysis 87
5.2 Labor Market Regression Model Results 95
5.3 Chapter Summary 117
CHAPTER 6: CONCLUSION AND FUTURE RESEARCH 120
6.1 Summary of Findings 120
6.2 Policy and Planning Implications 123
6.3 Research Limitation and Future Research 129
BIBLIOGRAPHY 132

v
LIST OF TABLES
Table 3.1 Job Accessibility and Control Variables 51
Table 4.1 Job Seekers and Jobs by Income Group in Each County 63
Table 4.2 T-test Results to Compare Low- and High-income Job Accessibility 74
Table 4.3 T-test Results to Compare Change of Job Accessibility between Low-
and High-income Job Seekers (2000) 82
Table 5.1 Labor Market Outcomes of Low-income Job Seekers with Distinct
Characteristics 91
Table 5.2 Descriptive Statistics of Variables in 1990 and 2000 93
Table 5.3.1 Correlation between Variables (1990) 96
Table 5.3.2 Correlation between Variables (2000) 97
Table 5.4 Estimation Results for the Full Sample 101
Table 5.5 Labor Market Outcomes of Low and High Job Accessibility
Bisections 106
Table 5.6 Effects of Job Accessibility on Labor Force Participation Rate by
Bisection of Job Accessibility and the Control Variables 111
Table 5.7 Effects of Job Accessibility on Employment Rate by Bisection of Job
Accessibility and the Control Variables 113
Table 5.8 Effects of Job Accessibility on Commute Time by Bisection of Job
Accessibility and the Control Variables 115
Table 5.9 Coefficients of Job Accessibility and T-scores to Compare the
Coefficients 117

vi
LIST OF FIGURES
Figure 3.1 Study Area 33
Figure 3.2 Job Accessibility Measure 42
Figure 4.1.1 Low-income Job Seekers 64
Figure 4.1.2 High-income Job Seekers 64
Figure 4.2.1 Low-income Jobs 64
Figure 4.2.2 High-income Jobs 64
Figure 4.3.1 Low-income Job Accessibility (2000) 69
Figure 4.3.2 High-income Job Accessibility (2000) 69
Figure 4.4.1 Low-income Job Demand (2000) 70
Figure 4.4.2 High-income Job Demand (2000) 70
Figure 4.5.1 Low-income Job Supply (2000) 71
Figure 4.5.2 High-income Job Supply (2000) 71
Figure 4.6 Differential of Job Accessibility (Low- vs. High-Income) (2000) 73
Figure 4.7.1 Change of Low-Income Job Accessibility 76
Figure 4.7.2 Change of High-Income Job Accessibility 76
Figure 4.8.1 Change of Low-Income Job Demand 78
Figure 4.8.2 Change of Low-Income Job Supply 78
Figure 4.9.1 Change of High-Income Job Demand 79
Figure 4.9.2 Change of High-Income Job Supply 79

vii
ABSTRACT
This dissertation tests whether changing urban structure has affected low-income job seekers’
labor market outcomes differentially by impacting their job accessibility. The relatively poor
labor market outcomes of minorities are well-documented in the Spatial Mismatch Hypothesis
literature which claims that the unequal labor market outcomes are partly caused by the spatial
barriers between minorities’ residences and their matching job opportunities. This research
aims to expand the demographic, geographic and temporal scopes of the Spatial Mismatch
Hypothesis by studying low-income job seekers’ job accessibility in the Los Angeles
metropolitan area in 1990 and 2000.
Using job accessibility which considers both job demand and job supply as the indicator of the
spatial barriers to competitive job opportunities, this research tests whether low-income job
seekers have lower job accessibility than the affluent majority in the polycentric urban structure,
and compares the changes of job accessibility between 1990 and 2000. Findings of this
research suggest that spatial mismatch still exists: low-income job seekers have lower job
accessibility than the affluent majority, and they are more disadvantaged in 2000 than in 1990.
However, mismatch is not only an issue in the inner cities, but also in parts of the suburbs.
Low-income job seekers are more disadvantaged in the inner-ring suburbs compared to
high-income job seekers mainly because of the lag of low-income job seekers’ residential
suburbanization.

viii
The second part of the empirical analysis is to test if spatial barriers, measured by job
accessibility, affect low-income job seekers’ labor market outcomes. This dissertation develops
a conceptual model to explain through spatial job search job accessibility affects low-income job
seekers’ labor market outcomes. Empirical results demonstrate that overall job accessibility has
positive impacts by increasing labor force participation rate and employment rate, and reducing
commute time. There are two noteworthy findings. First, the impacts are particularly
significant in census tracts with high accessibility which indicates relatively job rich areas, but
not in those with low accessibility which indicates a short supply of jobs. Second, job
accessibility tends to have larger impacts in census tracts with concentrations of disadvantaged
groups, who have limited monetary and transportation resources to overcome the spatial barriers.
This research is important for understanding the mechanism and consequences of spatial
transformation, and for facilitating planning and decision-making which address issues such as
equity in transportation investments, jobs-housing balance, and affordable housing provision.
1
CHAPTER ONE
INTRODUCTION

1.1 Research Background
This dissertation examines whether changing urban structure has affected low-income job
seekers’ labor market outcomes differentially by impacting their job accessibility. This
research is an extension of Kain’s Spatial Mismatch Hypothesis (SMH hereafter) under new
circumstances – urban spatial transformation and demographic changes. The second half of the
20
th
century has witnessed rapid suburbanization of population and employment. Polycentric
urban structure has emerged with multiple employment concentrations in the suburbs. In
addition to racial segregation, spatial segmentation by income has become increasingly evident
in metropolitan areas. Under the new circumstances, an important research question is raised:
have low-income people’s labor market outcomes been affected differentially in the evolving
urban structure?
The SMH was first developed by Kain (1968) to explain how spatial organization of employment
and population, in addition to human capital and socio-economic factors, affects minorities’
labor market outcomes. Kain’s SMH is based on observations of the changing urban structure
2
at that time: rapid suburbanization in a monocentric urban form. Two factors are of particular
importance. First, mainly because of racial discrimination in the housing market, minorities
(African Americans) had fewer residential choices and were constrained to the inner cities, while
the majority (Whites) did not face such constraints and rapidly suburbanized. Second, jobs also
suburbanized, especially those held traditionally by minority workers, such as manufacturing
jobs. These two factors combined resulted in enlarged spatial barriers between minorities’
residences in the inner cities and their job opportunities in the suburbs, and consequently affected
their labor market outcomes.
Kain’s seminal paper focuses on the geographic mismatch between African Americans’
residential and employment locations. Subsequent research found that other kinds of
mismatches contribute to minorities’ unequal labor market outcomes, including automobile
mismatch (Taylor and Ong 1995; Kawabata, 2003; Ong and Miller, 2005), skill mismatch
(Kasarda, 1985; Stoll 2005), and information mismatch (Kasinitz and Rosenberg, 1996;
Ihlanfeldt, 1999). In fact, all of these mismatch hypotheses are complementary explanations of
the enlarging spatial barriers between minorities’ residences and their matching job opportunities,
and why such spatial barriers harm minorities’ employment prospects.
Four decades after Kain’s (1968) seminal paper, many circumstances have changed. One of the
most important changes is urban structure transformation, which has significantly affected the
two premises upon which Kain’s SMH is based. On one hand, jobs have suburbanized more
3
extensively, with a greater share of jobs in the suburbs (e.g. the Los Angeles region studied by
Giuliano, et al. 2007). However, employment is not evenly distributed in the suburbs; instead,
there are multiple employment centers (Giuliano and Small, 1991; Anas, Arnott and Small,
1998). “Each of the centers in a polycentric city functions as a separate monocentric city,
producing a metropolitan area with separate urban realms or commuter sheds” (Weber, 2003,
P53). On the other hand, minorities have relatively more options in their housing locations than
before and have begun to move to the suburbs. However, the housing market is still segregated
by residents’ race and socioeconomic status (Kain, 1985; Schill and Wachter, 1995).
Particularly, spatial segregation by economic class has increased (Jargowsky, 1996).
Combining the two transformations, one may argue that suburbanization of both employment
and population has increased the proximity between minorities’ residences and job opportunities,
and thus has lessened the spatial mismatch (Gordon, Kumar and Richardson, 1989). But this is
not necessarily true. Most minorities are still concentrated in inner cities or limited to a few
places in the suburbs, while jobs are unevenly distributed with multiple concentrations.
Employment concentrations and housing segregation in a polycentric region might still create
great spatial barriers between minorities’ residences and their job opportunities. Then spatial
mismatch matters in the current urban structure.
During the last several decades, many metropolitan areas have experienced notable shifts in
demographics. Racial segregation between African Americans and Whites has been one of the
4
central issues in the U.S. Although African Americans as a whole still face great constraints in
the housing and labor market, racial segregation has been reduced (Wilson, 1980; Massey 2001).
Moreover, with the increasing size of the African American middle class, differences within
African Americans have enlarged in terms of their residential locations and socioeconomic status
(Wilson, 1987; Fischer, 2003). Furthermore, African Americans gradually have become a
relatively smaller minority group as Hispanics and Asians immigrate to major cities. The
original dichotomy of African Americans and Whites is no longer as crucial; rather,
segmentations of other racial/ethnic groups are also relevant. At the same time, poor people
become more segregated from the affluent majority over time (Massey and Eggers, 1993;
Abramson, Tobin, and VanderGoot, 1995). Economic segregation becomes increasingly
evident in the urban spatial transformation. Therefore, instead of examining racial/ethnic
minorities, this research focuses on low-income job seekers. Results of this research depict
low-income job seekers’ different labor market conditions with respect to the spatial
arrangements of low-income jobs and job seekers, and provide direct input to the planning and
policy efforts which aim to reduce poverty.
Many other factors implied in the SMH also have changed. Economic restructuring has shifted
employment from manufacturing to service jobs, which include low-skilled as well as
high-skilled professional jobs (Sassen, 1996; Muller, 2004). The shift has relatively reduced
the number of jobs traditionally held by low-income workers who usually possess lower-level
5
skill sets, and thus has exacerbated skill mismatch and impacted low-income job seekers’ labor
market outcomes (Kasarda, 1989). Mobility of job seekers has been enhanced with improving
transportation systems and increasing affordability of a car. With the improving automobile
access, low-income people are less dependent on public transit; they can travel longer distances
to reach more job opportunities. However, automobile mismatch still matters for those autoless
job seekers, most of whom are low-income.
These new circumstances – further suburbanization, changing demographics, economic
restructuring and enhanced mobility – combined may or may not have reduced the spatial
mismatch. It is not clear whether they have changed the underlying premise of the SMH –
some people face greater spatial barriers between their residences and relevant job opportunities
than others. The above-mentioned new circumstances have affected labor market outcomes for
everyone, but low-income people might be particularly disadvantaged in terms of the relative
changes in job accessibility and labor market performance compared to the affluent majority.
This research expands the SMH literature by providing a longitudinal study. Most SMH
literature focuses on cross sectional analysis of whether minorities have different labor market
outcomes at a given point in time. However, with changing circumstances, another equally, if
not more, important question raised is whether or not minorities’ job accessibility has improved.
To investigate this question, this research compares job accessibility between low- and
high-income job seekers, as well as changes in their job accessibility between 1990 and 2000 in
6
the Los Angeles metropolitan area.
This research focuses on job accessibility, a crucial factor that affects people’s labor market
outcomes. Job accessibility is a direct outcome of the spatial arrangement of jobs and
population, and it also significantly affects labor market outcomes, as job seekers with higher job
accessibility are more likely to participate in the labor force (Johnston-Anumonwo, 1996; Cooke
1997), to be employed (Ihlanfeldt and Sjoquist, 1990 and 1991; Raphael, 1998), or to have a
short commute (Kawabata and Shen, 2007), all else being equal. Although defined and
measured differently, accessibility is a crucial factor in the SMH literature. This research
applies a comprehensive job accessibility measure which captures the spatial distribution of jobs
and job seekers, as well as the travel impedance on transportation networks.

1.2 Research Statement
The original intention of the SMH is to examine whether urban spatial transformation, in
addition to social and economic factors, has contributed to the unequal labor market outcomes of
minorities. During the last four decades, urban structure has been transformed with continuous
suburbanization and decentralization of population and employment. Moreover, demographic
composition has shifted. Economic class has become a more relevant indicator of population
segments over time.
7
Under the new circumstances, I propose the research question: has the changing urban spatial
structure affected labor market outcomes for various income groups differently? Answers
to this question will demonstrate whether the SMH still matters in contemporary metropolitan
areas, and inform policy and planning what types of spatial arrangement and planning efforts can
improve or worsen low-income job seekers’ labor market outcomes.
Three hypotheses are proposed to answer the research question. First, low-income job seekers
still have lower job accessibility than the affluent majority in the polycentric urban structure.
As found in much of the empirical literature, low-income job seekers’ residential locations are
segmented and segregated; most of them are located in inner cities. Consequently, they might
face great spatial barriers between their residences and growing job opportunities in the suburbs.
Moreover, other pertinent factors might have impacted low-income job seekers’ accessibility,
such as their relatively limited mobility and the decreasing number of low-paid jobs in the
economic restructuring. This research offers an accurate estimation of low-income job seekers’
job accessibility, and compares that with the accessibility of high-income job seekers.
Second, the differential of job accessibility between income groups has increased as low-income
job seekers have decreasing job accessibility over time compared to high-income job seekers.
With the above-mentioned economic restructuring and recent immigrants, it is foreseeable that
demand/competition of low-income jobs grew faster than the relevant job supply. However, the
urban spatial transformation might have mitigated the disadvantages for some low-income job
8
seekers, especially those who live in the suburbs. Thus detailed examinations need to be done
to explore where and to what extent low-income job seekers’ job accessibility changed relative to
high-income job seekers in the polycentric urban structure.
Third, job accessibility significantly impacts low-income job seekers’ labor market outcomes,
namely labor force participation, employment status and commute time. Applying the spatial
job search theory, this research will develop labor market models to test whether job accessibility
actually affects the three labor market outcomes after their automobile ownership and
socioeconomic characteristics are controlled. Moreover, most existing research is static for
cross-sectional analysis. Fully utilizing available time-series data, this research explores
whether job accessibility has different impacts on labor market outcomes between 1990 and
2000.
Results of this research have great policy and planning implications for long-range planning,
which requires efficiency and equity considerations of capital investment and land use decisions.
Understanding the accessibility gaps between low- and high-income job seekers is an essential
step toward making land use and transportation investment plans to safeguard social equity and
environmental justice. Moreover, understanding to what extent policies to improve job
accessibility can actually benefit low-income job seekers facilitates planning to enhance social
benefits through targeting population segments and identifying their specific needs.
9
The remaining parts in the dissertation are organized in six chapters. Chapter 2 reviews
literature on the Spatial Mismatch Hypothesis, job accessibility, and how spatial barriers affect
labor market outcomes. Chapter 3 elaborates on the data and methodology applied in this
research. The following two chapters, Chapter 4 and Chapter 5, are the core of this dissertation;
Chapter 4 empirically tests Hypothesis 1 and 2, and Chapter 5 tests Hypothesis 3. Finally,
Chapter 6 concludes the dissertation with a discussion of findings, policy and planning
implications, and future research.
10
CHAPTER TWO
LITERATURE REVIEW
This chapter starts with a review of the SMH literature which explains how spatial mismatch
emerges from involuntary spatial barriers between minorities’ housing and jobs. Furthermore,
this chapter examines the related mismatch hypotheses, namely automobile mismatch, skill
mismatch and information mismatch, and explains how they are related to low-income job
seekers.
Two factors are particularly important for the SMH: housing market segregation and
employment suburbanization. In the urban spatial transformation, low-income job seekers face
residential segregation and employment suburbanization, like other disadvantaged population
segments, but with different characteristics. Specifically, spatial segmentation by class has
become more obvious with increasing economic segregation. The number and the location of
low-income job seekers’ job opportunities are affected rather adversely by economic
restructuring. This chapter reviews literature on the issues related to how low-income job
seekers are affected differently in the urban spatial transformation.
Job accessibility is a crucial indicator of the spatial barriers between job seekers and their
employment opportunities, and it directly impacts labor market outcomes (Wachs and Kumagai,
11
1973). This chapter reviews literature on the definition and measurements of job accessibility.
Finally, this chapter includes a review on the theoretical and empirical work which examines
how job accessibility affects labor market outcomes. The review will facilitate the
methodology development in Chapter 3.

2.1 Spatial Mismatch Hypothesis
The Spatial Mismatch Hypothesis was first developed by Kain (1968) to explain how spatial
distribution of jobs and population, in addition to human capital and socio-economic factors,
affects labor market outcomes. The original SMH has two important spatial factors. First,
because of racial discrimination in the housing market, minorities (African Americans) had
fewer residential choices and were constrained to the inner cities, while the majority (Whites) did
not have such constraints and rapidly suburbanized. Second, jobs also suburbanized, especially
those traditionally held by minority workers, such as manufacturing jobs. The SMH literature
argues that the two factors combined resulted in increased spatial barriers between the residences
of minorities in the inner cities and their potential job opportunities in the suburbs, consequently
affecting their labor market outcomes.
In addition to geographic mismatch, three other kinds of mismatches contribute to the unequal
labor market outcomes of different population segments. The first is automobile mismatch
12
which asserts that many of the differentials in labor market outcomes can be explained by
minorities’ lack of automobile ownership and inaccessibility to suburban jobs by public
transportation (Taylor and Ong 1995; Holzer, Ihlanfeldt and Sjoquist, 1994; Ong and Miller,
2005). The automobile mismatch hypothesis mirrors the original SMH in that if minority
people have adequate transportation resources to overcome spatial barriers, they will have better
employment outcomes. With increasing affordability of automobiles and improving transit
services, automobile mismatch has become a less significant factor. But automobile ownership
is highly correlated with income and many low-income job seekers are still autoless
1
, and their
labor market outcomes are particularly affected in places where transit services are inadequate.
However, the automobile mismatch hypothesis was challenged by the finding that jobs are
generally abundant in central cities, which are easily accessible for many minorities who live
there. Skill mismatch, which is the mismatch between low-skilled minority workers and
high-skilled jobs in inner cities, was brought into discussion. Researchers argued that because
of the changes in the urban and economic structure, jobs traditionally held by low-skilled
minority workers have been reduced; new and remaining jobs in central cities normally require
higher-level skill sets which minority workers often do not possess (Kasarda, 1985; Stoll 2005).
However, this hypothesis is mitigated by the finding that there are still many low-skilled jobs in
the inner cities, although they are paid less than the original production jobs (Wyly, 1990). The

1
Around 28% of low-income job seekers are autoless in the Los Angeles metropolitan area in 1990 and 2000, but
only 3% of high-income job seekers are autoless.

13
skill mismatch hypothesis is a specification of the SMH as spatial barriers between minority job
seekers and their matching job opportunities remain longer than that for other job seekers. To
address the skill mismatch hypothesis, this research differentiates jobs by workers’ income level,
and calculates job accessibility for low- and high-income job seekers respectively.
The automobile mismatch hypothesis and the skill mismatch hypothesis contribute to the
understanding of minorities’ labor market outcomes, but they do not explain many cases where
people cannot obtain matching jobs in their own neighborhoods (Kasinitz and Rosenberg, 1996).
Clearly, there must be other explanations, and social network theory provides one. Wilson
(1987) argued that people living in underclass neighborhoods have poor information about job
opportunities because they lack social contacts with individuals and/or institutions of the
mainstream society. The lack of access to information on relevant jobs adversely affects
minorities’ labor market outcomes (Ihlanfeldt, 1997). However, in empirical research, social
networks and information accessibility are very difficult to quantify. Some researchers (e.g.
O’Regan and Quigley, 1991; Parks, 2004) used employment status of other adults in the
household as a proxy for social networks. Meanwhile, Ihlanfeldt and Sjoquist (1990) suggested
that information on available job opportunities may decay rapidly with distance from one’s home
or neighborhood. This research controls for the effects of information mismatch in two aspects.
First, since information exchange has certain spatial characteristics, the job accessibility measure
that weights job opportunities by travel impedance partly captures information accessibility.
14
Second, social networks are not dramatically different within an income group, if job
accessibility and certain socioeconomic characteristics, especially race/ethnicity composition, are
controlled for. Household income can be used as a proxy for social networks as job seekers in
low-income households have relatively similar social networks and information about jobs.
The impacts of social networks are not addressed directly in this research, but are controlled to a
degree.
Most literature in the SMH literature focuses on African Americans because of their distinct
disadvantages in the labor market. Recently, the SMH literature has been extended to include
welfare recipients (Ong and Blumenberg, 1998; Blumenberg and Manville, 2004), Latino (Stoll,
2005; Ihlanfeldt, 1993), immigrants (Painter, et al. 2007), women (Kasarda and Ting 1996;
Johnston-Anumonwo, 1997), low-educated workers (Stoll, 2005), and low-skilled workers
(Kawabata, 2003). This research argues that low-income job seekers, another disadvantaged
group in terms of class, have also been affected differently in the urban spatial transformation.
This research focuses on income and aims to examine the differences between income classes.
Other individual attributes, such as education and skill levels which also contribute to income,
are examined in the SMH literature. The interrelationship between education, skills and income
are complicated, as education and skills significantly affect workers’ wage and earnings, which
compose a large portion of income. However, people might have other income, such as welfare
benefits and profits of investment, which combined with earnings influence their residential
15
locations and labor market behavior. Using only education and skills cannot fully capture those
income effects. Moreover, family/household income might be a better indicator of social
networks or social capital. It has been found that education and skills cannot fully explain the
younger generation’s earnings. Family or household income contributes to children’s earnings
through peer influence or role modeling of family members (Corcoran, et al. 1991). Therefore,
this research uses income instead of other individual attributes.
Results of the empirical tests of the SMH are mixed, partly because of the wide variety of
methods used in the literature. Three major methodological approaches are summarized by
Ihlanfeldt and Sjoquist (1998): 1) racial comparisons of commuting times or distances; 2)
correlations of wages, employment, or labor force participation with measures of job
accessibility; and 3) comparisons of the labor market outcomes of central city and suburban
residents (P852). This research falls in the second category of Ihlanfeldt and Sjoquist’s
classification. The empirical research in this category is reviewed in a later part of this chapter.

2.2 Low-income Residential and Employment Locations
This section reviews literature that addresses low-income job seekers’ residential locations and
their employment locations. In general, low-income job seekers also face residential
segregation and employment suburbanization. They are the same, if not more, depressed as
16
other disadvantaged groups in the urban spatial transformation.

Residential Segregation of Low-income Job Seekers
In classic residential location theories, residential location is determined primarily by a tradeoff
between commuting cost and land cost (Mills, 1972; Anas, 1982). Other factors affecting
residential location choices include people’s preferences, accessibility to other facilities and
amenities, as well as quality of life (McFadden, 1977). In addition to general residential
locations, residential segregation is of particular importance because it is one of the two major
spatial factors of the SMH.
The original intention of Kain’s SMH is to explain African Americans’ unequal labor market
outcomes, and racial segregation has been the focus. African Americans’ residential
segregation is well documented in the literature (e.g. Kain and Quigley, 1975). However,
recent literature provides much empirical evidence of declining segregation by race (e.g. Miller
and Quigley, 1990; Massey and Eggers 1993; Massey 2001; Massey and Denton 1993; Glaeser
and Vigdor, 2001), although to different extents.
Granted that race and class are interrelated, but class has become an increasingly important
factor in residential segregation. Over time, economic segregation has increased in contrast to
17
the decrease in racial segregation (Jargowsky, 1996; Pastor, 2001; Fischer, 2003). Low-income
people face great housing market discrimination as land-use regulations, especially exclusionary
zoning, leave them very few options regardless of their race or ethnicity. Economic reasons
also play significant roles in economic segregation mainly because low-income job seekers
cannot afford to live in many places where combined transportation and housing costs are high.
This is especially true for low-income job seekers who do not have access to automobiles and
have to live close to jobs and public transit. Furthermore, like other people, low-income people
prefer to live in neighborhoods of certain demographic composition or with certain amenities.
Most of the existing low-income neighborhoods are located in inner cities, which generally
further attract more low-income people.
Kain observed the difference of racial composition between inner cities and suburbs in his
seminal paper. Similarly, income disparities between cities and suburbs have increased.
“Cities and suburbs were more differentiated with respect to the economic variables – poverty
rates, per capita income, and unemployment rate – in 1990 than they had been in 1970” (Pack,
2002, P81-82). In the suburbanization process, low-income people are generally segregated
and concentrated in inner cities while high-income people are more likely to locate in the
suburbs. But a certain amount of low-income people still live in the suburbs. In 1990, 6.9
percent of the suburban population of Metropolitan Statistical Areas (MSAs) lived in poverty
(Glaeser, Kahn and Rappaport, 2000); that poverty rate increased to 7.5 percent in 2000 (Glaeser,
18
Kahn and Rappaport, 2008).
Overall, economic segregation has become more evident, suggesting the necessity for a study
that examines low-income people’s residential locations and consequently their job accessibility.

Low-income Jobs in Economic Restructuring
Four decades after Kain’s original SMH was proposed, the general trend of employment
suburbanization has not changed. With decreasing transportation and information costs, as well
as low land price and low wage rate in the suburbs, a greater share of jobs are located in the
suburbs (e.g. the Los Angeles region, studied by Giuliano, et al. 2007). But employment is not
evenly distributed in the suburbs; instead, there are multiple employment concentrations
(Giuliano and Small, 1991; Anas, et al. 1998). Agglomeration economies, proximity to
transportation nodes, existing built capital stock and physical infrastructure contribute to those
concentrations. Overall, traditional monocentric urban structure is gradually disappearing, and
polycentric urban structure has emerged with employment suburbanization and concentration
(e.g. Muller, 1981; Castells and Hall, 1994; Kloosterman and Musterd, 2001). “Each of the
centers in a polycentric city functions as a separate monocentric city, producing a metropolitan
area with separate urban realms or commuter sheds” (Weber, 2003, P53). Moreover, each
employment concentration has a distinct industry composition (Redfearn, 2006), and hence
19
provides a different supply of low-income jobs, which affects low-income job seekers’ labor
market outcomes.
An important factor related to employment distribution and the SMH is economic restructuring.
During the last several decades, cities have changed from the centers of production and
distribution of material goods to centers of service provisions (Kasarda, 1985; Wilson, 1980).
Compared to the conventional industries, many emerging new industries require higher skill-sets,
which low-income job seekers do not usually acquire. Furthermore, a reduction of
manufacturing jobs triggers an overall loss of job opportunities for less-educated, low-income
job seekers (Kasarda, 1989). Economic restructuring is an external cause of the increasing skill
mismatch.
Spatial distribution patterns of the new industries are different from those of the old ones.
Some of the new industries, especially those skill-intensive ones, such as FIRE (Finance,
Insurance, Real Estate) and information industries, value more of agglomeration economies
(Glaeser and Kahn, 2001; Giuliano, et al. 2007), because they need shared inputs of production
and consumption, and of specialized workers and human capital (Quigley, 1998). These types
of industries are more likely to form concentrations and clusters. Other new industries
encourage suburbanization and dispersion, as service-oriented industries tend to be dispersed
(Gordon and Richardson, 1996). On the other hand, the declining industries, especially
manufacturing, originally were located in inner cities. In addition to the shrinking total number
20
of manufacturing jobs, those jobs are further reduced from inner cities as decreasing
transportation cost and low land price in the suburbs encourage suburbanization of existing
manufacturing firms. Generally, economic restructuring affects low-income job seekers in two
ways: reducing the number of low-income jobs available, and shifting their locations.
It is also important to examine marginal changes of low-income jobs in addition to the
contemporary numbers of job opportunities. While significant concentrations of low-income
jobs in inner cities still exist, they might not be sufficient to meet the demand of the concentrated
low-income job seekers in inner cities. Moreover, continuing employment suburbanization
might have long-term impacts on low-income job seekers’ job accessibility, and further worsen
their labor market prospects. Whether low-income job supply meets job demand and whether
the redistribution and reallocation of low-income job seekers are synchronous with those of
low-income jobs in the spatial transformation are the critical questions that this research attempts
to answer.

2.3 Job Accessibility
In the urban spatial transformation, spatial barriers between low-income job seekers’ residences
and their job opportunities might have changed. Job accessibility is the key indicator of the
spatial barriers.
21
No matter what measurements are used, accessibility is always implicit or explicit in the SMH
literature since proximity implies higher levels of accessibility and distance implies lower levels
of accessibility (Perle, Bauder and Beckett, 2002). Handy and Niemeier (1997) provided a
sensible definition: “(a)ccessibility is determined by the spatial distribution of potential
destinations, the ease of reaching each destination, and the magnitude, quality, and character of
the activities found there” (P1175). Accessibility is place-based because it is highly related to
land use patterns and geographic locations of people and activities. However, different people
at the same location might have different accessibility. Available transportation resources are
important because the more resources a person has, the more activities that person can reach and
thus the greater his/her accessibility. Low-income job seekers generally have fewer resources
for travel. However, because of their limited resources, low-income job seekers have stronger
intentions to live close to different activities at the cost of enduring overcrowding and congestion
(Cervero, 1991). In this case, their accessibility might be high.
Many different measures of job accessibility are applied in the SMH literature. The simplest
proxies often applied are commute time and commute distance (e.g. in Ellwood, 1986; Ihlanfeldt
and Sjoquist, 1990). However, commuting patterns are a result of labor market outcomes and
limited to employed workers, and thus fail to include many unemployed job seekers. Commute
time is also affected by commute mode, as commute time is generally longer by public transit
than by private automobiles. Moreover, commuting behavior is endogenous with many other
22
factors. Workers, especially high-income workers, might trade longer commutes for a higher
wage and salary rate, or for spacious housing and amenities. Therefore, commuting time or
distance is not a sound indicator of job accessibility.
Other job accessibility indicators used in the SMH literature include job-worker ratio (Ellwood,
1986) and the number of jobs within a certain distance (Blumenberg and Ong, 1998). These
measures treat jobs within an arbitrary boundary equally, without considering the distance decay
of job opportunities. Gravity-based job accessibility measures are widely applied in the
literature (e.g. Raphael, 1998; Cervero, Sandoval, and Landis, 2002; Kawabata, 2003; Kawabata
and Shen, 2007), and a similar measure of gravity-based jobs-housing balance has also been
developed (e.g. Horner and Mefford, 2007). Following Shen (1998), this research applies a
relative gravity-based job accessibility model.
Accessibility is a measure of the ease of reaching potential destinations, in this research, relative
job opportunities. It is important to distinguish relative job opportunities which count both
supply and demand of job opportunities. With overall economic development, the absolute
number of job opportunities has increased. If only considered in terms of the absolute number
of job opportunities, low-income job seekers’ job accessibility would rise, although competition
for those jobs has increased also as a result of population growth. Therefore, a relative
accessibility measure weighting job supply by job demand is used. Another reason to include
job demand in the accessibility measure is that job supply is endogenous with job demand.
23
Labor force size changes with improvements in the supply side of job accessibility (Berechman
and Paaswell, 2001). A growing labor force creates high job demand and intensifies
competition. Therefore, places with high supply-side job accessibility do not necessarily
benefit all job seekers. Considering demand and supply of jobs at the same time addresses such
demand-supply endogeneity. Shen’s (1998) relative gravity-based model, which incorporates
job supply and competition, is applied in this research.
Accessibility is an important indicator of social development, as access to employment and urban
services constitutes an important measure of the quality of life (Wachs and Kumagai, 1973).
Since Ellwood’s (1986) early attempt, much research has quantitatively tested whether job
accessibility contributes to minorities’ labor market outcomes. The following section reviews
the theoretical and empirical literature which addresses this specific issue.

2.4 Job Accessibility and Labor Market Outcomes
Job accessibility measures spatial barriers. This section reviews the theoretical literature which
explains why and how spatial barriers affect labor market outcomes, as well as the empirical
literature which examines to what extent job accessibility affects labor market outcomes.

24
Spatial Job Search Theory
The theoretical exploration that has emerged to explain how accessibility affects job seekers’
labor market outcomes is job search theory, which has been studied since the 1960’s. Until
recently, job search theory had not been extended to incorporate spatial considerations in
searching behavior, nor been used to explain how spatial barriers affect people’s labor market
outcomes. In a theoretical overview on the mechanism of the spatial mismatch by Gobillon,
Selod and Zenou (2007), job seekers and potential employers’ spatial job search behaviors are
the major explanation.
From the employers’ perspective, they are more inclined to hire workers who live nearby
because of low commute compensation, all else being equal. Furthermore, employers might
estimate identical workers’ productivity based on their residential locations, with the assumption
that if workers commute longer distances, they are more stressed and less productive (Gobillon,
et al. 2007; Zenou, 2006). Zenou (2006) further pointed out that “this is particularly true in
large US Metropolitan Statistical Areas, where there is a lack of good public transportation,
especially from the central city to the suburbs” (P437). Los Angeles apparently is one of those
metropolitan areas. Moreover, employers’ job search ranges partly depend on the positions’
degree of specialization. Usually jobs requiring higher skill-sets entail more extensive spatial
searches (Simpson, 1992), and jobs requiring low skill-sets, mostly low-income jobs, entail much
smaller search ranges.
25
From job seekers’ perspective, job search arises when workers receive local wage offers which
they expect to be able to improve elsewhere. Job search costs consist mostly of transportation
costs from their residences to potential job sites (Simpson, 1992). Therefore, the spatial extent
of job search is closely related to the expected wage offer and the willingness to pay for the job
search. Low-income job seekers tend to have smaller search radii because their expected wage
offers are low, and they are less likely to afford extensive searches of distant job opportunities.
Information on potential job opportunities also significantly impacts the range of spatial job
search. The cost to acquire job opening information is important for job seekers, especially
low-income job seekers, who rely heavily on informal search methods for obtaining employment
(Holzer, 1987). Some researchers (e.g. Simpson, 1992) argued that information cost has
declining significance relative to screening costs because of technological developments in
communications. But low-income job seekers do not have equal access to the information and
communication technology (Goslee and Conte, 1998); they still heavily rely on information
exchange in social networks. At the same time, information accessibility also has spatial
characteristics, as suggested by Ihlanfeldt and Sjoquist (1990) that information on available job
opportunities decays with distance. The job accessibility measure applied in this research
captures the declining weight of job opportunities by distance, and in a way reflects the declining
availability of job information.
In job search behavior, employer discrimination sometimes plays an important role, particularly
26
against racial/ethnic minorities (Gobillon, et al. 2007). Race discrimination in labor markets
has become more complicated because of the class stratification within race/ethnicity groups
(Wilson, 1987). Focusing on low-income job seekers, this research does not intend to examine
labor market discrimination.
Job search range is highly correlated with spatial barriers because they impose searching cost
from one’s residence (Wasmer and Zenou, 2002). More extensive spatial job search leads to
better employment and wage offers (Schwartz, 1976; Stoll, 1999). Moreover, spatial barriers
impact job seekers’ other labor market behavior; they may refuse a job that it too far away
(Gobillon, et al. 2007; Brueckner and Martin, 1997), or quit the job more frequently (Gobillon, et
al. 2007; Zax and Kain, 1996).
Spatial job search theory explains through what mechanism spatial barriers affect labor market
outcomes. The following section reviews the empirical research which tests to what extent
spatial barriers, measured by job accessibility, have actual impacts.

Empirical Results
With limited data and technologies, it is very unlikely for researchers to build an equilibrium
model to fully capture the interrelationships between residential and employment locations,
27
commute, labor market outcomes and other factors. Most empirical research in the SMH
literature assumes fixed residential and employment location, and focuses on the effects of job
accessibility on labor market outcomes at one point in time.
In the SMH literature, employment rate (or probability of employment) and commute time are
the most commonly used indicators of labor market outcomes. Another important outcome is
labor force participation rate which represents low-income job seekers’ willingness to work.
Decisions as to participate in the labor force, no matter employed or not, are partly influenced by
the spatial arrangement of population and employment, and eventually affect job seekers’
socioeconomic status. Researchers have examined labor force participation of women (e.g.
Johnston-Anumonwo, 1996; Thompson, 1997; Cooke, 1997), and recently of immigrants
(Painter, Liu and Zhuang, 2007).
The empirical results are mixed. Some research found significant impacts of job accessibility
on labor market outcomes (Ihlanfeldt and Sjoquist, 1990, 1991; O’Regan and Quigley, 1991;
Raphael, 1998; Blumenberg and Ong, 1998; Kawabata, 2003, Painter, et al. 2007); others found
at most modest impacts (Ellwood, 1986; Cervero, Sandoval, and Landis, 2002). Job
accessibility measures and model specifications vary greatly in the empirical research.
First, the unit of analysis can be grouped into two categories, individual or aggregated. The
most commonly used individual level data are the PUMS (Public Use Microdata Sample) data
28
(e.g. in Ihlanfeldt and Sjoquist, 1990, 1991; McLafferty and Preston, 1992; Kawabata, 2003),
while others are based on other surveys (e.g. in Cervero, Sandoval, and Landis, 2002). Most
individual level data do not have spatial information at small geographic levels. On the other
hand, aggregated data commonly used are at the census tract level (e.g. in Ellwood, 1986;
Blumenberg and Ong, 1998), or the traffic analysis zone (TAZ) level (e.g. in Raphael, 1998;
Kawabata and Shen, 2007). A tradeoff has to be made when deciding on the unit of analysis.
Individual level data provide more accurate information on individual and household
characteristics, while aggregated data at a small geographic level provide more detailed
information on neighborhood and spatial characteristics. This research is more interested in
spatial details in a polycentric metropolitan area; therefore, the census tract is used as the unit of
analysis.
Second, measures of job accessibility vary, including commute time (e.g. Ellwood, 1986;
Ihlanfeldt and Sjoquist, 1990 and 1991), jobs to workers ratio (e.g. Ellwood, 1986; Blumenberg
and Ong, 1998; Ong and Blumenberg, 1998), and recent gravity-based job accessibility measures
(e.g. Raphael, 1998; Cervero, Sandoval, and Landis, 2002; Parks, 2004). However, competition
of jobs is also important for labor market outcomes. Some research included competing supply
as an explanatory variable in the models (e.g. Ralph, 1998; Painter, Liu and Zhuang, 2007).
This research applies the job accessibility measure which directly weights job supply by
competing demand. This measure was first developed by Shen (1998), and later was applied in
29
Kawabata (2003) and Kawabata and Shen (2007).
Third, theory and empirical work are integrated to various entents, and much of the SMH
literature is empirical analysis without theoretical models. Recently emerged spatial job search
theory is still in the improvement and refinement stage, let alone being applied to the empirical
work. Nevertheless, two important implications can be summarized from the theory. First, no
matter what the mechanism is, spatial barriers affect labor market outcomes. The effects are
expected to be particularly stronger for disadvantaged groups because of the small job search
ranges of them and their potential employers. Second, in addition to employment status and
commute time, wage is a crucial factor in both employers’ and employees’ labor market
decisions. Some research (e.g. McLafferty and Preston, 1992; Rogers, 1997) examined wage or
earnings, other research (e.g. Ellwood, 1986, Ihlanfeldt and Sjoquist, 1990, 1991; Raphael, 1998;
Kawabata, 2003) controlled for economic status by using household/family income or poverty
status, and implicitly used proxies for wage rates. Similarly, examining low-income job seekers,
this research controls wage rates to a certain degree.
In addition to space, other factors affect labor market prospects and lead to various empirical
results. Ellwood (1986) made a strong case which argues that race, instead of job accessibility,
impacts labor market outcomes. Ellwood’s study initiated the race versus space debate, and
was supported by other research (e.g. Harrison, 1972; Taylor and Ong, 1995). Recognizing that
racial difference in income might cause variation in residential location and job accessibility, this
30
research examines the effects of job accessibility on low-income job seekers’ labor market
outcomes while controlling race/ethnicity composition. Results of the empirical analysis will
contribute to the race versus space debate.

2.5 Chapter Summary
Recent research has shifted the focus more or less from African Americans to other demographic
groups, but has not yet addressed differences between classes. Detailed review of the literature
demonstrates that low-income people are disadvantaged in terms of residential segregation, as
well as the number and distribution of relevant job opportunities. Therefore, it is important to
examine their job accessibility.
Many different measures are developed to calculate job accessibility. This research adopts
Shen’s (1998) method which counts both job demand and supply. This job accessibility
measure indicates spatial barriers to relative job opportunities since higher job accessibility
means smaller barriers to competitive jobs. Moreover, the job accessibility measure can be an
indicator of spatial mismatch since it denotes whether the number and the location of jobs and
job seekers match each other. Variation of job accessibility can be directly shown across a
polycentric region; it helps to understand job accessibility differentials between inner cities and
suburbs, as well as within suburbs.
31
An important reason to focus on job accessibility is because of its expected impacts on labor
market outcomes. Particularly, recently emerged spatial job search theory provides theoretical
support for the significant impacts of job accessibility. However, the empirical results are
mixed. Learning from the existing literature, this research attempts to incorporate the theory
with the existing data. Moreover, targeting low-income job seekers, this research explores to
what extent spatial barriers matter for people within a certain class, and tries to fully describe
labor market performance of low-income job seekers by studying the three related but different
labor market outcomes: labor force participation, employment status and commute time.

32
CHAPTER THREE
DATA AND METHODOLOGY

To answer the research question, this dissertation will compare low- and high-income job
seekers’ job accessibility, and examine the impacts of job accessibility on labor market outcomes.
Chapter 3 first presents the datasets used in this dissertation, as well as the methods to identify
low- and high-income job seekers and jobs, which build the foundation for the empirical analysis.
Second, this chapter lays out the relative gravity-based job accessibility model which counts both
job demand and job supply for low- and high-income job seekers, respectively. Finally, this
chapter explains the labor market regression models to test the effects of job accessibility on
low-income job seekers’ labor market outcomes.
The Los Angeles metropolitan area, which includes the counties of Los Angeles, Orange,
Riverside, San Bernardino and Ventura, is chosen as the case study area. The study area is
restricted to the urbanized portion of the Consolidated Metropolitan Statistical Area (CMSA) in
the year 2000 to eliminate the vast tracts of mountains and deserts with little or no employment
or population. The geographic unit of analysis is the census tract. Figure 3.1 shows the study
area and the five-county region.

33
Figure 3.1 Study Area

3.1 Data
1990 and 2000 census tract level demographic data come from the U.S. Census Summary Tape
File 1 (STF1) which contains data from the 100% sample questionnaires used in the decennial
census, and the Summary Tape File 3 (STF3) which is generated from population sample of 1 in
6 households. The census tract level employment data are provided by the Southern California
Association of Governments (SCAG), the region’s Metropolitan Planning Organization (MPO).
The employment data are developed from wage and compensation data reported to the State
Economic Development Department (EDD) of the California Labor and Workforce
34
Development Agency. The employment data are classified by the 2-digit Standard Industry
Classification (SIC) codes. To maintain a consistent geography across the two analysis years,
2000 data are converted to the 1990 census tract geography. As a result, there are total 2493
census tracts in the study area.
The census tract is the basic unit of analysis in this research, but the census tract level STF1 and
STF3 data do not allow specifying socioeconomic characteristics of individuals or households
across different income groups, such as age and job types which are necessary for identifying job
seekers and jobs by income group. Therefore, the 5% Public Use Microdata Sample (PUMS)
data in both 1990 and 2000 are used to summarize those characteristics by income at the Public
Use Microdata Area (PUMA) level. The summarized characteristics by income are applied to
the census tract level aggregated data, based on the assumption that individuals or households in
a certain income group within a PUMA have similar characteristics. This practice is often used
in data preparation for transportation modeling purpose (Beckman, Baggerly and McKay, 1996).
There are 92 PUMAs in 1990 and 110 PUMAs in 2000 in the Los Angeles Metropolitan Area.
Using ArcGIS, the corresponding tables between the census tracts and the PUMAs are generated
in 1990 and 2000, respectively.
1990 and 2000 roadway network data are provided by SCAG. Network observed travel time is
used as travel impedance in the job accessibility measure. Network observed travel time
reflects real travel time using speed data of freeways and various levels of arterials in the region.
35
Sensitivity tests to compare the results of job accessibility using the observed travel time and
those using the AM peak travel time and the free-flow travel time in 2000 were performed. The
results were not significantly different with less than 5% variations for most census tracts.
It is noteworthy that roadway networks are used primarily by private automobiles. Available
1990 and 2000 transit networks are not fully calibrated, and do not reflect accurate transit
network travel time, which needs considerations such factors as walk/drive time to transit stops,
waiting time, and level of service for each transit route. Therefore, this research cannot
differentiate travel impedance of private automobile trips and public transit trips, although
recognizing that around 28% of low-income job seekers do not have an automobile in their
household
2
. Some levels of bias might be introduced, but it is not clear how significant the bias
is. Around 40-50% of autoless low-income workers still commute by private automobiles
3
.
Moreover, most autoless low-income job seekers are located in the inner cities, where transit
services are relatively better, and roadway networks in many ways reflect transit level of services
proportionally. Kawabata (2009) found that the disparity of job accessibility between workers
who commute by cars and those commute by public transit was relatively low in central areas but
relatively high in suburban areas, and such disparities were lessened in inner cities over time.
Job accessibility by transit will be examined in future research contingent on data availability.

2.

3.
Summarized from the PUMS data.

36
3.2 Methodology
Low-income and High-income Job Seekers
Job seekers are defined as potential workers of working age, 16 to 64 years old. Individuals
within this age range compose most of labor force and employment; summary of the PUMS data
shows that around 97% of labor force and of employment are within this age range in the Los
Angeles region in both 1990 and 2000. All working age people are included because their
intention to participate in the labor force and their employment outcomes are affected by spatial
organizations of jobs and housing.
Without the census tract level detailed personal income information available, household
income
4
is used as a proxy to define low-income and high-income job seekers. Households
below the 25
th
percentile household income of the region are defined as low-income households,
and people 16 to 64 years old in such households are low-income job seekers. Similarly, people
16 to 64 years old in households above the 75
th
percentile are defined as high-income job
seekers.
This definition is a compromise of data constraints and the research goal. The intention of this
research is to study job seekers by their income class in different locations, and thus detailed
information on individual income in small geographic units is ideal. However, such individual

4
In census, household income is defined as the total money income of all household member age 15+ during the
previous year.
37
data are not available. The decennial census provides the aggregated number of families or
households in certain income brackets. But “family” in census is defined as a group of two or
more related people, thus single-person households are not included. Potential job seekers in
those households are inappropriately excluded if the census tract “family” or “family income”
data are used. Therefore, household income is the most feasible income indicator available for
the purpose of this research.
Official definitions of “low-income”
5
by the Department of Housing and Urban Development
(HUD) and “poverty”
6
by the U.S. Census are based on family income and family size. In
addition to the exclusion of single-person households, another major obstacle to using those
definitions is that the Census does not provide family income by family size at the census tract
level; therefore, this research cannot follow those definitions exactly. Nevertheless, sensitivity
tests show that the 25
th
percentile household income, which is used to define low-income job
seekers, is very similar with the benchmarks to define very-low income family for the 4-person
family type by HUD in both 1990 and 2000. Granted that household income is not a very
accurate proxy for individual income, but it is a major indicator of family/household
characteristics and is directly associated with job seekers’ labor market outcomes.

5
US Department of Housing and Urban Development (HUD) defines low-income families as those with adjusted
incomes at or below 80 percent of the median income in their areas, with adjustments for smaller and larger families.
Very low-income families are those with adjusted incomes at or below 50 percent of their areas' median income.
6
Poverty thresholds in Census are defined according to size of the family and ages of the members.

38
To calculate the number of potential job seekers by income group for each census tract, the
census tract level number of households in each income quartile must be calculated first using
the Census STF3 data. Households with income below $19,800 in 1990 and/or $23,700 in 2000
are in the lowest income quartile, and thus are defined as low-income households. Households
with income greater than $61,200 in 1990 and/or $81,000 in 2000 are high-income households.
In addition, the total number of potential job seekers (persons 16-64 years old) for each census
tract can be generated from the number of persons in each age group from the Census STF1 data.
Next, the PUMS data are used to calculate the percentages of potential job seekers (persons 16 to
64 years old) in all four income groups for each PUMA. By applying the percentages to the
total number of job seekers for a specific PUMA aggregated from the census tracts level data, the
numbers of job seekers by income for that PUMA can be calculated. For each income group,
multiplying the total number of job seekers in the PUMA with the percentages of households in
each census tract within that PUMA, the tract level potential job seekers in that income group
can be calculated. This step assumes that within a PUMA, households in a certain income
group have a similar number of job seekers per household. Next, applying the Iterative
Proportional Fitting (IPF)
7
procedure, the numbers of job seekers are readjusted to fit both total
potential job seekers for each census tract and total potential job seekers for each income group

7
IPF is a mathematical procedure originally developed by Deming and Stephan (1940) as a way to make an un-biased
guess at individual cells of data, and the data in each cell converged to fit in marginal (row and column) totals.

39
in a PUMA. This practice is repeated for all 92 PUMAs in 1990 and 110 PUMAs in 2000 to
get the numbers of job seekers by income for all census tracts.

Low-income and High-income Jobs
Different industries tend to have different composition of labor force by income. For example,
manufacturing industry have higher share of jobs suitable for low-income workers than FIRE
(Finance, Insurance and Real Estate). This research first classifies eleven major industries for
both the census tract level data and the PUMS data. Then, for each industry, the PUMS data
are used to calculate the percentages of labor force in each income group within a PUMA. The
percentages are applied to the census tract level employment in that industry. Finally, the total
number of jobs suitable for each income group in that PUMA is calculated. Again, this practice
is repeated for all PUMAs in 1990 and 2000.
Others methods are also applicable to identify relevant job opportunities for each income group.
Occupation is a better indicator of income or wage rate, but the census tract level occupation data
are not available. Other indicators of job opportunities include new job openings (Shen, 2001)
and employment growth (Rogers, 1997; Painter, et al. 2007). However, those indicators are not
as relevant as the total number of jobs for potential job seekers, some of whom are not actively
looking for jobs. Moreover, the turnover rate (quits, layoffs, or discharges) of low-income jobs
40
tend to be higher (Doeringer, 1968; Simpson, 1992), therefore overall job supply is more
pertinent to this research.

Job Accessibility
Chapter 2 reviews many different types of job accessibility measures. This research applies the
measure developed by Shen (1998) which captures the two most important aspects of job
accessibility: travel impedance and spatial distribution of jobs, while considering spatial
distribution of job seekers as well. Different from Shen (1998) who applied his model to
low-wage workers, this research studies low- and high-income job seekers and compares their
job accessibility scores. More importantly, this research conducts comparative analysis over
time, and contributes to the understanding of the impacts of changing urban structure.
Furthermore, Shen used labor force (employed and unemployed workers) as his study subject,
while this research uses potential job seekers, no matter if they are in the labor force or not,
which is more pertinent to the SMH literature as many job seekers drop out of labor force
because of the spatial mismatch.
Travel Impedance
This research uses observed travel time on the roadway network as the travel impedance.
41
Census tract-to-census tract travel time matrix is calculated following the method in Redfearn, et
al. (2008). First, this research creates geographic centroids of all census tracts, and then links
the centroids to the existing nodes on the network through centroid connectors. Assuming the
average speed on those artificial links is 35 miles per hour (MPH), travel time from the centroid
of each census tract to the network can be calculated, and so can travel time between each pair of
census tracts based on their centroids.
Job Accessibility Model
Figure 3.2 visually depicts the concept of the job accessibility measure. Income group m job
seekers in census tract i can reach job opportunities in census tract j which is in i’s labor market
with certain travel impedance. However, since all job seekers in income group m within a
reasonable travel distance or time can potentially compete for the same jobs, job opportunities in
census tract j first have to be weighted by job demand/competition from income group m in
census tract k, which is in j’s labor market. Aggregating the demand/competition weighted job
supply in tract j by its travel impedance to tract i, a more accurate job accessibility can be
calculated for job seekers in tract i.
The labor market is defined by using commute shed at the 50
th
percentile of the actual commute
time in 1990 and 2000
8
. Extracted from the PUMS data, the commute shed was 30 minutes in

8
Sensitivity tests show that job accessibility results calculated using the 75
th
percentile of commute time, which is 60
minutes in 1990 and 2000, are not very different from those using the 50
th
percentile, because jobs and job seekers are
weighed down significantly by long commute time.

42
1990 and 35 minutes in 2000 for all automobile commute trips. It is recognized that the actual
commute time of low-income job seekers is different from that of high-income job seekers, but
the difference is partially caused by their differential job accessibility. To avoid such
endogeneity, this research uses the time distribution of all commute trips.

The following are the steps to calculate job accessibility. The first step is to calculate potential
job demand/competition around census tract j from relevant job seekers in census tract k within
j’s labor market. Equation (1) is a gravity-based model which uses impedance functions of
network travel time and weights relevant job seekers within the labor market of census tract j.
JD
j, m
= ∑
k
(W
k, m
f(C
kj
)) (1)
f(C
kj
) = exp (-bC
kj
) (2)
Figure 3.2 Job Accessibility Measure
Tract k
Job
Seekers
Tract j
Tract i
Job demand Job supply
Jobs
Job
Seekers
Tract j’s labor market
Tract i’s labor market

43
where JD
j, m
is demand of relevant jobs at census tract j from job seekers in income group m;
W
k, m
is the number of relevant job seekers in income group m at census tract k;
f(C
kj
) is a friction function for modes and links; it is a function of C
kj
, travel time from the
centroid of k to the centroid of j;
b is the impedance factor for trips, and its value is 0.10397, which is calculated through a
simple regression procedure. The regression has the natural log of the friction factors as the
dependent variable and travel time intervals as the independent variable
9
.
The second step is to calculate potential job supply around census tract i. Equation (3) is the
conventional gravity-based supply-side job accessibility model adapted from Hansen (1959) and
often used in research (e.g. in Levinson, 1998; Meyer and Miller, 2001)
JS
i, m
= ∑
j
E
j, m
f(C
ij
) (3)
f(C
ij
) = exp (-bC
ij
) (4)
where JS
i ,m
is supply of relevant jobs for census tract i for income group m;
E
j, m
is the number of relevant job opportunities at census tract j for income group m;
f(C
ij
) is a friction function of C
ij
, travel time from the centroid of i to the centroid of j;
Other notations are the same as those in the previous equations.
The third step is to calculate job accessibility of census tract i by summing up job opportunities
which are weighted by job demand (JD
j
) at each census tract j, and also weighted by travel
impedance f(C
ij
) between tract i and tract j.

9
Friction factors are calculated based on 1991 SCAG Household Travel Survey and 1990 census data. The factors
were used in the 2000 SCAG Regional Transportation Plan. This research uses the same factors for both 1990 and
2000.

44
A
i, m
= ∑
j
(E
j, m
f(C
ij
) / JD
j, m
)

= ∑
j
[ E
j, m
f(C
ij
) / ∑
k
(W
k, m
f(C
kj
) ] (5)

where A
i, m
is job accessibility of relevant job seekers in income group m who live in census
tract i;
Other notations are the same as those in the previous equations.
The job accessibility scores range from zero to infinity theoretically, and they have an important
characteristic that “the expected value of measured accessibility scores equals the ratio of the
total number of opportunities to the total number of people seeking the opportunities” (Shen,
1998, P350). A high job accessibility score indicates that the census tract is in a job rich area
and job seekers face small spatial barriers to reach abundant competitive job opportunities. On
the other hand, a low job accessibility score indicates jobs in short supply thus great spatial
barriers to relative job opportunities, and some job seekers need to search for and find jobs in
other job rich areas.
The job accessibility measure in this research is more place-based than people-based although it
distinguishes income groups. With the emphasis on urban spatial structure and the aggregated
nature of the census tract data, job seekers’ other characteristics such as automobile ownership,
travel behavior and trip chaining, are not considered in the job accessibility measure.

45
Labor Market Regression Models
One of the key questions in the SMH is to what extent spatial barriers affect people’s labor
market outcomes; in other words, whether identical job seekers with the same socioeconomic
characteristics achieve different labor market outcomes because of their different job
accessibility. Three labor market outcomes are examined: the labor force participation rate,
employment rate and commute time.
Unit of Analysis
This research uses the census tract, instead of the individual, as the unit of analysis. The reason
is that this research aims to capture spatial characteristics which most individual data (such as the
PUMS data) do not provide. Particularly, job accessibility is calculated based on the census
tracts. If one wants to calculate job accessibility for each individual, the census tract level job
accessibility scores in each PUMA must be averaged and then applied to all individuals in the
PUMA. The method is inaccurate and infeasible. PUMAs are generally too large and some
job seekers in a PUMA search for and obtain jobs within a PUMA. The aggregation eliminates
local variation of job accessibility which might be particularly significant for low-income job
seekers who have few resources to travel. Moreover, the boundaries of PUMAs changed
between 1990 and 2000. Therefore, the same job accessibility in one census tract in 1990 is
weighted differently in 2000 if the tract belongs to a different PUMA. Finally, the census tract
46
level data including job accessibility are better indicators of neighborhood characteristics, such
as residential segregation and social networks, which impose impacts on individual’s labor
market outcomes as well.
The individual level data, on the contrary, can provide much more details on individual job
seeker and household, but not enough details of geographic locations or neighborhood
characteristics. Neither aggregated nor individual data provide perfect information, but this
research chooses to focus on spatial characteristics.
In general, the choice of using the individual level data or using the aggregated data does not
seem to affect the empirical results of research which tests the effects of job accessibility on
labor market outcomes, as long as the geographic unit of the aggregated data is small enough (e.g.
TAZ or census tract). Research using either the aggregated or the individual data had mixed
results of job accessibility’s effects.
Conceptual Model
The research hypothesis is that job accessibility significantly impacts low-income job seekers’
labor market outcomes, all else being equal. In this research, job seekers’ locations are
identified by using the census tracts, which usually have between 2,500 and 8,000 persons and
are designed to be homogeneous with respect to population characteristics, economic status and
47
living conditions
10
. It is expected that in multivariate regression models, independent variables
calculated at the census tract level can represent overall characteristics in the relatively
homogenous census tracts, although variations of the individual characteristics within those
tracts are lost.
Residential and employment location decisions are dynamic. In reality, households and firms
can relocate to other census tracts to maximize their utilities or profits within certain constraints.
Those locational decisions are based on many considerations, notably commute cost. However,
both residential and employment relocations are costly, and usually happen with a time lag.
This research examines the study area at two points in time ─ 1990 and 2000, and assumes that
population and employment, in terms of their amount and locations, are exogenously given.
Much empirical research in the SMH literature made similar assumptions (e.g. Ellwood, 1986;
Ihlanfeldt and Sjoquist, 1990 and 1991; Raphael, 1998), while recognizing that many factors are
interrelated and they have significant impacts on labor market outcomes in the long run.
This research focuses on the labor force participation rate, employment status and commute time.
The wage rate is an important endogenous variable in a labor market, but it is controlled to a
degree. The subject matter is low-income job seekers, who live in households with income less
than $19,770 in 1990 and/or $23,691 in 2000. If a low-income household had one full-time
worker during the survey year, hourly wage rate for the full-time worker is lower than $9.5 and

10
http://www.census.gov/geo/www/cen_tract.html
48
$11.4 respectively in 1990 and 2000. Certainly, some low-income households have part-time
or seasonal workers who might have higher wage rates, but their employment status can be partly
reflected in the census tract level average labor force participation rate and employment rate.
Nevertheless, there is an upper bound of wage rates for low-income job seekers by definition,
and the upper bound is quite low. Therefore, this research assumes flat wage rates for
low-income job seekers.
It is assumed that transportation and communication costs are nontrivial between census tracts.
For labor market behaviors, such costs impose burdens for commute, job search and information
acquisition. A long commute is not preferable and many times a deterrent for low-income job
seekers who do not have commute compensation with their relatively flat wage rates. Because
of the costs, low-income job seekers are less likely to search for jobs further away or to have
relevant information of these jobs. Weighting jobs by network travel time between census
tracts, job accessibility, in a way, is an indicator of those costs. Higher job accessibility of a
census tract indicates lower costs for job seekers to search for and obtain competitive jobs.
Based on the above assumptions, the labor market regression models examine the effects of job
accessibility while controlling auto ownership and socioeconomic characteristics, which are the
two other major factors of labor market outcomes. All else being equal, lower job accessibility
in a census tract indicates a relatively job poor location for job seekers in that census tract. In
this situation, a lower percentage of job seekers in that census tract can acquire and secure jobs
49
nearby due to a relatively limited supply, hence a higher percentage of job seekers need to search
for jobs in further distance. Job search costs impose great burdens for low-income job seekers
in that census tract, and consequently the unemployment rate is expected to rise as eventually
some job seekers cannot afford extensive job searches and hence become unemployed. Others
who are able to search in a larger area may have to endure longer commute time when they are
employed, and they would increase the average commute time of the census tract. Because
there is an upper bound of low-income job seekers’ wage rates, and commute costs are relatively
higher for low-income job seekers than for others, there is no compensation for the longer
commute time. At the same time, when it is more difficult or costly to search for jobs, some job
seekers drop out of the labor market when they have other options, such as welfare dependence
or the pursuit of criminal activities which are substitutes for employment (Arnott, 1998). It
may cause the labor force participation rate to decline. From the above, the research argues that
job accessibility affects all three indicators of the labor market outcomes.
The above conceptual model might be particularly informative in predicting labor market
outcomes of the disadvantaged population segments (e.g. autoless, racial/ethnic minorities,
low-educated) who have fewer resources, including financial, transportation and time resources,
for job search and commute. Job accessibility measures spatial barriers to relative job
opportunities. Job seekers in places with lower job accessibility have less chance to be
employed locally because of the relatively limited job supply, and a higher share of these job
50
seekers face greater spatial barriers in their job search. Those with fewer resources cannot
easily and efficiently overcome those spatial barriers. Therefore, job accessibility is expected
to matter particularly to those disadvantaged population segments, rather than others.
Following are the labor market regression models to test the significance of job accessibility:
Labor Force Participation Rate: LF

= f (JA, Auto, S)
Employment Rate: EMP

= f (JA, Auto, S)
Commute Time: TIME

= f (JA, Auto, S, Mode)
Where JA indicates low-income job accessibility; control variables include Auto which denotes
auto ownership per low-income job seeker, S which denotes an index of socioeconomic variables
associated with each census tract, and Mode which denotes mode split of commute trips in the
census tract.
Dependent Variables
The labor force participation rate and employment rate are directly generated from the Census
STF3 data. In terms of the average commute time, the Census provides the number of
commuters in each commute time bracket. This research assumes that the average commute
time for all commuters in the same time bracket is the median value of that bracket, and then
calculates weighted average commute time for all commuters in each census tract.
51
Control Variables
The above labor market regression models are to test whether job accessibility has impacts on
low-income job seekers’ labor market outcomes. Table 3.1 provides the measures of job
accessibility and the control variables, as well as their expected signs.
Table 3.1 Job Accessibility and Control Variables
Expected Sign
(Labor Force
and Emp rate)
Expected Sign
(Commute
Time)
Job Accessibility
Low-income JA + -
Automobile Ownership
Autoless Job
Seekers
% Low-income job seekers who do
not have access to automobile
- ?
Socioeconomic characteristics
Low-income Job
Seekers
% Low-income job seekers
- -
Race % Hispanic ? ?
% Black - ?
% Asian - ?
Education % high-school diploma and below in
persons 25 years and older
- ?
Youth % young (16-24 years old) Job
Seekers
- -
Household
Structure
% of female-headed family (no
husband) with children under 18
- -
Commute pattern
Commute mode % workers commute by transit +
Those control variables are included as the indicators of the population segments who are very
likely to be disadvantaged in the labor market, because of their limited human capital, resources
52
to search for and find jobs, and labor market constraints. Job seekers in these segments are
generally referred to as the disadvantaged groups.
This section begins by describing how each variable is expected to affect labor force
participation rates and employment rates of the census tracts.
Automobile ownership must be controlled in the models since the job accessibility measure
assumes that all job seekers commute by automobile. More importantly, automobile ownership
has significant impacts on labor market outcomes. It is found to be significantly associated
with the probability to be employed (Ong, 2002), and to have short commute time (Kawabata
and Shen, 2007). Automobile ownership for low-income job seekers is defined based on the
availability of automobiles in their own households: job seekers in households with at least one
motor vehicle have access to automobiles. Similar to the method to calculate low-income job
seekers and jobs, this research summarizes the share of low-income autoless households in all
autoless households for each PUMA using the PUMS data, and applies that share to the census
tract level number of autoless households, then calculates the census tract level number of
low-income autoless households, and finally the number of job seekers in the low-income
autoless households.
Automobile ownership is significantly affected by household income, which is controlled to a
degree as this research focuses on low-income job seekers. The employment rate might be
53
endogenous with automobile ownership, as employed workers are in more needs and are more
likely to be able to afford a car, and job seekers with a car have higher mobility and thus are
more likely to be employed. However, after income is controlled, the endogeneity is not
expected to be significant
11
. In general, for the labor force participation rate and employment
rate, a higher share of autoless job seekers is expected to have negative impacts.
Focusing on low-income job seekers, this research uses the percentage of low-income job
seekers as a control variable in the regression models. Implicitly, job seekers in other income
groups are the reference group. A higher share of low-income job seekers are expected to be
associated with lower labor force participation rates and lower employment rates.
High concentrations of racial/ethnic minorities are generally expected to have negative impacts
on labor force participation rates and employment rates as most minorities face many residential
constraints. Most SMH literature focuses on African Americans and finds their distressed labor
market outcomes. However, results of research on Hispanics are mixed. Recent literature
found that Hispanics are an “exception” to the SMH as they might have relatively high
employment rates (Aponte, 1996), and/or high labor force participation rates (Light, 2006). No
literature has specifically addressed Asians so far, but they are expected to have lower labor
force participation rates and employment rates. Generally, racial composition has to be treated
carefully.

11
This research tried Two-stage Least Squares (2SLS) models to estimate employment rate and automobile ownership
endogenously, the regression results are not very different from those of the Ordinary Least Squares (OLS) model.

54
Lower educational attainments, indicating lower human capital, are expected to reduce labor
force participation rates and employment rates, especially when economic restructuring reduces
low-skilled jobs. Significant impacts of education on employment status are commonly found
in research (e.g. Ihlanfeldt and Sjoquist, 1998; Ellwood, 1986).
Youth (16-24 years old) is used in the models to control age effects. Youth usually has minimal
work experience as experience generally grows with the age of workers. Like education, work
experience is related with human capital, and is expected to impact labor market outcomes.
Moreover, youth might have other options instead of work, such as attending school or being
supported by parents. Therefore, census tracts with a higher share of youth are expected to
have lower labor force participation rates and lower employment rates.
A higher percentage of female-headed families with children under 18 would negatively impact
labor force participation rates and employment rates because of their constraints in the labor
market. Specifically, Blumenburg and Ong (1995) found that the percentage of female-headed
families with children under 5 increases welfare dependence consistently. Therefore census
tracts with a higher percentage of female-headed families with children are more likely to have a
higher share of welfare recipients and are impacted by welfare reform.
Effects of those control variables on commute time are mixed, and need careful examination. It
is uncertain whether census tracts with higher shares of autoless job seekers have longer
55
commute time, especially when commute mode is controlled in the models.
Census tracts with a higher share of low-income job seekers are expected to have shorter
commute times because those job seekers are more likely to be employed in low wage jobs
which do not provide commute compensation. Also, low-income job seekers tend to have
fewer resources for extensive job searches and longer commutes.
The effects of the share of each racial/ethnic group on commute time are uncertain. The
commute time of the racial/ethnic minorities, especially African Americans and Hispanics, is one
of the focuses in the SMH literature. However, the results are mixed. It is found that “racial
comparisons of commuting times frequently provide, at best, weak tests of the SMH” (Ihlanfeldt
and Sjoquist, 1998, P853). The mixed results are mainly caused by the endogeneity of some
factors such as travel mode and labor market discrimination.
The impacts of educational attainment on commute time are uncertain as well. Taylor and Ong
(1995) did not find education’s impacts on commute distance/time for all workers. But Shen
(2000) found that neighborhoods (measured by using Traffic Analysis Zone) with a higher share
of highly educated people have longer average commute time. Similarly, Kawabata and Shen
(2007) found lower educated individuals have shorter commute time by both transit and private
automobile modes. This research is leaning toward the expectation of shorter commute time for
lower educated job seekers because their jobs usually offer lower wage rates and they have fewer
56
resources or compensation for longer commutes.
A higher percentage of young job seekers are expected to reduce average commute time as youth
usually has shorter commutes (Ellwood, 1986; Taylor and Ong, 1995). A possible explanation
is that youth usually is employed in entry-level jobs typically with very low wage rates. In
addition, youth usually does not have the resources for longer commutes.
A higher share of female-headed families with children under 18 are expected to negatively
impact average commute time, mainly because of those job seekers’ family responsibilities
which put more time and space constraints on their commutes.
An additional control variable used in the commute time model, but not in the other two models,
is the mode split of workers in the census tracts. Although it is affected by automobile
ownership, household structure and other factors, commute mode cannot be sufficiently
explained by these control variables, and hence it is included in the commute time model.
Furthermore, commute mode and automobile ownership are interrelated but different.
Automobile ownership indicates potential capability, while commute mode indicates the actual
travel behaviors of the employed workers. The PUMS data show that 5-6% of low-income
workers who have automobiles in their households take public transit, while 40-50% of those
who do not have automobiles still commute by private automobiles. Meanwhile, around 40%
of workers who commute by transit actually have automobiles in their households, and the other
57
60% are autoless. Those percentages vary significantly across the study area, indicating
complicated travel behaviors in addition to automobile ownership. Therefore, automobile
ownership and commute mode are different control variables.
Model Limitation
It is very difficult to develop models which fully consider all variables and examine their
interrelationship in an equilibrium fashion. First, many factors cannot be quantified with
limited data. For example, much research has been done to explore spatial job search
mechanism, but job search patterns and hiring patterns, such as the reservation wage rates job
seekers perceive, the availability of child care, are very difficult to quantify. For another
example, residential segregation is one of the fundamental factors in the SMH, but it is very
difficult to assess discrimination and other constraints in a housing market.
Second, many variables are endogenous. Primarily, residential and employment locations are
endogenously determined. With a limited scope, this research assumes that residential and
employment locations are exogenously given. For the labor market outcomes, wage rate, labor
force participation, employment status and commute patterns are endogenous with each other.
Focusing on low-income job seekers and their job accessibility, this research assumes that such
endogeneity is controlled to an extent.
Moreover, residential endogeneity is an important issue in the SMH literature, and most research
58
tried to remedy the endogeneity by focusing on youth. This research does not intend to remedy
the endogeneity because it mostly is an issue for whites or high-income job seekers (Ihlanfeldt
and Sjoquist, 1998), who have many options in their residential locations. On the contrary,
low-income job seekers face more constraints in their residential locations, which many times are
not their voluntary choices. However, a comprehensive residential location choice model is
beyond the scope of this research.

3.3 Chapter Summary
This chapter describes data and methodologies for the empirical analysis of this research.
Using Los Angeles as the case study area, this research tracks urban spatial transformation over
time across a fairly large region.
To differentiate low- and high-income job seekers and their matching job opportunities, this
research combines two datasets, the census tract level data and the PUMS data, and applies the
summarized characteristics from the PUMS data to calculate the numbers of jobs and job seekers
by income group at the census tract level. The method is commonly applied in data preparation
for transportation modeling.
Job accessibility is measured at the census tract level, and it is an indicator of spatial mismatch
59
because it denotes the spatial barriers to reach relative job opportunities. The job accessibility
measure in this research is a function of network travel time, as well as of residential and
employment locations. Higher job accessibility scores indicate easiness o reach abundant
competitive job opportunities nearby, and job seekers in those census tracts have better chances
to acquire jobs. Lower scores indicate greater barriers to reach competitive job opportunities
within short distances, and job seekers have lower chances to easily acquire jobs.
Job accessibility is expected to affect low-income job seekers’ labor market outcomes. As
suggested by the conceptual models developed in this chapter, higher job accessibility is
expected to increase the labor force participation rate and employment rate, and reduce average
commute time in a census tract.

60
CHAPTER FOUR
LOW-INCOME JOB SEEKERS’ JOB ACCESSIBILITY
This chapter tests the first and the second hypotheses: 1. low-income job seekers still have lower
job accessibility than high-income job seekers in the polycentric urban structure; 2. the
differential of job accessibility between income groups has increased as low-income job
accessibility decreased compared to high-income job accessibility between 1990 and 2000.
Results of this chapter will demonstrate if low-income job seekers are better off or worse off in
the urban spatial transformation.
Spatial mismatch is generally an issue in inner cities as most literature implied. However, in
the polycentric urban structure, different places in the suburbs function differently. Kain’s
SMH indicates a temporal process: jobs and whites have suburbanized, and African Americans
are constrained in the inner cities. The redistribution and relocation processes continue in the
current polycentric urban form, but take different forms in various locations. The questions that
need to be answered are: where have the spatial changes occurred and how have the changes
affected the job accessibility of different population segments?
Job accessibility, capturing both job demand and supply, can be an indicator of spatial mismatch
since it denotes whether the number and the location of jobs and job seekers match each other, in
other words, if job seekers can reach abundant competitive job opportunities within a reasonable
61
commute shed. The accessibility scores capture a critical factor in the SMH − urban spatial
structure, which is determined by the locations of jobs and job seekers.
Moreover, with economic and social development, people might have improved job accessibility
no matter their socioeconomic status. Results of this chapter will show if low-income job
seekers’ accessibility increased or not compared to high-income job seekers, and more
specifically, where and to what extent the differences between their respective job accessibility
occur.

4.1 Descriptive Analysis
Table 4.1 shows the numbers of job seekers and jobs for low- and high-income groups, as well as
their respective growth rates between 1990 and 2000 by County in the study area.
By definition, the total numbers of low- and high-income households (not shown in the table) are
roughly the same, around 25% of total households. But because high-income households have
more job seekers per household than low-income households
12
, the numbers of high-income job
seekers are greater than those of low-income job seekers. The disparity is most significant in
Orange and Ventura counties. Between 1990 and 2000, growth rates of job seekers by income

12
Average number of job seekers (persons 16 to 64 years old) per low-income household is 1.8, and average number
of job seekers per high-income household is 2.4 for the metropolitan area in 1990 and 2000, summarized from the
PUMS data.

62
vary across the region. Riverside County had the fastest growth for both income groups, and
Los Angeles County had the slowest growth. Most pertinent to this research, low-income job
seekers grew faster than high-income job seekers in the study area, which is caused by recent
immigrants, who are generally low-income. Specifically, growth rates of low-income job
seekers were higher than those of high-income job seekers in Los Angeles, Orange and Riverside
Counties, while growth rates of low-income job seekers were lower in San Bernardino and
Ventura, two counties in the outer-ring suburbs. Suburbanization is obvious, but the extent of
suburbanization differs between income groups across the study area.
In terms of employment, all counties have significantly more high-income jobs than low-income
jobs. Between 1990 and 2000, because of the economic restructuring, the growth rates of
high-income job are higher than those of low-income jobs in all counties except Los Angeles
(-1.1% of low-income employment growth vs. -5.3% of high-income employment growth).
The significant reduction of high-income jobs in Los Angeles County was caused by the
withdrawal of aerospace and defense industries. Riverside County had the highest employment
growth rate for both income groups. Not coincidentally, Riverside County also had the highest
growth rate of job seekers in both income groups, which created increasing competition of the
growing jobs. Considering both the demand and supply of jobs is important for the job
accessibility measure.

63
Table 4.1 Job Seekers and Jobs by Income Group in Each County (in 1,000)

Low-Income
Job Seeker
High-Income
Job Seeker
Low-Income
Jobs
High-Income
Jobs
1990
Los Angeles 1,116 1,656 683 1,450
Orange 160 670 103 550
Riverside 94 135 42 78
San Bernardino 143 187 60 116
Ventura 37 155 19 106
Study Area 1,548 2,802 907 2,300
2000
Los Angeles 1,250 1,629 675 1,373
Orange 195 730 116 661
Riverside 124 163 60 120
San Bernardino 166 238 72 156
Ventura 38 178 21 134
Study Area 1,774 2,937 944 2,443
% Growth 1990-2000
Los Angeles 12.2% -1.7% -1.1% -5.3%
Orange 22.2% 8.9% 11.9% 20.1%
Riverside 32.4% 21.2% 45.3% 54.3%
San Bernardino 16.0% 27.1% 19.3% 35.4%
Ventura 4.7% 14.9% 12.4% 25.6%
Study Area 14.6% 4.8% 4.1% 6.2%
Distribution patterns of low- and high-income job seekers in 2000 are shown in Figures 4.1.1 and
4.1.2. Distribution patterns in 1990 are very similar. Most low-income job seekers are
concentrated in the inner cities; others are scattered in north Orange County and parts of the
Inland Empire (Riverside and San Bernardino counties). High-income job seekers have a very
different distribution pattern which is much more evenly distributed. There are some unobvious
clusters, particularly in the inner-ring suburbs, such as west Los Angeles along the coast line, and
parts of Orange County. High-income job seekers are relatively sparse in downtown Los
64
Angeles compared to low-income job seekers.
Figures 4.2.1 and 4.2.2 show the distributions of low- and high-income jobs in 2000. Although
at very different magnitudes, the distribution patterns of low- and high-income jobs are
somewhat similar. The most significant concentration is in downtown Los Angeles, and the
concentration stretches out along the Wilshire corridor to west Los Angeles County.
High-income jobs show more concentrations and clusters in suburban counties, such as north
Orange County and the central area of the Inland Empire (Riverside and San Bernardino counties)
along some freeways.
Figure 4.1.1 Low-income Job Seekers

Figure 4.1.2 High-income Job Seekers

Figure 4.2.1 Low-income Jobs

Figure 4.2.2 High-income Jobs

65
The above table and figures reveal how population and employment have suburbanized. As
expected, job seekers, including low-income job seekers, have moved to the suburbs.
Residential segregation still exists, now partly in the suburbs. Because of the changing
demographics, low-income job seekers grew faster than high-income job seekers, which
increased the demand of low-income jobs. Employment relocation and redistribution are more
complicated. Jobs suitable for low-income job seekers grew slower compared to those for
high-income job seekers. Geographically, both low- and high-income jobs grew the fastest in
the suburbs, and declined in Los Angeles County. But it is noteworthy that low-income jobs
are still largely concentrated in the inner cities, which most SMH literature does not explicitly
acknowledge.
The sequential changes of residential and employment locations are different for low- and
high-income job seekers. Jobs, both low- and high-income jobs, are redistributed and
reallocated in similar patterns in terms of the geographic extents of their suburbanization.
Riverside County had the fastest growth of low- and high-income jobs, followed by San
Bernardino and Ventura counties, although most jobs are still located in Los Angeles and Orange
counties. Regarding the residential distribution, the suburbanization pattern was similar for
high-income job seekers, who grew faster in San Bernardino, Riverside and Ventura counties,
the outer-ring suburbs of the region. However, the suburbanization pattern of low-income job
seekers was very different. The area with the highest growth rate was Riverside County in the
66
outer-ring suburbs, followed by Orange County, a part of the inner-ring suburbs. In other
words, low-income job seekers also suburbanized, but not to the same extent of overall jobs and
high-income job seekers. This clearly shows a lag in low-income job seekers’ residential
suburbanization. Because of the lag, in both time and space, for low-income job seekers to
follow their matching jobs, urban spatial transformation might have created involuntary spatial
barriers for low-income job seekers, who usually do not have equal mobility to overcome the
barriers and thus are particularly disadvantaged in terms of their job accessibility.
The spatial distributions of jobs and job seekers impact potential commute distance. The
average distance between all low-income jobs and job seekers was 29.5 miles in 1990 and it rose
to 31.8 miles in 2000. Similarly, the average distance between all high-income jobs and job
seekers increased from 33.1 to 35.0 miles in the study area. The shorter potential commute
distance of low-income job seekers can be explained by their relatively concentrated residential
locations. On the other hand, the increasing potential commute distance for both low- and
high-income job seekers are caused by the continuous suburbanization of employment and
population.
From the above illustrations, it is expected that low-income job seekers on average have lower
job accessibility because of their relatively limited job opportunities compared to the number of
job seekers. Moreover, it is expected that low-income job seekers have lower job accessibility
in 2000 than in 1990 because the growth of low-income job seekers outweighed growth of jobs.
67
The growth rate of low-income jobs was 4.13% during the 1990s, and the growth rate of
low-income job seekers was 14.6%. Shen (1998) has pointed out that “the expected value of
measured accessibility scores equals the ratio of the total number of opportunities to the total
number of people seeking the opportunities” (P350). The ratio of the total number of
low-income jobs to the total number of low-income job seekers is 0.570 in 1990 and 0.521 in
2000, and the ratio for high-income is 0.826 and 0.834, respectively. This research put
constraints of commute sheds in the job accessibility measure, and thus the actual weighted
average of job accessibility scores is not exactly at, but very close, to the ratios.
Although low-income job seekers are expected to be disadvantaged, the changing urban structure
might have mitigated the disadvantages for some of them, especially those who live in the
suburbs. Detailed examinations need to be done to explore where and to what extent
low-income job seekers are disadvantaged in terms of job accessibility in the polycentric urban
structure.

4.2 Job Accessibility Results
Following are the descriptive and statistical results to compare low- and high-income job
seekers’ job accessibility scores, as well as their changes from 1990 to 2000. From here on, job
accessibility of low-income job seekers is called low-income job accessibility, and similarly, job
68
accessibility of high-income job seekers is called high-income job accessibility.

Low- and High-Income Job Accessibility
This section provides cross sectional comparisons of low- and high-income job accessibility in
1990 and 2000. The following figures visually present job accessibility results in 2000 using
the job accessibility model described in Chapter 3. The geographic patterns are very similar in
1990.
Low-income job accessibility clearly forms several peaks in the region, mostly around certain
employment concentrations. Among them, downtown Los Angeles has the greatest
low-income job accessibility. Other peaks include north Orange County, east Ventura County
and the central area of the Inland Empire. Job accessibility decreases with increasing distance
from these peaks.
The distribution pattern of high-income job accessibility is similar as that of low-income job
accessibility, but at much larger magnitudes, which is expected as relatively more jobs are
available to high-income job seekers. Places with greater high-income job accessibility are
generally consistent with employment concentrations in the region.

69
Figure 4.3.1 Low-income Job Accessibility (2000)

Figure 4.3.2 High-income Job Accessibility (2000)

Many of the differences between low- and high-income job accessibility can be explained by
relevant job demand and supply. Following is the detailed analysis of the two sides of job
accessibility for both income groups. Because of the very similar distribution patterns between
1990 and 2000, this section focuses on the comparisons in 2000. Figures 4.4.1 and 4.4.2 exhibit
low- and high-income job seekers’ job demand, which represents the weighted number of job
70
seekers by their travel impedance to a census tract. As a result, job demand has a considerable
connection with overall job seekers’ residential locations.
Figure 4.4.1 Low-income Job Demand (2000)

Figure 4.4.2 High-income Job Demand (2000)

71
Figure 4.5.1 Low-income Job Supply (2000)

Figure 4.5.2 High-income Job Supply (2000)

Consistent with low-income job seekers’ residential locations, downtown Los Angeles has the
highest low-income job demand, which decreases rapidly with increasing distance from
downtown and stretches out to the central part of the Inland Empire. On the contrary,
high-income job demand shows a clear trend of suburbanization. The highest job demand
occurs in the areas to the west and to the south of downtown Los Angles, as well as in north
72
Orange County, and then spreads throughout the inner-ring suburbs with gradually declining
slopes. The different patterns of job demand are consistent with the observations that
low-income job seekers are concentrated in the inner cities, but high-income job seekers are
suburbanized.
The other side of the story is job supply, which is shown in Figures 4.5.1 and 4.5.2. Downtown
Los Angeles presents the highest job supply for low-income job seekers, and the supply
decreases with increasing distance from downtown and stretches out along freeways into Orange
County. Although low-income jobs grew very fast in the suburbs during the 1990s, Los
Angeles inner cities still have the greatest concentration of low-income jobs in either 1990 or
2000. The job supply for high-income job seekers shows a clear pattern of polycentricity with
peaks in both Los Angeles inner cities and Orange County (Santa Ana ─ Anaheim area).
Figure 4.6 shows the differences between low- and high-income job accessibility in the study
area in 2000. The distribution pattern cannot be easily explained by jurisdictional boundaries or
locations of economic activities; rather, it forms several circular rings even within the
asymmetric geographic boundary of the study area. As expected, high-income job seekers have
greater job accessibility in the whole study area, with very few exceptions. The greatest
differences can be found in the areas immediately around downtown Los Angeles, particularly in
west Los Angeles County and north Orange County. The differences are relatively smaller in
the inner cities and the three counties in the outer-ring suburbs ─ Riverside, San Bernardino and
73
Ventura. In general, low-income job seekers have lower job accessibility than high-income job
seekers, but they are more disadvantaged in the inner-ring suburbs.
Figure 4.6 Differential of Job Accessibility (Low- vs. High-Income) (2000)

To statistically measure the magnitudes of the differences between low- and high-income job
accessibility, paired-sample t-tests are applied for 1990 and 2000. Since the SMH argues that
low-income job seekers in the inner cities would be the most disadvantaged, Table 4.2 also
provides the t-test results by County in 2000 to compare the variations across the study area.
As expected, in both 1990 and 2000, the average scores of low-income job accessibility (0.547 in
1990 and 0.500 in 2000) are significantly lower than those of high-income job accessibility
(0.891 in 1990 and 0.916 in 2000). Across the five counties, the biggest gap of job accessibility
between income groups occurs in Orange County (-0.454), followed closely by Los Angeles
(-0.428), meaning that low-income job seekers in these two counties experience greater
74
disadvantages compared to high-income job seekers. Job seekers in the other three suburban
counties are not as disadvantaged. What might explain why Los Angeles County does not have
the biggest difference is that it is big and heterogeneous, therefore high low-income job
accessibility scores in the inner cities are reduced by low scores in other tracts in the County.
On the other hand, Orange County has the highest job accessibility for both low- and
high-income job seekers. Therefore, the big difference between income groups in Orange
County is relative, and it does not indicate low-income job seekers in Orange County are
particularly disadvantaged compared to low-income job seekers in other counties.
Table 4.2 T-test Results to Compare Low- and High-income Job Accessibility
Mean Std. Dev Mean
Diff
Std. Dev
Region Low 0.547 0.159
1990
High 0.891 0.236
-0.344 0.130
Region Low 0.500 0.163
High 0.916 0.240
-0.416 0.117
LA Low 0.496 0.167
High 0.924 0.219
-0.428 0.101
OR Low 0.543 0.142
High 0.998 0.254
-0.454 0.118
RV Low 0.449 0.174
High 0.770 0.304
-0.322 0.173
SB Low 0.438 0.139
High 0.750 0.241
-0.312 0.108
VN Low 0.504 0.162
2000
High 0.791 0.208
-0.286 0.072
(Notes: 1. means are the unweighted average; 2. all mean differences are significant at the 1%
significance level)
The above examinations clearly show notable differences between low- and high-income job
75
accessibility, as well as their respective job demand and supply. Two major observations can
be summarized.
1. Consistent with the expectations of the SMH, low-income job accessibility is significantly
lower than high-income job accessibility, indicating greater disadvantages of low-income job
seekers. Overall, the disadvantages in job accessibility are caused by the limited number of
low-income jobs available. Furthermore, the differences vary across the study area, which can
be partly explained by the changes in urban spatial structure and how each county functions
differently in the suburbanization process. The longitudinal study of the changes in job demand
and supply in the next section will delve into this issue.
2. At one point in time, 2000 in this case, low-income job demand and supply are the highest
in downtown Los Angeles, and both demand and supply decline with increasing distance from
downtown. In other words, low-income jobs and job seekers’ locations can still be largely
explained by the monocentric model, although the end results of job accessibility vary in the
suburbs. By contrast, high-income job demand and supply present obvious polycentric patterns,
with the highest demand and supply not only in the inner cities, but also in parts of the suburbs,
particularly in Orange County.

76
Change of Job Accessibility over Time
Figures 4.7.1 and 4.7.2 present changes of job accessibility from 1990 to 2000 for low- and
high-income job seekers, respectively.
Figure 4.7.1 Change of Low-Income Job Accessibility

Figure 4.7.2 Change of High-Income Job Accessibility

A more important question to be answered is whether low-income job seekers are better off or
worse off in the spatial transformation, compared to high-income job seekers. The distribution
77
patterns of changes in job accessibility are significantly different from those of job accessibility
in 1990 or in 2000.
For low-income job seekers, job accessibility decreased slightly in the inner cities with the
exception of west Los Angeles. Job accessibility decreased significantly in the northwest and
southeast parts of Los Angeles County and adjacent areas. Other areas generally showed
increases in low-income job accessibility. With respect to changes in high-income job
accessibility, accessibility increased in most parts of the inner cities, decreased slightly in some
areas immediately adjacent to the inner cities, and increased significantly in other areas,
especially in the most distant suburban communities.
The following section examines why the patterns of change in job accessibility differ between
low- and high-income groups, and why the changes happened in some specific locations.
Figures 4.8.1 and 4.8.2 exhibit changes in low-income job seekers’ job demand and supply from
1990 to 2000. Figure 4.8.1 indicates a clear trend of low-income job seekers’ suburbanization.
Specifically, low-income job demand decreased significantly from downtown and the inner cities,
and increased substantially in the inner-ring suburbs around inner cities. The significant growth
of job demand stretched out to the central area of the Inland Empire. In other words,
low-income job seekers have suburbanized to some suburban communities in Los Angeles and
Orange counties, as well as to some areas in Riverside and San Bernardino counties which are
further away.
78
At the same time, changes in low-income job supply (Figure 4.8.2) present a greater extent of job
suburbanization. Job supply declined significantly from the Los Angeles inner cities, declined
slightly from other parts of Los Angeles County and north Orange County, and increased
significantly in south Orange County and the central area of the Inland Empire, which are in the
outer-ring suburbs.
Figure 4.8.1 Change of Low-Income Job Demand

Figure 4.8.2 Change of Low-Income Job Supply

79
Figure 4.9.1 Change of High-Income Job Demand

Figure 4.9.2 Change of High-Income Job Supply

The changes in job demand and job supply can be used to explain many of the changes in
low-income job accessibility. In the inner cities, both low-income job demand and supply
decreased, and resulted in slight changes of job accessibility. In the inner-ring suburbs around
the inner cities, low-income job demand increased greatly while job supply declined slightly,
therefore job accessibility decreased. In the outer-ring suburban areas, both job demand and
80
supply grew, which resulted in a minor increase of job accessibility.
The changes in high-income job demand and supply present similar suburbanization and
clustering patterns (Figure 4.9.1 and Figure 4.9.2). Most parts of Los Angeles County and the
northern part of Orange County showed notable decreases in both job demand and supply, with
minor exceptions in south central Los Angeles. Increases happened in most parts of the Inland
Empire, the southern part of Orange County and Ventura County. The redistribution patterns in
job demand and supply of high-income job seekers were roughly synchronized, but job supply
increased at a bigger scale in the outer-ring suburbs. As a result, high-income job accessibility
increased significantly in the outer-ring suburbs.
The exhibitions above compared the changes in job demand and supply for both income groups
respectively. In general, low-income job seekers suburbanized with significant lag compared to
their matching jobs. Low-income job demand grew faster in the inner-ring suburbs than
anywhere else in the region. However, their job supply declined from the inner cities and the
inner-ring suburbs and increased in the outer-ring suburbs. Also, segregation occurred in
low-income job seekers’ suburbanization process as they grew faster in Riverside and San
Bernardino counties although both are further away from the inner cities. In contrast, Orange
and Ventura counties are geographically closer to the inner cities, but housing there is not usually
affordable to low-income job seekers. Comparing changes of job demand and supply, although
in both 1990 and 2000, most low-income jobs and job seekers are still located in the inner cities,
81
mismatch happened in their spatial and temporal changes. Therefore, it is more accurate and
important to examine the changes at the margin. Contrarily, high-income job seekers and jobs
suburbanized at a similar pace to similar areas. The different stages of suburbanization between
income groups are consistent with the expectations from the SMH literature. But in addition to
the locational differences between the inner cities and the suburbs, differences also occurred
within the suburbs as well.
Table 4.3 illustrates statistical differences of changes in job accessibility between income groups
by County. The greatest difference was found in Orange County (-0.155), followed closely by
Riverside County (-0.112), meaning that compared to high-income job seekers, low-income job
seekers in the two counties experienced the slowest growth, or the greatest loss, of job
accessibility. Surprisingly, Los Angeles County had the smallest difference (-0.043) of changes
in job accessibility between income groups. The results are interestingly different from the
expectations of the SMH literature, but reasonable from the examinations above. Los Angeles
is the only county with negative growth of job accessibility for both low- and high-income
groups, which results in a smaller difference in job accessibility between the two income groups.
Orange County had the largest increase in high-income job accessibility and the second largest
decrease of low-income job accessibility. It is a result of the relatively faster growth of
low-income job seekers and slower growth of high-income job seekers compared to their
respective job opportunities. Therefore, the difference in changes of job accessibility between
82
the two income groups is the greatest in Orange County. Riverside and San Bernardino are the
only two counties with positive growth of low-income job accessibility, although the growth is
small. Meanwhile, high-income job accessibility increased rather fast in these counties.
These observations indicate increasing job supply relative to demand for both income groups due
to the rapid employment suburbanization. Changes of job accessibility scores in Ventura
County are smaller compared to other counties.
Table 4.3 T-test Results to Compare Change of Job Accessibility between Low- and
High-income Job Seekers (2000)
Mean Std.
Dev
Mean
Diff
Std.
Dev
Region Low -0.048 0.068
High 0.023 0.119
-0.072 0.084
LA Low -0.061 0.060
High -0.018 0.080
-0.043 0.047
OR Low -0.042 0.069
High 0.114 0.138
-0.155 0.078
RV Low 0.002 0.093
High 0.113 0.187
-0.112 0.127
SB Low 0.015 0.046
High 0.102 0.095
-0.086 0.074
VN Low -0.018 0.086
1990-
2000
High 0.056 0.141
-0.073 0.183
(* All mean differences are significant)
Spatial mismatch still exits, but it extends beyond the boundary of the inner cities in the Los
Angeles area. Many original suburban communities, particularly those in Orange County,
began to function like the inner cities several decades ago when the SMH was first proposed.
Jobs and high-income job seekers have suburbanized faster in Riverside and San Bernardino
counties, while low-income job seekers suburbanized to the inner-ring suburbs close to the inner
83
cities. Spatial mismatch gradually becomes an issue in these areas as well, rather than just in
the inner cities.

4.3 Chapter Summary
Low-income job seekers have suburbanized, but mostly to the inner-ring suburbs immediately
adjacent to the inner cities. The geographic extent of their suburbanization is still much smaller
than that of high-income job seekers. At the same time, jobs, including low- and high-income
jobs, have been suburbanizing further out, especially to the Inland Empire. As a result,
low-income job seekers did not follow redistribution and reallocation of low-income jobs in a
timely manner. Moreover, the suburbs do not necessarily provide advantages for low-income
job seekers who live there. Orange County presents the greatest difference between low- and
high-income job accessibility in 2000, as well as the greatest difference of changes in job
accessibility. In other words, instead of Los Angeles, low-income job seekers become
relatively more disadvantaged in Orange County compared to high-income job seekers there.
Riverside County had the second largest difference of changes in accessibility between income
groups, indicating that low-income job seekers there are getting more disadvantaged compared to
high-income job seekers over time.
Detailed examinations of the changes in low-income job accessibility demonstrate that the
84
changes can be largely explained by the different suburbanization stages of jobs and job seekers.
Both low-income jobs and job seekers decreased from inner cities, thus job accessibility did not
change significantly. Low-income job seekers have suburbanized to the areas adjacent the
inner cities, but jobs have suburbanized further out. These two changes combined created
relatively excessive job demand and reduced low-income job accessibility in the inner-ring
suburbs. In the outer-ring suburbs, low-income jobs grew faster than low-income job seekers,
thus job accessibility increased.
There are many explanations of such mismatch in the suburbanization process over time. In
addition to employment suburbanization and residential segregation, causal relationship between
the relocation of jobs and job seekers can provide an answer. However, to my knowledge, little
empirical research so far has tested the causalities of residential and employment locations by
population segments and their relevant job types, with the exception of Greenwood and Stock
(1990). This research provides circumstantial evidence of the necessity for further research on
this topic. No matter the mechanism, compared to low-income job seekers, relocation and
redistribution of high-income job seekers are much more synchronous with those of jobs. This
has been true in the conventional monocentric urban structure as implied in Kain’s original SMH
study. This research shows that it is still true in the current polycentric urban structure, but in
different locations in the suburbs.
There are several interesting case study areas in the suburbs, such as the Santa Ana – Anaheim
85
area in Orange County. Before the 1980s, it had been a typical suburban area. However, over
time, this area became urbanized with relatively high density and mature communities. Jobs
and high-income job seekers began to move out, and the space was filled in by low-income job
seekers. A whole new process of outmigration of jobs and high-income job seekers, and
immigration of low-income job seekers is happening, but this time, in a suburban community,
instead of the inner cities. With more data, further case study in this area can provide a more
accurate picture of urban spatial transformation and its impacts on job accessibility by population
segments.
86
CHAPTER FIVE
IMPACT OF JOB ACCESSIBILITY
This chapter tests the third hypothesis that job accessibility significantly impacts low-income job
seekers’ labor market outcomes. Using job accessibility as an indicator of spatial mismatch in a
metropolitan area, Chapter 4 has demonstrated that spatial mismatch still exists: low-income job
seekers have lower job accessibility compared to high-income job seekers, and the differences
vary significantly across the study area. This chapter examines to what extent spatial mismatch
matters to the actual labor market outcomes: the labor force participation rates, employment rates
and commute time.
As reviewed in Chapter 3, other research has used different indicators of job accessibility and
tested the relationship between job accessibility and the labor market outcomes. This research
differs from most of other empirical research from three perspectives. First, this research
focuses on low-income job seekers and explores to what extent job accessibility matters to the
economically depressed people. Controlling income and job accessibility, this research also
tests if race/ethnicity still has differential impacts on the labor market outcomes within a certain
income group. Second, this research uses a more accurate job accessibility measure which
captures relative location of low-income jobs and job seekers in a polycentric metropolitan area.
Job accessibility is not only an indicator of spatial barriers between job seekers’ residences and
87
their relative job opportunities, but also an indicator of other related factors, such as costs of
spatial job searching and information acquisition. Third, fully utilizing time series data, this
research compares if the impacts of job accessibility on the three labor market outcomes changed
between 1990 and 2000.

5.1 Descriptive Analysis
Low-income Job Seekers’ Labor Market Outcomes
Job accessibility and its characteristics are closely examined in Chapter 4. As discussed in
Chapter 3, in addition to job accessibility, automobile ownership and socioeconomic variables
also are also expected to have impacts on the labor market outcomes.
Because of their expected distressed labor market outcomes, especially employment status, job
seekers with the characteristics measured by the control variables (autoless, low-income,
race/ethnicity minority, low-educated, young, in female-headed household with children, no
husband present) are generally referred to as the disadvantaged groups. Table 5.1 shows
different labor market outcomes of low-income job seekers by their distinct characteristics
(automobile ownership or socioeconomic characteristics) in the study area in 1990 and 2000
using the PUMS data. The labor market outcomes are shown in three categories: labor force
88
participation rate, employment rate, and the share of employed workers who have short
commutes. Short commutes are defined as commute time below the regional median, which is
30 minutes in 1990 and 35 minutes in 2000. To compare the distinct effects of automobile
ownership and socioeconomic characteristics, Table 5.1 also shows the outcomes of the
reference groups which are omitted from the regression models (see shaded rows).
In terms of automobile ownership, as expected, low-income job seekers with at least one
automobile in their households are more likely to have better labor market outcomes than those
without an automobile in their households, namely, a higher labor force participation, a higher
employment rate, and a higher share of workers with short commutes.
For the race/ethnicity groups, Whites, the reference group, always have better labor market
outcomes, except their labor force participation rate in 1990 when Hispanic low-income job
seekers had a higher labor force participation rate. Comparing minority groups, Hispanics have
the highest labor force participation rate, while Asians have the highest employment rate, and the
highest share of workers with short commute time in both 1990 and 2000. Blacks are the
groups consistently have the lowest employment rate and the lowest share of workers with short
commutes.
Low-educated job seekers have relatively poorer labor market outcomes compared to the others,
as low-educated job seekers have a lower labor force participation rate and a lower employment
89
rate. This has become especially evident as the economic restructuring has resulted in
relatively slower growth of jobs requiring low education and low skill-sets. Moreover,
low-educated workers have a slightly lower share of short commutes than the others. In other
words, a higher share of low-educated job seekers has long commutes. The reasons might be
low-educated job seekers’ slower commute modes, and the effects of spatial mismatch that their
suitable jobs are further away from their residences. Regression analysis controlling commute
mode and job accessibility can provide a more accurate estimate of educate on commute time.
As expected, youth (16-24 years old) does not perform as well as adults (25-64 years old) in the
labor market, with a lower labor force participation rate and a lower employment rate. This is
partly attributed to their limited work experience, but more so to their other options or
responsibilities such as attending school. Youth has fewer incentives as adults to search for
jobs, let alone to find jobs. A higher share of youth has short commutes than the others, and
this might be explained by the low-skilled and low-wage nature of youth’s jobs, as well as by
youth’s limited transportation resources.
Most job seekers in female-headed families with children under 18, no husband present, are
women
13
. Moreover, female-headed families with children are the group most heavily
dependent on welfare. From the PUMS data, in 1990, 44% of those families relied on welfare,
and they composed 30% of the welfare dependent households. In 2000, 36% of those families

13
95% of job seekers in female-headed families with children are women, summarized from the PUMS data in 1990
and 2000.

90
were welfare recipients, and they composed 54% of the welfare dependent households. Their
increasing share in the welfare dependent households is caused by the shrinking size of those
households from 1990 to 2000. The shift reflects changing legislations in the late 1990s,
particularly the pass of the Personal Responsibility and Work Opportunity Reconciliation Act
(PRWORA), also known as the Welfare Reform Act, in 1996. PRWORA requires welfare
recipients to begin working after two years of receiving benefits. Subsequently in 1998,
California passed Work Opportunity and Responsibility to Kids (CalWORKs) program to
implement the PRWORA, and CalWROKs similarly requires welfare recipients to participate in
welfare-to-work activities. The new legislations at that time provided strong incentives for job
seekers in welfare dependent families, especially those in female-headed families with children, to
participate in the labor force (Blank, 2002). However, there is no general conclusion on the
effects of welfare reform on employment rates.
Because of the external changes of the labor market conditions, the labor market performances of
the job seekers in female-headed families with children need to be carefully examined. The
labor force participation rate of those job seekers was lower than that of other job seekers in
1990, but was higher (59.0% vs. 49.6%) in 2000. One explanation is the above-mentioned
changing legislations. However, the employment rate did not change significantly for those job
seekers between 1990 and 2000, probably because of the increasingly larger size of their labor
force. Moreover, compared to other job seekers, those in female-headed families with children
91
still have a lower employment rate in both 1990 and 2000. In terms of their commute time, a
lower share of workers in female-headed families with children have short commutes in 1990,
but the share was roughly the same as other workers in 2000. The finding is also different from
the expectation that because of their family responsibilities and their limited transportation
resources, workers in female-headed families with children would have shorter commutes. The
lower share of short commutes might be a result of those job seekers’ relatively slower commute
mode by public transit, or of the effects of spatial mismatch that relevant jobs are further away.
Regression analysis in the next section will estimate actual effects of female-headed families
with children.
Table 5.1 Labor Market Outcomes of Low-income Job Seekers with Distinct
Characteristics

Labor Force
Participation Rate Employment Rate
Share of Workers
w/ Short Commutes
(%)
1990 2000 1990 2000 1990 2000
No automobile 49.9 45.0 77.1 76.1 69.1 70.8
With automobile 60.5 52.2 84.7 84.8 81.4 81.4
Hispanic 61.9 49.7 83.5 83.1 78.3 78.6
Black 49.7 49.2 72.2 72.5 74.0 75.7
Asian 50.1 45.6 87.2 87.3 79.8 81.1
White 59.5 54.8 86.1 85.7 82.4 81.7
Low Education 55.5 47.8 83.7 83.6 78.6 78.4
High Education 68.9 59.1 87.3 86.9 78.7 79.9
Youth (16-24) 54.4 47.3 78.4 77.2 82.3 81.3
Adult (25-64) 59.8 51.6 85.0 84.8 78.7 79.0
Female HH w/child 51.8 59.0 80.1 80.2 78.9 79.9
Other Household 59.2 49.6 83.8 83.5 81.6 79.5
Transit Mode 48.0 47.6
Private Auto Mode 81.1 80.9
Data source: 5% PUMS in 1990 and 2000
92
Comparing commute time between workers using public transit and private automobile, a higher
share of workers who commute by private automobile had short commutes than those
commuting by transit in both 1990 and 2000s.

Variables in the Labor Market Regression Models
The above analysis depicts labor market performance of low-income job seekers by their distinct
characteristics in the whole study area. This section delves into the labor market regression
models.
Table 5.2 provides descriptive statistics and the paired-sample t-test results to compare the mean
value of each variable between 1990 and 2000. The labor force participation rate declined
while the employment rate remained roughly the same. This is consistent with the expectations
since the study area has experienced faster growth of population than jobs, which changed
overall job demand and supply. The average commute time decreased slightly from 29.30 to
29.59 minutes, which is caused by errors in the unweighted average of the tract level commute
time.

93
Table 5.2 Descriptive Statistics of Variables in 1990 and 2000

1990

2000
Change
(2000-1990)
Mean Median StdDev Mean Median StdDev Mean StdDev
Dependent Variables
LF participation rate 0.67 0.68 0.10 0.61 0.62 0.09 -0.06 0.06
Employment rate 0.93 0.94 0.04 0.92 0.93 0.05 -0.01 0.04
Ave Commute Time 29.30 29.04 4.44 28.59 28.53 3.19 -0.71 3.35
Job Access
Low-inc Accessibility 0.55 0.56 0.16 0.50 0.48 0.16 -0.05 0.07
Automobile Ownership
Autoless 0.23 0.19 0.17 0.24 0.22 0.15 0.01 0.12
Socioeconomic
Low-inc Job Seekers 0.17 0.14 0.13 0.18 0.15 0.13 0.01 0.05
Hispanic 0.31 0.22 0.26 0.38 0.32 0.28 0.07 0.09
Black 0.08 0.02 0.16 0.08 0.03 0.14 -0.01 0.05
Asian 0.09 0.06 0.10 0.11 0.07 0.13 0.02 0.05
Low Education 0.49 0.47 0.21 0.47 0.46 0.23 -0.02 0.07
Youth 0.22 0.21 0.11 0.20 0.19 0.11 -0.02 0.13
Female HH w/Child 0.05 0.05 0.03 0.08 0.07 0.05 0.03 0.04
Commute Mode
By transit 0.05 0.02 0.08 0.06 0.03 0.08 0.01 0.03
(Notes: 1. Means and Medians are unweighted and hence are not consistent with some summary
numbers from the PUMS data. 2. All of the mean differences between 1990 and 2000 are
significant at the 1% significance level)
With respect to the independent variables, low-income job accessibility declined. The average
percentage of low-income job seekers who have no automobile in their households increased
slightly, which, however, is different from the expectation that automobile ownership rises with
increasing affordability of automobiles. This phenomenon might be explained by two reasons.
First, new immigrants and low-income households are less likely to afford automobiles, and they
weigh down the average automobile ownership. Second, the increasing density and mixed-land
94
uses in the Los Angeles region influence auto ownership decisions (Hess and Ong, 2001). For
socioeconomic variables, most of them remained roughly the same between 1990 an 2000,
except the increasing share of Hispanics, Asians, and female-headed families with children.
These changes are consistent with the expected shifts of demographics.
In terms of the distribution of each variable, the employment rate has relatively small standard
deviation because of the aggregated nature of the data. Many control variables are not strictly
normal distributions, namely percentages of Hispanics, Blacks, Asians, transit commuters, and to
a less extent percentages of autoless job seekers and low-income job seekers. The skewness is
mainly related to the residential segmentation and segregation in the housing market as
racial/ethnic minorities and low-income job seekers are more likely to be concentrated in certain
areas, and they are less likely to afford automobiles or to commute by automobiles.
Tables 5.3.1 and 5.3.2 provide Pearson correlation coefficients between all independent and
dependent variables in 1990 and 2000, respectively. Most variables are correlated with each
other at the 5% significance level. Between the two cross-section datasets, most correlations
are consistent. It is noteworthy that although the commute time has significant correlations
with most control variables, those correlations are generally small (below 0.2), indicating that
there is much noise in the correlations. Regression results in the next section can provide
accurate estimates of the specific impacts of job accessibility and the control variables on
commute time.
95
Moreover, in both 1990 and 2000, the share of Hispanics is highly correlated with the share of
low-educated people. Although there is a multicollinearity issue, this research keeps both
independent variables because of the large sample size (2473 observations) and the significance
of both variables for the hypothesis to be tested. Sensitivity tests were performed and omitting
either one of the two variables reduces the overall predictive power of the models.

5.2 Labor Market Regression Model Results

Cross-sectional Analysis in 1990 and 2000
Following are the results of the three labor market models. Table 5.4 shows the standardized
coefficient and the t-score of each variable in 1990 and 2000. The following section discusses
estimation results for each labor market outcome.

96
Table 5.3.1 Correlation between Variables (1990)

LF
partici
Emp rate Time Low JA Autoless Low Job
Seeker
Hispanic Black Asian Low Edu Young FemHH
w/ child
Transit
LF participation 1 0.336 0.144 -0.053 -0.376 -0.445 -0.129 -0.264 0.066 -0.340 -0.290 -0.209 -0.186
Employ rate 1 -0.126 -0.342 -0.410 -0.759 -0.601 -0.446 0.156 -0.717 -0.337 -0.517 -0.550
Commute Time 1 -0.151 -0.034 0.029 0.035 0.141 -0.038 0.066 -0.203 0.096 0.222
Low-inc Access 1 0.373 0.433 0.340 0.175 0.124 0.340 0.200 0.234 0.590
Autoless 1 0.638 0.225 0.261 -0.015 0.336 0.123 0.148 0.556
Low-inc JS 1 0.546 0.365 -0.087 0.681 0.370 0.333 0.707
Hispanic 1 0.004 -0.125 0.871 0.374 0.389 0.533
Black 1 -0.181 0.212 0.093 0.516 0.247
Asian 1 -0.121 -0.008 -0.079 0.029
Low Education 1 0.367 0.501 0.531
Young 0.182 0.241
FemHH w/ Chi 1 0.291
Transit mode 1
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

97
Table 5.3.2 Correlation between Variables (2000)

LF
partici
Emp rate Time Low JA Autoless Low Job
Seeker
Hispanic Black Asian Low Edu Young FemHH
w/ child
Transit
LF participation 1 0.399 -0.124 -0.192 -0.422 -0.516 -0.415 -0.245 0.021 -0.574 -0.306 -0.352 -0.360
Employ rate 1 -0.180 -0.273 -0.465 -0.697 -0.429 -0.344 0.135 -0.533 -0.302 -0.587 -0.473
Commute Time 1 -0.090 0.117 0.196 0.145 0.201 -0.078 0.186 0.042 0.218 0.326
Low-inc Access 1 0.428 0.470 0.284 0.177 -0.002 0.306 0.141 0.228 0.572
Autoless 1 0.635 0.304 0.281 -0.113 0.410 0.168 0.362 0.605
Low-inc JS 1 0.509 0.308 -0.108 0.632 0.370 0.601 0.723
Hispanic 1 -0.006 -0.264 0.904 0.401 0.543 0.515
Black 1 -0.200 0.170 0.077 0.468 0.204
Asian 1 -0.198 -0.095 -0.229 -0.092
Low Education 1 0.434 0.637 0.554
Young 1 0.304 0.192
FemHH w/ Chi 1 0.466
Transit mode 1
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

98
Labor Force Participation Rate
The predictive power of the labor force participation model is 0.39 in 1990 and 0.49 in 2000.
Low-income job accessibility has expected significant and positive impacts on labor force
participation rates in both 1990 and 2000. In other words, low-income job seekers living in
census tracts with higher job accessibility are more likely to participate in the labor market, all
else being equal.
The share of low-income job seekers without automobile has significant and negative impacts on
labor force participation rates, as expected. It confirms the anticipation that automobile
ownership affects costs of spatial job search, and those without an automobile in their
households would be less likely to actively search for jobs.
For other control variables, if a census tract has a higher percentage of low-income job seekers,
the labor force participation rate in that census tract is more likely to be lower. Race
(percentages of Hispanic, Black and Asian) has mixed effects on labor force participation rates.
The share of Hispanics has significant and positive impacts on labor force participation rates,
which is consistent with much of the existing literature as reviewed in Chapter 3. It is not
definite if Blacks and Asians have consistent impacts on labor force participation rates. The
share of Blacks has marginally significant and positive impacts in 1990, but negative impacts in
2000. The share of Asians does not have significant impacts in 1990, but has negative impacts

99
in 2000. After income is controlled, different racial compositions do not have consistent effects
on labor force participation.
The share of job seekers with low educational attainment has negative impacts on labor force
participation rates. It is noteworthy that its standardized coefficient is the second highest in
1990 and the highest in 2000, suggesting the strong effects of education on low-income job
seekers’ labor force participation rates. The share of youth has generally negative impacts on
labor force participation rates, as expected. The share of job seekers in female-headed families
with children has mixed effects: it has negative impacts in 1990, but positive impacts in 2000.
This might be explained by the new legislations passed in the late 1990s which encouraged labor
force participation of welfare recipients, many of whom are job seekers in female-headed
families with children.
Employment Rate
The predictive power of the employment rate model is 0.72 in 1990 and 0.62 in 2000.
Low-income job accessibility has expected significant and positive impacts on employment rates
in 2000, but not in 1990. In other words, job seekers with identical automobile ownership and
socioeconomic characteristics would not have significantly different employment status no
matter their job accessibility in 1990. However, it does not necessarily mean that job
accessibility does not affect employment rates. Detailed bisection analysis in the next section
will explore into where and to whom job accessibility matters.

100
Different from the expectation, the share of autoless low-income job seekers has significant and
positive impacts on employment rates in 1990, and insignificant impacts in 2000. This is
caused by the interaction of automobile ownership and the share of low-income job seekers in
explaining employment rates. Dropping the share of low-income job seekers from the
employment rate model results in the expected negative effects of the share of autoless job
seekers. In other words, the effects of the share of autoless job seekers might be overshadowed
by those of the share of low-income job seekers, which has the greatest effects on employment
rates than all of the other independent variables. There are several possible reasons for the
unexpected effects of automobile ownership. First, low-income autoless job seekers tend to
live close to jobs so that they do not need automobiles to commute to jobs. Second,
low-income job seekers would not participate in the labor force, that is, actively search for jobs,
if they do not have a good chance to be employed. Then automobile ownership does not have
expected impacts on employment rates which are partly based on the number of job seekers in
the labor force. On the other hand, automobile ownership significantly affects low-income job
seekers’ labor force participation rates, particularly for those unemployed autoless workers since
job searching costs are especially high for them. Therefore, low-income autoless job seekers
are very likely to drop out of labor force if they are unemployed.
Race/ethnicity (percentages of Hispanic, Black and Asian) has different effects on employment
rates. The share of Hispanics and the share of Blacks have significant and negative impacts on

101
employment rates. But the share of Asians has positive impacts, which indicates higher
employment rates among Asians.
Table 5.4 Estimation Results for the Full Sample (Standardized Coefficients and t-scores)
LF Participation Employment Rate Commute Time
1990 2000 1990 2000 1990 2000
Job Access
0.151 0.090 0.019 0.031 -0.420 -0.420 Low-income Job
Accessibility (8.11) (5.29) (1.49) (2.10) (-19.53) (-19.53)
Auto ownership
% Autoless -0.166 -0.088 0.049 0.026 -0.181 -0.117
(-7.68) (-4.51) (3.32) (1.52) (-7.67) (-4.88)
Socioeconomic
% Low-inc JS -0.278 -0.276 -0.485 -0.540 -0.256 -0.095
(-9.51) (-11.04) (-24.47) (-24.60) (-7.49) (-2.90)
% Hispanic 0.645 0.305 -0.095 -0.066 -0.171 -0.170
(17.22) (7.22) (-3.75) (-1.79) (-4.03) (-3.31)
% Black 0.039 -0.113 -0.153 -0.181 0.159 0.171
(1.71) (-5.98) (-9.93) (-11.00) (6.52) (7.61)
% Asian 0.010 -0.084 0.038 0.035 -0.001 -0.007
(0.57) (-5.24) (3.37) (2.50) (-0.05) (-0.39)
% Low Education -0.607 -0.810 -0.204 -0.162 0.276 0.182
(-14.87) (-17.89) (-7.35) (-4.12) (6.17) (3.36)
% Young -0.237 -0.027 -0.033 -0.034 -0.209 0.001
(-13.13) (-1.64) (-2.67) (-2.40) (-10.71) (0.03)
% FemHH w/Child -0.073 0.191 -0.146 -0.029 -0.018 0.010
(-3.18) (8.40) (-9.43) (-1.48) (-0.74) (0.35)
Commute Mode
% by transit 0.720 0.662
(24.70) (22.66)
Adj R
2
0.39 0.49 0.72 0.62 0.29 0.28
* Bold = sig at p < .05, Underscored Italic = sig at p < .10
(N=2446 in 1990 and N=2448 in 2000)

102
The share of job seekers with low educational attainment, the share of youth and the share of job
seekers in female-headed families with children have negative impacts on employment rates
consistently, and such effects are consistent with the expectation.
Commute Time
The predictive power of the commute time model is 0.29 in 1990 and 0.28 in 2000.
Low-income job accessibility has the expected significant and negative impacts on commute
time, and the impacts are the second largest among all independent variables, following the share
of transit commuters. It confirms the expectation that identical job seekers with the same
commute mode would have shorter commute time in census tracts with higher job accessibility.
Automobile ownership has negative impacts on commute time, which is different from the
expectation. This suggests that after commute mode (the share of transit commuters) is
controlled, autoless low-income job seekers are more likely to have shorter commute time. A
part of the reason is that they do not have many resources to afford extensive job searches in
further distances or longer commutes. Moreover, because of the inflexibility of their travels
without automobiles, autoless low-income job seekers are less likely to accept job offers which
require longer commutes, especially if they have family or other responsibilities.

103
The share of low-income job seekers has negative impacts on commute time. This is expected
as they are more likely to obtain low-wage and low-skilled jobs, which do not offer commute
compensation for long commutes.
Effects of each race/ethnic group on commute time are different. The share of Hispanics has
negative impacts on the average commute time, which suggests that they are more likely to work
in places close to their residences. However, the share of Blacks is positively correlated with
commute time, which is consistent with the SMH that Blacks tend to have longer commutes.
Combing the findings of the three labor market models, this research demonstrates that Blacks
generally have lower labor force participation rates, lower employment rates and longer
commute time, all else being equal. This suggests the significant disadvantages of Blacks in the
labor market compared to other race/ethnic groups, even after income is controlled. The share
of Asians has no significant impacts on commute time.
The share of job seekers with low educational attainment has significant and positive impacts on
commute time. This is different from the expectation, and it might be explained by the
suburbanization of low-skilled jobs, which reduced the possibility of low-educated job seekers to
find jobs near their inner-city residences. The share of youth has significant and negative
impacts on commute time in 1990, but not in 2000. Youth is expected to have shorter commute
time because of their low-wage jobs and their limited transportation resources. The share of job

104
seekers in female-headed families with children does not have significant impacts on commute
time.
Mode split has greater impacts on commute time than all other variables, and higher share of
transit commuters in the census tracts is very likely to increase the average commute time.
These are consistent with the expectations.

Bisection Analysis of Job Accessibility Coefficient Estimates
To test the robustness of the effects of job accessibility in the previous labor market regression
models, this section performs bisection analysis, particularly to test whether the impacts of job
accessibility are consistent across different sub-groups. Job accessibility measures spatial
barriers to competitive job opportunities. However, different population segments have
differential transportation and other resources to overcome such spatial barriers, thus job
accessibility affects their labor market outcomes to various degrees. Conceptually, job
accessibility would affect labor market outcomes of the disadvantaged groups more than other
groups.
This research uses a simple bisection method to divide the full sample in the study area – 2,473
census tracts – using the regional median of each variable. There are many different ways to
bisect the full sample. This research chooses to use the regional medians to ensure a reasonable

105
number of census tracts in each bisection, as well as some consistency in defining the
benchmarks to bisect the full sample using different variables at different times. Specifically,
census tracts with job accessibility higher than the regional median are high accessibility census
tracts, and those with job accessibility lower than the regional median are low accessibility
census tracts. High accessibility census tracts are generally in the inner cities and the inner-ring
suburbs, where low-income and other disadvantaged groups tend to concentrate. The labor
market outcomes are slightly different between the two groups of census tracts as shown in Table
5.5. Relatively, high job accessibility census tracts have slightly lower labor force participation
rates, lower employment rates and shorter commute time compared to low job accessibility
census tracts. The differences are caused by the distinctive socioeconomic characteristics
between the two groups of census tracts. High job accessibility tracts have higher
concentrations of the disadvantaged groups (autoless, low-income, race/ethnicity minorities,
low-educated, young, in female-headed family with children) who are expected to have lower
labor force participation rates and lower employment rates.
The full sample is also bisected based on each control variable, and job accessibility is examined
in these bisections to test if it has larger impact in census tracts with concentrations of those
disadvantaged groups. Census tracts with the share of each population segment (autoless,
low-income, Hispanic, Black, Asian, low-educated, youth, and female-headed families with
children) above their respective regional medians are tracts “with concentration” of these

106
disadvantaged population segments, and census tracts with the share below the regional medians
are those “without concentration” of these population segments.
Table 5.5 Labor Market Outcomes of Low and High Job Accessibility Bisections
1990 2000
Variables
Bi-
section Mean Median Std Dev Mean Median Std Dev
LF participation High 0.67 0.68 0.10 0.60 0.60 0.09
Low 0.68 0.69 0.10 0.61 0.63 0.09
Employ rate High 0.92 0.93 0.05 0.91 0.92 0.05
Low 0.94 0.95 0.03 0.92 0.94 0.05
Commute Time High 28.86 28.88 3.77 28.31 28.24 3.35
Low 29.72 29.31 4.98 28.60 28.66 3.19
Because of the potential interrelationships between job accessibility and each control variable, if
job accessibility has larger effects in census tracts with high job accessibility, the effects might
be a result of concentrations of certain disadvantaged groups who are more sensitive to the
change of job accessibility. Therefore, effects of job accessibility also need to be examined in
the “sub-bisections” which are defined based on both job accessibility and each control variable.
For example, to fully understand the distinctive impacts of job accessibility and auto ownership,
job accessibility will be examined in four sub-bisections: 1) census tracts with high job
accessibility and with concentration of autoless job seekers; 2) census tracts with high job
accessibility but without concentration of autoless job seekers; 3) census tracts with low job
accessibility and with concentration of autoless job seekers; and 4) census tracts with low job
accessibility and without concentration of autoless job seekers. The analysis based on the
“sub-bisections” will be repeated for each control variable. The average size of the

107
sub-bisections is around 620 (2,473 divided by 4), but the actual size ranges from 430 to 790
depending on the interrelationship between job accessibility and the control variable.
Generally, all sub-bisections have reasonable size for the regression analysis.
Labor Force Participation Rate
First, this research analyzes the impacts of job accessibility on labor force participation rates for
each bisection and sub-bisection. Table 5.6 shows the standardized coefficients and the t-scores
of job accessibility in both 1990 and 2000s. Effects of job accessibility in the bisections based
on either job accessibility or each control variable are shaded, and effects of job accessibility in
the sub-bisections based on both job accessibility and each control variable are not shaded.
Certainly, each control variable has different effects on labor force participation rates in different
bisections and sub-bisections, but control variables are not the focus of this research, and hense
their coefficient estimates are not shown. Moreover, effects of job accessibility in the
bisections based on one control variable might be influenced by other control variables which are
also bisected in a way because of the nonrandom relationships between the control variables.
However, there are 32 different sub-bisections based on job accessibility and eight control
variables, and thus analysis on those 32 different sub-bisections is comprehensive enough to
provide a general picture of job accessibility’s impacts on different population segments.

108
Job accessibility has significant positive impacts on labor force participation rates for the full
sample as examined in the previous section. But when the full sample is bisected into two
samples, one with high job accessibility and one with low job accessibility, impacts of job
accessibility are different between the two samples. In 1990, job accessibility has significant
impacts (β=0.168) in tracts with high accessibility, and marginally significant impacts (β=0.043)
in tracts with low accessibility. In other words, job accessibility is more likely to have
significant effects if it is above a certain benchmark. This is reasonable as job accessibility
measures spatial barriers to relative job opportunities. Basically, low job accessibility census
tracts provided relatively low job supply for the number of job seekers compared to the regional
level, thus a high percentage of job seekers need to search for and acquire jobs in high job
accessibility census tracts, where jobs are relatively abundant. The insignificant effects of job
accessibility on low-income job seekers’ labor force participation rates in the low accessibility
census tracts partly reflect the residential endogeneity of some job seekers who make choices to
trade commute for housing and other amenities, although low-income job seekers still face
residential segregation and housing affordability issues.
However, in 2000, job accessibility does not have significant impacts on labor force participation
rates in high accessibility census tracts, and has marginal significant impacts in low accessibility
tracts. This indicates that identical job seekers in different locations do not have significantly
different labor force participate rates. One possible explanation is that job accessibility does
not affect labor force participation rates in 2000 after automobile ownership and socioeconomic

109
variables are controlled. A more possible reason is that exogenous shocks of changing
legislations provided strong incentives for welfare recipients, many of whom are in
female-headed families with children, to participate in the labor force regardless of the spatial
barriers – job accessibility − they face. However, no data on welfare receipts are directly
available at the census tract level although the share of female-headed families with children can
be a proxy. Regression results show that the share of female-headed family with children has
significant and positive impacts on labor force participation rates in the full sample, as well as in
both bisections of low and high accessibility census tracts in 2000.
Table 5.6 also shows effects of job accessibility on labor force participation rates for each of the
bisection based on the control variables. Results are consistent that job accessibility has
significant and positive impacts in all bisections, except in census tracts without concentration of
Asians in 1990. However, this does not necessarily mean that job seekers are more likely to
participate in the labor force if they live in census tracts with higher job accessibility, because the
sample size of each bisection is still big and there are great variations within each bisection.
Impacts of job accessibility in the sub-bisections based on both job accessibility and each control
variable are examined to distinguish their respective effects.
Comparing job accessibility in the sub-bisections, which are un-shaded in Table 5.6, job
accessibility is more likely to have significant impacts in tracts with high job accessibility (the 4
th

column and the 7
th
column) than in tracts with low job accessibility (the 5
th
column and the 8
th

110
column) with very few exceptions. Again, this indicates that impacts of job accessibility on
labor force participation rates are more likely to be significant if job accessibility is above a
certain benchmark.
Job accessibility is more likely to have significant impacts in census tracts with concentrations of
the disadvantaged groups than in those without such concentrations. In 1990, although job
accessibility has significant impacts in all sub-bisections with high job accessibility, the
coefficients are greater in the sub-bisections with concentrations of the disadvantaged groups
than those without such concentrations. This suggests that job accessibility is especially
important for disadvantaged population segments, who have less transportation and other
resources to overcome spatial barriers.
As a result of the two reasons cited above, job accessibility constantly has significant impacts on
labor force participation rates in the sub-bisections with both high job accessibility and with
concentrations of disadvantage groups except Blacks and Asians in 2000. Moreover, job
accessibility does not affect labor force participation rates in the opposite sub-bisections ─ tracts
with low job accessibility and without concentrations of the disadvantaged groups, except
autoless job seekers in 1990 and Hispanics in 2000. It might be explained by the residential
endogeneity of job seekers in the majority groups, who choose to live in places with lower job
accessibility, mainly suburban communities, to have better quality of life and amenities.

111
Table 5.6 Effects of Job Accessibility on Labor Force Participation Rate by Bisection of Job
Accessibility and the Control Variables
1990 2000
Job accessibility Job accessibility
Control
Variables
Bi-
section
All High Low All High Low
All samples 0.151 0.168 0.043 0.090 0.025 0.040
(census tracts) (8.11) (6.90) (1.65) (5.29) (1.10) (1.71)
Autoless More 0.157 0.244 -0.005 0.104 0.062 -0.114
(5.97) (7.74) (-0.13) (4.33) (2.08) (-2.31)
Less 0.124 0.118 0.056 0.078 0.060 0.017
(4.90) (2.76) (1.72) (3.34) (1.52) (0.58)
Low-income More 0.180 0.215 0.073 0.126 0.079 0.089
(7.03) (7.37) (1.71) (4.79) (2.43) (2.11)
Less 0.066 0.164 0.004 0.054 0.002 0.006
(2.36) (3.56) (0.10) (2.14) (0.04) (0.19)
Hispanics More 0.218 0.231 0.022 0.135 0.141 0.026
(8.84) (8.09) (0.55) (5.47) (4.67) (0.68)
Less 0.112 0.109 0.024 0.098 -0.026 0.050
(4.86) (3.18) (0.81) (3.98) (-0.69) (1.69)
Blacks More 0.131 0.099 0.068 0.101 -0.03 0.120
(5.66) (3.33) (2.03) (4.32) (-0.98) (3.59)
Less 0.099 0.268 0.013 0.092 0.106 -0.008
(3.64) (7.36) (0.37) (3.94) (3.43) (-0.27)
Asian More 0.259 0.239 0.096 0.155 0.043 0.084
(9.61) (6.81) (2.56) (6.07) (1.20) (2.50)
Less 0.031 0.121 -0.055 0.048 0.039 -0.057
(1.19) (3.72) (-1.58) (2.08) (1.38) (-1.74)
Low-education More 0.202 0.231 0.028 0.174 0.143 0.025
(7.35) (7.05) (0.71) (6.21) (4.10) (0.61)
Less 0.112 0.131 0.048 0.068 -0.010 0.035
(4.28) (3.22) (1.40) (2.51) (-0.23) (1.07)
Youth More 0.196 0.224 0.060 0.110 0.105 0.061
(8.73) (8.66) (1.65) (4.51) (3.28) (1.71)
Less 0.097 0.105 0.025 0.092 -0.038 0.011
(3.93) (2.81) (0.77) (3.58) (-1.00) (0.36)
FemHHw/Child More 0.166 0.215 0.026 0.061 0.097 0.005
(7.01) (7.52) (0.66) (2.62) (3.08) (0.16)
Less 0.111 0.155 0.028 0.120 -0.043 0.049
(4.29) (3.94) (0.86) (4.79) (-1.12) (1.58)
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

112
Furthermore, effects of job accessibility are generally higher in 1990 than in 2000, indicating that
in 1990, spatial barriers are more likely to influence job seekers’ inclination to participate in the
labor force than in 2000. However, actively looking for a job does not guarantee finding a job.
Bisection analysis of employment rates and commute time will provide a comprehensive
depiction of job accessibility’s impacts on labor market performances.
Employment Rate
Table 5.7 shows the standardized coefficients and the t-score of job accessibility on employment
rates in different bisections and sub-bisections in 1990 and 2000.
Job accessibility has significant and positive impacts on employment rates for the full sample in
2000, but not in 1990. But when the full sample is bisected by job accessibility, impacts of job
accessibility are consistently significant in the bisection with high job accessibility, but not in the
bisection with low job accessibility. Particularly, impacts are negative, although insignificant,
in the bisection of low job accessibility in 2000. Similar as the effects of job accessibility on
labor force participation rates, job accessibility is more likely to affect employment rates if it is
above a certain benchmark. In other words, in the bisection of low accessibility census tracts,
employment rates are not significantly different regardless of the spatial barriers to relative job
opportunities. It suggests that a higher share of workers in low accessibility census tracts
acquire jobs in further distances. The examination of job accessibility’s impacts on commute
time in the next section can test this assumption.

113
Table 5.7 Effects of Job Accessibility on Employment Rate by Bisection of Job Accessibility
and the Control Variables
1990 2000
Job accessibility Job accessibility
Control
Variables
Bi-
section
All High Low All High Low
All samples 0.019 0.069 0.004 0.031 0.057 -0.027
(census tracts) (1.49) (3.91) (0.24) (2.10) (2.76) (-1.36)
Autoless More 0.041 0.128 -0.015 0.019 0.093 -0.034
(2.12) (4.96) (-0.53) (0.86) (3.12) (-1.14)
Less -0.006 0.001 0.001 0.033 0.029 -0.030
(-0.35) (0.04) (0.03) (1.55) (0.84) (-1.15)
Low-income More 0.048 0.128 0.015 0.066 0.100 -0.054
(2.17) (4.96) (0.38) (2.51) (3.05) (-1.32)
Less -0.050 -0.022 -0.001 -0.008 0.078 -0.031
(-2.33) (-0.58) (-0.03) (-0.34) (2.00) (-1.07)
Hispanics More 0.043 0.108 0.024 0.051 0.078 0.037
(2.07) (4.09) (0.73) (2.22) (2.56) (1.09)
Less 0.001 0.044 -0.015 0.024 0.02 -0.053
(0.04) (1.35) (-0.52) (1.08) (0.60) (-1.86)
Blacks More 0.070 0.112 0.045 0.049 0.051 -0.061
(3.66) (4.36) (1.59) (2.00) (1.49) (-1.94)
Less -0.047 0.013 -0.009 0.002 0.079 0.002
(-2.64) (0.53) (-0.35) (0.12) (2.97) (0.09)
Asian More 0.002 0.033 -0.026 0.081 0.05 -0.046
(0.08) (1.06) (-0.92) (3.43) (1.40) (-1.69)
Less 0.012 0.078 -0.006 -0.009 0.072 -0.001
(0.69) (3.29) (-0.27) (-0.45) (2.78) (-0.02)
Low-education More 0.068 0.149 0.010 0.035 0.064 0.052
(2.94) (5.31) (0.29) (1.35) (1.90) (1.38)
Less -0.049 -0.023 -0.004 0.042 0.117 -0.061
(-2.15) (-0.62) (-0.14) (1.75) (3.22) (-2.1)
Youth More 0.069 0.126 0.043 0.080 0.067 0.000
(3.44) (5.07) (1.24) (3.20) (2.16) (0.01)
Less -0.057 -0.002 -0.024 -0.026 0.019 -0.057
(-2.68) (-0.06) (-0.85) (-1.24) (0.58) (-2.06)
FemHHw/Child More 0.063 0.123 -0.024 0.043 0.061 0.026
(3.71) (5.32) (-0.96) (1.90) (2.01) (0.86)
Less -0.038 -0.030 0.032 0.004 0.004 -0.058
(-1.79) (0.92) (1.08) (0.19) (0.13) (-1.97)
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

114
In the bisections based on each control variable, job accessibility is more likely to have
significant impacts on employment rates in those with concentrations of the disadvantaged
groups than in those without such concentrations, with the exceptions of Asians in 1990, and
autoless and low-educated job seekers in 2000. This is consistent with the conceptual
framework that spatial barriers are more likely to affect the employment status of the
disadvantaged groups because of their limited resources for spatial job searches and commutes.
Comparing the effects of job accessibility in the sub-bisections based on job accessibility and
each control variable, job accessibility is more likely to affect the sub-bisections with high job
accessibility than those with low job accessibility, to affect those with concentrations of each
disadvantage group than those without such concentrations. This again confirms the repeated
findings. First, job accessibility is more likely to have impacts on employment rates if it is
above a certain benchmark. The reason that job accessibility does not matter as much in census
tracts with lower job accessibility might be related with some low-income job seekers’
residential endogeneity. They choose to live in low accessibility census tracts, usually in the
suburbs, and endure longer commutes. Second, the impacts of job accessibility are especially
important for the disadvantaged groups, who have fewer transportation and other resources to
overcome spatial barriers.

115
Table 5.8 Effects of Job Accessibility on Commute Time by Bisection of Job Accessibility
and the Control Variables
1990 2000
Job accessibility Job accessibility
Control
Variables
Bi-
section
All High Low All High Low
All samples -0.420 -0.215 -0.309 -0.420 -0.308 -0.223
(census tracts) (-19.53) (-7.21) (-11.76) (-19.53) (-11.26) (-7.61)
Autoless More -0.245 -0.220 -0.208 -0.240 -0.237 -0.114
(-7.61) (-5.41) (-4.66) (-7.82) (-6.55) (-2.31)
Less -0.438 -0.254 -0.338 -0.483 -0.398 -0.258
(-17.43) (-6.14) (-10.11) (-18.44) (-9.40) (-7.10)
Low-income More -0.241 -0.217 -0.258 -0.195 -0.236 -0.110
(-7.49) (-5.54) (-5.54) (-5.79) (-5.99) (-2.05)
Less -0.433 -0.261 -0.321 -0.512 -0.414 -0.260
(-17.54) (-5.98) (-9.87) (-20.36) (-9.67) (-7.46)
Hispanics More -0.423 -0.296 -0.290 -0.283 -0.240 -0.147
(-12.32) (-7.48) (-6.96) (-8.44) (-6.10) (-3.12)
Less -0.375 -0.184 -0.315 -0.496 -0.407 -0.270
(-14.73) (-4.53) (-9.89) (-18.54) (-10.42) (-7.56)
Blacks More -0.380 -0.149 -0.304 -0.268 -0.260 -0.193
(-12.12) (-3.67) (-7.89) (-7.88) (-5.79) (-4.42)
Less -0.480 -0.341 -0.335 -0.550 -0.321 -0.258
(-16.07) (-7.53) (-9.28) (-19.83) (-8.96) (-6.54)
Asian More -0.380 -0.237 -0.278 -0.461 -0.366 -0.171
(-12.69) (-5.42) (-7.67) (-15.65) (-8.92) (-4.35)
Less -0.414 -0.177 -0.260 -0.382 -0.246 -0.182
(-13.51) (-4.29) (-7.02) (-12.60) (-6.89) (-4.15)
Low-education More -0.331 -0.214 -0.273 -0.201 -0.204 -0.102
(-9.12) (-4.97) (-6.29) (-5.64) (-4.86) (-2.05)
Less -0.395 -0.217 -0.310 -0.521 -0.439 -0.262
(-15.52) (-5.16) (-9.69) (-20.26) (-11.58) (-7.48)
Youth More -0.336 -0.292 -0.205 -0.287 -0.214 -0.185
(-11.05) (-7.60) (-5.14) (-8.87) (-5.72) (-3.96)
Less -0.410 -0.162 -0.340 -0.500 -0.445 -0.229
(-14.78) (-3.49) (-10.13) (-18.25) (-11.30) (-6.19)
FemHHw/Child More -0.305 -0.271 -0.193 -0.274 -0.238 -0.147
(-9.29) (-6.27) (-4.34) (-8.11) (-6.05) (-3.02)
Less -0.444 -0.184 -0.330 -0.520 -0.431 -0.260
(-15.95) (-4.18) (-9.97) (-18.97) (-10.16) (-7.16)
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

116
Temporal Analysis
This research tests whether job accessibility has the same effects on the labor market outcomes
between 1990 and 2000. Table 5.9 shows the unstandardized coefficients of job accessibility,
standard errors, and the t-scores to compare the coefficients by the bisections of job accessibility.
For labor force participation rates, job accessibility has significantly smaller impacts in 2000
than in 1990, based on the coefficients and the t-test results. The smaller impacts in 2000 are
particularly caused by the insignificance of job accessibility’s impacts in the high job
accessibility bisection in 2000, but the impacts in the low job accessibility bisection remained
constant. It is noteworthy that most disadvantaged groups, especially those in female-headed
families with children, tend to locate in census tracts with high job accessibility, which might be
particularly affected by the exogenous factor of welfare reform in the 1990s.
Effects of job accessibility on employment rates remain constant between 1990 and 2000, in the
full sample, as well as in the two bisections, although job accessibility has significant impacts
only in high accessibility bisections.
Job accessibility has smaller effects on commute time in 2000 than in 1990, and such difference
mainly occurred in the low accessibility bisection. In other words, in 1990, job accessibility
shortens commute time to a greater extent than in 2000. This can be explained by changes in
the variation of commute time; low accessibility census tracts have greater standard deviation in

117
1990 than in 2000, indicating that the suburbanization in the 1990s relatively improves
jobs-housing balance in the suburbs in 2000.
Table 5.9 Coefficients of Job Accessibility and T-scores to Compare the Coefficients
Unstandardized
Coefficient
T-test to compare
coefficients
Dependent
Variables
Bi-
section
Year
B Std.Err Diff t-value
LF Participation All 1990 0.083 0.010
2000 0.047 0.009
0.036 2.68
High 1990 0.153 0.022
2000 0.018 0.017
0.135 4.86
Low 1990 0.036 0.022
2000 0.037 0.022
-0.001 -0.03
Employment Rate All 1990 0.005 0.004
2000 0.009 0.004
-0.004 -0.71
High 1990 0.035 0.009
2000 0.025 0.009
0.01 0.79
Low 1990 0.001 0.006
2000 -0.013 0.010
0.014 1.20
Commute Time All 1990 -11.475 0.587
2000 -8.145 0.417
-3.330 -4.62
High 1990 -8.181 1.134
2000 -8.468 0.752
0.287 0.21
Low 1990 -14.938 1.270
2000 -7.898 1.037
-7.04 -4.29
* Bold = sig at p < .05, Underscored Italic = sig at p < .10

5.3 Chapter Summary
This chapter shows that in both 1990 and 2000 job accessibility has significant impacts on
low-income job seekers’ labor market outcomes. Places with high job accessibility are more
likely to have higher labor force participation rates and/or higher employment rates, and shorter

118
commute time, all else being equal. The impacts are particularly significant in census tracts
with higher accessibility and with concentration of the disadvantaged groups. On one hand, job
accessibility is more likely to affect low-income job seeker’s labor market performance if it is
above a certain benchmark. It might be associated with the residential endogeneity of some
low-income job seekers that they choose to live in low accessibility areas to trade for other
amenities. On the other hand, the disadvantaged groups have limited resources in terms of
money, transportation and time, to overcome the spatial barriers. Therefore, job accessibility
tends to have more significant impacts on their labor market performances. On the contrary,
the affluent majority has many means to overcome spatial barriers, thus their labor market
outcomes are not as related to job accessibility.
Moreover, job accessibility generally has smaller impacts on labor force participation rates in
2000 than in 1990. This suggests that other socioeconomic characteristics and exogenous
shocks in the labor market play more important roles in 2000. Job accessibility has smaller
impacts on commute time in 2000 than in 1990, particularly in low accessibility census tracts,
which are generally in the suburbs.
This research finds that some variables have distinctive effects on different labor market
outcomes. Particularly, Hispanics tend to have higher labor force participation rates but lower
employment rates. This suggests that all three labor market outcomes should be examined to
fully describe the labor market performances of different population segments.

119
It is also important to examine a large study area in greater details by different samplings. This
research applies a simple version of the bisection method, using the regional medians of job
accessibly and each control variable, to select samples from the study area based on different
characteristics. The results are very constructive to tell a detailed story and to confirm the
conceptual framework that job accessibility affects low-income job seekers’ labor market
outcome.
But job accessibility is not the only significant factor in explaining labor market outcomes.
After income is controlled, educational attainment consistently has great impacts on all three
labor market outcomes. The impacts of race composition are mixed, but in general, African
Americans are still more likely to be disadvantaged with lower labor force participation rates,
lower employment rates and longer commutes, even within the same income class.
Spatial barriers, measured by job accessibility, still matter for low-income job seekers’ labor
market outcomes. Therefore, understanding the importance of spatial barriers is critical for
planners and policy makers who seek to better allocate resources to address the efficiency and
equity issues.

120
CHAPTER SIX
CONCLUSION AND FUTURE RESEARCH

6.1 Summary of Findings
This dissertation answers the research question and confirms that changing urban spatial
structure has affected low-income job seekers’ job accessibility and their labor market outcomes.
Job accessibility is the key throughout this dissertation since it is a direct indicator of spatial
mismatch between jobs and residences in the urban structure and it impacts labor market
outcomes.

Urban Spatial Transformation
This research shows distribution and allocation patterns of population and employment by
income groups. Particularly, high-income job seekers and jobs suburbanized at a similar pace
to similar areas in the region, which might be caused by the tendency of jobs following
high-income people, or by their similar location preferences. It is interesting to find that
high-income job seekers, who have more resources for travel, are better off in terms of job
accessibility.

121
On the other hand, low-income job seekers clearly are lagged behind in the redistribution process.
The mechanism of their slow suburbanization is not examined in this research. Possible
explanations include housing market discrimination, unaffordability of large houses in the
suburbs, or simply low-income people’s preferences to live in certain neighborhoods.
Moreover, low- and high-income jobs have much similarity in their suburbanization process.
Comprehensive location choice models for low-income job seekers and jobs will be very helpful
to explain the spatial transformation. In general, this research clearly tells a consistent story of
spatial mismatch, even in a polycentric metropolitan area. However, the mismatch extends
beyond the boundary of the inner cities. Now, it is an issue in the inner-ring suburbs as well.

Low-income Job Accessibility and Labor Market Outcomes
In the overall urban spatial transformation, low-income job accessibility has been affected
differently. This research expands the demographic, geographic and temporal scope of the
SMH. First, different from most of the existing literature on the SMH, this research focuses on
low-income job seekers to delve into the issue of economic class and finds that low-income job
seekers generally have lower job accessibility and poorer labor market performances than other
job seekers.

122
Second, this research expands the geographic scope by recognizing the polycentric pattern of the
study area, and finds that job accessibility differs not only between inner cities and suburbs, but
also within suburbs. It is obvious that places with high job accessibility are usually the places
with employment concentrations. Specifically, in the polycentric Los Angeles region, some
suburban areas, especially those in the inner-ring suburbs, function similarly to inner cities in the
original monocentric model. Low-income job seekers started to move into those areas, but jobs
and high-income job seekers moved further out into the outer-ring suburbs. As a result,
compared to high-income job seekers, low-income job seekers in the inner-ring suburbs are
particularly disadvantaged in terms of job accessibility as well as the change of job accessibility
over time. Spatial mismatch still exists, but now partly in the suburbs.
Third, this research also finds that over time, low-income job seekers are worse off than
high-income job seekers in terms of job accessibility, primarily because of the relatively slower
growth of low-income jobs. This is caused by the combination of a decline in jobs requiring
lower skill-sets as a result of economic restructuring, and an increase in immigrants with lower
educational attainments that increased the number of low-income job seekers. Moreover,
low-income job seekers lagged behind jobs and high-income job seekers in the suburbanization
process, which is a major reason for the spatial mismatch of low-income job seekers, particularly
in the inner-ring suburbs.

123
Finally, to explain why and how spatial barriers affect labor market outcomes, this research
developed a conceptual model utilizing spatial job search theory and examined to what extent job
accessibility impacts low-income job seekers’ labor market outcomes. Results suggest that
overall job accessibility has significant and positive impacts on labor market outcomes by
increasing labor force participation rates and employment rates, and reducing commute time.
There are two noteworthy findings. First, the impacts are particularly significant in census
tracts with higher accessibility. Second, job accessibility tends to have larger impacts in census
tracts with concentrations of the disadvantaged groups, who have limited resources to overcome
the spatial barriers.
In general, urban spatial structure has affected low-income job seekers’ job accessibility
differently, and job accessibility impacts their labor market outcomes.

6.2 Policy and Planning Implications
This research has significant policy and planning implications. Much research has been done to
thoroughly discuss the three streams of strategies to promote minorities’ employment prospects:
moving jobs closer to the workers (inner-city development strategy), moving people closer to the
jobs (desegregation strategy), and making it easier for workers to get to existing jobs (mobility
strategy) (Ihlanfeldt and Sjoquist, 1998). This section does not intend to address each strategy,

124
as much literature has already done so. Rather, this section discusses some specific planning
and policy implications derived from this research.

Employment and Housing
Low-income job seekers face two major challenges with respect to their relevant jobs in the Los
Angeles region: the number of low-income jobs has been decreasing because of the economic
restructuring, and low-income jobs suburbanized faster than low-income job seekers. However,
the “inner-city development strategy” is unlikely to work for the benefits of low-income job
seekers. Employment location choices are decided by many factors, such as land rent,
commute costs and agglomeration economics, which play different roles together. Among
these factors, the effects of local development policies to promote employment growth in certain
areas are uncertain (Agarwal, 2009).
Jobs-housing balance is a popular planning topic, especially in regional planning. However, it
is important to carefully distinguish the types of jobs with the types of housing. As explained,
low-income job seekers have totally different job demand from high-income job seekers.
Therefore, it is not efficient to consider jobs-housing balance of all population segments and all
job types together. Rather, it is important to treat population segments differently and
understand their specific needs. Another issue is to balance jobs with housing in what

125
geographic areas. It is impossible to match jobs with housing at small geographic scales as
many people, especially high-income job seekers, have sufficient transportation resources and
prefer amenities and quality of life over short commutes. Realistic analysis has to be based on
reasonable commute sheds and count in distance decay of both people and jobs.
Another issue specifically related to low-income job seekers is affordable housing. Where and
how to provide affordable housing is a sensitive planning issue for local governments, and it is
required by law: “there is need to . . . expand the supply of rental housing that is affordable to
very low income and low-income families . . .” (National Affordable Housing Act of 1990,
Public Law 101-625 Section 202). As much of the research in public policy proposes to solve
the issue by aspatial fiscal policies (e.g. housing vouchers), this research contributes to the
discussion by potentially identifying critical locations of affordable housing. Of course,
decisions as to the location of affordable housing depend on many factors, but employment
locations should be a major consideration as they have great effects on low-income job seekers’
overall socioeconomic status. Findings of this research suggest that at a certain point in time,
low-income job accessibility is relatively high in the inner cities, north Orange County, and some
parts of the outer-ring suburbs. However, over time, low-income job accessibility declined in
the inner cities and the inner-ring suburbs, and increased in the outer-ring suburbs. Therefore,
the decisions on the locations of affordable housing depend on whether policy intends to promote
low-income job seekers’ employment prospects in the short- or in the long-run. The most

126
strategic locations appear to be in the Inland Empire where job accessibility is relatively high,
and consistently increases over time.

Automobile Ownership
Findings of this research show that improving job accessibility affects low-income job seekers’
labor market outcomes. In addition to the locations of jobs and housing, another way to
improve job accessibility is to enhance mobility, which is mainly decided by transportation
networks and transportation resources available. Without further technological breakthroughs,
marginal benefits of improvements on the current transportation networks in the study area are
not expected to be significant. Therefore transportation resources, especially automobile
ownership, of low-income job seekers are particularly important.
In the study area, around 28% of low-income job seekers have no automobile in their households.
But how and to what extent automobile ownership affects their labor market outcomes is
uncertain. Job seekers adjust their labor market behaviors and their commute behaviors in very
flexible ways. A higher share of auto-less job seekers in a census tract have expected negative
effects on labor force participation rates, but not on employment rates or commute time, which
are more significantly impacted by the share of low-income job seekers. This indicates that, all
else being equal, because of the difficulties in spatial job searches particularly associated with

127
automobile ownership, job seekers would not actively looking for jobs unless they are likely to
be employed. Similarly, after the mode split is controlled, census tracts with higher shares of
autoless job seekers are more likely to have shorter commute time because of their limited
transportation resources for extensive job searches and longer commutes.
Automobile ownership is not the determining factor of workers’ commute behaviors. Although
only around 5-6% of low-income workers who have at least an automobile in their households
take public transit, it is surprising to find that around 40-50% of those who do not have an
automobile in their households still commute by private automobiles
14
. Meanwhile, around
40% of workers who commute by transit have at least an automobile in their households, and the
other 60% are autoless. Automobile ownership is an indicator of convenience and ease to travel,
but not necessarily the determining factor of low-income job seekers’ labor market outcomes.
Many researchers argue that a better way to assist low-income job seekers is to assist them to
have the same mobility as others. The most straightforward way is to subsidize private
automobile ownership so that job seekers can have similar resources to overcome spatial barriers.
However, the problem is that the combined costs of an automobile, maintenance and insurance
are still too high for low-income job seekers. Other planning and policy strategies which aim to
improve low-income job seekers’ mobility need to be explored.

14
Calculated from PUMS 1990 and 2000 data for the study area.

128
Other strategies
Related to improving automobile ownership, enhancing transit services for job seekers is also a
popular research topic. Kawabata and Shen (2007) found significant impacts of job
accessibility by transit on commute time, but they did not test its impacts on employment status.
However, from a broader perspective of transportation planning, transit service is regressive in
nature, especially those expensive rail transit systems. Moreover, transit systems have been
heavily subsidized from gas taxes, property taxes and sales taxes, which are regressive as well.
Most transit serves the majority, especially high-income suburban residents who work in the
inner cities. Therefore, low-income job seekers have been disadvantaged greatly, and probably
will continue to be by future transit investments.
Also found in this research, other factors, especially education, have greater impacts on
low-income job seekers’ labor market outcomes. Some strategies, in addition to the above
mentioned, such as job training and education services, might be beneficial in promoting
low-income job seekers’ labor market outcomes. However, the effects of those training and
education programs remain highly controversial (Card and Sullivan, 1998).

129
6.3 Research Limitation and Future Research
Uniqueness of the Los Angeles Region
Certainly, spatial trends and job seekers’ labor market outcomes in this research are
circumstantial based on the Los Angeles metropolitan area, which is unique in many ways.
From the aspatial perspective, Los Angeles has experienced great external shocks which
influence both population and employment distributions. The demographics in the Los Angeles
region have been changed with rapidly increasing immigrants, particularly Hispanics, who are
generally low-income. On the other hand, in additional to the economic restructuring most
metropolitan areas in the U.S. are experiencing, Los Angeles was hit particularly severely by the
withdrawal of the defense and aerospace industries in the 1990s, causing many job opportunities
to disappear.
From the spatial perspective, the Los Angeles region is widely recognized as a dispersed region
with great share of employment and population in the suburbs. However, research also found
that Los Angeles is not an outlier with respect to employment decentralization (Gordon and
Richardson, 1996). Comparative analysis of Los Angeles and other metropolitan areas will
provide an answer.
Because of the intent of this research to analyze the effects of job accessibility on labor market
outcomes, which are affected by changing numbers of jobs and job seekers, this research does

130
not differentiate the spatial and aspatial factors. But future research will explore solely the
spatial factors in the urban spatial transformation, which will provide a better flat form for
comparative analysis across metropolitan areas as well.

Transit Accessibility
With limited data, this research does not delve into job accessibility by transit, which is a major
research topic in transportation planning. Pending data availability, future research will be
conducted to compare job accessibility by automobile and by transit for low-income job seekers,
as well as their respective impacts on labor market outcomes. Based on the existing research
and common knowledge, it is expected that job accessibility by transit is much lower than that by
automobile. With expanding transit networks and increasing traffic congestion, levels of
services by transit might be able to start catching up with that of private automobile in some
areas. However, future research also needs to address how many financial and other resources
have to be put in transit to reach those levels of services. Moreover, various transit services
have to be distinguished in the future study. Obviously, commuter rail is unaffordable to most
low-income job seekers; therefore, expanding and improving commute rails are not very relevant
for aiding disadvantaged groups.

131
2010 Census Data and Other Study Areas
2010 census data will be available in the near future. The new data allow expansion of this
research and examination of low-income job seekers’ labor market performance over a longer
time span, thus providing more reliable results. Using data in two decades, from 1990 to 2010,
can help to tell a better story of the transformation of various places in the suburbs, especially
changing employment and residential locations in the inner-ring suburbs. In addition, not only
the urban structure, but also the demographic and socioeconomic characteristics of the study area
have changed significantly after 2000. The Los Angeles metropolitan area continues to be an
interesting case study area for future research.

132
BIBLIOGRAPHY
Agarwal, A. (2009) Testing the Entrepreneurial City Hypothesis: A Study of the Los Angeles
Region, Ph.D. Dissertation, University of Southern California
Abramson, A., M. S. Tobin, and M. R. VanderGoot (1995) “The changing geography of
metropolitan opportunity: The segregation of the poor in U.S. metropolitan areas, 1970 to 1990,”
Housing Policy Debate Vol. 6, 45-72.
Anas, A., R. Arnott, and K. A. Small (1998) “Urban spatial structure,” Journal of Economic
Literature, Vol. 36, No. 3, 1426-1464.
Aponte, R. (1996) “Urban employment and the mismatch dilemma: Accounting for the
immigration exception,” Social Problems, Vol. 43, 268–283.
Arnott, R. (1998) “Economic theory and the spatial mismatch hypothesis,” Urban Studies, Vol.
35, No. 7, 1171-1118.
Berechman, J. and R. Paaswell (2001) “Accessibility improvements and local employment: an
empirical analysis,” Journal of Transportation and Statistics, 4(2/3), 49-66.
Beckman, R. J., K. A. Baggerly and M. K. McKay (1996) “Creating synthetic baseline
populations,” Transportation Research Part A, Vol. 30, No. 6, 415-429.
Blank, R. M. (2002) “Evaluation welfare reform in the United States,” Journal of Economic
Literature, Vol. 40, 1105-1166.
Blumenberg, E. and M. Manville (2004) “Beyond the spatial mismatch: welfare recipients and
transportation policy,” Journal of Planning Literature, Vol.19, 182-205.
Blumenberg E. and P. Ong (1998) “Job Accessibility and Welfare Usage: Evidence from Los
Angeles,” Journal of Policy Analysis and Management, Vol. 17, No. 4, 639-657.
Brueckner, J. and R. W. Martin (1997) “Spatial mismatch: An equilibrium analysis,” Regional
Science and Urban Economics, Vol. 27, 693-714.
Card, D. and D. Sullivan (1988) “Measuring the effect of subsidized training programs on
movements in and out of employment,” Econometrica, Vol. 56, No. 3497-3530.

133
Cervero, R. (1991) “Land uses and travel at suburban activity centers,” Transportation
Quarterly, Vol. 45, 479–491.
Cervero, R., O. Sandoval, and J. Landis (2002) “Transportation as a stimulus of
Welfare-to-Work: Private versus public mobility,” Journal of Planning Education and Research,
Vol 22, 50-63.
Castells, M. and P. G. Hall (1994) Technopoles of the World: The Making of 21
st
Century
Industrial Complexes. New York: Routledge
Cooke, T. J. (1997) “Geographic access to job opportunities and labor-force participation among
women and African Americans in the greater Boston metropolitan area,” Urban Geography, Vol.
18, Issue 3, 213-227.
Corcoran, M., R. Gordon, D. Laren and G. Solon (1991) “The association between men’s
economic status and their family and community origins,” Journal of Human Resources,
575-601.
Doeringer, P. (1968) “Ghetto labor markets ─ problems and programs,” in Conference on
Transportation and Poverty Report, American Academy of Arts and Sciences (AAAS)
Ellwood, D. T. (1986) “The spatial mismatch hypothesis: Are there teenage jobs missing in the
ghetto?,” in Richard B. Freeman and Harry J. Holzer, eds., The Black Youth Employment Crisis,
Chicago: University of Chicago Press, ch. 4. 147-185.
Fischer, M. (2003) “The relative importance of income and race in determining residential
outcomes in U.S. urban areas, 1970-2000,” Urban Affairs Review, Vol. 38, 669-696.
Giuliano, G. and K. Small (1991) “Subcenters in the Los Angeles region,” Regional Science and
Urban Economics, Vol. 21, No. 2, 163-182.
Giuliano, G., C. Redfearn, A. Agarwal, C. Li, and D. Zhuang (2007) “Employment
concentrations in Los Angeles, 1980-2000,” Environment and Planning A, volume 39,
2935-2957.
Glaeser, E. L. and M. E. Kahn (2001) “Decentralized employment and the transformation of the
American city,” NBER Working Paper Series, Working Paper 8117
Glaeser, E. L., M. E. Kahn and J. Rappaport (2000) “Why do the poor live in cities?” NBER
Working Paper Series, Working Paper 7636

134
Glaeser, E. L., M. E. Kahn and J. Rappaport (2008) “Why do the poor live in cities? The role of
public transport,” Journal of Urban Economics, Vol. 63, 1-24.
Glaeser, E. L., and J. L. Vigdor (2001) Racial segregation in the 2000 census: Promising news.
Washington, DC: Brookings Institution.
Greenwood, M. J. and R. Stock (1990) “Patterns of change in the intrametropolitan location of
population, jobs, and housing: 1950 to 1980,” Journal of Urban Economics, Vol. 28, 243-276
Gobillon, L., H. Selod, and Y. Zenou (2007) “The Mechanisms of Spatial Mismatch”, Urban
Studies, Vol. 44, Issue 12, 2401-2428.
Gordon, P., A. Kumar, and H. Richardson (1989) “The Spatial Mismatch Hypothesis: Some New
Evidence,” Urban Studies, Vol. 26, 315-326.
Gordon, P. and H. W. Richardson (1996) “Employment decentralization in U.S. metropolitan
areas: is Los Angeles on outlier or the norm?” Environment and Planning A, Vol. 28, 1727-1743.
Goslee, Susan and C. Conte (Ed.) (1998) Losing Ground Bit by Bit: Low-Income Communities in
the Information Age. What's Going On Series, Benton Foundation
Handy, S. L. and D. A. Niemeier (1997) “Measuring accessibility: an exploration of issues and
alternatives,” Environment and Planning A, Vol. 29, Issue7, 1175-1194.
Harrison, B. (1972) “The intrametropolitan distribution of minority economic welfare,” Journal
of Regional Science, Vol. 12, 23-43.
Hess, D. B. and P. M. Ong (2001) “Traditional Neighborhoods and Auto Ownership,” The Ralph
and Goldy Lewis Center for Regional Policy Studies at UCLA, Working Paper Series #37
Holzer, H. J. (1987) “Informal job search and black youth employment,” The American
Economic Review, Vol. 77, Issue 3, 446-452.
Holzer, H. J., K. R. Ihlanfeldt, and D. L. Sjoquist (1994) “Work, search, and travel among white
and black youth,” Journal of Urban Economics, Vol. 35, 320-345.
Horner, M. W. and J. N. Mefford (2007) “Investigating urban spatial mismatch using
job-housing indicators to model home-work separation,” Environment and Planning A, Vol. 39,
1420-1440.

135
Ihlanfeldt, K. R. (1993) “Intraurban job accessibility and Hispanic youth employment rates,”
Journal of Urban Economics, Vol. 33, Issue 2, 254-271.
Ihlanfeldt, K. R. (1997) “Information on the spatial distribution of job opportunities within
metropolitan areas,” Journal of Urban Economics, Vol. 41, 218-242.
Ihlanfeldt, K. R. (1999) “Are poor people really excluded from jobs located in their own
neighborhood?: Comments on Reingold and some additional evidence from the multi-city study
of urban inequality,” Economic Development Quarterly, Volume 13, No. 4, 307-314.
Ihlanfeldt, K. R. and D. L. Sjoquist (1990) “Job accessibility and racial differences in youth
employment rates,” American Economic Review, Vol. 80, 267-276.
Ihlanfeldt, K. R. and D. L. Sjoquist (1991) “The effect of job access on black and white youth
employment: A cross-sectional analysis,” Urban Studies, Vol. 28, No. 2, 255-265.
Ihlanfeldt, K. R. and D. L. Sjoquist (1998) “The spatial mismatch hypothesis: A review of recent
studies and their implications for welfare reform,” Housing Policy Debate, Vol 9, Issue 4,
849-892.
Jargowsky, P. A. (1996) “Take the money and run: Economic segregation in U.S. metropolitan
areas,” American Sociological Review, Vol. 61, 984-998.
Johnston-Anumonwo, I. (1996) “Location, Race, and Labor Force Participation: Implications for
Women of Color,” in Women’s Travel Issues, U. S. Department of Transportation, Ch 8, 339-354.
Johnston-Anumonwo, I. (1997) “Race, gender and constrained work trips in Buffalo, NY, 1990,”
Professional Geographer, Vol. 49, Issue 3, 306-317.
Kain, J. F. (1968) “Housing segregation, negro employment, and metropolitan decentralization,”
Quarterly Journal of Economics, Vol. 82, 175–197.
Kain, J. F. (1985) “Black suburbanization in the Eighties: a new beginning or a false hope?” in J.
M. Quigley and D. Rubinfeld (Ed.) American Domestic Priorities: An Economic Appraisal,
Berkeley: University of California Press, 253-282.
Kain, J. F. and J. M. Quigley (1975) “Housing markets and racial discrimination: A
microeconomic analysis,” Urban and Regional Studies Number 3, National Bureau of Economic
Research

136
Kasarda, J. D. (1985) “Urban change and minority opportunities,” in: P. E. Peterson (Ed.) The
New Urban Reality, Washington, DC: The Brookings Institution, 33–67.
Kasarda, J. D. (1989) “Urban industrial transition and the underclass,” The Annals of the
American Academy of Political and Social Science, 501-526.
Kasarda, J. D.and K. Ting (1996) “Joblessness and poverty in America’s central cities: causes
and policy prescriptions,” Housing Policy Debate, Vol 7, Issue 2, 387-419.
Kasinitz, P. and J. Rosenberg (1996) “Missing the connection: social isolation and employment
on the Brooklyn waterfront,” Social Problems, Vol. 43, No. 2, 180-187.
Kawabata, M. (2003) “Job access and employment among low-skilled autoless workers in US
metropolitan areas,” Environment and Planning A, Vol. 35, 1651-1668.
Kawabata, M. (2009) “Spatiotemporal dimensions of modal accessibility disparity in Boston and
San Francisco,” Environment and Planning A, Vol. 41, 183-198.
Kawabata, M. and Q. Shen (2007) “Commuting inequality between cars and public transit: the
case of the San Francisco Bay Area, 1990-2000,” Urban Studies. Vol. 44, 1759-1780.
Kawabata, M. (2003) “Job accessibility and labor market outcomes for low-skilled autoless
workers in US metropolitan areas,” Environment and Planning A, Vol. 35, Issues 9, 1651-1668.
Kloosterman, R. and S. Musterd (2001) “Polycentric urban region: as a research concept,” Urban
Studies, Vol. 38, No. 4, 619-629.
Levinson, D. M. (1998) “Accessibility and the journey to work,” Journal of Transport
Geography, Vol. 6, Issue 1, 11-21.
Lewis, O. (1959) Five Families: Mexican Case Studies in the Culture of Poverty, New York:
Basic
Lewis, O. (1998) “The culture of poverty,” Society, Vol. 35, Issue 2, 7-9.
Light, I. (2006) Deflecting Immigration Networks, Markets, and Regulation in Los Angeles, New
York: Russell Sage Foundation
Losch, A. (1954) The Economics of Location (translation), New Haven, CT: Yale University
Press

137
McFadden, D. (1977) “Modeling the Choice of Residential Location,” Cowles Foundation
Discussion Paper, No. 477
Massey, D. (2001) “Residential Segregation and Neighborhood Conditions in U.S. Metropolitan
Areas,” Chapter 13 in N. J. Smelser, W. J. Wilson, and F. Mitchell (Ed.) America Becoming:
Racial Trends and Their Consequences, Volume 1, National Research Council, 391-434
Massey, D., and N. Denton (1993) American Apartheid: Segregation and the Making of the
Underclass. Cambridge: Harvard University Press
Massey, D. S. and M. L. Eggers (1993) “The spatial concentration of affluence and poverty
during the 1970’s,” Urban Affairs Quarterly, Vol.29, 299-315.
McLafferty, S. and V. Preston (1992) “Spatial Mismatch and labor market segmentation for
African-American and Latina women,” Economic Geography, Vol. 68, No. 4, 406-431.
Meyer, M. D. and E. J. Miller (2001) Urban Transportation Planning: A Decision-oriented
Approach, McGraw-Hill Series in Transportation.
Mills, E. S. (1972) Studies in the Structure of the Urban Economy, Baltimore, MD: Johns
Hopkins University Press
Miller, V. P. and J. M. Quigley (1990) “Segregation by racial and demographic group: Evidence
from the San Francisco Bay Area,” Urban Studies, Vol. 27, 3-21.
Muller, P. O. (1981) Contemporary Suburban America. Englewood Cliffs, N.J.: Prentice-Hall
Muller, P. (2004) “Transportation and urban form: Stages in the spatial evolution of the
American metropolis,” in S. Hanson and G. Giuliano, (eds.) The Geography of Urban
Transportation, 3
rd
edition. New York: Guilford Press, 59-85.
O’Regan, K. and J. Quigley (1991) “Labor market access and labor market outcomes for urban
youth,” Regional Science and Urban Economics, Vol. 21, 277-293.
Ong, P. (2002) “Can ownership and Welfare-to-Work,” Journal of Policy Analysis and
Management, Vol. 21, No. 2, 239-252.
Ong, P. and E. Blumenberg (1998) “Job access, commute and travel burden among welfare
recipients,” Urban Studies, Vol. 35, Issue 1, 77-93.

138
Ong, P. and D. Miller (2005) “Spatial and transportation mismatch in Los Angeles,” Journal of
Planning Education and Research, Vol. 25, 43-56.
Pack, J. R. (2002) Growth and Convergence in Metropolitan America, Washington, D.C.:
Brookings Institution Press.
Painter, G., C. Y. Liu and D. Zhang (2007) “Immigrants and the spatial mismatch hypothesis:
labor market outcomes among immigrant youth in Los Angeles,” Urban Studies, Vol. 44, No.
13, 2627-2649.
Parks, V. (2004) “Access to work: the effects of spatial and social accessibility on
unemployment for native-born black and immigrant women in Los Angles,” Economic
Geography, Vol. 80, Issue 2, 141-172.
Pastor, M. (2001) “Geography and opportunity,” Chapter 14 in N. J. Smelser, W. J. Wilson, and
F. Mitchell (Ed.) America Becoming: Racial Trends and Their Consequences, Volume 1,
National Research Council, 435-467.
Perle, E. D., H. Bauder and N. Beckett, 2002 “Accessibility measures in spatial mismatch
models,” The Professional Geographer, Vol. 54, Issue 1, 106-110.
Quigley, J. (1998) “Urban diversity and economic Growth,” Journal of Economic Perspectives,
Vol. 12, 127-138.
Raphael, S. (1998) “The spatial mismatch hypothesis and black youth joblessness, evidence from
the San Francisco Bay Area,” Journal of Urban Economics, Vol. 43, 79-111.
Redfearn, C. (2006) “The Manifestation of Agglomerative Economies: The Role of Employment
Centers in the Los Angeles Metropolitan Area,” Unpublished Manuscripts
Redfearn, C., G. Giuliano, A. Agarwal, S. He and L. Hu (2008) “Network accessibility and the
evolution of urban employment”, Metrans project, 06-16.
Rogers, C. L. (1997) “Job search and unemployment duration: Implications for the spatial
mismatch hypothesis,” Journal of Urban Economics, Vol. 42, 109-132.
Sassen, S. (1996) “Economic restructuring and the American city,” Annual Review of Sociology,
Vol.16, 465-490.

139
Schwartz, A. (1976) “Migration, age, and education,” Journal of Political Economy, Vol. 84,
701-719.
Shen, Q. (1998). “Location characteristics of inner-city neighborhood and employment
accessibility of low-wage workers,” Environment and Planning B, Vol. 25, 345–365.
Shen, Q. (2000) “Spatial and Social Dimensions of Commuting,” Journal of the American
Planning Association, Vol. 66, No. 1, 68-82.
Shen, Q. (2001) “A spatial analysis of job openings and access in a US metropolitan area,”
Journal of the American Planning Association, Vol. 67, 53-68.
Schill, H. and S. M. Wachter (1995) “Housing market constraints and spatial stratification by
income and race,” Housing Policy Debate, Vol. 6, Issue 1, 141-167.
Simpson, W. (1992) Urban Structure and the Labor Market, Oxford: Clarendon Press
Stoll, M. A. (1999) “Spatial Job Search, Spatial Mismatch, and the Employment and Wages of
Racial and Ethnic Groups in Los Angeles,” Journal of Urban Economics, Vol. 46, 129-155.
Stoll, M. A. (2005) “Geographical skills mismatch, job search and race,” Urban Studies, Vol. 42,
695–717.
Taylor, B. and P. Ong (1995) “Spatial mismatch or automobile mismatch? An examination of
race, residence and commuting in US metropolitan areas,” Urban Studies, Vol. 32, 1453–1473.
Thompson, M. A (1997) “The impact of spatial mismatch on female labor force participation,”
Economic Development Quarterly, Vol. 11, 138-145.
Wachs, M. and T. G. Kumagai (1973) “Physical accessibility as a social indicator,”
Socio-Economic Planning Science, Vol. 7, 437–456.
Wasmer, E. and Y. Zenou (2002) “Equilibrium search unemployment with explicit spatial
frictions,” Labour Economics, Vol. 13, 143–165.
Weber, J. (2003) “Individual accessibility and distance from major employment centers: An
examination using space-time measures,” Journal of Geographical Systems, Vol. 5, 51-70.
Wilson, W. (1980) The Declining Significance of Race: Blacks and Charging American
Institutions (2
nd
Edition) Chicago, IL: University of Chicago Press

140
Wilson, W. (1987) The Truly Disadvantaged, Chicago, IL: University of Chicago Press
Wyly, E. K. (1990) “Race, gender, and spatial segmentation in the Twin Cities,” The
Professional Geographer, Vol. 48, Issue 4, 431-444.
Zax, J. S. and J. F. Kain (1996) “Moving to the suburbs: Do relocating companies leave their
black employees behind,” Journal of Labor Economics, Vol. 14, No. 3, 472-504.
Zenou, Y. (2006) “Urban labor economic theory,” in R. J. Arnott and D. P. McMillen (ed.) A
Companion to Urban Economics, Blackwell Publishing, 418-439.

Abstract (if available)

Abstract This dissertation tests whether changing urban structure has affected low-income job seekers’ labor market outcomes differentially by impacting their job accessibility. The relatively poor labor market outcomes of minorities are well-documented in the Spatial Mismatch Hypothesis literature which claims that the unequal labor market outcomes are partly caused by the spatial barriers between minorities’ residences and their matching job opportunities. This research aims to expand the demographic, geographic and temporal scopes of the Spatial Mismatch Hypothesis by studying low-income job seekers’ job accessibility in the Los Angeles metropolitan area in 1990 and 2000.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Testing the entrepreneurial city hypothesis: a study of the Los Angeles region

PDF

Beyond spatial mismatch: immigrant employment in urban America

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

Population and employment distribution and urban spatial structure: an empirical analysis of metropolitan Beijing, China in the post-reform era

PDF

Structural transformation and globalization

PDF

Structure, agency, and the Kuznets Curve: observations and implications for sustainability planning in U.S. cities

PDF

From structure to agency: Essays on the spatial analysis of residential segregation

PDF

Spatial and temporal expenditure-pricing equity of rail transit fare policies

PDF

A megaproject matrix: ideology, discourse and regulation in the Delhi Metro Rail

PDF

Redlining revisited: spatial dependence and neighborhood effects in mortgage lending

PDF

Urban air pollution and environmental justice: three essays

PDF

Narrowing the achievement gap and sustaining success: a qualitative study of the norms, practices, and programs of a successful high school with urban characteristics

PDF

Spaces of market-culturalism: the case of Chungking Mansions, Hong Kong

PDF

Humanitarian aid and the production of spatial knowledge and practices in the U.S.-Mexico borderlands

PDF

Making the connection: The California urban superintendent and parent and community involvement

PDF

Informal consent: the complexities of public participation in post-civil war Lebanon

PDF

A spatial analysis of human trafficking in Greater Los Angeles

PDF

Walkability as 'freedom': the ecology of school journey in inner city Los Angeles neighborhoods

PDF

Children’s travel behavior in journeys to school

PDF

Outsourcing for school outcomes: a multi-case study examination of outsourced and in-house literacy coaching models

Asset Metadata

Creator Hu, Lingqian (author)

Core Title Urban spatial transformation and job accessibility: spatial mismatch hypothesis revisited

School School of Policy, Planning, and Development

Degree Doctor of Philosophy

Degree Program Policy, Planning, and Development

Publication Date 02/09/2010

Defense Date 10/29/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag job accessibility,OAI-PMH Harvest,spatial mismatch hypothesis,urban structure

Place Name California (states), Los Angeles (city or populated place)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Giuliano, Genevieve (committee chair), Moore, James (committee member), Redfearn, Christian L. (committee member)

Creator Email lingqiah@usc.edu,lingqian.hu@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2844

Unique identifier UC1137469

Identifier etd-Hu-3392 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-301711 (legacy record id),usctheses-m2844 (legacy record id)

Legacy Identifier etd-Hu-3392.pdf

Dmrecord 301711

Document Type Dissertation

Rights Hu, Lingqian

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu