University of Southern California Dissertations and Theses

Governing public goods: how representation and political power in local and regional institutions shape inequalities

(USC Thesis Other)

Governing public goods: how representation and political power in local and regional institutions shape inequalities

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Governing Public Goods: How Representation and
Political Power in Local and Regional Institutions
Shape Inequalities

By
Yeokwang (Brian) An

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
OF UNIVERSITY OF SOUTHERN CALFORNIA

In Partial Fulfillment of the
Requirement of the Degree

DOCTOR OF PHILOSOPHY
PUBLIC POLICY AND MANAGEMENT

May 2019

Dissertation Committee:
Shui-Yan Tang (Chair)
Morris E. Levy
Michael Thom

1

“By now, all this talk about airy academic research may seem disconnected from a world
in which so many people labor so hard at getting ahead or keeping others down. But in
business and government, in law and medicine, in politics and international diplomacy, no
skill is valued more highly than the ability to recognize a problem, then to articulate it in a
way that convinces others both to care about it and to believe it can be solved, especially
by you. If you can do that in a class on medieval Tibetan rugs, you can do it in an office
on Main Street, Wall Street, or on Queen’s Road in Hong Kong.”

Booth, W. C., Colomb, G. G., & Williams, J. M. (2003). The Craft of Research.
University of Chicago press. page 65.

For Hyewon

2
Acknowledgements
I never thought that I would pursue PhD degree. I was a person who was not interested in
research. I was green regarding the research enterprise when I arrived at USC. In fact, an
academic career was not in my life plan until I came to the United States in 2012 for a
master’s degree in international relations at Georgetown. My journey in the PhD program
at USC was fun, exciting, and of course, sometimes boring and frustrating. One day,
however, I figured out the meaning of “re-search”: searching for answers persistently,
which meant not giving up along the way even in the face of frustration.
With respect to this, I am so grateful to my former advisor, Raphael Bostic. His name does
not appear in my dissertation committee, as he left the USC in summer 2017 to become the
President and CEO of the Federal Reserve Bank of Atlanta, but he was my primary advisor
in my first three years in the doctoral program. The training I received from him was not
about sophisticated econometric techniques—which I initially expected—but rather, it was
more about academic attitude and persistent efforts regarding “re-search.” I still remember
the days he taught me how to pick up the phone and make phone calls to government
agencies to obtain data that I thought would be too hard to find.
Having entered the PhD program without an organized research agenda or interest, I was
very lucky to meet my dissertation committee members to develop my own. In particular,
I am indebted to Morris Levy, who has been an incredible resource for me in developing
my research interest in local public goods early in the program. My connection with him
started from my email request for his course syllabi, and our research collaboration began
after that. Chapter 3 in this dissertation would not have been possible without his
mentorship.
I also want to thank Shui-Yan Tang for providing valuable insights and advising me as the
new chair after Raphael Bostic left. Due to this transition, I was able to receive balanced
training between public administration (with Yan Tang as my new chair) and urban
development and policy (with Raphael Bostic as my former chair). In fact, my interest in
special districts in Chapters 4 and 5 originates from the final class I took from Yan’s course.
I also thank Michael Thom for helping me develop my interest in public finance and
management.
Last, everything from the beginning to the end of my PhD degree completion would have
been impossible without the support of my spouse, Hyewon, who, as a fellow doctoral
student at USC, has been together with me every single day. With the birth and upbringing
of our two children—Jisu and Jiho—it was Hyewon who encouraged my progress along
the way.

3

Abstract
This dissertation examines how representation and political power in local and regional
institutions influence the quality, quantity, and spatial allocation of public goods in the
United States. Specifically, this dissertation focuses on the impact of three major
governance and institutional forces: institutional rules in regional governance, racial
structure of income inequality and type of local government. Each of these factors is
examined at different levels of geographic scope including urban, rural, and metropolitan
areas and from the communities in unincorporated areas to school districts, cities,
metropolitan areas, and states. The findings of the four essays suggest that each of these
forces significantly influence the allocation of local public investments. Unraveling what
governance and institutional forces and to what extent they impact the local provision of
public goods provides important implications for policy makers to manage their
performance and to design better institutional arrangements for efficient and equitable
allocation of public resources.

Keyword
local public goods; institutional design; representation; political power; local and
regional governance

4
Table of Contents
Acknowledgment…………………………………………………………………………... 2
Abstract…………………………………………………………………………………...... 3
Table of Contents………………………………………………………………………...... 4
List of Figures……………………………………………………………………………… 6
List of Tables………………………………………………………………………………. 7

Chapter 1. Introduction…………………………………………………………………...... 8
1. Dissertation Overview………………………………………………………………... 8
2. Summary of the Four Essays……………………………………………………….... 9

Chapter 2. What Determines Where Public Investment Goes? Regional Governance
and the Role of Institutional Rules and Power…………………………………
15
1. Introduction……………..……………..……………..……………..………………... 16
2. Political Power and Public Goods Allocation across Localities in a Region………... 19
3. Metropolitan Planning Organizations………………………………………………... 22
4. Theory and Hypotheses……………..……………………………………………….. 24
5. Data……………..……………..……………..……………..………………………... 27
6. Methods……………..……………..……………..………………………………...... 32
7. Results……………..……………..……………..……………..……………………... 37
8. Discussion and Conclusion……………..……………..……………..………………. 43
9. References……………..……………..……………..……………..………………..... 46
10. Appendix……………………………………………………………………………... 55

Chapter 3. It’s Not Just Welfare: Racial Inequality and the Local Provision of Public
Goods in the United States……………………………………………………..

60

1. Introduction…………………………………………………………………………... 61
2. Group Attitudes and Support for Public Goods Spending…………………………… 63
2.1. Just Welfare?........................................................................................................... 65
2.2. Which Types of Public Goods?............................................................................... 67
2.3. Political Mechanism……………………………………………………………… 68
2.4. Independent Effects of Racial Diversity and Income Inequality…………………. 69
3. Data and Variables………………………………………………………………........ 70
3.1. Between Group Share of Total Income Inequality……………………………….. 73
3.2. Empirical Model………………………………………………………………….. 76
4. Results………………………………………………………………………………... 78
4.1. The Effects on Total Public Goods Spending……………………………………. 78
5
4.2. Which Public Goods?.............................................................................................. 81
4.3. Political Mechanisms……………………………………………………………... 85
5. Conclusion………………..……………..……………..……………..………………. 86
6. References……………..……………..……………..……………..…………………. 88
7. Appendix……………..……………..……………..……………..…………………... 99

Chapter 4. Capitalizing on Collective Action: Creation of Community Services Special
District and Property Value Appreciation……………………………………..

105

1. Introduction…………………………………………………………………………... 106
2. Background…………………………………………………………………………... 109
3. Data and Methods……………………………………………………………………. 113
4. Results………………………………………………………………………………... 119
5. Concluding Discussion………………………………………………………………. 122
6. References……………………………………………………………………………. 126
7. Appendix…………………………………………………………………………….. 134

Chapter 5. Community’s Governing Capacity and the Success of Special District as an
Instrument for Neighborhood Governance …………………………………….

135

1. Introduction…………………………………………………………………………... 136
2. Background…………………………………………………………………………... 138
3. Creation of Local Governing Institutions and Property Value Change……………… 141
4. Conditional Effect of a Special District by a Community’s Governing Capacity…… 142
5. Methods and Data…………………………………………………………………… 145
6. Descriptive Data…………………………………………………………………....... 146
7. Quantitative Analysis………………………………………………………………... 147
8. Qualitative Case Study Findings…………………………………………………….. 148
8.1. Success Story of Community Services District in Helendale……………………... 149
8.2. Failure Story of Community Services District in Los Osos………………………. 151
9. Discussion and Conclusion…………………………………………………………... 153
10. References…………………………………………………………………………… 158
11. Appendix……………………………………………………………………………. 164

Chapter 6. Conclusion and Policy Implications……………..…………………………….. 168
1. Conclusion……………..……………..……………..……………..…………………. 168
2. Policy Implications……………..……………..……………..……………………….. 170
3. Contribution to the Literature……………..……………..……………..……………... 174

6
List of Figures

Figure 1-1. Structure of Dissertation………………………………………………….. 9
Figure 2-1. Summary Information on MPO Projects and Local Voting Power……… 51
Figure 2-2. Distribution and Concentration of MPO Local Voting Power…………… 52
Figure 2-3. Conditional Logit Model Setup…………………………………………... 52
Figure 2-4. Conditional Logit Regression Results for the Houston MPO…………….. 53
Figure 2-5. Conditional Logit Results on Power in All Four Regions………………... 53
Figure 2-6. Conditional Logit Results on Austin MPO by Period……………………. 54
Figure A2-1. Conditional Logit Results for the Four MPO Regions…………………. 55
Figure 4-1. Governing Structure Under County and CSD system……………………. 130
Figure 4-2. Special District Sampling Process………………………………………... 130
Figure 4-3. Map of Special District Communities, Adjacent Cities, and Nearest
County Service Areas……………………………......................................

131
Figure 4-4. Mean Statistics and T-test of Home Characteristics……………………… 132
Figure 4-5. Difference-in-Difference Estimate of CSD Effect on Property Values
Over Time…………………………………………………………………

133
Figure 4-6. Difference-in-Difference Estimate of CSD Effect on Property Values
Over Time (Using Alternative Control Group)…………………………... 133
Figure 5-1. Administrative Process to Create Community Services District…………. 162
Figure 5-2. Impact of Special Districts on Property Values…………………………... 162
Figure 5-3. Robustness of Analysis Using County Service Area as Control Group….. 163

7
List of Tables

Table 2-1. 60 MPO Policymakers Interviewed……………………………………….. 50
Table 2-2. Effects of Independent Variables by Standard Deviation…………….…… 50
Table A2-1. Data and Source………………………………………………………….. 56
Table A2-2. Summary Statistics for City-level Independent Variables 2001-2010…... 57
Table A2-3. Summary Statistics for Project-level Independent Variables 2001-2010.. 58
Table 3-1. 10 Most and Least Racially Unequal and Heterogeneous Cities in 2000…. 93
Table 3-2. Effect of Racial Inequality on Total Public Goods…………….………….. 94
Table 3-3. Results at the School District Level…………….…………………………. 95
Table 3-4. Budget Cut Effect of Racial Inequality on Public Goods…………………. 96
Table 3-5. Results with Political and Institutional Variables…………….…………… 98
Table A3-1. Mean Statistics for Key Variables in the Four Samples, 1980-2000……. 99
Table A3-2. Data Sources…………….…………….…………….…………………… 100
Table A3-3. Alternative Specification for Effect of Racial Inequality………………... 101
Table A3-4. Summary of the Findings…………….…………….……………………. 102
Table 4-1. List of Special Districts in the Sample…………….…………….………… 128
Table 4-2. Difference-in-Differences Hedonic Regression Results…………………... 129
Table A4-1. Number of Housing Transactions Between Treated and Control Group... 134
Table 5-1. Mean Statistics and T-test Results of Housing Characteristics……………. 161
Table A5-1. Demographic Profiles of Communities …………………………………. 164
8
Chapter 1. Introduction

Dissertation Overview

This dissertation examines the local provision of public goods in the United States. Public
goods and services that citizens in local jurisdictions receive and enjoy are essentially a
measure for quality of life. The processes and decisions allocating local public goods often
contribute to disparities in the livability and general appeal of communities in the region.
Among various potential factors influencing these allocative decisions, this dissertation
studies three major determinants that have received little attention from the academic and
policy community.
Focusing on representation and political power of local actors from governance and
institutional perspectives, these determinants are 1) regional institutional design that allocate
voting power to the member city jurisdictions in metropolitan councils; 2) urban
demographic structure, especially the racial structure of income inequality among the citizen
groups; 3) type of local government, specifically whether the entity is a special-purpose
district or a general-purpose government.
How these forces influence the distribution of local public goods is examined at
different levels of local jurisdiction and political geography, from the communities in
unincorporated areas to school districts, cities, metropolitan areas in urban and rural areas.
Figure 1-1 summarizes the broader agenda and approach of this dissertation.

9
Figure 1-1: Structure of Dissertation

Summary of the Four Essays
The first essay (chapter 2) studies how institutional rules in collaborative regional
organizations affect the equitable allocation of public goods across cities. It specifically
examines the rules that allocate voting power to cities in the regional governing policy
boards. Using a mixed-methods approach, the essay first shares the views of practitioners
that were obtained from interviews regarding the role of such institutional rules in
resource allocation. Interestingly, the policymakers interviewed believed that
organizations make data-driven decisions that are independent of institutional and
political factors. The essay then develops an argument about how institutionalized power
influences resource allocation even in a setting where policymakers do not explicitly
recognize its role. While previous studies argue that such factors should matter in
resource allocation, empirical studies that support this claim have been rare in the
literature.
10
By using novel data on the geocoded transportation projects that were planned and
approved by the metropolitan planning organizations in Texas, the analysis finds that
local voting power in the regional policy council is consistently a major factor that is
associated with the observed geographic distribution of the projects. Moreover, the results
suggest that the degree of power concentration of the dominant city in the region
influences whether the remaining cities’ power is relevant. These results were far
different than what was predicted by the policymakers we interviewed, suggesting that
institutional governance rules may be more important than previously recognized.
This essay is the first study to examine how institutional rules and political factors
determine resource allocation outcomes in collaborative interlocal-governmental
organizations. The arguments and findings contribute to both the collaborative
governance scholarship and the distributive politics literature.
The second essay (chapter 3) measures the impact of a broader urban demographic
structure on the investments local governments make for various public goods.
Specifically, the study quantifies the extent to which income inequality falls along racial
lines and whether this measure of racial inequality dampens the local provision of public
goods. This study has implications for the political power of racial groups and the
collective impacts on public goods investment in localities.
The impact of demographic structure, particularly social heterogeneity, has been
studied by scholars in various social science disciplines. Two factors that have often been
scrutinized by researchers are income inequality and racial diversity. The attention to
these two factors and the ways in which they erode public support for collective
11
investments is a timely topic that is relevant to the United States more than ever as the
country is becoming a highly racially diverse and more economically unequal society.
While previous studies have found a negative relationship between racial diversity
and local public spending and a similar pattern between racial income inequality and state
welfare programs, no studies have examined whether racial inequality influences the local
provision of public goods. This study is the first to examine this relationship. It argues
that there is a redistribution effect in the financing of local public goods and that the
heterogeneity of public goods users fosters such redistribution when their characteristics
are divided along racial and income lines. One consequence is the shrinkage of local
public investment in public goods, suggesting a more profound role of racial inequality
than has been previously recognized in the literature.
The third essay (chapter 4) compares the performance of special-purpose districts
to general-purpose governments in local public goods provision and service delivery.
Despite the fastest growth in specialized governments and their prevalence in American
local governance, few studies have compared their effectiveness relative to that of
general-purpose local governments. This chapter focuses on the community services
districts in California that are created as an instrument for neighborhood governance in
unincorporated areas.
While the literature and media often have characterized special districts as shadow
governments that promote the private interests of developers and affluent citizens, their
flexibility and innovation that many local communities take advantage of have been
overlooked in the literature. The cases of community services districts add to our
understanding of specialized governance, as they provide a broad range of neighborhood-
12
level services while allowing for a better representational system. Interestingly, the
creation of these districts replaces the existing county authority’s service power, and as a
result, the citizen residents can elect their district governing body from their respective
communities. Under the county’s governing system, however, the communities need to
have a county board of supervisors as their governing authority. Hence, the creation of a
community services district is a governance redesign process that involves extensive
collective action for grassroots democracy.
The analysis uses difference-in-difference hedonic regressions to evaluate the
performance of these community services districts relative to the services provided by
county authorities. Utilizing information on seven special districts in Southern California
that were created between 1988 and 2012, the analysis finds that, on average, the creation
and services of community services district increase residential home prices by 12 percent
compared to the prices in the surrounding areas under county management. While there is
a growing body of literature on (quasi) private governments such as homeowner
associations and business improvement districts, the use of special districts as an
instrument for neighborhood governance has been given little attention. This chapter
introduces this use of special districts to the literature and shows that property values can
be enhanced via community services districts, which are created to address collective
action problems by citizen residents.
The last essay (chapter 5) builds on chapter 4 and studies the role of community
governing capacity in the success of special district management. While the quantitative
hedonic analysis in the first few essays establishes the positive impact of community
services districts on property values, it is silent on the local contextual factors with which
13
communities can achieve such an impact. Of course, creating a new governing institution
such as a community services district itself does not guarantee improved services. The
literature on citizen and community governance suggests that local communities need to
develop the governing capacity and managerial skills necessary to take advantage of such
institutional innovations in special districts.
This chapter proposes three factors that are critical to the success of a community
services special district, including a clear problem statement, prepared local managerial
leadership, and strong support from citizens. To test this conditional theory of special
district performance, the essay uses mixed methods. First, as in the previous chapter, the
quantitative analysis uses a hedonic model and difference-in-difference methodology to
measure the impact of community services districts on property values. An upgraded
approach here is that the analysis is executed for each of the seven communities in the
sample. The findings emphasize that two communities had a consistently negative impact
on property values over time as a result of the creation of community services districts,
while the other five communities saw a consistently positive impact over time.
This essay then presents a qualitative case study analysis to obtain a deeper
understanding of the determinants of such divergent trajectories. Case study data are
collected from site visits and field interviews with the managers of these special districts.
The qualitative case analysis reveals that the presence (or absence) of a community’s
governing capacity explains the subsequent success (or failure) of special district
management. A community that features a positive impact of special districts (on property
values) had a clear problem statement, prepared local managerial leadership, and strong
support from citizens, whereas a community that shows a negative impact lacked these
14
conditions. Further, the community that features a negative impact failed to coordinate
with a county and state authority on environmental regulation. Hence, intergovernmental
coordination also contributed to the low effectiveness of special districts. This study
introduces and situates the context of special districts, specifically community services
districts, within the literature on citizen and community governance and institutional
change. By doing so, this work contributes to a richer understanding of the managerial
implications following the creation of special districts.

15
Chapter 2. What Determines Where Public Investment Goes?
Regional Governance and The Role of Institutional Rules and Power

Summary
As an embodiment of collaborative governance model, metropolitan planning organizations in
the United States allocate federal, state, and local funds to member municipalities for
transportation projects across their regions. To examine how institutional rules and power
shape where public investment goes, we examine the extent to which the allocation of local
voting power in regional governing policy boards influences the spatial allocation of
transportation investments. Our analysis shows that the power structure of regional policy
boards is consistently a major factor associated with the observed geographic distribution of
investments. Moreover, the results suggest that the degree of power concentration of the
dominant city in the region influences whether the remaining cities’ power matters. These
results were far different than what was predicted by the policymakers we interviewed,
suggesting that institutional governance rules may be more important than previously
recognized.

16
Introduction
Local governments in the United States have increasingly chosen collaborative approaches to
resolve institutional collective action problems and to achieve better regional and local policy
outcomes. Much of the recent research on interlocal collaboration asks what rules and incentives
are available to local jurisdictions so that public managers can overcome transaction costs
embedded in joint actions and maximize benefits (Song, Park, and Jung 2018; Yi et al. 2017;
Lubell et al. 2017). One strand of scholarly inquiry focuses on what forms of governance should
be adopted to integrate respective institutional collective action problems. Depending on the
characteristics and scale of the policy problem and the degree of transaction costs entailed in
collaboration, localities may join informal networks such as voluntary associations, create cross-
jurisdictional special districts, or form regional councils of governments (Feiock 2013). This
essay focuses on the third of these.
Regional councils have been found to minimize transaction costs for high risk
cooperation (Kwon, Feiock, and Bae 2014), which is particularly useful when the collaboration
involves many actors and they need to make long-term investments jointly (Olson 1971). One
such policy area is transportation planning, which, by its nature, carries interjurisdictional
externalities that cannot be addressed by local governments’ independent actions. In the United
States, the regional council for transportation planning is the metropolitan planning organization
(MPO). MPOs decide how to allocate federal, state, and (often) local funds for transportation
investments at the regional level. As of 2019, there were 420 MPOs in the Unites States, and
they collectively allocate hundreds of billions of federal dollars every year to member
municipalities on regionally significant transportation projects (NARC 2019). Despite their
17
prominent role in American public policy, MPOs have not been a subject of empirical
assessment in the scholarship with respect to what guides their investment allocation decisions.
In order to understand how MPOs manage their collective decision-making process of
resource allocation, we interviewed 60 representatives in four large metropolitan areas including
elected officials from member jurisdictions, professional managers, and the agency directors and
staff. Despite some variations between the regions, the collective view we obtained from these
semi-structured interviews was that the organizations’ funding decisions were primarily data-
driven and collaborative, such that no single jurisdiction’s interests prevailed over others.
Notably, the MPO policymakers and managers in the process believed that differences in voting
power held by member jurisdictions did not affect their collaborative governance outcomes.
Our empirical analysis, which builds on institutional theories of urban governance and
political power, indicates otherwise. Leveraging a unique dataset of geocoded transportation
projects programmed and approved by the four largest MPOs in Texas, we show that the extent
of cities’ voting power in governing board is an important factor that explains the distribution of
the projects across the local jurisdictions. This remains true even when we consider other
external factors that were described to be more important by the interviewees, such as traffic and
road pavement conditions, demographics, and the employment environment.
We also argue that the internal power structure of MPOs has consequences for whether
the funding will be more equitably allocated to their member cities in the region or not. Our
analysis lends support to this claim. In regions where local voting power is heavily concentrated
in a primary dominant city, marginal shifts in power among the remaining localities do not
significantly affect funding allocation outcomes. By contrast, when the rules establish a more
18
even distribution of voting power, a marginal increase in power for a non-primary city translates
into a significantly higher likelihood that a project located in the city will receive funding.
The results are far different than what was predicted by the policymakers and
practitioners we interviewed, suggesting that institutional governance rules may be more
important than previously recognized. This essay provides some avenues to enrich the
discussions on the role of institutions on regional development from an equity perspective. First,
the findings add to the discussions on ideal forms of government in metropolitan economies.
While scholars have long debated whether a consolidated government form would enhance the
efficiency and equity of public services, one area that often has been neglected in the literature is
the role of existing regional institutions and how they can contribute to equitable development.
Second, in keeping with recent arguments on the distribution of power and institutional
design in collaborative governance scholarship, our results suggest a connection between
research and actions that policymakers are recommended to take. We suggest that policymakers
and public managers must not only give attention to the institutional rules and distribution of
power in intergovernmental organizations. They must also take actions to develop a more
inclusive representative system. Developing such a system requires empowering less powerful
localities in decision-making so that the funding allocation process is not dominated by a few
powerful actors but rather is shaped by collective inputs across a range of participants. In doing
so, policymakers would be tasked to strike a right balance between such inclusive representation
and both efficiency and equity considerations of resource allocation.
The paper proceeds as follows: The next two sections provide a brief background on the
literature on public goods provision as well as information on regional planning organizations
19
and the role played by them in public investment allocation. It then builds on relevant theories to
develop our arguments and hypotheses. The following section presents the data, methods, and
empirical analysis. The paper is then concluded by discussing our contribution to the scholarly
debates on metropolitan economy as well as policy implications.

Political Power and Public Goods Allocation across Localities in a Region
A large literature has identified multiple factors that shape the provision and spatial allocation of
public goods across localities in a region. Tiebout (1956)’s seminal theory identified residential
choice dynamics as a key consideration. By voting with their feet and allocating themselves
across jurisdictions that vary in the bundle of public goods local governments provide, citizens
reveal their demand.
Building on Tiebout (1956), Peterson (1981) introduces a City Limit model in which a
competitive urban environment causes local governments to prefer development-oriented public
investment, such as in highways and transportation infrastructure, to redistributive programs.
While Peterson’s theoretical argument has found some empirical support, both in terms of
mayoral preferences (Saiz 1999) and city spending patterns (Schneider 1989; Minkoff 2012;
Jimenez 2014), many have argued that the model reduces the complex structure of city decision-
making to economic imperatives and constraints and overlooks the role of other factors,
including political and institutional considerations (Basolo and Huang 2001; Einstein and Glick
2018; Hajnal and Trounstine 2010).
Indeed, a growing body of recent research finds that political ideology influences the
allocation of public goods even at the local level, with a particular ideology being associated
20
with lower levels of local public spending (de Benedictis-Kessner and Warshaw 2016; Gerber
and Hopkins 2011), stronger support for redistributive programs (Einstein and Glick 2018), and
greater collaboration through interlocal agreements (Song, Park, and Jung 2017; Gerber, Henry,
and Lubell 2013). Yet, despite the prominence of theories on urban governance that have focused
on political power (e.g., Dahl 1961; Stone 1989), relatively few empirical studies explicitly
incorporate political power considerations into resource allocation decision-making process for
urban public goods (Hochschild 2008). Our research adds to this limited literature by exploring
how political power affects resource allocation decision-making in the context of regional
planning organizations.
This essay is also related to both distributive politics and collaborative governance
scholarship. Political scientists have long argued that individual legislators in a national
government care mainly about the public projects that flow into their districts (Weingast 1979).
It is widely known that unequal distribution of power yields to an inequitable distribution of
public spending in national politics (Ansolabehere, Gerber, and Synder 2002; Snyder, Ting, and
Ansolabehere 2005). This pattern has also been observed in state politics for the allocation of
major highway construction across counties (Nall 2018). Further, at the regional level, Gerber
and Gibson (2009) argue that local competition in MPOs may shape their resource allocation
decisions to reflect some aspects of distributive politics with a balance of power between local
and regional interests.
The processes in regional planning organizations in principle, however, seem to more
closely reflect collaborative decision-making than a zero-sum game (Deyle and Wiedenman
2014; Innes and Gruber 2005). As their core function is to establish a fair and impartial setting
for effective regional decision making, member jurisdictions engage in a consensus-oriented and
21
deliberative process to prioritize region-wide transportation investments that will benefit all in
the area (Deyle and Wiedenman 2014), a feature frequently described as collaborative
governance process in the literature.
1
Public administration scholars have noted that the
distribution of power and resources across participants in collaborative governance is a critical
factor for sustainability and success of the forum (Emerson, Nabatchi, and Balogh 2012; Ansell
and Gash 2008; Tang and Mazmanian 2010; Choi and Robertson 2013). They argue that power
disparities among participants hinder the pursuit of a joint course of action. These studies also
suggest that such negative consequences of power can be diminished by sharing or redistributing
power and resources so that weaker or underrepresented groups can be empowered in the
decision-making process.
In this essay, we show that, consistent with the distributive politics literature, a similar
power dynamic among members of regional organizations causes higher shares of public dollars
to shift to the jurisdictions with greater voting power (Shapley and Shubik 1954). We also show
that the design of the rules, and the resultant degree of voting power concentration, matters in
regional decision making, as scholars of collaborative governance suggest. These findings are
particularly striking because the policymakers and managers in these institutions did not
acknowledge such an explicit role for power when they were asked. Perhaps power is more
implicitly embedded in the institutions and shapes the decisions leading up to political choice,
similar to a dynamic put forward by Stone (1980).

1
While the focus of this essay is city’s MPO voting power in the governing policy board, the MPO
process extensively involves nonprofit and community interests during technical advisory committee
meetings and public hearing. Such process fits to the definition of collaborative governance process
described by Emerson, Nabatchi, and Balogh (2012).
22
Metropolitan Planning Organizations
MPOs are a class of regional organizations that focuses on transportation planning. Though they
date to the 1930s, MPOs became much more prevalent after the Federal Aid Highway Act of
1962, which mandated that any urbanized area with a population greater than 50,000 should
establish an MPO. The MPO role in allocating transportation funds was advisory until the 1991
enactment of the Intermodal Surface Transportation Efficiency Act (ISTEA) (Solof 1998; Lewis
and Sprague 1997). Before ISTEA, regional transportation decision-making was a top-down
management enterprise, with state and federal governments making the resource allocation
decisions (Sanchez 2006). ISTEA reversed this by empowering MPOs to be programming bodies
and not just planning agencies. The legislation gave MPOs authority and discretion over the
project selection process, which to that time had primarily been managed by state highway
departments (now called departments of transportation).
The jurisdictions in a metropolitan area generally comprise an MPO. The decision-
making body of an MPO is its governing policy board.
2
The members of the governing policy
board are representatives drawn from the MPO’s member jurisdictions. These representatives are
typically elected officials, such as mayors, city council members, county judges and
commissioners, and often also include professional officials from transportation authorities. To
more deeply understand how MPOs manage the collective decision-making process with various
actors engaged, in the summer of 2017, we conducted site interviews with 60 policymakers and
managers in the four largest metropolitan areas in Texas. All interviews were semi-structured

2
The nomenclature can vary by states and agencies. Some refer to them as “MPO Policy Board,”
“Regional Transportation Committee,” or “Transportation Policy Council.”
23
and each lasted for between 30 and 60 minutes.
3
Table 2-1 shows that interviewees included
elected officials from cities and counties, professional managers such as directors of public
works and transportation, and MPO directors and staff.

[Table 2-1]

The interviews provided important insights on the MPO’s decision process management.
First, an MPO’s project selection process involves multiple stages. Member jurisdictions submit
transportation project proposals to the MPO for consideration. These projects typically represent
long-standing priorities as articulated in the jurisdiction’s comprehensive plan. Given the
extensive public participation involved in developing such a comprehensive plan, the projects are
likely to reflect something akin to jurisdictional consensus. Indeed, city councils will often attach
resolutions or letters of support for the project(s) described in the submitted proposal.
For this reason, the interviewees indicated that their interests and approaches are not
different whether they are elected officials or professional managers. This finding contrasts with
the assumption used in Gerber and Gibson (2009) that elected officials would focus on local
interests whereas public managers would pursue broad-based regional interests. Similarly,
whether officials were elected at-large or from districts, the interviewed governing board
members indicated that parochial interests regarding MPO projects were unimportant, given that
they had been appointed to seats representing a whole city or county.

3
The list of all interviewees is available in the Appendix B.
24
The MPO then assesses each submitted project proposal considering the needs and the
likely impact of the project on their region. An MPO will often establish a scorecard to grade
each proposal according to a specific set of factors. The three most common factors cited by the
interviewees were the existing level of congestion on the road, physical road conditions, and
traffic safety. Based on the scoring system, MPO staff rank order projects and present the list to
the governing policy board, which then makes final allocation decisions.
4
Our interviews suggest
that this list is generally accepted by the governing policy board and approved without any
changes being made.

Theory and Hypotheses
The process detailed above suggests that allocation decisions are data-driven and independent of
power and institutional factors. However, this simple story may not be correct. In studying
governance, Stone (1980) highlights the notion of “systemic power,” where all processes leading
up to a collective decision are influenced by power considerations because of a recognition
among the less powerful that those with power could exercise it to give or take away resources.
This reality can shade decisions, regardless of whether those with more power intend to use it or
if there is explicit competition or conflict between actors. Stone (1980) argues that local public
officials are subject to this dynamic in their dealings with business enterprises and other upper-
strata interest groups, even without the groups’ overt political maneuvering.

4
Depending on the size of the MPO, the MPO may establish policy subcommittees on a particular theme
(e.g., pedestrian and bicycle, transit, air pollution). These subcommittees review relevant transportation
project proposals and make recommendations about what projects should be funded. However, even in
these cases, the governing policy board retains ultimate decision-making authority.
25
Admittedly, Stone’s power represents invisible and underlying influence that individuals
or groups in dominant positions have; his theory does not focus on institutional decision-making.
Other theories, however, such as institutional analysis and development framework in Ostrom
(2009) and institutional collective action framework in Feiock (2013), scale up individual and
inter-group problems to the institutional level. Further, Moe (2005) claims that such power
relationships exist within and between governmental institutions. Applying these views to MPO
context, we argue that if a kind of systematic power has been institutionalized into formal voting
power via organizational rules, one should observe consequences for resource allocation. Indeed,
Stone’s systematic power to formal power construct suggests that institutionalized power should
matter even in settings where policymakers within organizations do not explicitly acknowledge
its role.
An empirical study conducted by Pfeffer and Salancik (1974) illustrates that this Stone-
type power dynamic plays a critical role in resource allocation decisions in universities, which
share some bureaucratic similarities with MPOs. They measure the power held by departments
both from interviews with its heads (systematic type) and from the analysis of archival records of
departmental representation on major university committees (formalized type), and show it is
possible to distinguish between systematic and formalized power empirically. Their analysis
finds that the departments with greater systematic power are consistently allocated larger budgets
than others, even after controlling for workload, number of faculty and national rank.
Thus, there is reason to study power relationships in the context of regional decision-
making even when participants in the process do not acknowledge so and understand more
clearly their role in the allocation of resources across jurisdictions. Our baseline research
question is a straightforward one: to what extent does the allocation of MPO local voting power
26
influence the spatial allocation of transportation investments? We start with the following simple
hypothesis:
H1) Institutional Power: A local government with more voting power on the governing policy
board will see more resources flow to its jurisdiction even though local representatives do not
intend to exercise their power in the decision-making process.
The perspective here is informed by the systemic power concept of Stone (1980), and so
this hypothesis does not rely upon evidence that cities with more voting power sought to use it. If
this hypothesis is confirmed, the fact that technical and advisory groups’ recommendation of
project selection is rubber stamped by the policy board suggests that the initial project list to be
considered may have been already tailored to the power distribution of the ultimate decision-
making authority. It then would validate Stone (1980)’s systematic power concept in this
institutional setting.
Further, the influence of power regarding decisions about which projects to fund may
depend on the degree to which power is concentrated or evenly distributed on the governing
policy board, a feature we describe as power structure. Jones (2010) argues that a model of
regional organization that relies upon transaction costs and voluntary cooperation frameworks
(e.g., Feiock 2007) may be incomplete unless it incorporates elements of power relationships. He
argues that special emphasis should be given to bargaining among governmental units at
different levels of power in studies of regional organizations.
Given this, our second hypothesis considers these relative power relationships. If voting
power is disproportionately distributed on the board, with one or a few primary cities having
most of the power, those cities could exert a dominant influence on the board’s decision-making.
27
If true, we would expect the power held by the remaining local jurisdictions to be ineffectual in
shaping the spatial distribution of transportation investments such that a particular jurisdiction
received funds.
By contrast, if the rules establish a more even distribution of voting power on the
governing policy board, individual jurisdictions may be more able to exercise power and see
investments located within their boundaries. Here, cities with largest voting power may still be in
primary power positions, but they may be short of possessing dominant power to
disproportionately control the allocation decisions. In this case, the remaining cities would be
able to collectively influence investment allocation outcomes.
Concentration of power has been identified as a critical consideration in collaborative
governance research. Imbalances between the resources or power of different stakeholders
threaten the legitimacy of collaborative process and thus can lead to policy outcomes in favor of
powerful actors (Emerson, Nabatchi, and Balogh 2012; Ansell and Gash 2008; Choi and
Robertson 2013). In these instances, institutions need to be (re)designed such that the rules
accommodate a balance of power among the actors for the sustainability of the governance
regime (Crosby and Bryson 2005; Tang and Mazmanian 2010). Thus, our second hypothesis is
as follows:
H2) Power Structure: a more even distribution of voting power among cities in the MPO
governing policy board is likely to result in more equitable allocation of public resources.

Data
28
This essay uses unique data on transportation investment by the four largest MPOs in Texas:
Dallas-Fort Worth, Houston-Galveston, Austin and San Antonio. We combine information from
three sources to construct the MPO project database. First, data on every MPO-funded
transportation roadway project built in Texas during 2001-2010 was obtained from the Texas
Department of Transportation (TxDOT). These data include comprehensive information on each
MPO-funded project, including the type of project, the sources and the amount of funding
provided by federal, state and local governments, and the highways and roads on which the
project was built. Regarding project categories, the data identifies 33 different categories.
TxDOT categorizes projects using different levels of aggregation, and we use their highest levels
of aggregation. This aggregation collapses the 33 into seven categories.
5
Second, TxDOT also
provided detailed geospatial data, which allowed us to geocode each project and identify the city
or cities in which it was built. Third, we collected spatial traffic information, such as annual
average daily traffic flow, the level of congestion, speed limit, and road pavement quality
(International Roughness Index) for every highway and road that are managed by TxDOT. All
data sources are documented in Table A2-1 in the Appendix A.
The full Texas MPO dataset included 7,350 projects in 25 metropolitan areas in the state.
However, the geospatial database did not include 3,179 of these projects, and we dropped these
projects from the working dataset as a result. The projects for which there was no geospatial data
were typically small projects, and they comprised a small percentage of MPO total expenditures.
The remaining 4,171 projects were merged with the data on traffic and roads to create an
augmented project dataset. Among the 25 MPO planning areas, we chose the largest four MPO

5
These seven categories are 1) Bridge 2) Freeway 3) Restoration 4) Traffic and Safety 5) Construction 6)
Scenic 7) Miscellaneous.
29
regions as a study frame for two reasons. First, our interviews were conducted only in these four
metropolitan areas, so we wanted a consistency. Second, we faced sample size issues, as the
other regions did not have many MPO project observations. Supporting this point, the four areas
we decided to study had 61 percent of all geocodable MPO projects built in Texas during the
period of this study. The choice of these four regions also provides useful variation in the
concentration of cities’ voting power in the policy boards, a key element to test the second
hypothesis about power structure and the allocation of public investment.
A measure of city-level voting power was added to this MPO project dataset. To
calculate city’s power share, the rosters of governing policy boards were obtained from the four
MPOs for the years between 2001 and 2010. We also tracked the bylaw documents and any
revisions made on the rules for allocating voting seats in this period. Cities’ voting power shares
were calculated such that each seat had a proportional amount of power. For example, if the
policy board had 10 seats, each seat would represent 10 percent of the power. This 10 percent
would then be allocated to the city or cities that decide which representative will occupy the seat.
For MPO policy board seats allotted to county or state district representatives, the interviews
indicated that it would be reasonable to distribute the representative’s voting power among the
constituent cities, based on each city’s voting-eligible population share of the total in that county
or state congressional district. If multiple cities were jointly assigned a seat, their shares were
divided based on their population.
6
Finally, based on the interviews, the votes held by state or
regional transportation or transit-related authorities were dropped from the denominator when

6
This incidence was only the case in Dallas-Fort Worth MPO. We also coded in an alternative way in
which we allocate a full single seat to one city that represents its group. The results did not change under
this approach.
30
calculating voting power share, unless these agencies had a clear local political jurisdictional
boundary and constituency.
7

Figure 2-1 shows the spatial distribution of MPO projects and cities in the four regions
with information on their numbers, mean and median of cities’ voting power shares, average
project duration years, and their size. An average MPO project took two to three years to
complete and cost more than a million dollars. Alluding to the rank order of power
concentration, the mean city voting power share was the lowest in the Houston metro area,
followed by Dallas-Forth Worth, Austin, and San Antonio.

[Figure 2-1]

Figure 2-2 visually illustrates the degree of power concentration among the four regions
by showing the distribution of voting power held by each city in the region. The rank order of
power concentration among the four metropolitan areas—with the Houston-Galveston metro
being the least concentrated, followed by Dallas-Fort Worth metro, Austin metro, and San
Antonio metro as being the most concentrated—reflects the underlying differences in their
institutional rules on voting seat allocation.
The rules in the Houston-Galveston and Dallas-Fort Worth areas establish a relatively
even distribution of the voting seats, whereas the rules in the Austin and San Antonio areas
allocate more power to its primary cities. As an example of how the rules operate, consider the

7
Our results do not change when we include these agencies into the denominator.
31
Houston-Galveston MPO for 2006. Of its 25 seats, one seat was allocated for each of the 7
largest cities and each of the 8 counties in the region. The City of Houston received additional
two seats and its county (Harris) additional one seat. Further, three seats were allocated to
smaller cities in three large counties and the remaining four seats were filled by representatives
of transportation agencies.
The Dallas-Fort Worth area MPO policy board comprised 40 seats. In their allocation,
smaller cities were grouped together to exceed the population threshold for having a seat and for
region-wide representation. The three largest cities in the region – Dallas, Fort Worth and
Arlington – were allotted six, three, and two seats, respectively. Fourteen seats were reserved for
individual cities or clusters of cities that had populations of 50,000 or greater. The two counties
that include Dallas and Fort Worth (Dallas and Tarrant) received two seats apiece, and the other
four counties in the metropolitan area were allocated one seat apiece. The remaining seven seats
were allocated to transportation agencies and DFW international airport.
For the Austin area MPO as of 2006, of the 23 seats of its governing board, four seats
were allocated to the City of Austin and three seats to its county (Travis). Each one seat was
allocated to two other counties (Williamson and Hay) in the region and 10 seats were further
allocated to state legislative districts that cover Austin either entirely or partially.
8
Among the
cities, Round Rock was the one of two cities that received a seat in the region as their population
exceeded 50,000. The other seat was allocated to a representative from small cities and City of
West Lake Hills was the representative in that year. The remaining two seats were allocated to
transportation agencies. Lastly, in San Antonio area MPO, among the 17 seats, the City of San

8
In Austin area MPO, state legislators were allotted ten seats between 2003 and 2007. Since 2008, the
rule changed the allotment to three seats and it abolished their representation since 2011.
32
Antonio and its county (Bexar) received six and four seats, respectively. The smaller cities in the
region were divided into three groups, with each receiving a seat. The remaining four seats were
distributed to transportation agencies.
The more inclusive representation in the Houston-Galveston region was possible because
its MPO did not designate a high population threshold, which allowed representation for several
other cities. It also included outlying counties in the region, which helped alleviate the power
concentration of the city of Houston. The Dallas-Fort Worth MPO also adopted an inclusive
representative system by clustering smaller municipalities with larger ones in groups so that they
could jointly exceed the designated population threshold and be allocated a seat. Such
considerations were less prominent in the Austin MPO and virtually absent in the San Antonio
MPO. The Herfindahl index associated with each MPO in Figure 2-2, which measures the degree
of power concentration, confirms this rank order.
9

[Figure 2-2]

Methods
To examine if cities with more voting power see more investment flow to their jurisdictions, we
estimate how the voting power share that a city holds on the MPO governing policy board
predicts the likelihood that a city gets a project in its boundary. In choosing a model, we start by
assuming that a project could be allotted to any city in the MPO planning area, but ends up being
located in a certain city (or cities). Such a structure is conducive to the use of a conditional

9
The value of Herfindahl index closer to 0 indicates less concentration of power while the value closer to
1 indicates more concentration of power.
33
logistic regression technique (i.e. fixed effects logistic regression). Notationally, we estimate the
following model:
Pr ⁡ ( 𝑦 𝑖𝑗 𝑡 ) =
exp ⁡ ( 𝛽 𝑃𝑜 𝑤𝑒 𝑟 𝑗𝑡
+ 𝛾 𝐶 𝑖 𝑡 𝑦 𝑗𝑡
+ 𝛿𝑃𝑟 𝑜 𝑗𝑒 𝑐𝑡 𝑖𝑡
∗ 𝑃𝑜 𝑤𝑒 𝑟 𝑗𝑡
)
∑ exp ⁡ ( 𝛽 𝑃𝑜 𝑤𝑒 𝑟 𝑘𝑡 𝑘 ∈ 𝐶 (𝑖𝑡 )
+ 𝛾 𝐶 𝑖 𝑡 𝑦 𝑘𝑡
+ 𝛿𝑃𝑟 𝑜 𝑗𝑒 𝑐𝑡 𝑘𝑡
∗ 𝑃𝑜 𝑤𝑒 𝑟 𝑘 𝑡 )

The unit of analysis is a project-city-year pair, and Pr(yijt) represents the latent probability
that project i will be located in city j in a year t, given all other city options in the MPO’s choice
set of cities (C(it)) at time t. Operationally, the dependent variable is an indicator variable coded 1
if a project is located in a city and 0 otherwise. Thus, if a project in a city gets funded by the
MPO, the project-city pair for that city in a given year t would be coded as 1 and all the other
city observations involving that project in year t would be coded 0. Consider Figure 2-3 for an
illustration of this model set up. In 2010, there were 125 cities in Houston-Galveston region.
Suppose MPO project i was allocated to city A in the region in that year (denoted as City 1 in
Figure 2-3). Then, that MPO project-city A pair is coded 1 and all the remaining 124 MPO
project-city pairs are coded 0. The model allows a project to be located in multiple cities to
account for the cross-jurisdictional nature of transportation investment as shown in Figure 2-1.
For instance, if the allocated MPO project cross three cities (say city A, B, and C), three project-
city pairs are coded 1 for this MPO project and the remaining 122 pairs are coded 0. Each MPO
project in a given year is coded in this way.

[Figure 2-3]

34
The key independent variable of interest is a city’s voting power share on the MPO
governing policy council (Power in equation 1). To isolate the effect of power, the model
includes city population, other city-level demographic variables, and measures of the highway
demand and travel behavior of residents (collectively denoted as City in equation 1). The city-
level demographic controls include measures of local economic vitality such as the
unemployment rate, the share of people with incomes below 100% of the federal poverty line,
and median household income. We expect that all three variables would be negatively associated
with a city’s likelihood of receiving an MPO project. Regarding median household income,
studies often suggest that more affluent citizens are better able to exercise power, and so can
successfully challenge and block projects that might introduce disamenities, such as the
congestion that could arise with a significant upgrade in road capacity or quality (Glaeser and
Ponzetto 2017; Brinkman and Lin 2019; Altshuler and Luberoff 2004).
We include a number of control variables that proxy for intensity of car use, based on the
premise that higher car use will be associated with higher demand for transportation
improvements. Three are demographic measures correlated with car use: the share of families
with children younger than 6 years old, the share of the population with at least a bachelor’s
degree, and the percentage of the population older than 64 years old. Since families with children
and adults with more education tend to rely on car use more than older people (Hanson and
Hanson 1981; Shen 2000; Boarnet 2011), we expect that the first two variables will be positively
associated with a city’s likelihood of receipt of an MPO project and the third will show a
negative relationship.
Two controls – the average commute time of residents and the percentage of residents
who use their vehicles to commute – seek to capture intensity of car use, which we take to be a
35
proxy for the demand for highways and roads. We expect that these will be positively associated
with the likelihood a city receives an MPO project within its boundary.
We also include as controls the number of business establishments and their number of
employees, which are measures of each city’s economic capacity and could be indicators of
whether roads are needed.
10
Data for all these controls were obtained from the Census. Summary
statistics for all city-level variables are presented in Table A2-2 in the Appendix A.
Project-level highway and road characteristics should also affect MPO investment
decisions, as the MPO practitioners we interviewed described them to be the most important
factors. The most significant were congestion relief, mobility, connectivity, and safety. For
example, projects to improve a highway with more congestion should be more likely to be
pursued by the MPO than projects focused on less congested highways, all else equal. Thus, the
model includes project type, road pavement quality (using the International Roughness Index),
average annual daily traffic flow, road speed limit, current level of congestion, and future level
of congestion projected by state DOT as control variables (denoted as Project in equation 1).
To this project-level vector, we also include an MPO project size. While the model
considers every city in a region as possible location(s) for the allocation of projects, its major
drawback is that the binary nature of the dependent variable does not consider project size. Such
an approach fails to capture a possibility that cities with greater voting power may systematically
receive greater scale projects (measured by expenditure). To address this, we construct a binary
variable that indicates if an MPO project’s size in dollars falls in or above the 90
th
percentile of

10
We report the results without the measures of business activities as they are highly correlated with
city’s population size. Substituting city’s population for business measures or simply adding these
measures, however, do not change the results in any substantive way.
36
the project expenditure distribution.
11
Because project-level fixed effects are embedded in the
conditional logistic regression, we incorporate these time-invariant project-level variables by
creating interaction terms for each characteristic with the power variable. These interaction terms
allow us to examine if the effects of power are moderated by project size considerations and road
characteristics. Summary statistics for all project-level variables are presented in Table A2-3 in
the Appendix A.
For a given MPO area, observations are pooled and a single regression is run with robust
standard errors clustered by projects, since MPO projects are long-term investments that are
implemented over multiple years. Note that, in the analysis, while cities’ MPO voting power
shares and populations are yearly available for the entire study period from 2001 through 2010,
other city-level demographic controls are matched to the project-city-year pairs to the extent that
data permits. Namely, the controls from 2000 Census are matched to the project-city-year pairs
in 2001-2005 and the controls from 2006-2010 American Community Survey are matched to
those in 2006-2010. The project-level highway and road characteristics are matched to all pairs
as interaction terms. Separate estimates are obtained for each MPO area, because an MPO’s
choice set (i.e., the set of cities it could choose to invest in) is mutually exclusive of the choice
set of other MPOs. This permits a comparative analysis to assess the second hypothesis
regarding power concentration, as the four MPOs show varying degrees of voting power
distribution.

11
The project expenditure size at the 90
th
percentile in each region is $3.3MM (Dallas-Fort Worth),
$6.0MM (Houston-Galveston), $3.6MM (Austin), and $4.9MM (San Antonio).
37
Results
We first describe the results in Houston-Galveston region, where the voting power is least
concentrated, and later compare them to the other three regions. All coefficients are presented in
odds-ratio to ease interpretation.
Regarding the power variable, our main variable of interest, panel B of Figure 2-4 shows
that power is consistently associated with the likelihood that a city receives an MPO project in its
jurisdiction, and in the expected direction. A one percentage point increase in a city’s power is
associated with a 45% greater likelihood of receiving a project within its boundary. Interestingly,
while one might have expected population to be a key driver of the distribution of resources,
given its central role in establishing the distribution of seats on MPO policy boards, the analysis
suggests otherwise. Population is negatively associated with the likelihood of having a project
within a particular city’s jurisdiction.
Other control variables have associations that conform with general expectations laid out
earlier. Cities with higher median incomes were less likely to have an MPO project located
within their boundaries. Higher city unemployment and higher poverty rates are associated with
decreases in the likelihood of that city receiving a project. An increase in the percentage of
families with young children in a city is associated with a greater likelihood of receiving an MPO
roadway project. We also see that the percentage of a city’s population that is elderly is
negatively associated with that city’s likelihood of receiving a project, whereas the share of a
city’s residents with a bachelor’s degree or graduate degree is positively associated with the
likelihood of having a project in its city’s boundary.
38
While the demographic relationships largely conformed to expectations, the relationships
for the commuting variables did not. Cities with residents who spend more time commuting were
less likely to get an MPO project, which runs counter to our expectation that commuting is a
signal of the salience of road investment. Similarly, the finding that the share of workers who
commute with their private autos was not related to the likelihood of getting a project. Lastly,
there are some moderating effects of road and highway conditions on power, but the power
variable itself remains as a strongly significant predictor.

[Figure 2-4]

Some might be concerned that the power and population variables are closely related,
given the role that population plays in some MPO seat allocation rules. To explore this, we rerun
the analysis including either power or population to see if results change in important ways. The
results of this exercise are shown in panels B and C of Figure 2-4. In panel B, which excludes
population, the odds-ratio on power diminishes somewhat, but the results remain virtually the
same. Similarly, when power is omitted from the model (panel C), the population variable
remains negatively associated with the likelihood of receipt of an MPO project.
12

Together, our analysis suggests that cities’ voting power still significantly predicts the
distribution of resources even after controlling for population, other city-level demographic and

12
We also repeated this exercise (panel C) for the other three regions. The coefficient for population was
negative in the Dallas-Fort Worth metro, and positive in the Austin and San Antonio metro, and none of
them was statistically distinguishable from zero, making a stronger case that it is power rather than
population that is associated with the geographic distribution of MPO projects. Also, the results for the
other three regions (panel A) remained robust when the analysis excluded population (panel B).
39
car use intensity, and project-level characteristics such as project size and road and highway
conditions. The analysis thus supports the first hypothesis regarding institutional power. Further,
the interaction term between project size and power is not significant, suggesting that cities with
greater power did not receive larger scale projects. Rather, it is cities’ influence through power
that impacts the likelihood of bringing the projects their home.
Figure 2-5 reports the results on power for all four MPO regions in specifications
including all controls. Recall that the second hypothesis posits that power should be a less
important factor in explaining the MPO project distributions as one moves from regions where
power is least concentrated to the region where power is most concentrated. Consistent with
expectations, the coefficient associated with power decreases as one moves from the Houston-
Galveston area to the San Antonio area. A percentage point increase in a city’s MPO voting
power in Houston-Galveston, Dallas-Fort Worth, Austin, and San Antonio area is associated
with increases in the likelihood that city receives an MPO project by 45%, 39%, 18%, and 6%,
respectively. Moreover, the power relationship is not statistically significant in San Antonio,
which is the MPO area with the greatest power concentration. Further investigation reveals that
an interaction term between project size and power in San Antonio area is positive and
significant, unlike the other three regions (See Figure A2-1 in the Appendix A). This indicates
that the city of San Antonio consistently received bigger scale projects, a key driver of the
weaker power result. San Antonio’s dominant position for having large scale projects within its
boundaries makes it virtually impossible for an increase in power among the remaining cities to
affect allocation outcomes. Compared with the Houston-Galveston region, the effect of power in
the Dallas-Fort Worth area is not statistically different, which suggests that the inclusive
representational rules that both MPOs adopted affects the equity of their resource allocation in a
40
similar way. As expected, the effect of power in Austin area is weakly different from the
Houston-Galveston area (p<0.1), whereas the power effect in the San Antonio region is
significantly different from that in the Houston-Galveston area (p<0.001).

[Figure 2-5]

Across the three regions, the results for the control variables, which we report in the
Appendix Figure A2-1, generally mirror those seen for the Houston-Galveston case, with some
exceptions. The coefficient on population for the Austin and San Antonio MPOs is not
statistically different from zero, making a stronger case that it is power rather than population
that is key. Similarly, the coefficient on elderly population is not statistically different from zero
for these two MPOs. Second, the relationship between project location and a city’s
unemployment rate is not consistent across the regions, suggesting that one should be cautious
drawing conclusions about the nature of this relationship. Lastly, the coefficient on the
percentage of a city’s population that commutes using a personal automobile is positive in the
Dallas-Fort Worth and Austin MPO, which offers some support for the salience argument
introduced above.
While the coefficients on power decline consistent with our expectations, we cannot
easily draw conclusions about the relative magnitude of effects across factors from the odds
ratios since the units of the independent variables are different. One way to make magnitude
comparisons is to quantify the size of the effect with a constant one standard deviation change in
a variable. Table 2-2 shows such a comparison. In the Houston-Galveston region, the magnitude
41
of the power effect is substantially greater than that of any other variables. Specifically, a one
standard deviation increase in a city’s power share in the region is associated with more than a
doubling of the likelihood it receives an MPO investment in its jurisdiction. Other demographic
variables explain the distribution of MPO projects more modestly. The other MPO regions are
something of a mixed bag. Power is important in all but the most concentrated region (San
Antonio), but the magnitude of the relationship does not vary monotonically as predicted.

[Table 2-2]

We also acknowledge that examining these four MPOs may not be an apples-to-apples
comparison from the viewpoint of their differences in jurisdictional fragmentation. The varying
degrees of metropolitan fragmentation may be a culprit that explains the underlying differences
in an MPO’s institutional rules on voting seat allocation across these four areas. From a policy
perspective, for example, it would be interesting to know if a more equitable allocation of
resources would have occurred if the San Antonio MPO had distributed its voting power more
evenly and inclusively.
To address this question, we investigate an MPO that saw significant changes in their
rules over time and test whether this impacted the distribution of resources. Among the four
regions, the Austin MPO had a substantial change in their rules and subsequently their
distribution of voting power.
13
As a result, the power dominance in the city of Austin fell from

13
By contrast, the other three regions did not see such significant changes in voting power. For example,
the power share of city of Dallas, Fort Worth, Houston, and San Antonio in 2001 was 21.3, 14.4, 21.1,
42
60 percent in 2001 to 57 percent in 2005 and to 45 percent in 2009. In Figure 2-6, we repeat the
analysis for the Austin MPO region with the period split into two: the 2001-2005 period when
the city of Austin still had dominance with its majority voting status (panel A) and 2006-2010
period when the city lost its dominant power position (panel B).
The results affirm our second hypothesis. The power variable is positive and significant
in the later period when the distribution of voting power became more even among localities
with Austin losing its dominance (panel B). In contrast, it is not significant in earlier years when
power was more concentrated in the city of Austin (panel A). Further, in these earlier years, the
interaction term between project size and power is positive and weakly significant. Recall that
this pattern was similarly observed in the San Antonio MPO. When the city of Austin maintained
its dominant power position, it consistently received bigger scale projects. Compared with the
later years, this helps explain why an increase in the remaining cities’ power in those earlier
years did not impact the likelihood of receiving an MPO project within their boundaries.
Hence, the analysis suggests that our claim on the second hypothesis is not only confined
to the between-metropolitan comparisons that are in different levels of fragmentation, but it is
also observed from a within-metropolitan change, holding the level of fragmentation constant.
The finding illustrates that changes in an MPO’s rules that induce a more even distribution of
voting power can achieve a more equitable allocation of public resources across cities throughout
the region.
[Figure 2-6]

and 81.4 percent, respectively. The corresponding share in 2010 was 19.5, 13.5, 21.2, and 75. 5 percent,
respectively.
43
Discussion and Conclusion
This essay examines regional planning organizations that plan and allocate public investment in
transportation and analyzes the extent to which the internal power structure of the governing
board of those organizations explains the geographic distribution of its public investment across
local jurisdictions. Focused on the four largest metropolitan areas in Texas, our analysis shows
that the power structure is consistently a major factor. Moreover, the degree of power
concentration of the dominant city in the region influences whether the power held by the other
cities matters; if power concentration is sufficiently great, the importance of the distribution of
the remaining power disappears. Aside from the finding that higher concentration of power
mitigates the power relationship, we are unable to say more about this relationship for
metropolitan areas with less intensive concentrations. Among the four regional planning areas, a
monotonic power effect was not observed, which points to a more complex dynamic. More
research will be needed to understand this better.
Interestingly, the results did not conform with the expectations of practitioners and
policymakers working inside these organizations, who claimed that power was at best a minor
factor shaping decision-making. Indeed, the findings revealed that the importance of the
institutional rules on public investment allocation is much greater than professionals believe
them to be.
Our study provides both theoretical and practical implications. On theoretical side, we
extend Stone (1980)’s systematic power concept to institutionalized settings such as local voting
power in regional governing policy boards. By extending Stone’s theory, we argue that
institutionalized power should matter even in a setting where policymakers do not acknowledge
44
its explicit role. While our empirical investigation focuses on regional organizations, our
perspective on institutionalized power may be relevant to other regional and interregional
organizations that also play important roles in urban and regional governance and public goods
provision.
Our work also provides a new perspective to the debate on the optimal structure and form
of metropolitan governance. Advocates of metropolitan consolidation have argued that city-
county consolidation can promote efficiency, equity, and accountability, a view many public
choice scholars eschew. Such structural reform can mitigate growing inequalities between central
cities and suburbs in principle, but consolidation referendum have not been popular in practice.
As a result, several scholars have focused on alternatives such as municipal annexation,
interlocal agreements and service contracts, and the creation of intergovernmental special
districts (Carr 2004).
Along this line, many have started looking at the role that regional coalitions and
governments can play to reduce the disparities across the metropolitan community (Dreier,
Mollenkopf, and Swanstorm 2004; Orfield 2011; Rusk 1993). Most work on regional institutions
have focused on whether they can facilitate coordination and regionalism (Sciara 2017; Kwon
and Park 2014; Kwon, Feiock, and Bae 2014; Gerber and Loh 2011). Some scholars even
suggest that existing regional bodies be granted more authority above their current control on
transportation polices to scale up their influence in other areas, such land use and affordable
housing, to advance equitable development (Orfield 2011).
Often neglected in these debates, however, is a discussion on how to design these
institutions and strike a balance in power among the constituent local jurisdictions to achieve
45
regional equitable outcomes. Savitch and Vogel (2004), for example, highlights that the power
dimension in city-county consolidation, which holds a key to a successful reform, has been
overlooked both by public choice scholars and consolidation reformers. From a governance
reform perspective, our work contributes to this line of thought by suggesting that policymakers
need to attend to the issue of balance of power in any existing collaborative institutions such as
regional bodies for successful governance reforms.
Our key point is that the design and structure of rules allocating power in regional
institutions matters for the subsequent distribution of resources across member jurisdictions. To
allow for cities to meaningfully contribute to the collective decision-making process, we
recommend that policymakers and public managers should develop representation systems in
which voting power is shared more inclusively among the local actors.
The rules adopted in the Houston-Galveston and Dallas-Fort Worth MPOs provide some
insights on this recommendation. In the Dallas-Fort Worth region, the organization adopted a
cluster-based representation approach in which small municipalities were paired with larger ones
in a group, so that seats represent the voices of smaller localities. The Houston-Galveston MPO
did not establish a population threshold for seat allocation that was too high, and the board
consequently included a broad range of cities. While designating a threshold that is proportional
to the distribution of cities’ populations is common in other MPOs, and may be inevitable to
some extent, it can institutionally exclude smaller jurisdictions from the decision-making
process. Also, both MPOs allocated seats to their outlying counties in metropolitan areas, which
further helped alleviate the concentration of power in their central cities. These instances
illustrate how various rules can help establish a collaborative governance model that results in
equitable allocation of resources.
46
References

Altshuler, Alan A., and David E. Luberoff. 2004. Mega-projects: The Changing Politics of
Urban Public Investment. Brookings Institution Press.
Ansell, Chris, and Alison Gash. 2008. Collaborative Governance in Theory and Practice. Journal
of Public Administration Research and Theory 18(4): 543-571.
Ansolabehere, Stephen, Alan Gerber, and Jim Snyder. 2002. Equal Votes, Equal Money: Court-
ordered Redistricting and Public Expenditures in the American States. American Political
Science Review 96(4): 767-777.
Basolo, Victoria, and Chihyen Huang. 2001. Cities and Economic Development: Does the City
Limits Story Still Apply? Economic Development Quarterly 15(4): 327-339.
Boarnet, Marlon G. 2011. A Broader Context for Land Use and Travel Behavior, and a Research
Agenda. Journal of the American Planning Association 77(3): 197-213.
Brinkman, Jeffrey, and Lin, Jeffrey. (2019). Freeway Revolts. Working paper, Federal Reserve
Bank of Philadelphia.
Carr, Jered B. 2004. Perspectives on City-County Consolidation and Its Alternatives. City-
County Consolidation and Its Alternatives: Reshaping the Local Government Landscape pp.1-
24.
Choi, Taehyon, and Peter J. Robertson. 2013. Deliberation and Decision in Collaborative
Governance: A Simulation of Approaches to Mitigate Power Imbalance. Journal of Public
Administration Research and Theory 24(2): 495-518.
Crosby, Barbara C., and John M. Bryson. 2005. Leadership for the Common Good: Tackling
Public Problems in a Shared-Power World. Vol. 264. John Wiley & Sons.
Dahl, Robert A. 1961. Who Governs?: Democracy and Power in an American City. Yale
University Press.
de Benedictis-Kessner, Justin, and Christopher Warshaw. 2016. Mayoral Partisanship and
Municipal Fiscal Policy. Journal of Politics 78(4): 1124-1138.
Deyle, Robert E., and Ryan E. Wiedenman. 2014. Collaborative Planning by Metropolitan
Planning Organizations: A Test of Causal Theory. Journal of Planning Education and Research
34(3): 257-275.
Dreier, Peter, John H. Mollenkopf, and Todd Swanstrom. 2004. Place Matters: Metropolitics for
the Twenty-First Century. University Press of Kansas.
Einstein, Katherine Levine, and David M. Glick. 2018. Mayors, Partisanship, and Redistribution:
Evidence Directly From US Mayors. Urban Affairs Review 54(1): 74-106.
47
Emerson, Kirk, Tina Nabatchi, and Stephen Balogh. 2012. An Integrative Framework for
Collaborative Governance. Journal of Public Administration Research and Theory 22(1): 1-29.
Feiock, Rrichard C. 2007. Rational Choice and Regional Governance. Journal of Urban Affairs
29(1): 47-63.
Feiock, Richard C. 2013. The Institutional Collective Action Framework. Policy Studies
Journal 41(3): 397-425.
Gerber, Elisabeth R., Adam Douglas Henry, and Mark Lubell. 2013. Political Homophily and
Collaboration in Regional Planning Networks. American Journal of Political Science 57(3): 598-
610.
Gerber, Elisabeth R., and Clark C. Gibson. 2009. Balancing Regionalism and Localism: How
Institutions and Incentives Shape American Transportation Policy. American Journal of Political
Science 53(3): 633-648.
Gerber, Elisabeth R., and Daniel J. Hopkins. 2011. When Mayors Matter: Estimating the Impact
of Mayoral Partisanship on City Policy. American Journal of Political Science 55(2): 326-339.
Gerber, Elisabeth R., and Carolyn G. Loh. 2011. Prospects for Expanding Regional Planning
Efforts in Michigan. Urban Studies 48(11): 2303-2319.
Glaeser, Edward L., and Giacomo AM Ponzetto. 2017. The Political Economy of Transportation
Investment. Economics of Transportation. Published electronically. Nov 22. doi:
10.1016/j.ecotra.2017.08.001.
Hajnal, Zoltan L., and Jessica Trounstine. 2010. Who or what governs?: The Effects of
Economics, Politics, Institutions, and Needs on Local Spending." American Politics
Research 38(6): 1130-1163.
Hanson, Susan, and Perry Hanson. 1981. The Travel-activity Patterns of Urban Residents:
Dimensions and Relationships to Sociodemographic Characteristics. Economic
Geography 57(4): 332-347.
Hochschild, Jennifer L. 2008. Clarence N. Stone and the Study of Urban Politics. Power in the
City: Clarence Stone and the Politics of Inequality.
Innes, Judith E., and Judith Gruber. 2005. Planning Styles in Conflict: the Metropolitan
Transportation Commission. Journal of the American Planning Association 71(2): 177-188.
Jimenez, Benedict S. 2014. Separate, Unequal, and Ignored? Interjurisdictional Competition and
the Budgetary Choices of Poor and Affluent Municipalities. Public Administration Review 74(2):
246-257.
Jones, Bryan D. 2010. Conflict, Power and Irreconcilable Preferences: Some Limits to Self-
organizing Mechanisms. Self-Organizing Governance: Collaborative Mechanisms to Mitigate
Institutional Collective Action Dilemmas pp.73-90.
48
Kwon, Sung-Wook, and Sang-Chul Park. 2014. Metropolitan Governance: How Regional
Organizations Influence Interlocal Land Use Coordination. Journal of Urban Affairs 36(5): 925-
940.
Kwon, Sung-Wook, Richard C. Feiock, and Jungah Bae. 2014. The Roles of Regional
Organizations for Interlocal Resource Exchange: Complement or Substitute? American Review
of Public Administration 44(3): 339-357.
Lewis, Paul George, and Mary Sprague. 1997. Federal Transportation Policy and the Role of
Metropolitan Planning Organizations in California. Public Policy Institute of California.
Lubell, Mark, Jack M. Mewhirter, Ramiro Berardo, and John T. Scholz. 2017. Transaction Costs
and the Perceived Effectiveness of Complex Institutional Systems. Public Administration Review
77(5): 668-680.
Minkoff, Scott L. 2012. The Proximate Polity: Spatial Context and Political Risk in Local
Developmental Goods Provision. Urban Affairs Review 48(3): 354-388.
Moe, Terry M. 2005. Power and Political Institutions. Perspectives on Politics 3(2): 215-233.
Nall, Clayton. 2018. The Road to Inequality: How the Federal Highway Program Polarized
America and Undermined Cities. Cambridge University Press.
National Association of Regional Councils. Retrieved April 1, 2019, from narc.org.
Olson, Mancur. 1971. The Logic of Collective Action. Vol. 124.
Orfield, Myron. 2011. Metropolitics: A Regional Agenda for Community and Stability.
Brookings Institution Press.
Ostrom, Elinor. 2009. Understanding Institutional Diversity. Princeton University Press.
Peterson, Paul E. (1981). City Limits. University of Chicago Press.
Pfeffer, Jeffrey, and Gerald R. Salancik. 1974. Organizational Decision Making as a Political
Process: The case of a university budget. Administrative Science Quarterly 19(2): 135-151.
Rusk, David. 1993. Cities Without Suburbs. Woodrow Wilson Center Press.
Sanchez, Thomas W. 2006. An Inherent Bias? The Brookings Institution Series on
Transportation Reform.
Saiz, Martin. 1999. Mayoral Perceptions of Developmental and Redistributive Policies: a Cross-
national Perspective. Urban Affairs Review 34(6): 820-842.
Savitch, Hank V., and Ronald K. Vogel. 2004. Suburbs Without a City: Power and City-County
Consolidation. Urban Affairs Review 39(6): 758-790.
49
Schneider, Mark. 1989. The Competitive City: The Political Economy of Suburbia. University of
Pittsburgh Press.
Sciara, Gian-Claudia. 2017. Metropolitan Transportation Planning: Lessons From the Past,
Institutions For the Future. Journal of the American Planning Association 83(3): 262-276.
Shapley, Lloyd S., and Martin Shubik. 1954. A Method for Evaluating the Distribution of Power
in a Committee System. American Political Science Review 48(3): 787-792.
Shen, Qing. 2000. Spatial and Social Dimensions of Commuting. Journal of the American
Planning Association 66(1): 68-82.
Snyder Jr, James M., Michael M. Ting, and Stephen Ansolabehere. 2005. Legislative Bargaining
Under Weighted Voting. American Economic Review 95(4): 981-1004.
Solof, Mark. 1998. History of Metropolitan Planning Organizations. Newark: North Jersey
Transportation Planning Authority, Inc.
Song, Minsun, Hyung Jun Park, and Kyujin Jung. 2018. Do Political Similarities Facilitate
Interlocal Collaboration? Public Administration Review 78(2): 261-269.
Stone, Clarence N. 1980. Systemic Power in Community Decision Making: A Restatement of
Stratification Theory. American Political Science Review 74(4): 978-990.
Stone, Clarence N. 1989. Regime Politics: Governing Atlanta, 1946-1988. Univ Press of Kansas.
Tang, Shui-Yan, and Daniel A. Mazmanian. 2010. Understanding Collaborative Governance
from the Structural Choice-Politics, IAD, and Transaction Cost Perspectives. Unpublished
Manuscript. USC Bedrosian Center Working Paper Series.
Tiebout, Charles M. 1956. A Pure Theory of Local Expenditures. Journal of Political Economy
64(5): 416-424.
Weingast, Barry R. 1979. A Rational Choice Perspective on Congressional Norms. American
Journal of Political Science 23(2): 245-262.
Yi, Hongtao, Liming Suo, Ruowen Shen, Jiasheng Zhang, Anu Ramaswami, and Richard C.
Feiock. 2017. Regional Governance and Institutional Collective Action for Environmental
Sustainability. Public Administration Review. Published electronically 29 June. doi:
10.1111/puar.12799.

50
Table 2-1: 60 MPO Policymakers Interviewed

Type of MPO
Policymakers
Dallas-Fort
Worth MPO
Houston
MPO
Austin
MPO
San Antonio
MPO
Total (60) 21 18 12 9
Mayor 6 1 2 2
Council Member 2 1 1 2
City Manager - 1 1 1
City Director - 6 3 1
County Judge 2 - 1 -
Commissioner 4 2 2 1
MPO/State Staff 4 6 2 2
Others 3 1 - -
Notes: - indicates no respondent in that position. Others include representatives from transit authority,
business interest group, and county precinct administrator. The list of interviewees is available in the
Appendix B.
Table 2-2: Effects of Independent Variables by Standard Deviation
Notes: The effects were calculated based on the results in Figure A2-1 and standard deviations of the
variables are reported in Table A2-2 and A2-3 in the Appendix A. Bolded are the effects that are
statistically significant at least at 0.05 level.
Variables
Houston-
Galveston
metro
Dallas-Fort
Worth
metro
Austin
metro
San
Antonio
metro
Power (%) 107% 91% 156% 80%
Population (thousands) -27% -32% -27% -6%
Project size x power 0% 0% 4% 12%
Unemployment rate (%) -26% 4% -29% 17%
Poverty rate (%) -18% -6% -38% -81%
Median household income
($1000)
0% 0% 0% 0%
Family household with children
(%)
40% 9% 66% 27%
Elderly population (%) -34% -41% -3% 8%
Bachelor’s degree holder (%) 31% 42% 68% -51%
Commute time (mins) -15% -18% -89% -27%
Commuter with private auto (%) 6% 14% 85% -13%
Project type x power -2% -1% 22% -33%
Road pavement quality (IRI) x
power
4% 6% 7% 17%
Traffic flow x power 23% 10% -4% -23%
Road speed limit x power -45% -16% -161% 45%
Current congestion x power 84% 17% 44% 134%
Future congestion x power -70% -18% 52% -101%
51
Figure 2-1: Summary Information on MPO Projects and Local Voting Power
Dallas-Fort Worth Region
Population: 6.3 million (2010)
N of cities: 139
N of cities with non-zero power share: 136
Mean (0.75%) and median (0.13%) of cities’ voting power share
N of MPO projects built in cities: 945
N of MPO project-year observations: 2701
Average project duration year: 2.86
Average project size in a year: $1.5 million
Houston-Galveston Region
Population: 6.0 million (2010)
N of cities: 125
N of cities with non-zero power share: 125
Mean (0.69%) and median (0.04%) of cities’ voting power share
N of MPO projects built in cities: 649
N of MPO project-year observations: 1942
Average project duration year: 2.99
Average project size in a year: $1.9 million
Austin Region
Population: 1.7 million (2010)
N of cities: 46
N of cities with non-zero power share: 46
Mean (2.11%) and median (0.08%) of cities’ voting power share
N of MPO projects built in cities: 152
N of MPO project-year observations: 511
Average project duration year: 2.33
Average project size in a year: $1.0 million
San Antonio Region
Population: 1.8 million (2010)
N of cities: 28
N of cities with non-zero power share: 27
Mean (3.58%) and median (0.1%) of cities’ voting power share
N of MPO projects built in cities: 279
N of MPO project-year observations: 917
Average project duration year: 3.29
Average project size in a year: $1.0 million
Notes: All unincorporated areas in each county is considered as a single place and is counted in the
number of cities.

52
Figure 2-2: Distribution and Concentration of MPO Local Voting Power

Notes: Herfindahl index was calculated from the cities’ average voting power shares in 2001-2010.

Figure 2-3: Conditional Logit Model Setup

MPO
City 1-MPO
project i pair
Coded 1
City 2-MPO
project i pair
Coded 0
City 3-MPO
project i pair
Coded 0
City 4-MPO
project i pair
Coded 0
............
…………
Coded 0
City 124-MPO
project i pair
Coded 0
City 125-MPO
project i pair
Coded 0
53
Figure 2-4: Conditional Logit Regression Results for the Houston MPO

Panel A Panel B Panel C
Notes: Unit of analysis is project-year-city pair. A total of 1942 project-year pairs were matched to 124
(2001, 2003-2005) or 125 cities (2006-2010), resulting in 242,016 observations for the analysis in all
three panels. The data for 2002 was not available. *** p<0.001, ** p<0.01 * p<0.05.

Figure 2-5: Conditional Logit Results on Power in All Four Regions

Notes: The figure summarizes four models with full controls.
*** p<0.001, ** p<0.01 * p<0.05
54
Figure 2-6: Conditional Logit Results on Austin MPO by Period
Panel A Panel B

Notes: *** p<0.001, ** p<0.01 * p<0.05. + p<0.1

55
Chapter 2 Appendix A. Figures and Tables
Figure A2-1: Conditional Logit Results for the Four MPO Regions

Notes: *** p<0.001, ** p<0.01 * p<0.05.

56
Table A2-1: Data and Source

Data/Variables Description Source
Voting Power
The share of voting power that cities have in a MPO
governing board. The number of eligible voters in
2000 census blocks is used to translate power from
counties/state congressional districts to cities.

Joint Power
Agreement or Bylaws
documents obtained
from each MPO; U.S
Census; National
Historical Geographic
Information Systems
(NHGIS); Texas
Legislative Council
MPO project list with
geospatial and expenditure
information
All MPO roadway projects implemented in Texas
during 2001-2010. All geocoded with expenditure
information
Texas Department of
Transportation
Annual Average Daily
Traffic (AADT) count
Total volume of vehicle traffic of a highway or
roadfor a year divided by 365 days (vehicles/day)
Texas Department of
Transportation
Road Speed Limit Road Speed Limit (mph).
Texas Department of
Transportation
Road International
Roughness Index
A value that describes the amount of roughness
measured in inches per mile, with readings taken at
intervals of 0.1 miles by a profiler van driving the
roadway. Measurements are independent of weather
conditions (temperature, sunlight, or wind). IRI
values range from 1 (smoothest) to 950 (roughest).
Unit: Inch/mile
Texas Department of
Transportation
Level of Current
Congestion
Vehicle density on the roadway (vehicles/mile)
Texas Department of
Transportation
Level of Future
Congestion
Vehicle density on the roadway projected for 2032
(vehicles/mile)
Texas Department of
Transportation
Demographic Control
Variables
City’s population (both total population size and %
of population in the region), unemployment rate,
poverty rate, median household income, % of
bachelor’s degree holders among the population 25
years older, % of the elderly population, % of Family
household with own children under 6 years,
commute time, and % of commuters using private
vehicle
U.S. Census
Decennial Census
2000 & American
Community Survey
2006-2010
Employment Environment
Variables
Number of business establishments, number of paid
employees
U.S. Census
Zip Code Business
Patterns 2001 to 2010

57
Table A2-2: Summary Statistics for City-level Independent Variables 2001-2010
VARIABLES Dallas Fort-Worth Area Houston-Galveston Area
CITY-LEVEL Obs Mean
Std.
Dev
Min Max Obs Mean
Std.
Dev
Min Max
Power (%) 1310 0.75 2.33 0 21.51 1049 0.69 2.35 0.00 21.24
Population
(in thousands)
1310 39.93 125.06 0.03
1202.7
2
1049 27.98 187.79 0.05 2108.95
Population (%) 1310 0.72 2.25 0.00 23.94 1049 0.52 3.48 0.00 41.03
Unemployment Rate (%) 1310 5.16 3.13 0 20.95 1049 6.38 5.02 0 58.05
Poverty Rate (%) 1310 8.63 6.67 0 38.69 1049 11.89 8.28 0 43.47
Median Household
Income (in thousands)
1310 75.96 36.53 28.75 261.71 1049 67.59 37.64 20.50 250
% of Elderly Population 1310 9.13 4.55 2.25 27.78 1049 11.42 4.49 2.25 31.21
% of Population with
Bachelor’s Degree
1310 28.42 17.16 0 82.39 1049 23.82 19.85 0 85.29
% of Family with Own
Children under 6 Years
1310 11.19 4.60 0 26.92 1049 9.39 4.70 0 23.95
Commute Time (mins) 1310 29.24 5.29 16.19 46 1049 28.46 6.48 17.95 59.44
% of Commuters with
Private Autos
1310 92.67 4.01 67.68 100 1049 93.63 3.88 76.46 100
VARIABLES Austin Area San Antonio Area
CITY-LEVEL Obs Mean
Std.
Dev
Min Max Obs Mean
Std.
Dev
Min Max
Power (%) 410 2.11 8.57 0 59.98 265 3.58 15.15 0 84.52
Population
(in thousands)
410 26.62 112.69 0.2 795.53 265 52.63 235.82 0.43 1334.41
Population (%) 410 1.85 7.87 0.01 55.63 265 3.29 14.71 0.03 80.23
Unemployment Rate (%) 410 5.15 3.18 0 14.91 265 4.71 3.13 0.44 13.59
Poverty Rate (%) 410 8.91 7.80 0 36.86 265 8.91 7.90 0 34.44
Median Household
Income (in thousands)
410 72.42 30.83 26.73 160.98 265 77.76 35.49 26.63 172.5
% of Elderly Population 410 11.88 7.09 1.82 37.40 265 14.47 7.48 3.72 36.57
% of Population with
Bachelor’s Degree
410 31.31 20.44 1.94 86.57 265 36.24 21.85 3.71 79.17
% of Family with Own
Children under 6 Years
410 10.39 5.37 0 26.04 265 8.94 5.33 0.91 28.17
Commute Time (mins) 410 29.93 5.88 17.78 47.26 265 24.43 4.24 14.72 33
% of Commuters with
Private Autos
410 91.36 4.87 76.95 100 265 92.63 4.19 76.50 98.75

58
Table A2-3: Summary Statistics for Project-level Independent Variables 2001-2010
VARIABLES Dallas Fort-Worth Area Houston-Galveston Area
PROJECT-LEVEL Obs Mean
Std.
Dev
Min Max Obs Mean
Std.
Dev
Min Max
Project size
(90
th
percentile)
2701 0.10 0.30 0 1 1942 0.10 0.30 0 1
Project type 2701 3.97 1.52 1 7 1942 4.12 1.43 1 7
Road pavement
quality
2701 116.13 39.66 0 314 1942 114.39 36.35 0 274
Average annual daily
traffic flow
2701 37188 29164 260 135852 1942 42179 37675 0 144005
Road speed limit 2701 56.29 8.10 0 70 1942 56.66 8.26 0 70
Current congestion 2701 32.55 24.00 1.14 160.35 1942 29.81 23.89 0.56 131.18
Future congestion 2701 48.04 34.44 1.95 224.49 1942 45.69 37.76 1.19 234.17
Project size x Power 2701 0.05 0.55 0 21.30 1942 0.04 0.35 0 6.20
Project type x Power 2701 1.88 5.33 0 97.29 1942 3.27 10.56 0.001 141.62
Road pavement
quality x Power
2701 54.21 152.71 0 2785.22 1942 86.92 261.73 0 3122.74
Average annual daily
traffic flow x Power
2701 16718 51182 0 1372463 1942 31265 114067 0 2072264
Road speed limit
x Power
2701 26.29 71.88 0 1290.65 1942 45.29 139.88 0 1380.80
Current congestion
x Power
2701 14.72 40.64 0 893.45 1942 23.08 87.21 0.003 1495.34
Future congestion
x Power
2701 21.86 61.26 0 1312.83 1942 35.19 132.11 0.004 2093.46
VARIABLES Austin Area San Antonio Area
PROJECT-LEVEL Obs Mean
Std.
Dev
Min Max Obs Mean
Std.
Dev
Min Max
Project size
(90
th
percentile)
511 0.10 0.30 0 1 917 0.10 0.30 0 1
Project type 511 3.21 1.14 1 5 917 3.99 1.46 1 7
Road pavement
quality
511 88.68 30.09 0 190.76 917 101.16 37.48 0 299.46
Average annual daily
traffic flow
511 41173 26282 249 94265 917 40347 28779 1897 111335
Road speed limit 511 60.14 7.34 36.81 75 917 59.38 10.35 29.36 70
Current congestion 511 34.19 24.36 1.68 111.50 917 30.15 15.47 2.60 83.67
Future congestion 511 52.65 35.59 2.35 167.14 917 47.56 24.08 4.22 117.14
Project size x Power 511 0.24 2.64 0 56.67 917 0.34 3.95 0 84.52
Project type x Power 511 7.99 30.36 0.003 283.34 917 14.97 62.28 0 528.71
Road pavement
quality x Power
511 214.84 756.21 0 5955.81 917 393.28 1693.71 0 25310.41
Average annual daily
traffic flow x Power
511 101961 429965 3.85 4813865 917 159801 771274 0 7304841
Road speed limit
x Power
511 150.23 538.89 0.16 3843.58 917 224.34 893.65 0 5700.03
Current congestion
x Power
511 78.59 337.22 0.03 5196.99 917 113.27 496.45 0 4678.34
Future congestion
x Power
511 120.50 502.09 0.05 7638.60 917 179.82 793.57 0 7344.15
59

Chapter 2 Appendix B. MPO Practitioners Interviewed (60)

San Antonio Metropolitan Area (9)

MPO and TxDOT Staff
Isidro Martinez Director of Alamo Area Metropolitan Planning Organization
Mario Jorge District Engineer of TxDOT San Antonio District
Transportation Policy Board member
Kevin A. Wolff County Commissioner Precinct 3, Bexar County
Renee Green Director of Public Works/County Engineer, Bexar County
Ron Reaves Council Member District 3, City of New Braunfels
Chris Riley Mayor, City of Leon Valley representing Greater Bexar County
Council of Cities
Ray Lopez Council Member District 6, City of San Antonio
Ron Nirenberg Mayor, City of San Antonio
Scott Wayman City Manager, Live Oaks representing Northeast Partnership

Austin Metropolitan Area (12)

TxDOT Staff
Terry McCoy District Engineer, TxDOT Austin District
Kevin Dickey Deputy District Engineer, TxDOT Austin District
Transportation Policy Board member
Ann Kitchen Council Member District 5, City of Austin
Dale Ross Mayor, City of Georgetown
Trey Fletcher Assistant City Manager, City of Pflugerville
Craig Morgan Mayor, City of Round Rock
Gerald Daugherty County Commissioner Precinct 3, Travis County
Brigid Shea County Commissioner Precinct 2, Travis County
Sarah Eckhardt County Judge, Travis County
Technical Advisory Committee
Bob Daigh County Engineer, Williamson County
Gary Hudder Director of Transportation, City of Round Rock
Meredith Johnson Planner II, City of Buda representing small cities in Hays County

Houston Metropolitan Area (18)

Houston Galveston Area Council and TxDOT Staff
David Wurdlow Transportation Program Manager
Nicholas Williams Transportation Coordinator
Rick Guerrero Intergovernmental Relations Manager
Hans-Michael Ruthe Principal Planner
Bill Brudnick Director of Transportation Development and Planning, TxDOT
Houston District
Tucker Ferguson District Engineer, TxDOT Beaumont District
60
Chapter 3. It’s Not Just Welfare: Racial Inequality and the Local
Provision of Public Goods in the United States

Summary
Recent research shows that inequality between racial groups is a critical determinant of
redistributive policy in the U.S. Using various measures of local and state spending and examining
multiple levels of geographical and political jurisdictions, we extend this research to government
spending on local public goods. Specifically, we examine (1) whether the extent to which income
inequality falls along racial lines dampens local and state government spending on public goods,
(2) which types of public goods are most affected by the racial structure of inequality, and (3)
whether political variables such as local leaders’ racial identities and party affiliation mediate the
relationship between racial inequality and spending on public goods. The findings reaffirm the
need to consider racial diversity and inequality jointly as influences on policy.

61
Introduction
Classical public choice models of redistributive policy in democracies assume that the median
voter opts for a rate of redistributive taxation that maximizes her own post-transfer income
(Romer 1975; Meltzer and Richard 1981; cf. McCarty, Poole, and Rosenthal 2006). These
models predict that the median income voter will choose a higher level of redistribution as pre-
tax income inequality rises.
But the power of group identities and attitudes to shape voters’ preferences about
redistribution complicates this prediction (Tajfel and Turner 1986; Alesina and Glaeser 2004;
Hero 1998; Lee 2017). If voters judge redistributive policies on the basis of how they affect
groups they identify with or dislike, they will often oppose policies that transfer income away
from their own group to others, even if these policies make them individually better off (Alesina
and Glaeser 2004; Luttmer 2001; Gilens 1999). Group stereotypes may also color the way that
policy makers and program administrators explain economic inequality, assess its urgency and
formulate plans for dealing with it (Soss et al. 2008, 2011).
14

These insights have focused growing attention on the racial and ethnic “structure” of
inequality as a critical determinant of redistributive policy (Lind 2007; Baldwin and Huber 2010;
Hero and Levy 2016, 2017).
15
By racial “structure,” scholars have meant the extent to which
total disparities in income are a result of inequality among individuals within salient social

14
Scholars have proposed a number of other reasons that diversity may lower redistribution and the
provision of public goods, including preference heterogeneity between groups (Alesina, Baqir, and
Easterly 1999), the efficiency of imposing social sanctions (Miguel and Gugerty 2005), and the ease of
coordination (Habyarimana et al. 2007). Our focus here is on explanations that invoke group affinity and
accompanying stereotypes as a factor in redistributive policy choices (Luttmer 2001; Vigdor 2000;
Alesina, Baqir, and Easterly 1999; Hero 1998).
15
Lupu and Pontusson (2011) also examine the structure of inequality across class groups.
62
groups or of inequality between members of different groups. Under standard assumptions,
16
and
holding the total level of income inequality and racial diversity constant, redistribution entails
greater inter-group transfers as intergroup inequality rises (Lind 2007; Hero and Levy 2016,
2017). If the pivotal influence on policy is drawn from the wealthier group and feels animus
toward the subordinate group or stereotypes it as “undeserving,” more intergroup inequality is
expected to dampen redistributive generosity. A large body of research explores the influence of
total income inequality and racial diversity on redistribution and has reached mixed conclusions
(Benabou 2000; Alesina and Glaser 2004; though see Milanovic 2000; Hopkins 2011; Boustan et
al. 2013). Research on the racial structure of inequality has obtained more consistent results by
attending to the interdependence of these two factors: inequality can affect redistribution in
different ways depending on whether it straddles or parallels ethnic group boundaries (Lind
2007), and diversity has a more pronounced negative impact on redistribution when between-
group income differentials are wide (Hero and Levy 2017; see also Matsubayashi and Rocha
2012).
Research on this topic in the U.S. has focused on welfare policy as a dependent variable
(Lind 2007; Hero and Levy 2017). Welfare is arguably an “easiest case” for theories linking the
racial structure of inequality to redistribution because it is a racialized policy domain that entails
pure transfers of income. But this paper argues that the racial structure of inequality influences a
considerably broader range of public policies that involve some measure of redistribution. We
focus on the local provision of public goods, a policy area that has been found to be influenced

16
Models often assume a linear tax and equal allocation of tax revenue net of deadweight loss across
individuals (Romer 1975; Meltzer and Richard 1981; McCarty, Poole, and Rosenthal 2006).
63
by variation in the level of racial diversity (Alesina, Baqir, and Easterly 1999; though see
Hopkins 2009 and Boustan et al. 2010).
Following Hero and Levy (2016, 2017), we measure the racial structure of local income
inequality using the Theil Index, which is an entropy-based measure of inequality that can be
decomposed into between and within group components. Using a variety of data sources that
span multiple geographic and political jurisdictions and encompass an era in which American
political leaders often linked fiscal distress with immigration and demographic change, we
explore whether variation in the between-race component of income inequality predicts variation
in the local provision of public goods. We also theorize and examine which types of public
goods would be most affected by retrenchment in response to rising between-group inequality.
We argue that public goods that admit private substitution are most likely to be affected by racial
inequality. Finally, we analyze political factors that would be expected to mediate the
relationships between demographics and policy, including the party affiliation of local leaders,
descriptive representation, and the degree of metropolitan fragmentation.

Group Attitudes and Support for Public Goods Spending
Research has found strong relationships between group animus and conservative fiscal attitudes,
including opposition to redistribution. Notably, these findings often hold even when the policy in
question has no overt connection to race or ethnicity. The most consistent findings link a variety
of measures of White prejudice toward Blacks to opposition to welfare (Sears et al. 2000; Gilens
1999; Lind 2007; Luttmer 2001; Bobo and Hutchings 1996). Some have also linked conservative
fiscal preferences to opposition to immigration or anti-Latino sentiment (Hajnal and Rivera
2014). Racial diversity may also erode interpersonal trust, participation, and other forms of
64
engagement with the public sphere (Putnam 2007), a dynamic that is stronger among those high
in racial animus (Costa and Kahn 2003; Alesina and La Ferrara 2002). Issue framing and elite
messaging also play an important role in policy racialization. Gilens (1999) traces the rise of
imagery of Black Americans in news stories about welfare and argues that the increasing
prominence of racialized portrayals of welfare beneficiaries led to diminished support for the
program among American Whites after the 1960s.
Consistent with theories positing in-group bias in the allocation of resources, researchers
have often found links between racial/ethnic context and local or state economic policy. States
and localities with higher Black populations tend to have lower expenditures on welfare and
public goods (Hero 1998; Key 1949; Blalock 1967; Luttmer 2001; Vigdor 2000; Fellowes and
Rowe 2004; McGuire and Merriman 2006; Fording 2003; Johnson 2003; Wright 1976; for
extensions to multi-racial contexts with mixed results, see Fox 2004; Hero and Tolbert 1996;
Hero and Preuhs 2007; Abrajano and Hajnal 2014).
17

Recent scholarship has convincingly demonstrated relationships between racial inequality
and welfare spending at the state level. Matsubayashi and Rocha (2012) find that the impact of
states’ Black population shares on welfare spending depends on whether Black-White income
inequality is high (less spending) or low (more spending). Lind (2007) and Hero and Levy
(2017) find that between- and within-race inequality have different effects on welfare effort and
generosity in the states. Lind (2007) finds that within-race inequality is positively correlated with
spending whereas between-race inequality appears to have no net effect. This makes sense
because more between-race inequality has two potentially offsetting effects: it raises inequality

17
For more on the determinants of social policy-making in the states, see Plotnick and Winters 1990;
Erikson et al. 1993; Brown 1995; Barilleaux 1997; Lieberman and Shaw 2000; Berry et al. 2003.
65
overall, which may lead to more demand for redistribution, but it also accentuates average
inequality between groups, which may have the opposite effect. Hero and Levy (2017) propose
that a critical variable is the share of total income inequality that can be explained by inter-racial
disparities (as opposed to by disparities within racial groups). Controlling for the total level of
income inequality and the level of racial diversity, they show that this share, which isolates the
racial structure of inequality from the level of inequality, is powerfully negatively correlated
with welfare spending after the 1996 national welfare reform that gave more discretion to the
states but not before it.
Just Welfare?
A critical question is to what extent the racial structure of inequality influences policy domains
other than welfare. Research has documented associations between intergroup inequality and
public goods such as education, where older Whites may reject using “our” tax dollars to finance
“their” schooling (Poterba 1997; Myers 2007) and even “developmental” spending (Peterson
1981) on goods such as transportation and sanitation (Alesina, Baqir, and Easterly 1999).
Trounstine (2016) finds that racially segregated locales are most prone to such effects,
presumably because reductions in spending on public goods can be targeted so that they
minimally affect the services available to the dominant group.
Even though sales and property taxes that finance many public goods tend to be
regressive, most public goods tend to be financed more by wealthier citizens and consumed more
by poorer citizens, who are more likely to lack private alternatives to public goods. Thus, as in
the case of welfare spending, we hypothesize that, a given level of total income inequality and
ethno-racial diversity, larger racial disparities in income will reduce support for public spending
generally, including on goods that are not usually classified as redistribution.
66
However, we note three reasons that the applicability of social affinity models of
redistribution to the provision of public goods should not be taken for granted. First, political
pressure to reduce taxes or bond issues that fund spending on local public goods may be strongly
counteracted by structural forces inherent in competitive federalism. Local governments must
compete for high tax-paying individuals and produce public goods needed to support an
employer base (Peterson 1981). Unlike cutting welfare spending and other programs targeted at
the poor, cutting funds for infrastructure, education, and policing may conflict with these
imperatives. Second, and relatedly, local public officials may find ways to circumvent public
opposition to redistributive taxation (Rugh and Trounstine 2011). Third, whereas welfare policy
is known to be ‘racialized’ (Gilens 1999), it is less clear to what extent citizens’ judgments about
bond levies or taxes collected to fund public goods such as education, police protection, parks,
and sanitation are as well. Thus, group inequalities may foster less resistance to spending on
these goods than they do on welfare. Indeed, empirical research on the links between racial
diversity and public goods spending have turned up mixed results at best (Hopkins 2011;
Boustan et al. 2013), though this research has not zeroed in on the racial structure of inequality,
instead treating income inequality as a separate potential influence on policy.
On the other hand, neither changes in public opinion nor responsiveness to them are
necessary for racial inequality to influence policy outcomes. In the same way that policy-makers’
conceptions of poverty in particular are shaped by the racial influences and stereotypes,
perceptions of economic inequality more generally may be informed by the local structure of
inequality across racial groups, as in Soss et al.’s (2011) “Racial Classification Model.” This, in
turn, may feed into evaluations of the usefulness of taking steps to shore up equality of
opportunity or equal access to a wide range of public services.
67
Which Types of Public Goods?
Economic theories on public goods provide a basis for conjecturing about which public goods
between-race inequality should be more or less likely to affect, and under what circumstances.
Not all public goods share the same characteristics nor they are financed in the same way, and
those that strike citizens as more clearly redistributive, or more likely to transfer income between
groups having different social affinity, would seem most likely to be affected by the core factor
we have highlighted here. Efforts to typologize public goods inevitably face challenges.
Peterson (1980), for example, struggled with the issue of whether education should be
considered a “developmental” or a “redistributive” good. In reality, of course, it is both.
One especially useful framework is provided by Besley and Coate (1991), who argue that
the redistributive consequences of the public provision of goods are larger for goods that admit
private substitution. They reason that if the quality of such a good provided in the public sector is
low, rich households will be willing to pay for a higher-quality good in the private sector while
lower-income households will still consume the publicly provided good. Education is probably
the most widely scrutinized example of such a good, but health care and hospitals (e.g. Peterson
1981; Schneider 1989; Jimenez 2014), parks and recreational spaces, and even policing and
security (see, e.g., Bergstrom and Goodman 1973; Borcherding and Deacon 1972) are subject to
similar dynamics. By contrast, it is more difficult to imagine private substitutes for highways,
sanitation, fire protection, and libraries.
18

18
With the rise of online services offering book purchase or lending, libraries arguably admit more private
substitution nowadays than they used to. However, most of this change occurred after our period of study.
Here, we conceptualize the services provided by libraries more broadly with a focus on their functioning
as public open space. Libraries offer spaces to study, read and various community services programs in
which the users interact with other neighbors. In this broader sense, we classify libraries as goods that are
less privately substitutable.
68
Applying this reasoning, we hypothesize that the public goods most affected by racial
structure of inequality will be those that admit private substitution more easily by their nature.
The policy areas that most clearly have these characteristics are education, health or hospital
services, parks and recreation, and police protection. Those who can afford are more likely to
educate their children in private schools, use private hospitals with higher quality care, and enjoy
more amenities in private parks and recreational spaces through their membership. The high-
income households also hire private securities to protect their properties and neighborhoods. Of
course, we acknowledge that this is only one dimension that would affect the degree to which
racial inequality would be expected to dampen spending on any given public good.
Political Mechanisms
How does the structure of inequality influence these policy outcomes? Prior research suggests
several possible political mechanisms. Whites living in economically unequal areas may tend to
elect more conservative and Republican leaders, who in turn enact fiscal retrenchment (Carsey
1995; Gelman 2009). They may also be more reluctant to vote for minority elected officials
because they expect that these leaders will pursue programs that induce more intergroup transfers
(Citrin, Green, and Sears 1990). In contrast, more conservative and Republican leaders may
respond to the majority group’s preference for lower taxation and less spending on public goods
that entail redistribution. They may also share the view that higher spending on public goods is
illegitimate or wasteful because it tends to benefit “undeserving” members of minority groups
who will not use it productively. Thus, the partisan and racial identities of local and state elected
officials may mediate the relationship between racial inequality and spending on public goods.
Institutional factors may also come into play as moderators of these effects. Some
research finds that direct democracy fosters the adoption of policies that harm minority group
69
interests and rights (Hajnal, Gerber, and Louch 2002). Political fragmentation in metropolitan
areas is asserted to produce similar dynamics. The social stratification-government inequality
(SSGI) thesis, developed by Hill (1974) and Neiman (1976), suggests that highly fragmented
metropolitan areas are characterized by entrenched income segregation with the affluent majority
concentrated in high income suburban municipalities with more ample tax bases. The slack
resources in these areas are more likely to flow to members of the same group and those in
nearby municipalities, not to minorities who reside disproportionately in central-cities. Lowery
(1999), for example, calls the fragmented arrangements “institutionally racist in their profound,
albeit unintended, consequences.” Intensified interjurisdictional competition in highly
fragmented metropolitan areas may therefore moderate any dampening effects on public goods
investment (Peterson 1981; Minkoff 2009, 2012; Jimenez 2014). Consequently, we expect that
metropolitan fragmentation may moderate the effects of racial income inequality on local public
goods spending. To the extent that available data permit, we test each of these possibilities in
turn.

Independent Effects of Racial Diversity and Income Inequality
Although our focus is on the impact of the racial structure of inequality, it is worth briefly
addressing what relationships we would expect to emerge for racial diversity per se and total
income inequality once the relative magnitude racial component of inequality is statistically
controlled. Here we must outline competing expectations rather than clear hypotheses.
Prominent research by Alesina, Baqir, and Easterly (1999) argues that racial diversity could
weaken support for public goods spending due to the heterogeneous preferences that members of
different groups have over the form that public goods should take (what curriculum a school
70
should offer, say, or what neighborhoods a road should run through). If this mechanism is
correct, we might expect that more diverse locales would produce lower amounts of public goods
even when the racial structure of inequality is held constant. However, intuitively, racially
diverse locales in the U.S. also contain larger shares of non-White voters who may have more
economically liberal views and pressure on government for more spending on public goods.
This would lead to a positive relationship between diversity per se and local public goods
spending.
Turning to inequality, a standard expectation from the public choice literature is that
more inequality leads to greater public support for redistributive taxation (e.g. Meltzer and
Richard 1981), which may in turn lead to more ample provision of public goods. However,
research going back to Dye (1969) finds that inequality is sometimes negatively related to
redistribution. Hero and Levy (2017) find that total inequality, controlling for the racial structure
of inequality, is not consistently related to welfare spending in the states. They speculate that
inequality has many overlapping group dimensions, such as class, gender, urban versus rural,
suburb versus city, and immigration or citizenship-status that may dampen any positive effect of
total inequality on social spending even when racial inequality is taken into account.

Data and Variables
Our analytic approach examines the relationship between the racial structure of local income
inequality and government spending on public goods and services.
19
We begin by constructing
four panel data sets at different levels of political geographic unit—cities, metropolitan areas,

19
In keeping with prior research (e.g. Alesina, Baqir, and Easterly 1999), our focus is on urban public
goods. The dynamics of rural public goods provision may follow different dynamics.
71
states, and school districts— in the U.S. over three Census years (1980, 1990, and 2000).
20
There
are two reasons for choosing these different governmental jurisdictions. First, under the fiscal
federal system of the United States, there are various types of local governments which share
financial responsibilities for the provision of local public goods. The divisions of fiscal
responsibilities are not comparable across cities (Peterson 1981); this complicates the
interpretation of findings in studies of local public goods with aggregated finance data, because
we cannot tell at which levels of jurisdiction the relationship pertaining to a specific public good
is observed. For this reason, we start our analysis with a fiscally standardized set of cities for
local finance.
Second, our analyses examine whether the local provision of public goods is determined
by the level of between-race income inequality within a jurisdiction. However, Tiebout (1956)
sorting of residents into communities is a potential threat to our analysis at the city level. If
wealthy White residents select out of jurisdictions that are highly diverse or racially unequal and
into homogeneous locales, bias is introduced and our measures of between-race inequality and
racial diversity would not be exogenous to the model. However, since sorting is arguably more
likely to occur within the jurisdictions that comprise a metropolitan area than between the metro
areas, using a metro sample should mitigate such potential Tiebout bias at least to some degree.
We further extend the testing of our hypothesis with a state sample, in which the Tiebout sorting
is least likely to occur, and we also analyze school districts for public spending on education. If
the results from testing the hypothesis across the several different jurisdictional levels are
essentially similar, then our argument is better supported.
21
However, we would expect that the

20
The necessary data from the 2012 Census of Governments was not available at the time the analysis
was done.
21
The same approach, for instance, is taken by Alesina, Baqir, and Easterly (1999).
72
results would be less precisely estimated in the larger units since there is no direct
correspondence between the aggregate diversity in these units and the jurisdictional levels across
which spending on most of the public goods we analyze are being aggregated.
Our city sample comes from the Fiscally Standardized Cities (FiSCs), a publicly
available database at the Lincoln Institute of Land Policy. FiSCs database specifically allows for
comparisons of local government finances across the nation’s 112 largest urban cities by
accounting for differences in the structure of local governments.
22
For our metro sample, we use
a consistent Metropolitan Statistical Area (MSA) definition given by Office of Management and
Budget as of June 2003. There are 362 MSAs by this definition which are made up of 1086
urban counties. Our city and MSA sample indicates that the analysis will essentially be an
examination about urban public goods provision. For school districts, we construct a sample with
independent jurisdictions with populations of at least 2,500 across the nation, whether they are
elementary, secondary, or unified. The data sources for all our independent variables are listed in
Table A3-2 in the Appendix A.
The dependent variables are ten categories of local spending on public goods: hospitals,
health, police protection, parks and recreation, housing and community development, highways,
sanitation (sewerage and trash pick-up), fire services, libraries, and education. The finance data
are from the direct expenditure category in the Census of Governments (1982, 1992, and 2002).
As mentioned, however, FiSCs provides its own finance data at the city level. All our spending
data are measured in constant dollars (2011$), per household. For education, it is per pupil

22
The construction of FiSCs involves adding expenditures for the city government plus an appropriate
share from overlying counties, school districts, and special districts. Due to this adjustment, the numbers
of city populations are different from those of the U.S. Census. Our city results reflect this adjustment for
the number of households but non-adjustment does not change the results. A reader curious about the
methodology used for FiSCs may see Langley (2013) for further details.
73
expenditures also in constant dollars (2011$). Since our interest is the local provision of public
goods, for our metro and school district samples, we deduct the intergovernmental revenues of
the federal and state governments from the local direct expenditures, following Peterson’s (1981)
suggestion.
23
It should be noted, however, that there is essentially no intergovernmental revenue
from higher levels of government for the local provision of police protection, fire services, parks
and recreation, and libraries in the Census of Governments datasets.
Between-Race Group Share of Total Income Inequality
While income inequality measures such as the Gini Coefficient, inter-quantile income ratios, and
the ratio of mean to median income have been commonly used in applied social science research,
one of the limitations of these measures is that they are not decomposable between and within
groups. An additively decomposable inequality measure is defined as a measure such that the
total inequality of a population can be expressed as a weighted average of the inequality within
subgroups of the population and the inequality between them. Bourguignon (1979) shows that
Theil’s following T index (1967) meets these criteria.
𝑇 𝑏 = ⁡ ∑ 𝐼 𝑔 ∗ ln ( 𝐼 𝑔 𝑃 𝑔 ⁄ ) ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡ ⁡
𝑔 𝑖 = 1

where T
b
is the inequality between the g groups, I
g
is group g’s income share, and P
g
is group g’s
population share.

23
Peterson (1981) points out that controlling for the transferred intergovernmental revenue from higher
level government as a control variable may result in spurious correlations between dependent variable and
independent variables due to an already high correlation between the aggregated expenditures and
intergovernmental revenues. He instead proposes deducting the intergovernmental revenues from local
total spending in order to explore variations in local spending. We also controlled for intergovernmental
revenue from higher level government for each public good. Our primary results are not sensitive to this.
74
To understand the Theil Index, consider the ln ( 𝐼 𝑔 𝑃 𝑔 ⁄ ) ⁡ term. It expresses how large each
group’s income share is relative to its population share. If all groups’ income shares and
population shares are exactly equal, there is no between-group inequality in that population, for
the groups indexed by g. This makes the fractions all equal to one, and their natural log is zero,
leading to a between-group T of zero. However, if a group possesses more income share than
their population share, it positively contributes to income inequality. If a group has less income
share than their population share, it negatively contributes to income inequality. But since each
of these contributions is weighted by the group’s income share, the positive deviations from
parity outweigh the negative deviations in the sum, and the greater the deviations, the greater the
between-group T becomes.
Note that this between-group inequality measure accounts for the total inequality between
groups indexed by g. It is not simply the mean difference in incomes between groups in society,
a variable that could be quite large even if a society is highly heterogeneous but has a small
minority population that is extremely disadvantaged relative to the majority. Instead, this
measure of between-group inequality accounts for both this “depth” of inequality and its breadth
– how much of the society’s total income does it pertain to. Modest disparities between groups
may nevertheless contribute substantially to explaining total inequality if more than one group
holds a significant share of the total income. But even very large disparities between groups may
contribute little to a society’s total income inequality if almost all of the income resides in one
group, which can happen in sharply unequal but still quite racially homogeneous settings.
We first calculate the total income inequality by applying this formula to Theil’s T index
with the “groups” simply being household income brackets according to the U.S. decennial
Census and call this total income inequality. Note that this process underestimates the total
75
amount of income inequality because it captures inequality only between groups defined by
income brackets, not between all individuals or households in the society. We also calculate
Theil’s T index with the groups being race/ethnic groups and label this between-race inequality.
For the race and ethnic groups, we use “Non-Hispanic White”, “Black”, “Asian”, “Native
American”, and “Hispanic.” Our main explanatory variable racial inequality is the share of
between-race inequality of total income inequality, measuring the extent to which total income
inequality is accounted for by between-race disparities. Table A3-1 in the Appendix A presents
the summary statistics for key variables of our four samples.
Table 3-1 lists the 10 most and least racially unequal and heterogeneous cities in 2000
from our city database. Of particular note is that the cities in which racial inequality explains the
highest portion of total inequality do not overlap greatly with the cities with the highest level of
diversity per se. Predictably, there is greater overlap between the least racially unequal and the
least diverse settings. Intuitively, this is because places with very little racial diversity must also
be ones in which between-group inequality contributes very little to income disparities.

[Table 3-1]

An illustrative comparison can help underscore the nature of our theoretical expectations
about racial inequality and urban public goods provision. Consider, for example, that in 2000,
both New Orleans and Cleveland were highly racially diverse cities, similar in population size,
unemployment rates, poverty rates, and average household incomes. And New Orleans had a
higher level of income inequality overall, which according to standard public choice models
76
would promote a higher level of income redistribution. However, between-race income
disparities explained a considerably larger share of income inequality in New Orleans than in
Cleveland. In New Orleans, more than 13% of the total income inequality was due to the
disparity between the city’s racial groups, a percentage that was two times higher than the
national average. Despite its racial diversity, Cleveland’s between-race income gaps explained
merely 3% of total inequality. In keeping with this paper’s theoretical framework, New Orleans
spent much less on public goods ($3000 per household in 2011 constant dollars) than the
national average ($4600 per household in 2011 constant dollar), whereas Cleveland spent much
more ($6800 per household in 2011 constant dollar). Of course, there are myriad other
potentially influential differences between these two cities, underlining the need for the
systematic analysis we turn to next.

Empirical Model
The relationship between racial inequality and local public finances is estimated using the
following panel model specification:
∆ 𝑦 𝑖𝑡 + 1
= ∆ 𝛼 + 𝛽 ( ∆ 𝑅 𝑎 𝑐𝑖 𝑎 𝑙 ⁡𝐼𝑛 𝑒 𝑞𝑢𝑎𝑙 𝑖 𝑡 𝑦 )
𝑖𝑡
+ 𝛾 ∆ 𝑇 𝑖𝑡
+ 𝛿 ∆ 𝐷 𝑖𝑡
+ 𝜏 ∆ 𝑋 𝑖𝑡
+ ∅ ∆ 𝑍 𝑖𝑡
+ 𝜋 ∆ 𝑃 𝑖𝑡
+ ∆ 𝑢 𝑖𝑡

where i indexes a city, or MSA, or state, or school district in Census year t; y is a local provision
of public goods such as direct expenditures per household (constant 2011$). Note that all our
right hand side variables are lagged by one year. For the metro and state sample, all local
government spending in metro and state boundary are aggregated.
77
Our main independent variable Racial Inequality is the share of between-race income
disparities of total inequality, T is a vector of total inequality, D is a vector of racial diversity,
24

and X contains a set of demographic characteristics of the sample units, known to affect local
public goods spending from the literature including log of total population, average household
income, the share of the population over 25 years of age with a college degree, the share of the
population 65 years old or more, the unemployment rate, the poverty rate under 100% of the
federal poverty line. Z represents local public goods-specific conditions such as crime rate for
the police spending model and hospital bed rate, physician rate, and the number of hospitals for
the health and hospital spending models. P is a set of political and institutional variables
measuring if a mayor or governor is racial minority and their political party.
25
For the state-level
analysis, we also use citizen ideology score, which is borrowed from Berry et al. (1998). The
degree of political fragmentation–the number of municipalities in a metropolitan area–is
controlled in the metropolitan area-level analysis.
The coefficient β estimates the effect of racial inequality on the local provision of public
goods while the coefficient 𝛾 estimates the effect of income inequality per se; and the coefficient
𝛿 estimates the effect of racial diversity per se. Note that our estimate β, with the estimates of 𝛾
and 𝛿 , disentangles the effect of the racial structure of income inequality from the effects of

24
The racial Diversity index is defined by 1 − ∑ 𝑅 𝑎𝑐 𝑒 𝑖 2
𝑖 , where 𝑅 𝑎𝑐 𝑒 𝑖 denotes the share of household
population as race i of householder and i = {Non-Hispanic White, Black, Asian and Pacific Islander,
Native-American, and Hispanic}.
25
Ideally, we also want to control for racial composition of the city councils and the state legislatures. We
obtained city council’s racial composition data between 1986 and 2010 from Jessica Trounstine and 1981
survey data from International City/County Management Association, but we could only match 82, 74,
and 55 cities to our city sample for 1980, 1990, and 2000. We obtained the state legislatures’ racial
composition data from Preuhs (2005, 2006) and Juenke and Preuhs (2012), but none of the state had
either upper or lower chamber with the racial minorities being a majority during the period of our study.
78
income inequality and racial diversity that might otherwise absorb β, had we not included a
vector of racial inequality.
26

We take the advantage of the longitudinal data structure in two ways. First, even though
the evidence presented here cannot cleanly identify a causal relationship, using conservative
fixed-effects specifications that limit the analysis to variation within units over time rather than
random-effects eliminates unobserved time-invariant confounders. This partly addresses the
concerns on the possibility of omitted variable bias. We also include Census year fixed effects to
capture the association between unobserved variables and local spending on public goods to
account for unexplained time trends.

Results
The Effects on Total Public Goods Spending
Table 3-2 presents the results with demographic control variables from the three different
samples of jurisdictional responsibility. The dependent variable is sum of local and state
spending on the nine public goods we examine. What stands out is that the effects of racial
inequality on total public goods spending are consistently negative and statistically significant,
whereas the effects of racial diversity and total income inequality are not. Consistent with the
social affinity models described above, the extent to which inequality is a byproduct of
disparities between races is a key determinant of aggregate spending on public goods. The total

26
In our model, we do not include an interaction term between racial inequality and racial diversity,
because our racial inequality measure already accounts for group sizes by taking a weighted average
based on each group’s income share. However, we separately estimated the interactions between the
racial diversity and racial inequality measure for each model and found no significant coefficients on the
interaction terms.
79
amount of income inequality or the level of ethnic diversity per se does not exert a consistent
effect.
[Table 3-2]

The city and metro results in Table 3-2 show the negative signs for both racial inequality
and diversity, but only racial inequality is negative and significant (p<0.05). The same pattern
obtains in the state results, although racial inequality falls short of statistical significance (p =
0.11). But even this weak relationship is arguably notable given that states have little direct
financial responsibilities for local public goods.
Comparing the results at three different jurisdictional levels, the effect of racial inequality
is amplified as the political unit increases in size, consistent with the possibility that estimate at
lower levels of aggregation are biased toward zero due to residential selection effects. Because
the state-level results should be least susceptible to Tiebout sorting, the effect of racial inequality
should be largest. But they are also playing out over a larger jurisdiction in which some
municipalities and school districts may be relatively less affected by racial inequality in the state
as a whole. This explains why these units show the largest effects but the least precisely
estimated relationships (i.e. large standard errors).
While the Herfindahl-Hirschman index is a measure that is commonly used for racial
diversity in the literature, its potential drawback is that the index mechanically treats
homogenous Black, Hispanic, and White cities alike. Because our study taps a variation of
income differentials across groups, differentiating these cities needs to be considered. In Table
A3-3 in the Appendix A, we report the results with percentage of Black and Hispanic household
instead than Herfindahl index of racial diversity (Panel A). This alternative model specification
80
does not change the main findings reported in Table 3-2. Rather, it seems to reinforce the
significance of racial inequality effects on total public goods spending.
To allay potential concerns about outlier effects and high leverage points, in Panel B in
Table A3-3 in the Appendix A, we also report the results with our dependent variables logged.
The results show that racial inequality still manifests a negative effect in the city sample
(p<0.05), whereas in the metro sample, the standard error gets larger but the magnitude and sign
of the estimates remains consistent with the theory that a higher between-race share of total
inequality dampens spending on local public goods.
These results strongly corroborate our expectation that such effects are not confined to
welfare redistribution. Clearly, there is a link between local-level inequality and the provision of
public goods that are not as well known to be racialized and whose stated purpose is not to
redistribute income. Notably, most of the categories of public goods we analyze are not classified
as “redistributive,” instead falling under headings such as developmental or allocative (Peterson
1981).
By contrast, the independent effects of racial diversity and income inequality are quite
varied and never statistically significant. To speculate, it may be that the competing expectations
we outlined for the independent effects of racial diversity and income inequality do off-set to
some extent. Racial diversity may make it more challenging to reach agreement on how public
funds should be spent and the form that goods should take, weakening support for the necessary
compromises. But it may also tend to increase support for spending on public goods because
more diverse areas tend to have more non-White voters, who also tend to have more left-leaning
attitudes toward government spending and taxation. Inequality may produce more demand for
taxation but if there are between-group dimensions of inequality that our analysis has not
81
considered (e.g. class or immigration status) and are not fully captured by the racial inequality
component, these inter-group disparities might push in the opposite direction. However, why the
effects would off-set differently at different levels of aggregation remains puzzling, and with the
caveat that causal effects cannot be perfectly identified in such models, a second possibility is
that these variables tend not to matter, on balance, in and of themselves except inasmuch as they
overlap. We will have more to say about these effects below, where we disaggregate effects on
total public goods spending by category.
Which Public Goods?
Which public goods in particular are being reduced? Our theory led us to expect negative
impacts to be largest in several categories: education, health or hospitals, parks and recreation,
and police protection. Recall that these goods allowed the greatest substitutability to privately
provided services and hence seemed more prone to being cut.
We start by describing the education spending results at the school district level. Table 3-
3 shows negative effect of racial inequality on school district spending and the results are robust
across seven different model specifications.

[Table 3-3]

In column 1, we analyze a fully balanced school district sample (n=21,267), in column 2
we display the results excluding outliers, and in column 3, we further eliminate the districts with
negative expenditures under our definition of local provision. In column 4, the model accounts
for school finance reform that occurred during the period of our study. School districts had
82
substantial changes in their finance sources – a shift from local funding toward greater state
funding, fueled in part by court-ordered school finance reform (Murray, Evans, and Schwab
1998; Hoxby 2001; Card and Payne 2002; Corcoran and Evans 2008). Controlling for this factor
is meant to account for the changes in the structure of school finance. The reform was in place to
equalize local school finance by redistributing the resources from the wealthier districts to the
poorer districts. We use average household income as a reasonable proxy for the wealth of the
district (see, e.g., Card and Payne 2002) and through the interaction term, we allow the effect of
district-level average household income to vary by a state’s school finance regime.
Because intergovernmental revenue from states constitutes to be a main funding source in
school district spending, we do not deduct intergovernmental revenue from the total district
spending in column 5, but instead control for intergovernmental transfer as an independent
variable. Either approach yields similar results, however. In column 6 and 7, we further sort out
the districts with high discretion over the use of their funds from those with low discretion. Our
expectation is that we would see stronger negative effect of racial inequality for those with high
discretion over their funds, whereas the effect would be weaker for the districts with low
discretion. The results in column 6 and 7 confirm this prediction as well. Next, turning to city,
metro, and state-level analyses, we examine if these predictions are corroborated for the other
public goods we highlighted, after controlling for the goods-specific condition variables. Table 3-
4 shows these results. The broad patterns we observe are consistent with our theoretical
expectations, though predictably also imperfectly so.

[Table 3-4]

83
Our city-level results in Panel A show that racial inequality triggers the municipal budget
cuts for the investment on hospitals, police protection, and parks and recreation. On the other
hand, there is no substantial effect of racial inequality on city spending on highways, sanitation,
fire, libraries, and housing and community development. The metro-level results in Panel B
illustrate that the budgets of health and police protection are similarly cut as racial inequality
grows. However, the budgets of parks, highways, sanitation, fire, libraries, and housing and
community development are not significantly affected by racial inequality in the areas. In state-
level analysis (Panel C), the budgets are also cut for police protection, park and recreation
services. Housing and community development is also one of the budget cut policy areas that
racial inequality influences. The sign of racial inequality on housing and community
development spending is negative in all three samples (Panel A through C), but only significant
at the state level, possibly because local revenues only constitute 25% of total state and local
spending in this policy area.
The consistency of the results across the public goods largely corroborates our theoretical
expectation that the effects of racial inequality would be strongest for privately substitutable
goods. In Table A3-4 in the Appendix A, we summarize an array of evidence examined at
different levels of jurisdictions and across a variety of public goods. It shows that racial
inequality is a strong predictor of variation in government spending on local public goods, with
clear negative impacts.
27

27
Of the 28 coefficients in our findings (see Table A3-5 in the Appendix A), racial inequality shows
negative signs for 21 and 9 of these negative relationships are statistically significant. In contrast, of the
28 coefficients, racial diversity shows negative signs for 18 and only one of these negative relationships
are statistically significant.
84
Turning again to the independent effects of racial diversity and income inequality, the
picture is quite a bit murkier. Racial diversity is statistically significant no more often than we
would expect by sheer chance in estimating many separate regression models, and each of the
two significant coefficients is differently signed. This in fact parallels mixed findings in the
research literature (Lee et al. 2015; cf. Alesina, Baqir, and Easterly 1999).
28

The positive relationship between diversity and education spending at the school district
level is consistent with our speculation that in more diverse settings, controlling for racial
inequality, more spending would be achieved as a function of the more fiscally liberal
preferences of non-Whites. However, this conjecture does not find clear support for any other
public good.
To the extent that we obtain significant results for total inequality, they are negative,
which goes against the standard expectation in public choice models that more income inequality
will lead to more support for redistributive taxation and hence, presumably, to more ability to
spend on public goods. This may reflect the influence of other between-group dimensions of
inequality. Or perhaps the median income voter models from the public choice literature simply
do not pertain well to the U.S. case, either because policy-making is frequently insulated from
mass opinion or because voters’ appraisal of trends in inequality does not match economic
reality.

28
Lee et al. (2015) theorize that the effect of racial diversity will be positive if the demand for a public
good is price inelastic and negative if the demand is price elastic. They further argue that racial diversity
leads to substitution of local public expenditures on some public goods for expenditures on other public
goods. In a panel analysis, Lee et al. (2015) also find a negative relationship between racial diversity and
city spending on sanitation and a positive association between racial diversity and school district
expenditure on education. They point to the public finance literature in support of their claim and findings
that the empirical estimate of the price elasticity of demand for education is lower than for that of
sanitation.
85
Political Mechanisms
To examine if the political mechanisms we highlighted mediate the relationship between racial
inequality and local public goods spending, in Table 3-5, we add a set of political and
institutional variables with a focus on racial identities and party affiliation of those who control
the local and state political institutions and the citizen ideology. Because the number of cities in
the sample becomes smaller with these political variables, we first report the baseline results
with the same smaller sample size. Then, we examine if adding the political variables mediates
the relationship reported in the baseline results. For the metro-level analysis, we investigate if
metropolitan fragmentation moderates the effect of racial inequality on local public goods
spending. To save space, in Table 3-5, we focus on the results on racial inequality with these
political variables.

[Tale 3-5]

The results overall suggest that the political and institutional variables do not mediate or
moderate the relationship between racial inequality and local public goods spending. Whether
the cities and states elect White mayors and governors or Republican leaders does not mediate
the effect of racial inequality on local public goods provision (Panel A and C). The null
mediation effects from Berry et al.’s (1998) citizen ideology score further suggest that the
citizens’ perceptions of economic inequality between groups and their unwillingness to pay for
other groups’ greater benefit may be so profound that the set of political, institutional, and public
opinion variables we consider here do not mediate the relationship at all (Panel C). Or, as Soss et
86
al.’s (2011) “Racial Classification Model” suggests, policy-makers’ conceptions of economic
inequality may be informed by the local structure of inequality across racial groups, regardless of
their racial identities or party affiliation. Contrary to expectations, the metropolitan fragmented
environment does not moderate the relationship at the metro level, either (Panel B).

Conclusion
The results of extensive analyses across four different jurisdictional levels and many areas of
local government responsibilities broadly support the core theoretical claim: the more that
income inequality is attributable to inequities between racial and ethnic groups, rather than
between individuals within groups, the less investment localities make in public goods. By
contrast, we find little evidence that the overall level of income inequality is consistently linked
to local investment in public goods, and there is no consistent negative or positive relationship
between racial diversity per se and public goods spending.
These results are generally consistent with Lind (2007) and Hero and Levy (2017), who
find a negative association between the extent of between-race inequality and welfare
redistribution at the state level. That is, the racial ‘structure of inequality,’ rather than aggregate
inequality or social heterogeneity in and of themselves, is a critical factor. Thus, it seems that it’s
not just welfare policy which is affected by between-race inequality. Its impacts are also evident
regarding a range of ostensibly non-redistributive policies at the local level as well (cf. Hersh and
Nall 2016).
Much clearly remains to be illuminated in future research. For one, it would be useful to
link between-race inequality to public opinion or vote choice or to local elected officials’
87
interpretation of public opinion, as this would provide more direct evidence of the mechanism
most researchers have proposed (Hajnal and Rivera 2014, Hersh and Nall 2016). However, this
bottom-up mechanism is not mutually exclusive of the alternative explanation we noted, namely
the application of racial biases by policy makers in developing understandings of the nature of
local inequality and utility of investing in expansive public services. For another, it would be
useful to ascertain if these results might help explain some of the inconsistency in findings
concerning the relationship of diversity with public goods spending. While Alesina et al. (1999)
report fairly robust results, others (e.g. Hopkins 2011, Boustan et al. 2013) have not corroborated
their findings. Hopkins (2011) also shows that the cross-sectional relationship Alesina et al.
present appears to diminish over time. If the growth in local racial diversity has outpaced the
growth in between-group inequality, and between-group inequality rather than diversity is the
proximate causal agent, this is the pattern we would expect.
In the contemporary context, the political significance of race may well be most clearly
manifest through between-race economic inequality and its impacts on explicitly redistributive
policies. However, our analyses of public goods provision suggest that the impact of racial
inequality on public policies may go well beyond welfare and thus be more pervasive and more
profound than previously recognized. In any case, the findings of between-race economic
inequality and public policies identified here and elsewhere underscore the importance of further
and careful inquiry to better understand how, how much, and why those relationships exist and
persist within and across levels of the American political system and to the importance of
considering inequality and diversity jointly as influences on political outcomes rather than as
separate variables.

88
References
Abrajano, Marisa, and Zoltan L. Hajnal. 2015. White Backlash: Immigration, Race, and
American Politics. Princeton University Press.
Alesina, Alberto, and Edward Glaeser. 2004. Fighting Poverty in the US and Europe: A World of
Difference. New York: Oxford University Press.
Alesina, Alberto, and Eliana La Ferrara. 2002. “Who Trusts Others?” Journal of Public
Economics 85 (2): 207-234.
Alesina, Alberto, Reza Baqir, and William Easterly. 1999. “Public Goods and Ethnic Divisions.”
Quarterly Journal of Economics 114 (4): 1243-1284.
Baldwin, Kate, and John D. Huber. 2010. “Economic Versus Cultural Differences: Forms of
Ethnic Diversity and Public Goods Provision.” American Political Science Review 104 (04):
644-662
Barrilleaux, Charles. 1997. “A test of the independent influences of electoral competition and
party strength in a model of state policy-making.” American Journal of Political Science Oct 1
1462-1466.
Benabou, Roland. 2000. “Unequal societies: Income distribution and the social contract.”
American Economic Review 90 (1): 96-129.
Bergstrom, Theodore. C., and Robert P. Goodman. 1973. “Private demands for public
goods.” American Economic Review 63 (3): 280-296.
Berry, William D., Evan J. Ringquist, Richard C. Fording and Russell L. Hanson. 1998.
“Measuring Citizen and Government Ideology in the American States, 1960-93.” American
Journal of Political Science 42 (1): 327-48.
Berry, William D., Richard C. Fording, and Russell L. Hanson. 2003. “Reassessing the “race to
the bottom” in state welfare policy.” Journal of Politics 65 (2): 327-349.
Blalock, Hubert M. 1967. Toward A Theory of Minority-Group Relations. NY: John Wiley and
Sons.
Bobo, Lawrence, and Vincent L. Hutchings. 1996. “Perceptions of racial group competition:
Extending Blumer's theory of group position to a multiracial social context.” American
Sociological Review 61 (6): 951-972.
Borcherding, Thomas E., and Robert T. Deacon. 1972. “The demand for the services of non-
federal governments.” American Economic Review 62 (5), 891-901.
Bourguignon, Francois. 1979. “Decomposable income inequality measures.” Econometrica 47
(4): 901-920.
Boustan, Leah Platt, Fernando Ferreira, Herman Winkler, and Eric Zolt. 2013. “The effect of
rising income inequality on taxation and public expenditures: Evidence from US municipalities
and school districts, 1970–2000.” Review of Economics and Statistics 95 (4): 1291-1302.
89
Brown, Robert D. 1995. “Party cleavages and welfare effort in the American states.” American
Political Science Review 89 (1): 23-33.
Card, David, and A. Abigail Payne. 2002. “School finance reform, the distribution of school
spending, and the distribution of student test scores.” Journal of Public Economics 83 (1): 49-82.
Carsey, Thomas M. 1995. “The contextual effects of race on White voter behavior: The 1989
New York City mayoral election.” Journal of Politics 57 (1): 221-228.
Citrin, Jack, Donald Philip Green, and David O. Sears. 1990. “White reactions to black
candidates: When does race matter?” Public Opinion Quarterly 54 (1): 74-96.
Costa, Dora L. and Matthew E. Kahn. 2003. “Civic engagement and community heterogeneity:
An economist's perspective.” Perspectives on Politics 1 (1): 103-111.
Corcoran, Sean, and William N. Evans. 2010. “Income inequality, the median voter, and the
support for public education.” No. w16097. National Bureau of Economic Research.
Dye, Thomas R. 1969. “Income inequality and American state politics.” American Political
Science Review 63 (1): 157-162.
Erickson, Robert S., Gerald C. Wright, and John P. McIver. 1993. Statehouse Democracy:
Public Opinion and Policy in the American States. Cambridge: Cambridge University Press.
Fellowes, Matthew C., and Rowe, Gretchen. 2004. “Politics and the New American Welfare
States.” American Journal of Political Science 48 (2): 362-373.
Fording, Richard C. “Laboratories of democracy or symbolic politics.” 2003. Race and the
politics of welfare reform 298-319.
Fox, Cybelle. 2004. “The Changing Color of Welfare? How Whites’ Attitudes Toward Latinos
Influence Support for Welfare.” American Journal of Sociology 110 (3): 580-625
Gelman, Andrew. 2009. Red state, blue state, rich state, poor state: Why Americans vote the way
they do. Princeton University Press.
Gilens, Martin. 1999. Why Americans Hate Welfare: Race, Media, and the Politics of
Antipoverty Policy. Chicago: University of Chicago Press.
Habyarimana, James, Macartan Humphreys, Daniel N. Posner, and Jeremy M. Weinstein. 2007.
“Why Does Diversity Undermine Public Goods Provision.” American Political Science Review
101 (4): 709-725.
Hajnal, Zoltan L., Elisabeth R. Gerber, and Hugh Louch. 2002. “Minorities and direct
legislation: Evidence from California ballot proposition elections.” Journal of Politics 64 (1):
154-177.
Hajnal, Zoltan, and Michael U. Rivera. 2014. “Immigration, Latinos, and white partisan politics:
The new democratic defection.” American Journal of Political Science 58 (4):773-789.
Hero, Rodney E. 1998. Faces of Inequality: Social Diversity in American Politics. New York:
Oxford University Press
90
Hero, Rodney E., and Morris E. Levy. 2016. “The Racial Structure of Economic Inequality in the
United States: Understanding Change and Continuity in an Era of “Great Divergence”.” Social
Science Quarterly 97 (3): 491-505.
Hero, Rodney E., and Morris E. Levy. 2017. “The Racial Structure of Inequality:
Consequences for Welfare Policy in the U.S.” Social Science Quarterly. doi:10.1111/ssqu.12427
Hero, Rodney E., and Caroline J. Tolbert. 1996. “A Racial/Ethnic Diversity Interpretation of
Politics and Policy in the States of the U.S.” American Journal of Political Science (40) 4: 851-
871.
Hero, Rodney E., and Robert R. Preuhs. 2007. “Immigration and the evolving American welfare
state: Examining policies in the US states.” American Journal of Political Science 51 (3): 498-
517.
Hersh, Eitan D., and Clayton Nall. 2016. “The Primacy of Race in the Geography of Income-
based Voting: New Evidence from Public Voting Records.” American Journal of Political
Science 60 (2): 289-303.
Hill, Richard Child. 1974. “Separate and unequal: governmental inequality in the
metropolis.” American Political Science Review 68 (4): 1557-1568.
Hopkins, Daniel J. 2011. “The Limited Local Impacts of Ethnic and Racial Diversity.”
American Politics Research 39 (2): 344-379.
Hoxby, Caroline M. 2001. “All School finance Equalizations Are Not Created.” Quarterly
Journal of Economics 116 (4): 1189-1231.
Jimenez, Benedict S. 2014. “Separate, unequal, and ignored? Interjurisdictional competition and
the budgetary choices of poor and affluent municipalities.” Public Administration Review 74 (2):
246-257
Johnson, Martin. 2003. “Racial context, public attitudes, and welfare effort in the American
states.” Race and the politics of welfare reform 151-170.
Juenke, Eric Gonzalez, and Robert R. Preuhs. 2012. “Irreplaceable legislators? Rethinking
minority representatives in the new century.” American Journal of Political Science 56 (3): 705-
715.
Key, V.O. 1949. Southern Politics in State and Nation. Knoxville: University of Tennessee
Press.
Langley, Adam H. 2013. “Methodology Used to Create Fiscally Standardized Cities Database.”
Lincoln Institute of Land Policy Working Paper. Cambridge, MA: Lincoln Institute of Land
Policy.
Lee, Alexander. 2017. “Ethnic Diversity and Ethnic Discrimination: Explaining Local Public
Goods Provision.” Paper presented at the Midwest Political Science Association Conference.
Lee, Soomi, Dongwon Lee, and Thomas E. Borcherding. 2015. “Ethnic Diversity and Public
Goods Provision Evidence from US Municipalities and School Districts.” Urban Affairs Review
52 (5): 685-713.
91
Lieberman, Robert C., and Greg M. Shaw. 2000. “Looking inward, looking outward: The politics
of state welfare innovation under devolution.” Political Research Quarterly 53 (2): 215-240.
Lind, Jo Thori. 2007. “Fractionalization and the Size of Government.” Journal of Public
Economics 91 (1): 51-76.
Lowery, David. 1999. “Sorting in the fragmented metropolis: Updating the social stratification-
government inequality debate.” Public Management Review 1 (1): 7-26.
Lupu, Noam, and Jonas Pontusson. 2011. “The Structure of Inequality and the Politics of
Redistribution.” American Political Science Review 105 (2): 316-336.
Luttmer, Erzo F.P. 2001. “Group Loyalty and the Taste for Redistribution.” Journal of Political
Economy 109 (3): 500-528.
Matsubayashi, Tetsuya, and Rene R. Rocha. 2012. “Racial diversity and public policy in the
states.” Political Research Quarterly 65 (3): 600-614.
McCarty, Nolan, Keith T. Poole, and Rosenthal, Howard. 2006. Polarized America: The Dance
of Ideology and Unequal Riches. Cambridge, MA: The MIT Press.
McGuire, Therese J. and Merriman, David M. 2006. “Has Welfare Reform Changed State
Expenditure Patterns?” National Poverty Center. Policy Brief #7.
Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.”
Journal of Political Economy 89 (5): 914-927.
Miguel, Edward and Mary Kay Gugerty. 2005. “Ethnic Diversity, Social Sanctions, and Public
Goods in Kenya.” Journal of Public Economics 89 (11): 2325-2368.
Milanovic, Branko. 2000. “The Median Voter Hypothesis, Income Inequality and Income
Redistribution: An Empirical Test with the Required Data.” European Journal of Political
Economy 16 (3): 367-410.
Minkoff, Scott L. 2009. “Minding your neighborhood: The spatial context of local
redistribution.” Social Science Quarterly 90 (3): 516-537.
Minkoff, Scott L. 2012. “The proximate polity: Spatial context and political risk in local
developmental goods provision." Urban Affairs Review 48 (3): 354-388.
Murray, Sheila E., William N. Evans, and Robert M. Schwab. 1998. “Education-finance reform
and the distribution of education resources.” American Economic Review 88 (4): 789-812.
Myers, Dowell. 2007. Immigrants and boomers: Forging a new social contract for the future of
America. Russell Sage Foundation.
Neiman, Max. 1976. “Social stratification and governmental inequality.” American Political
Science Review 70 (1): 149-154.
Peterson, Paul E. 1981. City Limits. Chicago: University of Chicago Press.
Plotnick, Robert D., and Richard F. Winters. 1990. “Party, political liberalism, and
redistribution: An application to the American states.” American Politics Quarterly 18 (4): 430-
458.
92
Poterba, James. 1997. “Demographic Structure and the Political Economy of Public Education.”
Journal of Policy Analysis and Management 16 (1): 48-66.
Preuhs, Robert R. 2005. “Descriptive representation, legislative leadership, and direct
democracy: Latino influence on English only laws in the states, 1984–2002.” State Politics &
Policy Quarterly 5 (3): 203-224.
Preuhs, Robert R. 2007. “The conditional effects of minority descriptive representation: Black
legislators and policy influence in the American states.” Journal of Politics 68 (3): 585-599.
Putnam, Robert D. 2007. “E Pluribus Unum: Diversity and Community in the Twenty First
Century.” Scandinavian Political Studies 30 (2): 137-174.
Romer, Thomas. 1975. “Individual Welfare, Majority Voting, and the Properties of a Linear
Tax.” Journal of Public Economics 4 (2):163-185.
Rugh, Jacob S., and Jessica Trounstine. 2011. “The provision of local public goods in diverse
communities: Analyzing municipal bond elections.” Journal of Politics 73 (4):1038-1050.
Schneider, Mark. 1989. The competitive city: The political economy of suburbia. University of
Pittsburgh Press, 1989.
Sears, David O., Jim Sidanius, and Lawrence Bobo. 2000. Racialized politics: The debate about
racism in America. University of Chicago Press.
Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2008. “The color of devolution: Race,
federalism, and the politics of social control.” American Journal of Political Science 52 (3): 536-
553.
Soss, Joe, Richard C. Fording, and Sanford Schram. 2011. Disciplining the Poor: Neoliberal
Paternalism and the Persistent Power of Race. University of Chicago Press.
Tajfel, Henri and John C. Turner. 1986. “The Social Identity Theory of Intergroup Behavior.”In
S. Worchel and L.W. Austin (eds.) Psychology of Intergroup Relations. Chicago: Nelson-Hall.
Theil, Henri. 1967. Economics and Information Theory. Amsterdam: North Holland.
Tiebout, Charles M. 1956. “A Pure Theory of Local Expenditures.” Journal of Political
Economy 64 (5): 416-424.
Trounstine, Jessica. 2016. “Segregation and Inequality in Public Goods.” American Journal of
Political Science 60 (3): 709-725.
Vigdor, Jacob. 2002. “Interpreting Ethnic Fragmentation Effects.” Economic Letters 75 (2): 271-
276.
Wright, Gerald C. 1976. “Racism and Welfare Policy in America.” Social Science Quarterly 57
(1): 718-730.

93
Table 3-1: 10 Most and Least Racially Unequal and Heterogeneous Cities in 2000
10 Most Racially Unequal Cities
Racial
Inequality
10 Most Racially Heterogeneous
Cities
Racial
Diversity
1 Atlanta, GA 0.242 1 Oakland, CA 0.704
2 Washington, DC 0.210 2 Stockton, CA 0.691
3 Dallas, TX 0.149 3 New York, NY 0.686
4 Houston, TX 0.146 4 Houston, TX 0.685
5 Montgomery, AL 0.142 5 Los Angeles, CA 0.681
6 New Orleans, LA 0.132 6 Long Beach, CA 0.670
7 Los Angeles, CA 0.123 7 Chicago, IL 0.664
8 Shreveport, LA 0.120 8 Dallas, TX 0.652
9 Baton Rouge, LA 0.120 9 San Jose, CA 0.645
10 Miami, FL 0.119 10 Sacramento, CA 0.637

National Average 0.063 National Average 0.481
10 Most Racially Equal Cities
10 Most Racially Homogenous
Cities

1 Warren, MI 0.0002 1 Warren, MI 0.126
2 Detroit, MI 0.003 2 Spokane, WA 0.139
3 Dayton, OH 0.009 3 Hialeah, FL 0.143
4 Gary, IN 0.010 4 Lincoln, NE 0.149
5 Knoxville, TN 0.011 5 Madison, WI 0.233
6 Huntington Beach,
CA
0.014 6 Des Moines, IA 0.255
7 Kansas City, KS 0.016 7 Portland, OR 0.285
8 Rochester, NY 0.017 8 Lexington KY 0.294
9 Hialeah, FL 0.018 9 Gary, IN 0.294
10 Spokane, WA 0.022 10 Knoxville, TN 0.304
Notes: Racial inequality is calculated with Theil’s T index whereas racial diversity is calculated with an
inverse of Herfindahl-Hirschman index. The universe of the sample is 112 largest cities in the U.S.

94
Table 3-2: Effect of Racial Inequality on Total Public Goods
Dependent Var: City Metro State
per HH Expenditure
(constant 2011$)
𝜷 SE

𝜷 SE

𝜷 SE
Racial Inequality -2589.10** 1136.35 -3559.87** 1467.20 -10794.34 6660.73
Total Income Inequality 231.54 2681.19 2034.32 2306.54 -4259.87 7043.76
Racial Diversity -1847.67 1175.53 -237.77 2018.27 -4104.88 8101.75
Log of Population -518.18 426.18 135.44 341.70 -1757.75 1355.83
Average HH income 0.07*** 0.02 0.02* 0.01 0.14* 0.07
Share of College Graduate -4357.81 2893.65 -2143.40 2978.21 -18856.96 11231.4
2
Share of Elderly Population -358.87 3499.81 -3794.30 4280.05 -6473.09 11190.3
7
Poverty Rate 4479.25 3010.36 -2027.55 3623.54 20704.15 17844.1
2
Unemployment Rate -6985.29 5114.92 -2117.83 2739.15 -6891.40 6098.36
Constant 7351.80 5358.00 -161.51 4359.32 26197.13 19303.3
6
Fixed Effects City and Year MSA and Year State and Year
N 336 1,086 150
R
2
-within 0.576 0.213 0.660
Notes: Robust standard errors clustered by each sample unit in parentheses. R-squared-within reports
variation within units explained by the covariates as a % of total within-unit variation.
*** p < 0.01 ** p <0.05 * p<0.1.

95
Table 3-3. Results at the School District Level
Sample: School
Districts
With Population >=
2,500
Dependent variable:
School District
Spending on K-12
education
(per pupil)
Full
Balanced
Sample
(1)
Excluding
Outliers
(2)
Excluding
Outliers &
Negative
spending
(3)
Excluding
Outliers &
Controlling
for SFR
(4)
Excluding
Outliers &
Controlling
for IGR
(5)
Excluding
Outliers &
Controlling
for IGR for
the districts
with high
discretion
over funding
(6)
Excluding
Outliers &
Controlling
for IGR for
the districts
with low
discretion
over funding
(7)
Racial Inequality -1113.97** -2941.70*** -2308.07*** -2720.22*** -2454.83*** -3257.38** -1192.99
(517.44) (754.05) (771.95) (747.30) (751.39) (1451.32) (742.48)
Total Inequality -1366.23*** -846.58** -914.76*** -785.14** -615.85* -84.76 -1073.35***
(382.38) (338.47) (339.65) (341.26) (330.36) (659.79) (329.87)
Racial Diversity 2569.75*** 2685.02*** 2534.58*** 2445.79*** 1563.52*** 3610.01*** 325.43
(389.73) (353.39) (361.43) (353.93) (345.11) (588.54) (351.13)
Log of Total
Population -229.54** -288.48*** -257.82*** -283.05*** -520.28***
-889.22*** 10.59
(112.88) (90.65) (92.74) (89.30) (90.97) (176.50) (89.23)
Average HH income 0.07*** 0.07*** 0.07*** 0.07*** 0.06*** 0.06*** 0.05***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Share of College
Graduate 5225.75*** 5898.05*** 5957.21*** 5974.81*** 4424.75***
6765.36*** 29.13
(1193.93) (882.89) (898.71) (878.33) (878.22) (1320.92) (942.80)
Share of the Elderly
10332.48**
*
10433.55**
*
10538.63**
*
10224.57**
* 7834.82***
7713.53*** 6660.59***
(1133.70) (963.57) (968.95) (980.96) (905.68) (1413.03) (1078.28)
Poverty Rate 3069.82*** 2361.37*** 2597.66*** 2566.92*** 1099.50* -469.94 1482.48**
(710.18) (643.43) (660.70) (651.53) (665.22) (1521.95) (638.98)
Unemployment Rate -1054.11*** -1007.64*** -989.98*** -1022.28*** -1084.52*** -1608.92*** -868.80***
(162.75) (138.58) (130.03) (138.24) (136.19) (311.05) (141.21)
School Finance
Reform (SFR) -468.40**

(183.39)
Avg HH income x SFR 0
(0.00)
Intergovernmental
Revenue from State
(IGR)
0.57*** 0.36*** 0.88***
(0.02)
(0.02) (0.03)

Constant -1547.25 -795.29 -702.77 -641.15 4160.43*** 8547.05*** -1366.31
(1073.16) (856.41) (874.09) (850.62) (875.01) 1714.61 (844.99)
School Districts FE Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes
N 21,267 20,424 20,227 20,424 20,424 10,212 10,212
R
2
-within 0.252 0.301 0.311 0.304 0.697 0.666 0.776
Notes: We define outliers as the school districts with the largest 1% and smallest 1% of changes in either Racial Inequality
index or in the school district spending on education by decade. For the models (1) through (4), the dependent variables
deduct the intergovernmental revenues from higher level government to measure local provision. For the models (5)
through (7), the dependent variables do not deduct the intergovernmental revenues. Deducting the intergovernmental
revenue for the models (5) through (7), however, does not change the results. For the court-ordered school finance reform
coding, we follow Card and Payne (2002)’s taxonomy. The school districts with high discretion over their finance are
defined as those whose share of state intergovernmental revenue in their total local education direct spending is below the
median (56.13%) as of 1980. The districts with low discretion are above the median in 1980. Robust standard errors
clustered by school districts in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
96
Table 3-4: Budget Cut Effect of Racial Inequality on Public Goods

Racial
Inequality
Total
Inequality
Racial
Diversity
Hospital
Bed Rate
Physician
Rate
# of
Hospitals
Crime
Rate
Obs
R
2
-
within
Panel A: City
(1) Hospitals -1385.51** 2450.69 -172.68
NA NA NA

336 0.10
(692.61) (1863.85) (722.13)
(2) Health
52.88 -85.10 -439.09
NA NA NA

336 0.39

(292.80) (521.86) (287.29)

(3) Police -530.66*** 172.73 -53.33 -0.00
330 0.76
(190.88) (389.66) (204.21) (0.01)
(4) Parks -557.26** 549.53 321.03
336 0.20
(253.93) (539.36) (248.38)
(5) Housing -492.58 -915.85 222.84

336 0.29
(414.77) (963.61) (387.43)

(6) Highways -80.24 439.47 -117.58

336 0.06
(372.67) (678.25) (346.06)

(7) Sanitation 473.28 -1725.92* -1362.65***

336 0.10
(464.86) (880.2) (431.10)

(8) Fire -109.96 -187.15 -167.76

336 0.51
(136.92) (250.24) (142.36)

(9) Libraries 34.36 -435.05*** -62.68

336 0.25
(70.85) (131.27) (57.93)

Panel B: Metro
(1) Hospitals -1719.47 1891.33 -569.67
1,086 0.02
(1332.47) (2122.20) (1986.95)
(2)-1 Health -892.34** -439.35 16.84
1,086 0.09
(343.79) (545.98) (328.98)
(2)-2 Health -1311.95*** -437.09 262.31 0.00 -0.34 1.75*
770 0.10
(414.48) (738.43) (405.19) (0.03) (0.37) (0.93)
(3)-1 Police -607.02*** 363.64* 95.00
1,086 0.61
(182.43) (204.23) (198.86)
(3)-2 Police -567.96*** 391.28* 40.71 -0.00
1,007 0.60
(197.90) (220.89) (199.21) (0.00)
(4) Parks -72.61 -77.08 -15.31
1,086 0.22
(217.29) (303.35) (215.44)
(5) Housing -521.78 -430.36 81.56
1,086 0.06
(342.37) (290.87) (221.75)
(6) Highways 347.88 583.23 -151.67
1,086 0.11
(229.14) (394.43) (307.33)
(7) Sanitation 30.80 76.20 369.41
1,086 0.15
(387.81) (585.10) (497.61)

(Cont’d) Racial
Inequality
Total
Inequality
Racial
Diversity
Hospital
Bed Rate
Physician
Rate
# of
Hospitals
Crime
Rate
Obs
R
2
-
within
(8) Fire -148.32 145.22 19.52 1,086 0.41
97

(121.53) (184.41) (113.77)
(9) Libraries 22.98 -78.50 -83.46
1,086 0.30

(58.96) (69.01) (63.43)
Panel C: State
(1) Hospitals
2053.50 6885.41** -838.80
1.34* -2.16 -1.14
150 0.28
(1889.81) (3026.20) (1762.80) (0.67) (2.15) (0.87)
(2) Health -524.23 -114.09 -523.92 0.29 -2.09 -1.30
150 0.72
(1106.43) (2292.05) (1391.83) (0.46) (1.31) (0.78)
(3) Police -1457.34* -285.84 407.21 0.02
150 0.69
(863.87) (1133.48) (850.89) (0.01)
(4) Parks -2186.40* -2199.05** -336.09
150 0.43
(1116.33) (1020.76) (969.73)
(5) Housing
-2574.08* -2986.94** -1223.15
150 0.38
(1385.51) (1325.31) (1239.27)
(6) Highways -1878.77 -1714.98 720.10

150 0.45
(2030.51) (2369.29) (2140.64)
(7) Sanitation -1495.04 -2617.84 -1424.07

150 0.48
(1301.39) (1603.84) (1725.75)
(8) Fire -529.61 -78.00 -145.66

150 0.62
(340.83) (431.66) (403.97)
(9) Libraries -223.50 -221.49 -72.89

150 0.53
(252.20) (192.25) (228.11)
Notes: Robust standard errors clustered by each sample unit in parentheses. R-squared-within reports
variation within units explained by the covariates as a % of total within-unit variation. Demographic variables are
controlled (not shown). NA notes data Not Available. *** p<0.01, ** p<0.05, * p<0.1.

98
Table 3-5: Results with Political and Institutional Variables
Dependent Var: Spending on Total Public Goods (per HH, 2011$)
Panel A: City (1) (2) (3) (4)
Racial Inequality -2367.28** -2357.21** -1845.60 -1746.37
(1151.94) (1154.776) (1491.03) (1467.22)
Mayor’s Race
(1=White)

-17.92
(125.58)

Mayor’s Party
(1=Republican)

-173.21
(113.59)
N 318 296
Panel B: Metro (1) (2) (3)
Racial Inequality -3559.87** -3576.51** -3404.48**
(1467.20) (1494.12) (1556.06)
Fragmentation -0.25 -0.80
(3.84) (4.28)
Fragmentation -9.61
x Racial Inequality (15.38)
N 1,086
Panel C: State (1) (2) (3)
Racial Inequality -10794.34 -10973.00* -9766.22
(6660.73) (6527.14) (6459.57)
Governor’s Race -472.98*
(1=White) (263.94)
Citizen Ideology 14.09 10.63
(Berry et al.) (12.27) (11.76)
Governor’s Party -52.33
(1=Republican) (80.32)
N 150
Note: Robust standard errors clustered by each sample unit in parentheses.
Demographic variables are controlled (not shown). The model estimates each column separately.
*** p<0.01, ** p<0.05, * p<0.1.

99
Chapter 3 Appendix A. Tables
Table A3-1. Mean Statistics for Key Variables in the Four Samples During 1980-2000
Jurisdictional Level City Metro State School District
Independent Variables
Racial Inequality 0.10 0.06 0.05 0.02
(0.07) (0.04) (0.03) (0.04)
Total Income Inequality 0.34 0.29 0.30 0.30
(0.08) (0.05) (0.05) (0.08)
Racial Diversity 0.50 0.35 0.33 0.25
(0.13) (0.16) (0.14) (0.19)
Log of Total Population 13.65 14.24 15.87 10.77
(1.23) (1.47) (0.88) (1.24)
Average HH income 63,906 71,844 68,251 71,649
(in 2011 dollars) (12751) (14275) (11850) (25712)
Share of College Graduate 0.22 0.22 0.21 0.17
(0.07) (0.07) (0.05) (0.10)
Share of the Elderly 0.11 0.12 0.12 0.12
(0.03) (0.03) (0.02) (0.05)
Poverty Rate 0.18 0.12 0.13 0.12
(0.05) (0.04) (0.03) (0.08)
Unemployment Rate 0.08 0.06 0.06 0.08
(0.03) (0.02) (0.01) (0.11)
Crime Rate 8576 5714 *1007 5302
(2747) (1786) (1412)
Hospital Bed Rate 390 *776 405
(201) (160)
Physician Rate 252 *1017 232
(109) (74)
# of Hospitals 39 *776 207
(51) (143)
Dependent Variables (per HH in 2011$)
Spending on Hospitals 618 375 1021
(631) (685) (437)
Spending on Health 324 102 536
(275) (136) (247)
Spending on Police 846 614 650
(324) (253) (224)
Spending on Parks 382 279 298
(218) (159) (128)
Spending on Housing & 580 98 321
Community Development (407) (135) (183)
Spending on Highways 480 326 1233
(219) (185) (370)
Spending on Sanitation 723 557 560
(307) (224) (176)
Spending on Fire 398 299 268
(120) (110) (97)
Spending on Libraries 102 86 83
(48) (42) (33)
Spending on K-12 education
(per pupil)
4292
(3406)
State Intergovernmental
Revenue on local education
(per pupil)

4625
(2138)
N of Observations 336 1,086 150 21,267
Notes: All variables are weighted by total population. Standard deviation in parenthesis.
* denotes the sample size for specific variables in the metro sample.
100
Table A3-2. Data Sources
Variables/Sample Source(s)
Racial Inequality, Total Income Inequality, Racial
Diversity, and Demographic control variables
U.S. Census Bureau; Decennial Census of
Housing and Population 1980; 1990; 2000

Gov’t Expenditure on Public Goods U.S. Census Bureau; Census of Governments
1982; 1992; 2002

Fiscally Standardized City-level Finance Data Lincoln Institute of Land Policy

School District-level Data Inter-university Consortium for Political and
Social Research (1980) and National Center for
Education Statistics (1990 and 2000)

Crime Rate, Hospital Bed Rate, Physician Rate,
and Number of Hospitals
County and City Data Books 1983; 1991; 2000;
2004

Mayor Race and Ethnicity Municipal Government Websites

Mayor Political Party Municipal Government Websites

Governor Race and Ethnicity State Government Websites

Governor Political Party State Government Websites

Metropolitan Fragmentation Rate National Historical Geographic Information
System (NHGIS) Geographic Boundary Shape
Files 1980; 1990; 2000

101
Table A3-3. Alternative Specification for Effect of Racial Inequality on Total Public Goods
Panel A City Metro State
Dependent Var:
per HH Expenditure
𝜷 SE

𝜷 SE

𝜷 SE
Racial Inequality
-3587.35*** 1107.22 -3485.83** 1586.53 -11728.21* 6699.52
Total Income Inequality
1026.95 2708.34 2146.84 2298.65 -3233.83 6598.11
% Black Household
-2359.37 2021.38 -68.20 5497.25 -10272.64 7276.62
% Hispanic Household
-1313.51 2446.84 1192.71 3032.83 -425.95 9192.87
Demographic Controls
Yes Yes Yes
Fixed Effects
City and Year MSA and Year State and Year
R
2
-within
0.579 0.213 0.661
Panel B City Metro State
Dependent Var: per HH
Expenditure in Log
𝜷 SE

𝜷 SE

𝜷 SE
Racial Inequality
-0.6773** 0.2944 -0.7596 0.5153 -1.0294 0.9886
Total Income Inequality
-0.4971 0.5966 -0.1600 0.7969 0.0879 1.1985
Racial Diversity
-0.1208 0.3020 -0.4712 0.6196 -0.2126 1.1949
Demographic Controls
Yes Yes Yes
Fixed Effects
City and Year MSA and Year State and Year
R
2
-within
0.565 0.218 0.763

102
Table A3-4. Summary of the Findings

Racial
Inequality
Total
Inequality
Racial
Diversity
Total Number of coefficients 28 28 28
Negative coefficients 21 18 18
Significant negative coefficients 9 5 1
Positive coefficients 7 10 10
Significant positive coefficients 0 1 1
City-level
Number of coefficients 9 9 9
Negative coefficients 6 5 7
Significant negative coefficients 3 2 1
Positive coefficients 3 4 2
Significant positive coefficients 0 0 0
Metro-level
Number of coefficients 9 9 9
Negative coefficients 6 4 4
Significant negative coefficients 2 0 0
Positive coefficients 3 5 5
Significant positive coefficients 0 1 0
State-level
Number of coefficients 9 9 9
Negative coefficients 8 8 7
Significant negative coefficients 3 2 0
Positive coefficients 1 1 2
Significant positive coefficients 0 1 0
School District-level
Significant negative coefficients 1 1 0
Significant positive coefficients 0 0 1
Notes: All results are based on the models with demographic control variables as in Panel A, Table 3-4.
“Significant coefficients” are statistically significant at least at 10% level.
Robust standard errors clustered by each sample unit in parentheses.

103
Chapter 3 Appendix B
Data
Our city-level data is available at https://www.lincolninst.edu/subcenters/fiscally-standardized-
cities/search-database. FiSCs database does not provide the estimates for the number of households. Also,
the city populations are different between FiSCs database and the estimates of the decennial Census due
to the calculation methodology of FiSCs. Since the public expenditures in our dependent variable are per
household, we calculate the ratio of the populations between the two data sources and adjust the number
of households from the decennial Census with the ratio to FiSCs data. All city-level results presented in
Table 3-3 are with the adjusted number of households but the results are not sensitive to the non-
adjustment.
Our school district-level data for 1980 is Summary Tape File (STF) 1 and 3. STF 3 is collected from
Inter-university Consortium for Political and Social Research (ICPSR) at
http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/3518. STF 1 is collected from
http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/3517
School district data for both 1990 and 2000 are compiled from the National Center for Education
Statistics (NCES) at http://nces.ed.gov/programs/edge/tables.aspx?ds=census&y=1990 and
http://nces.ed.gov/programs/edge/tables.aspx?ds=census&y=2000. For 1990 and 2000, only STF 3 is
available. 1990 data is called School District Data Book (SDDB). Compared to 2000 data, 1990 data
provides a more comprehensive set of variables. For 1990, we used the variables of both all households
and all persons, depending on our choice of the variables. The code book for 1990 SDDB is available at
http://www.nber.org/sddb/. For 2000 data in NCES database, the names of subcategory for each variable
are missing. But they are listed in the same order as the variables in 2000 decennial Census of Population
and Housing. We used the codebook of 2000 decennial Census to match the names of variables.
Variables-Total Inequality, Racial Inequality and Racial Diversity
In constructing total income inequality measure, we calculated the income shares of the household groups
by taking the midpoints of the income brackets. The numbers of household income brackets in 1980,
1990, and 2000 are 17, 25, and 16. We dealt with the top code issue by estimating the average household
income of the highest income group with aggregate household income, assuming that the incomes of the
rest of the groups are uniformly distributed around their midpoints of the income brackets.
For our racial inequality variable, we could not use individual income for the following reason: In order to
use individual measure, we need income variable on how many individuals 15+years are in various
income brackets for the measurement of total income inequality. However, in Census 1990, this variable
is not available at the county-level (which is necessary to create metro-level sample) and in Census 1980,
it is only available for unrelated individuals 15+ years. Only the Census 2000 provides useful
information. Thus, we chose household for the measurement of racial inequality and this is why we
measured racial diversity and local public spending by households as well.
Instead, we checked the correlation for our three primary variables of interest–racial diversity, total
income inequality, and racial inequality—between individual measure and household measure for 1398
104
urban counties in 2000, the year in which these measures are all available. For individual race, we used
non-Hispanic White, non-Hispanic Black, non-Hispanic Asian and Pacific Islander, non-Hispanic Native
American. For individual income, we excluded those with no income among population 15+ years. The
numbers of income brackets between the two measures are also different. Despite these discrepancies, the
correlations for diversity, total inequality, and racial inequality are 0.98, 0.87, and 0.79, respectively.
The “average household income by race” variable, which is essential in measuring the between-race
group income inequality, is not available in the Census 1980. Instead, “average family income by race” is
available. Due to this data limitation, for the year 1980, we used the family in measuring between-race
group income inequality, but we still used the household for total income inequality and racial diversity
for the consistency with other years.
Also, while the average household income of non-Hispanic White is available for Census 2000, this
variable is not available for Census 1990 and Census 1980. Assuming that the average household income
of Hispanic who reports themselves as White is not much different from that of Hispanic as a whole, we
deducted the aggregate household income of Hispanic White from White and this is what we get for the
aggregate/average household income for Non-Hispanic White in 1990. We also repeat this process for
Non-Hispanic White family with Census 1980.
We checked our assumption from the IPUMS-USA database for the three decennial years with the
national average: In 2000 with 1% national sample, the average household income of the Hispanic White
group was $45,919 whereas that of the Hispanic group as a whole was $43,589 in 2000 dollars. In 1990
with 1% national sample, they were $31,809 and $29,981 in 1990 dollar (in the same order). In 1980 with
1% metro sample, they were $16,148 and $16,054 in 1980 dollars (in the same order). We considered
them as not being substantively different since the percentage of the average income difference in the
average income of whole Hispanic group is very small at 6% and 5%, respectively.
While the categories of household race are Non-Hispanic White, Black, Asian, Native American, and
Hispanic for our state, metro, and city sample. These categories for school district are slightly different as
follows. For school district data, in 2000, non-Hispanic White, Black, Asian, Native American, two or
more races (others), and Hispanic are used. For 1990, non-Hispanic White, non-Hispanic Black, non-
Hispanic Asian, non-Hispanic Native American, non-Hispanic Others, and Hispanic are chosen. For
1980, non-Hispanic White, non-Hispanic Black, Asian, Native-American, non-Hispanic Others, and
Hispanic are used.
In calculating the aggregate/average family income of non-Hispanic White family in school district 1980
data, we attempted to use the average family size of non-Hispanic White in each district in 2000 (2000 is
the only year that provides this information in our data sets) but the calculated figures look unreliable for
some school districts when they were compared to the family income of Hispanic in 1980. For this
reason, we instead used White family income for non-Hispanic White family in order to measure racial
inequality. But, for racial diversity at the school district-level in 1980, the share of non-Hispanic White
household was still used.
105
Chapter 4. Capitalizing on Collective Action: Creation of Community
Services Special District and Property Value Appreciation

Summary
When citizens are not satisfied with the quality of public services from the existing municipal
authorities, what actions can they take? This essay examines community services districts, a
class of special districts created in California as an instrument for neighborhood governance.
Citizen residents create these districts to reshape a representational structure in their governing
authority in order to achieve better services. Using the seven recently created districts in
Southern California as a study case, this paper measures their impacts. Housing price is used
as an evaluation metric, and the quasi-experimental research design utilizes hedonic regressions.
The findings suggest that the creation and service provision of new community services
districts increase residential property values by 12 percent compared to those in the surrounding
areas. Hence, a local governing structure that aligns the scope of representation with the
community boundary seems more likely to improve the service quality than a structure that
does not.

106
Introduction
Special districts—a type of local government—are much more prevalent in American
communities than the general public and scholars realize. In 2017, there were 90,075 local
governments in the United States, and 43 percent of these were special-purpose governments.
29

These districts have been the fastest-growing institutional form among all types of local
governments, at least since 1942, the earliest year in which the census reported its numbers.
30
In
California, the focus of this paper, special districts constitute nearly half of all 4,425 municipal
governments.
Despite their prevalence, we have a very limited understanding of their variation in
institutional type and service scope and, more importantly, of their governance impacts. Rather,
most practitioners and academics associate these institutions with negative images of shadow
governments that have been characterized by media and anecdotal narratives. As noted by Mullin
(2009), however, “special districts are not all the same” (p. 184); some are created to address
compelling local and community problems, and they offer a wide range of neighborhood-level
services.
The purpose of this essay is to introduce community services districts (CSDs) that have
proliferated in California as an instrument for neighborhood governance in unincorporated areas.
This research then assesses their impacts on residential property values. One of the unique
characteristics of CSDs is that they are created by citizen residents to address collective action

29
This count excludes the number of school districts, following the U.S. Census Bureau’s definition and
classification. Throughout the essay, the term “special district” and “special-purpose government” are
used interchangeably.
30
Since 1942, they have grown by an average of 10 percent every 5 years, resulting in in increase from an
initial 12,319 to 38,542 by 2017. These numbers exclude school districts (Author’s calculation, 1942-
2017 Census of Governments).
107
problems in communities. This institutional type is more flexible than a typical special district,
which focuses on one or two functions. A CSD provides various neighborhood-level services,
such as water, sewer, fire protection, parks and recreation, and road maintenance. Under
California state law, CSDs can provide up to 31 services (Senate Bill 135, 2006). CSDs also add
innovation to the representation of constituencies by enabling residents to elect governing board
members from their respective communities rather than from a larger county jurisdiction. As a
result, residents do not need to share their resources with other communities in the county as
before; instead, they can invest in resources exclusively for their own community planning and
development.
One of the cases this essay examines helps illustrate this point. Helendale, home to 6,000
residents, is an unincorporated community in San Bernardino County. As it is located in the
desert area, water infrastructure has been a significant resource in this community. Until 2006, its
water policies were planned and developed by the county as one of its planning and service areas
(called the “county service area”). The residents, however, regularly experienced leakage
problems from the aged infrastructure under the county’s planning and management. While
paying for 18 full-time county employees, the taxpayers did not frequently experience timely
responses, repairs were temporary, and preventive maintenance services were lacking. This long-
lasting problem, which was not acceptable from the standpoints of either resource efficiency or
representation, sparked collective action among the resident themselves: They first dissolved the
existing county service area in their jurisdiction. Then, the community had an election to form a
CSD as their own local government, featuring a new governing body from local professionals
who reside in their communities.
108
With eight full-time employees, the first action the district took was investing $2 million
to replace and fix the aged water infrastructure. They also contracted with a waste hauler
company to customize trash collection services to the community’s needs. Further, the Helendale
CSD recently purchased $4 million in water rights from a local provider as part of its strategic
future water planning and management. These community-focused solutions might not have
been considered under the county’s planning system because the county is also responsible for
other communities and policy areas.
The motivation and purpose of CSD creation suggest that this type of special district
could improve the efficiency of neighborhood services. To test this question, this essay uses a
quasi-experimental research design to measure the effect of CSD creation and services on
residential property values. The findings indicate that the creation of a CSD increases single-
family home prices by 12 percent compared to the prices of other homes in the surrounding
county area. This evidence suggests that when the rules allow, citizens can create public
institutions to improves their neighborhood-level service quality.
This study contributes empirical evidence to the discussion of the impact of special
districts on local communities. Scholars often have studied other private or quasi-private
organizations such as homeowner associations and business improvement districts and examined
how they impact residential and commercial property values. No studies, however, have
examined whether special districts can be an instrument for neighborhood governance and
enhance property values. With digitization in special district boundaries, academics can now
examine the impacts of these less studied institutions that play an important role in local and
community governance.
109
This question is also of importance to practitioners. Knowing whether the creation of
special districts improves neighborhood services is a first step for them to determine whether
these institutions are necessary at all. Obtaining a more nuanced understanding of what district
types and characteristics can then enhance property values constitutes a second step that helps us
assess their effectiveness and determine when we will need them. Further, practitioners can
consider how the innovative institutional features of CSDs can be applied to other contexts of
community planning and development.

Background
Special districts—also referred to as special-purpose governments—deliver focused services that
their constituents want. Most of these districts provide a single service. This service
specialization contrasts with the broader range of services provided by general-purpose
governments, such as cities, towns, and counties. This difference prompted a scholarly debate on
whether specialization makes districts provide services more efficiently and responsively than
general-purpose local governments. Most existing studies examine whether areas with more
special districts are associated with higher expenditures for services. The evidence has been
mixed, though more studies have generally found a positive relationship, potentially signaling
the low effectiveness of special districts compared with cities and counties.
Two major limitations, however, prohibit us from drawing a conclusion on their
performance from this class of studies. First, many special districts are different from each other
in their function, service scope, and governance characteristics. As such, studies of special
districts need a more nuanced approach that distinguishes these differences. Second, a criticism
110
has been that expenditure outcomes are not the best metric to evaluate special districts’
performance. For example, a higher number of special districts and their association with higher
expenditures can also mean that residents are willing to pay more for quality services.
31
Perhaps
because of these issues and combined with sparsely available data, special districts have received
less attention from urban scholars than other institutions have, such as business improvement
districts, innovation districts, and homeowner and/or neighborhood associations. As a result, the
importance and relevance of special districts have largely been isolated from mainstream
discussions of urban planning and development.
Existing dominant views on special districts originate from Nancy Burns’s (1994)
influential work. Burns argues that private developers initiate and dominate the process of
creating special districts to finance their infrastructure activities. For wealthy homeowners, these
business interest groups serve as their organizational bases to further the characteristics and
qualities of communities that they desire. This combination allows both developers and affluent
citizens to extract private values from public institutions. Others have also proposed similar
political economy perspectives on the creation of special districts. MacManus (1981), for
example, claims that local officials have incentives to create special districts as a property-tax
relief mechanism. Because local governments often face statutory restrictions on taxing and
borrowing powers, officials may consider special districts as a way of circumventing these
constraints, especially when they face fiscal stress. Recent scholars (e.g., Deslatte, Scott, and
Carter 2019) also argue that when private developers drive the special district formation

31
A major exception to this approach is Mullin’s (2008) work, which compares the service
responsiveness of independent water districts to that of cities and counties by examining the adoption of
progressive rates.

111
processes, they rarely face opposition from local political officials because it contributes to the
growth of their areas without the costs for local governments, at least in the short term.
Unsurprisingly, in most of these views, there is no public who receives the benefits or
bears the costs of services from the districts and hence plays an equally important role. When
special districts are examined from such a micro lens of neighborhood governance, their
institutional diversity suggests that they are not all the same and, hence, that a more nuanced
perspective is needed. With regard to their utility in neighborhood-level governance, one such
example is CSDs, whose formation process is initiated and led by residents themselves to
address their collective action problems. As illustrated in the example of the Helendale
community, most of these CSDs in California are located in unincorporated areas. By virtue of
their locations, residents in these communities initially receive essential utility services from
their county governments as one of their planning and service areas.
The planning and institutional structure under the county system as shown in Figure 4-1
suggests that it may not be an optimal form of service provision for the respective communities
from an efficiency perspective. Under the county’s governing system, each local community in
unincorporated areas is one of the planning and service areas that a county is responsible for. The
same county supervisors act as a governing body for all these communities, but they may
selectively allocate their attention to more politically salient issues in their county jurisdiction
(Besley and Coate 2003; Mullin 2009). As in Helendale, when the residents perceive that they
receive insufficient attention and services given the taxes they pay, the creation of a CSD can
then be considered as a community’s new governing model.

[Figure 4-1]
112

In this vein, residents have to dissolve the existing county service area through a local
agency formation commission (LAFCO) in their county. They then submit a proposal to create a
new CSD to the LAFCO for review. After approval, the community can hold an election for its
formation. For the CSD’s governing board members, the residents should elect local
professionals who reside within their community boundary. An advantage of having such a local
body, as compared with the county system, is that infrastructure policies can be planned and
developed exclusively for residents’ own community, perhaps with a long-term perspective.
Does neighborhood governance reform such as establishing CSDs improve services as
residents wish? To test this, this essay uses residential property values as an evaluation metric.
Using this metric overcomes a common methodological issue in previous studies, which failed to
directly measure the benefits of services. Other scholars have used this metric to examine the
impacts of such institutions as homeowner associations, neighborhood associations, and business
improvement districts. Given that CSDs are created to improve neighborhood-level services, we
should expect that their creation and service provision will be capitalized into the property values
of residential homes as well.
Along similar lines, Meltzer and Cheung (2014) and Craw (2017) report that houses in
homeowner associations are sold at a 6 percent premium over their counterparts in the same
census tracts or block groups. Craw (2017) also reports that home sale prices in neighborhood
associations are 9 percent higher than those without in the same block groups. Similarly, Ellen et
al. (2007) find that the creation of business improvement districts increases commercial property
prices by 15 percent. Notably, a more recent study that examines new city incorporations (from
the status of a previously unincorporated area) reports a 12-13 percent increase in housing prices
113
(Patrick and Mothorpe 2017). Given that CSDs are formal local governments and provide
essential public services such as water, sewer, trash collection, and fire protection, their effect on
residential property values is expected to be positive and greater than that of homeowner
associations or neighborhood associations.

Data and Methods
Data were gathered from two main sources. First, GIS shapefiles of special districts were
collected from LAFCOs in the following five counties in Southern California: Los Angeles,
Orange, San Bernardino, San Diego, and San Luis Obispo. Second, the property transaction data
were supplied from DataQuick Information Systems. This database, a collection of public
records on property characteristics and transactions from the county assessor and register of
deeds officers, includes all real estate property transactions in California during the period from
1988 to 2012. The data were cleaned to retain only arms-length, fair market value transactions on
single-family residential, detached houses.
In 2017, 248 special districts were active in these five counties. As shown in Figure 4-2,
three restrictions were applied for the sampling process for this study. 1) Housing transaction
data are available before and after the creation of districts with populations of at least 1,000. 2)
The districts were formed through collective action by local residents to obtain governmental
power after they dissolved the existing county service areas. 3) Each district serves a single
jurisdiction but not multiple ones, as this accords with the concept of community and
neighborhood governance.

114
[Figure 4-2]

The sample districts are listed in Table 4-1 with the type of services they provide, the
urban-rural classification in their location, and the year they were formed. In all seven
communities, county authorities previously provided comparable services, either directly or
through county service areas. Six of them are CSDs that provide a mix of services, from water,
sewers, and solid waste collection to street lighting, road maintenance, and parks and recreational
services. The remaining Chino Valley Fire District provides a single service specializing in fire
protection and emergency services. Turning to the district financing, a major source in the five
CSDs is user fees, while the other two districts rely entirely on property taxes. Notably, all these
seven districts are the sole entities that provide the listed services, and they do not overlap with
other types of local government entities (e.g., other special districts, cities, or counties).

[Table 4-1]

Figure 4-3 presents a map of the seven special districts, adjacent cities, and nearest
county service areas that are still active and provide comparable services. The analysis assigns
treatment to all arms-length transactions of single-family houses that occurred within the special
district communities. The control group is the transactions of single-family homes in the areas
surrounding the district jurisdictions. The distances for the surrounding areas were determined
115
based on the volume of home sales between the two groups, as illustrated in Table A4-1 in the
Appendix A.
32

[Figure 4-3]

Figure 4-4 shows the mean statistics and t-test results of home characteristics between the
treated and control groups from the transaction records before and after the districts were
formed. Before the districts were formed, an average home in the treated group sold at a discount
compared to that in the control group (p<0.01), even though the treated homes show higher
amenities in their property structural characteristics, such as larger lot size and living area, more
bedrooms and baths, and newer properties (p <0.01 for all these characteristics). After the
districts were formed, however, houses in the treated group sold at a premium (p<0.01).
The price discount of treated homes before the creation of special districts suggests that
the home values might have been suppressed in these communities due to poor service quality
from the county authorities. In contrast, the price premium of treated homes after the creation of
special districts indicates that the districts would have provided better services. Of course, this
simple test of mean difference in housing prices and other characteristics does not tell us much
about the impact of special district formation and their service provision. It does not account for
other potentially influential differences between these two groups. To address this, the next
section turns to a systematic analysis.

32
In constructing a control group, the house transactions that occurred in adjacent incorporated cities
were excluded if the treated groups are the transactions in unincorporated communities. This approach is
used to avoid duplicate transactions counts in the analysis.
116

[Figure 4-4]

The analysis applies the difference-in-differences (DID) method to a hedonic regression
model to estimate the net benefits of services granted by a CSD. This method differs from
existing empirical approaches on special district research in two ways. First, the DID method
allows us to explore the counterfactual scenario of whether the service efficiency would have
been different if decisions were made by county authorities. Not exploring this dimension has
been one of the major limitations in previous research. Second, the analysis quantifies the
benefits that the two governing forms impart for their communities as well as the costs. Existing
empirical studies have focused on the costs of service provision (i.e., service expenditures) but
not on the benefits, in part due to measurement difficulty. The DID hedonic regression
overcomes this limitation by quantifying the net benefits of services. The estimated model is as
follows:
ln ( 𝐻𝑜𝑚𝑒 ⁡𝑆𝑎 𝑙 𝑒 ⁡𝑃𝑟 𝑖 𝑐𝑒 )
𝑖𝑡
= ⁡ 𝛼 0
+ 𝜗𝐻
𝑖𝑡
+ 𝛽 𝑆𝐷
𝑖 ∗ 𝑃𝑜 𝑠𝑡 𝑡 + 𝛾 𝑆𝐷
𝑖 + 𝛿 𝑃 𝑜 𝑠𝑡
𝑡
+ 𝜑 𝑗 𝑇 𝑟 𝑎 𝑐𝑡 ⁡ 𝐹𝐸
𝑗 + 𝜔 𝑞 𝑄 𝑢 𝑎 𝑟 𝑡 𝑒 𝑟 ⁡𝐹𝐸
⁡ 𝑡 + 𝘀 𝑖𝑞 𝑡 (1)
where i, j, t, and q are a property, a census tract, a year, and a quarter, respectively.
The dependent variable is the log of the sale price of a single-family home. H is a vector
of housing structural characteristics that are commonly used in hedonic models (Rosen 1974),
which includes the lot size, the size of the living area, the number of bedrooms and baths, and the
property age at the time of sale. The quadratic terms of bedrooms, baths, and age are also
included to capture their potentially nonlinear effects on house prices. SD is an indicator variable
117
of treatment that equals 1 for the treated homes and 0 for the control homes (see Table A4-1 and
Figure 4-3). 𝑃𝑜 𝑠𝑡 𝑡 is an indicator vector that equals 1 for the years after districts were formed
and 0 for the years before they were formed. The coefficient 𝜗 captures the value of hedonic
features that people place on the housing sale in the market, 𝛾 measures the price difference
between houses located within CSD community boundaries and the control houses in the
surrounding areas, and 𝛿 captures the price difference before and after district formation.
𝛽 —the coefficient on 𝑆𝐷
𝑖 ∗ 𝑃𝑜 𝑠𝑡 𝑡 —is the difference in these differences, which captures
the overall impact of services (i.e., net benefits) granted by the creation of CSDs and their
services. The model also includes census tract fixed effects and quarter-year fixed effects to
account for the systematic differences across neighborhoods as well as any unobservable
influences that are correlated with both seasons and years and home sale prices. Notably, while
the model does not include any political and administrative variables such as a community’s
voting turnout rate and residents’ political ideology, the before-and-after model cancels out any
differences that exist between the CSD communities and their surrounding areas. As long as
these factors do not change over time such that they are correlated with the district formations
and subsequent changes in housing prices, they do not bias the main effect estimates. 𝘀 𝑖𝑞 𝑡 is an
idiosyncratic error that is allowed to be serially correlated among the home sales within the
clusters but is independent across them.
The analysis uses a pooled sample to capture the overall impact across the districts. To
obtain a robust and consistent estimate of the impact, four years are considered as a preferred
time window after the formation of CSDs. Public services are not established overnight, and
even after the districts are formed, communities have a transition period during which services
are still provided through a contract with their county authority. Further, it would take some time
118
for citizens to perceive differences in service quality even after the new CSD system is set up.
Additionally, for a before-and-after analysis, the model needs to use the same duration for all
seven districts. Each community had at least four years of property transaction records after the
CSD creation.
The choice of this time window is also justified by methodological concerns. Research
shows that using a DID strategy with many years of data leads to inconsistent standard errors due
to serial correlations (Bertrand, Duflo, and Mullainathan 2004). For example, the sale prices of
nearby homes may be serially correlated among themselves, and such serial correlation can be
higher with more years of data. This results in small standard errors that would lead researchers
to overreject the null hypothesis. The model therefore limits the time window to four years at
most for 𝑃𝑜 𝑠𝑡 𝑡 and estimates equation (1) with standard errors clustered by districts—the highest
level of geography with which we suspect the errors are substantially correlated (Cameron and
Miller 2015).
A critical question in the DID research design is whether a control group is a valid
counterfactual of a treated group in the absence of a policy shock. To address this concern, the
analysis uses an alternative control group to estimate equation (1). The alternative control homes
are drawn from the county service areas nearest to the CSD communities. The county authorities
in these areas still provide services comparable to those of CSDs.
Because each CSD community in this study was also one of the county service areas
before it was formed, these control areas would serve as valid counterfactuals.
33
There are also

33
The descriptive statistics also lend support to this claim. Similar to the observed pattern in Figure 4-4,
the selling prices of the treated homes were lower on average by 11 percent than those of alternative
control homes before the districts were formed (p<0.01) even though they had larger lots and were newer
119
other reasons to choose the control homes from the nearest county service areas as a robustness
check. The baseline model relies on special district boundaries to identify the impacts. In this
case, the potential presence of spillover effects across borders and endogenous border formation
are among the major threats to causal inference. In principle, the presence of mutual aid
agreements between the district communities and the surrounding areas can benefit residents in
the area who are just across the border from the districts. This would bias the estimate of the
CSD effect. Alternatively, if the district communities drew their boundaries purposefully,
excluding some properties while including others, this would also bias the estimate. Therefore, if
the results obtained from both approaches look similar, this strengthens the validity of the
research design.

Results
Table 4-2 reports the main DID hedonic regression results. Column 1 shows the results with
standard errors clustered by property, and column 2 presents the results with errors clustered by
district community. In column 1, the result for SD indicates that homes within the special district
jurisdictions were transacted at systematically lower prices, on average, by 11.1 percent,
compared to their counterparts in the surrounding areas. This result indicates a likely inefficiency
of services that have persisted in these communities, which explains why the citizen residents
took collective action to create special districts. The Post variable shows that all houses,
regardless of their treatment status, saw an average 8.4 percent appreciation of prices during the
first four years after the districts were formed. Our main variable of interest—SD x Post—shows

properties. After the districts were formed, however, the treated home prices were higher on average by
20 percent than these alternative control homes.
120
that the average effect of the CSD formation on home sale prices during the first four years is an
increase of 12.1 percent compared to those for the control homes, which offsets the existing
discount. As shown in column 2, when a more conservative test is used for statistical
significance, the variables SD and Post are not significant at all. The SD x Post DID interaction
term, however, remains significant at the 0.05 level, showing the robustness of this main effect.
34

[Table 4-2]

The coefficients on the hedonic controls generally show the expected results with the
right signs and magnitudes in reasonable ranges. Larger living area, lot size, and number of baths
are positively associated with home sale prices, indicating the greater value people place on these
amenities in the local housing markets. Home age is also positively associated with price. Given
that many homes in the study areas are newer properties with an average age of 13 years at the
time of sale, consistent with hedonic theory, age seems to represent attractive vintage
construction (a positive sign) more than depreciation of the structure (a negative sign). The
variable for the number of bedrooms, however, is not statistically significant. Rather, in these
communities, the number of baths appears to be a stronger determinant of home prices. The
results for squared control variables indicate nonlinear relationships between home sale prices
and age and the number of baths. As expected, when standard errors are clustered by district
(column 2), most hedonic controls lose statistical significance. This result suggests that district-

34
The results do not change even though the analysis excludes Chino Valley Fire District, making the
sample consist entirely of CSDs in unincorporated areas. Alternatively, an additional analysis included a
binary variable that controls for the homes in the Chino Valley Fire District to separate the effects of
CSDs. This analysis also yielded similar results.
121
level clustering is in effect a more conservative test in dealing with potential issues related to
serial correlation.
Figure 4-5 presents the results of this analysis over time from the year the districts were
formed to the next four years. Note that the results represent cumulative effects over time.
Consistent with expectations, there is no effect at the time the CSDs were formed (Post_0), as
shown by the nearly zero coefficient that is not statistically distinguishable from zero. In the
following years, the housing markets put a premium on the home sales within the special district
boundaries. If these results were found only for the first and second years immediately after the
district formation, we would reason that the observed effect might be due to the initial
expectations in the housing markets, not necessarily because of the service provision from the
special districts. In such cases, if the districts did not perform well in the subsequent years, we
should see a declining effect over time in the third and fourth years. This, however, is not the
case, as illustrated in Figure 4-5. The main effect remains stable in the first and second years
after the districts were created. Additionally, the effect increases somewhat in the third and four
years, from 10 percent to 11-12 percent.

[Figure 4-5]

Figure 4-6 presents the analysis using the alternative counterfactuals drawn from the
nearest county service areas. The results remain virtually unchanged from the findings reported
in Figure 4-5.
35
At the time the districts were created, there is no effect on property values. The

35
The results also remain robust when two districts (Chino Valley Fire and Ground Squirrel Hollow) that
do not have comparable county service areas are excluded.
122
magnitude of the coefficient is very small, at approximately 2 percent, and it is not significantly
different from zero. The effect is then quickly capitalized into the property values in the next
year, with a 12.6 percent premium in home sale prices (p<0.1). The effect then rises to 16.4
percent in the second year (p<0.01) and then remains stable in the third and fourth years at
approximately 12-13 percent (p<0.05). Importantly, the point estimates do not change whether
the control homes are drawn from the areas surrounding the district communities or from the
nearest county service areas. This result indicates that potential confounding factors, such as
spillover of benefits across borders and endogenous border formation, have not biased the main
estimate.

[Figure 4-6]

Concluding Discussion
The analysis presented in this essay suggests that CSDs can provide more efficient
neighborhood-level services than county authorities do and can therefore enhance residential
property values in communities. This empirical finding contributes to the literature, given that
virtually no study has examined the impacts of special districts on property values.
36
More
generally, the role of special districts as an instrument for neighborhood governance has not been
considered by urban scholars. Since the representative characteristics of CSDs contrast with

36
The exception to this claim is Billings and Thibodeau (2011, 2013). Their work, however, does not
consider different types of special districts. As such, all special districts in their data are considered to be
the same entities.
123
those of other special districts as well as other types of neighborhood institutions, the results
have several implications for community planning and development.
First, unlike a typical special district that focuses on a single service, CSDs are a much
more flexible form of governance model that can provide a broad range of services in
unincorporated areas, just as city and county governments do. The only service power they lack
is land use and economic development. Despite this lack of power, a number of neighborhood-
level services they provide can enhance residential property values as the analysis suggests.
Perhaps because of this advantage, these institutions have proliferated in the unincorporated
areas in California. In 2018, 1985 special districts were active in the state, and 315 of them were
CSDs.
In addition to the multiple services the CSDs can provide, this institutional type presumably
improves the efficiency of neighborhood-level services by aligning their representation with the
service scope in those communities. By virtue of their location, the local communities in
unincorporated areas receive services from their county authorities. Hence, their governing body
represents the broader interests of constituencies throughout the county jurisdiction but not the
particular needs of each of these communities. While one might argue that the county has more
resources and technical competencies than what these respective unincorporated communities
can supply, there is a discrepancy between their representation and service scope.
The creation of CSDs can correct this problem, with residents electing local professionals
from their respective communities. Under the CSD, citizen residents can even invest their
resources in infrastructure development with a long-term perspective. They retain the authority
to replace their governing board members through elections, and this serves as a mechanism for
performance accountability. This grassroot democracy aspect contrasts with what citizens often
124
face in other (quasi) private neighborhood governance institutions, such as business
improvement districts (BIDs) and homeowner associations (HOAs).
Schaller and Modan (2005) note that BIDs unilaterally grant power to a narrow set of
stakeholders such as property and business owners. Although BIDs have been known to boost
commercial property values and activities (e.g., Ellen et al. 2007; cf. Sutton 2014) and prevent
crimes in areas (e.g., Hoyt 2005; Brooks 2008), low-income residents and small business actors
are typically excluded from decision-making processes. The benefits that result from urban
revitalization, such as the adoption of BIDs, have also raised issues of gentrification and its
potentially negative consequences for low-income residents in such neighborhoods. The
institutional and representational structure of CSDs sheds light on these democratic aspects in
neighborhood governance.
It is also important to highlight that CSDs are formal governmental authorities, and as such,
they provide essential utility services that replace those provided by general-purpose
governments such as cities and counties. In contrast, other (quasi) private neighborhood
institutions such as BIDs and HOAs provide services that are supplemental to local government
services. In addition, by their private nature, these services usually target specific groups, which
can also raise concerns regarding the equity of place. By contrast, being a public institution,
CSDs have services that reach everyone in the communities in the same manner, placing them on
equal footing. Hence, just as their establishment was sparked by extensive collective action
processes, their subsequent planning and development should also reflect the main interests of
their constituency and its overall impact on the welfare of the community.
While CSDs exist primarily in unincorporated areas, whether they are urban and rural, their
similarity and differences compared with other neighborhood institutions such as BIDs, HOAs,
125
and neighborhood councils provide a number of interesting educational points. In particular, by
focusing on the representational structure in CSDs, scholars and practitioners may think about
how such innovative institutional features can help them incorporate the democratic aspect of
collective action in urban revitalization efforts.

126
References
Besley, Timothy, and Stephen Coate. 2003. Centralized versus Decentralized Provision of Local
Public Goods: A Political Economy Approach. Journal of Public Economics 87: 2611-2637.
Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004. How Much Should We Trust
Differences-in-Differences Estimates? Quarterly Journal of Economics 119: 249-275.
Billings, Stephen, and Thomas G. Thibodeau. 2011. Intrametropolitan Decentralization: Is
Government Structure Capitalized in Residential Property Values? Journal of Real Estate
Finance and Economics 42(4): 416-450.
Billings, Stephen B., and Thomas G. Thibodeau. 2013. Financing Residential Development with
Special Districts. Real Estate Economics 41(1): 131-163.
Brooks, Leah. 2008. Volunteering to Be Taxed: Business Improvement Districts and the Extra-
governmental Provision of Public Safety. Journal of Public Economics 92(1-2): 388-406.
Burns, Nancy. 1994. The Formation of American Local Governments: Private Values in Public
Institutions. Oxford University Press on Demand.
Cameron, A. Colin, and Douglas L. Miller. 2015. A Practitioner’s Guide to Cluster-robust
Inference." Journal of Human Resources 50(2): 317-372.
Carter, David P., Aaron Deslatte, and Tyler A. Scott. 2018. The Formation and Administration of
Multipurpose Development Districts: Private Interests Through Public Institutions. Perspectives
on Public Management and Governance.
Craw, Michael. 2017. Institutional Analysis of Neighborhood Collective Action. Public
Administration Review 77: 707-717.
Ellen, Ingrid Gould, Amy Ellen Schwartz, Ioan Voicu, Leah Brooks, and Lorlene Hoyt. 2007.
The Impact of Business Improvement Districts on Property Values: Evidence from New York
City [with Comments]. Brookings-Wharton Papers on Urban Affairs pp. 1-39.
Hoyt, Lorlene M. 2005. Do Business Improvement District Organizations Make a Difference?
Crime in and around Commercial Areas in Philadelphia. Journal of Planning Education and
Research 25(2): 185-199.
MacManus, Susan A. 1981. Special District Governments: A Note on Their Use as Property Tax
Relief Mechanisms in the 1970s. Journal of Politics 43(4): 1207-1214.
Meltzer, Rachel, and Ron Cheung. 2014. How are Homeowners Associations Capitalized into
Property Values? Regional Science and Urban Economics 46: 93-102.
Mullin, Megan. 2008. The Conditional Effect of Specialized Governance on Public
Policy. American Journal of Political Science 52(1): 125-141.
127
Mullin, Megan. 2009. Governing the Tap: Special District Governance and the New Local
Politics of Water. MIT Press.
Patrick, Carlianne, and Christopher Mothorpe. 2017. Demand for New Cities: Property Value
Capitalization of Municipal Incorporation. Regional Science and Urban Economics 67: 78-89.
Schaller, Susanna, and Gabriella Modan. 2005. Contesting Public Space and Citizenship:
Implications for Neighborhood Business Improvement Districts. Journal of Planning Education
and Research 24(4): 394-407.
Sutton, Stacey A. 2014. Are BIDs Good for Business? The Impact of BIDs on Neighborhood
Retailers in New York City. Journal of Planning Education and Research 34(3): 309-324.

128
Table 4-1. List of Special Districts in the Sample
Special
District
Popul
ation
Urban
/Rural
County Services
Previously
Provided
by
County?
Previously
County
Service
Area?
Major
Finance
Source
Year
Formed
Chino
Valley Fire
District
16,000 Urban
San
Bernardino
Fire protection,
emergency
services, building
inspection and
permits
Yes No
Property
tax
1990
Phelan-
Pinon Hills
CSD
21,000 Urban
San
Bernardino
Water, solid waste,
street lighting,
parks
Yes Yes User fee 2008
Helendale
CSD
6,000 Rural
San
Bernardino
Water, sewer, solid
waste, graffiti
abatement, street
lighting
Yes Yes User fee 2006
Los Osos
CSD
15,000 Urban
San Luis
Obispo
Water, drainage,
fire protection
Yes Yes User fee 1998
Heritage
Ranch CSD
3,000 Rural
San Luis
Obispo
Water, sewer, solid
waste, parks
Yes Yes User fee 1990
San Miguel
CSD
2,500 Rural
San Luis
Obispo
Water, sewer, solid
waste, street
lighting, parks, fire
protection
Yes Yes User fee 2000
Ground
Squirrel
Hollow
CSD
1,100 Rural
San Luis
Obispo
Road maintenance,
solid waste
Yes No
Property
tax
2004
Notes: CSD is Community Services District. Population estimates are from American Community Survey
2011-2015 5-year estimates.

129
Table 4-2. Difference-in-Differences Hedonic Regression Results
Dependent Var: Log of Sale Price of Single-family home
RHS variables
(1)
Clustered by Properties
(2)
Clustered by Districts
SD x Post_4 0.121*** 0.121**
(0.013) (0.045)
SD -0.111*** -0.111
(0.024) (0.067)
Post_4 0.084*** 0.084
(0.015) (0.103)
Log of living area 0.480*** 0.480***
(0.018) (0.128)
Log of lot size 0.030*** 0.030
(0.005) (0.027)
Age 0.017*** 0.017
(0.001) (0.011)
Age squared -0.0002*** -0.0002
(0.00002) (0.0002)
Bedrooms -0.013 -0.013
(0.009) (0.039)
Bedrooms squared -0.0001 -0.0001
(0.0006) (0.002)
Baths 0.163*** 0.163
(0.020) (0.090)
Baths squared -0.023*** -0.023***
(0.003) (0.005)
Intercept 7.30*** 7.30***
(0.179) (0.873)
Tract FE Yes
Quarter-Year FE Yes
Observations 78,087
R-squared 0.3465

Notes: The table shows effect of special district (SD) formation on home sale price in the first
four years, capturing public values. Robust standard errors in parentheses.
*** p < 0.01, ** p < 0.05, * p < 0.1.

130
Figure 4-1. Governing Structure Under County and CSD system

Figure 4-2. Special District Sampling Process

Community Services District:
A more locally representative
body
Sole Community
County Board
of Supervisors
County Service
Area 1
County Service
Area 2
County Service
Area 3
131
Figure 4-3. Special District Communities, Adjacent Cities, and Nearest County Service
Areas

San Bernardino County

San Luis Obispo County

Notes: Orange color indicates special districts in the sample. Blue color indicates the nearest county
service areas. Light green color indicates cities and census-designated places.

132
Figure 4-4. Mean Statistics and T-test of Home Characteristics

Note: *** p<0.01

133
Figure 4-5. The Difference-in-Difference Effect of CSD Creation
on Property Values Over Time

Notes: The figure shows the cumulative effects of the creation of CSDs on house prices, over
time from the year the districts formed (Post_0) to the four years after formations (Post_4). The
results were estimated from the equation (1) with robust standard errors clustered by districts. The
coefficients on the DID interaction term (i.e. SD*Post) are reported in solid line and 95%
confidence intervals are reported in dashed lines.

Figure 4-6. The Difference-in-Difference Effect of CSD Creation
on Property Values Over Time (Using Alternative Control Group)

Notes: The control group was drawn from the county service area nearest to each community
services district community.

134
Chapter 4 Appendix A. Tables

Table A4-1. The Number of Housing Transactions Between Treated and Control Group
Group
Before
/After
Observation
All Districts

Treated Before 26,117
Control Before 23,788
Treated After 70,782
Control After 57,869
Chino Valley Fire District

Control homes drawn from the
within a 1.5 mile ring
Treated Before 7,860
Control Before 9,088
Treated After 60,925
Control After 47,993
Phelan-Pinon Hills CSD

Control homes drawn from the
within a 2.5 mile ring
Treated Before 7,521
Control Before 6,484
Treated After 1,799
Control After 1,239
Helendale CSD

Control homes drawn from the
within a 10 mile ring
Treated Before 6,654
Control Before 4,210
Treated After 1,947
Control After 2,003
Los Osos CSD

Control homes drawn from the
within a 5 mile ring
Treated Before 2,962
Control Before 2,550
Treated After 3,501
Control After 3,110
Heritage Ranch CSD

Control homes drawn from the
within a 9 mile ring
Treated Before 290
Control Before 290
Treated After 1,752
Control After 2,936
San Miguel CSD

Control homes drawn from the
within a 5.5 mile ring
Treated Before 286
Control Before 574
Treated After 605
Control After 377
Ground Squirrel Hollow CSD

Control homes drawn from the
within a 2 mile ring
Treated Before 544
Control Before 592
Treated After 253
Control After 211

135
Chapter 5. Community’s Governing Capacity and the Success of Special
District as an Instrument for Neighborhood Governance

Summary

Building on the previous chapter, this essay argues that the community’s governing capacity
should condition the impacts that special district creation brings to places. Three factors are
examined as critical determinants of the success of specialized governance: the clarity of a problem
statement, prepared local managerial leadership, and support from the public. Using a hedonic
model, the quantitative analysis shows that the creation of a community services district yields
both positive and negative impacts on residential property values. Qualitative case study analysis
complements the heterogeneous findings of quantitative analysis by identifying why some
communities have failed in the implementation of special-purpose governance while others have
succeeded. Further analysis reveals that a community’s governing capacity is a precondition for
the success and failure of specialized governance.

136
Introduction
Special districts have often been characterized as shadow governments that generally lack
visibility and accountability to the public. While a considerable number of studies have
examined what institutional environment and other related factors drive their creation, attention
to their performance has been relatively scant. This lack of attention is partly due to limited data
availability for these entities compared with other state and local governments and partly to our
very limited understanding of their variation in institutional type; unlike cities and counties, the
specialized functions of districts vary significantly among themselves. Hence, research on
special districts also needs to consider the local contexts within which these districts emerge and
operate.
Among the various types of special districts, this essay focuses on community services
districts (CSDs) in California that are created by citizen residents’ collective action to use them
as an instrument for neighborhood governance.
37
One of their interesting features is that these
districts provide a broad range of neighborhood-level services such as water, sewer, trash
collection, fire protection, and road maintenance that cities and counties tend to provide. Hence,
CSDs are quite different from other special districts, which typically offer a single service.
More interestingly, these districts do not simply emerge out of an institutional vacuum.
Given the extensive collective action and administrative processes that residents have to muddle
through to create such districts, communities take such serious actions and investments when
they are frustrated with the existing services from their county governments. They first dissolve
the existing county service areas in their jurisdictions and then can create these districts as new

37
As this study focuses on California, it adopts the term used in the state. The nomenclature, however,
may vary in other states.
137
and sole governing authorities for neighborhood-level services. As a result, communities achieve
better representation, which allows them to elect local governing bodies from their own places
rather than relying on officials from the larger county jurisdictions.
When communities propose these new governance models that fit their local conditions
better than existing institutions did, do their efforts translate into improved services that create
greater impacts on their places? Citizen governance and institutional change scholars have
argued that establishing a new institutional rule for citizen involvement per se would not grant
the positive policy outcomes that would be expected (Cooper, Bryer, and Meek 2006; Ho and
Coates 2004; Irvin and Stansbury 2004; Box 1997). In other words, creating a new institutional
rule or form is one thing, and managing their performance is often another task (Hong 2015;
Bovaird 2007). What factors then make a major difference in the performance outcomes during a
community’s governance re-design process is a question this study addresses.
Motivated by this research question and the innovative institutional features of CSDs, this
study provides empirical findings that contribute to the knowledge of both practitioners and
academics. The managers of CSDs who were interviewed for this research claimed that these
institutional forms, with the decision-making authority directly granted from their communities,
should provide more cost-effective services compared with those provided by county authorities.
However, the managers generally lacked evidence supporting this claim. To investigate this
claim, this essay uses massive administrative records of housing transactions and assesses the
consequences of special districts through their impact on property values. As illustrated in Craw
(2017), housing sale price is an appropriate evaluation metric for sub-local governance
organizations. Further, using the field work and case study approach, this research unravels what
factors explain the divergence of special districts’ impacts. Obtaining such nuanced
138
understanding can help practitioners consider when they should advocate for these emerging
community governance models and whether they should approve their proposals for their
formation.
For academics, this study illustrates that special districts are not necessarily the
institutions of private values, which often serve the interests of developers and affluent citizens
(Burns 1994; Carter, Deslatte, Scott 2018). Depending on the local contexts and the interests and
composition of the actors dominating their formation process, these districts can also be flexibly
used as a model of neighborhood governance. Further, with more digitization in special district
information, this essay shows that scholars can study their effectiveness. Such an open data trend
enriches future avenues of research on special districts and their performance compared with that
of other institutions, as Skelcher (2006) notes. This chapter showcases one of these examples by
focusing on the districts that serve as an institutional vehicle for grassroots democracy in many
local communities in California.

Background
Research on special districts thus far has primarily relied on aggregate data released by the U.S.
Census Bureau. As noted by Foster (1997), Most of these studies equate the number of special
districts in a particular jurisdiction with their significance. Special districts, however, are one of
the most complex institutions among the local governments; the functions of services they
provide vary significantly among themselves, from utilities to very specialized services such as
cemetery management and mosquito abatement. Despite their institutional diversity, it has been
widely believed that these institutions lack visibility and accountability to the general public; few
139
citizens know whether they exist, and as a result, these institutions typically do not face the
political pressure that general municipal governments such as cities, towns, and counties do.
Existing scholarly evidence also suggests that it is often developers and affluent citizens who
initiate and dominate their formation process for private interests (Burns 1994; Cater, Deslatte,
and Scott 2018). Due to their political invisibility, citizens who live in several overlapping
districts can be charged higher tax rates due to the shared fiscal commons (Berry 2008).
These dominant views notwithstanding, according to public choice theory, however, at
least some special districts may provide quality services that their constituents desire (Foster
1997; Mullin 2009). This theoretical lens focuses on the positive side of metropolitan
fragmentation in which citizens sort themselves into a jurisdiction that provides the best bundle
of services according to their preferences (Ostrom, Tiebout, Warren 1961; Tiebout 1965). Along
these lines, special districts might serve as an institutional vehicle that provides service suited to
local conditions (Mullin 2008). Indeed, often overlooked in the previous studies is the
differences between district types, service scope, and governance characteristics.
Eger II (2006), Foster (1997), and Bourdeaux (2004), for example, argue that a
typological approach should be established to better understand these nuanced differences within
special district entities. Hence, it is not surprising that scholars often find positive managerial
practices in some special districts, in contrast to the existing dominant views on them. Heikkila
and Isett’s (2007) field work, for example, demonstrates that citizen involvement and
performance management in the Texas districts that specialize in housing, water, or health
services were higher than what critics expected. Further, research has shown that managers in
large special districts often have commitment to public values that is similar to that of city
140
governments (Berman and West 2012), and district managers employ various strategies such as
ethics management to foster community building and engagement (Chen et al. 2013).
Hence, recent scholarship indicates that studies of special districts need to consider
distinctions between the district types and the local contexts within which these entities are
created and operate (Mullin 2009; Deslatte, Scott, Carter 2019; omitted for review purposes).
This essay introduces multipurpose CSDs within this context. In California, these districts
primarily exist in unincorporated areas, and their creation replaces the existing services from
county authorities. Notably, the process of creating these districts is initiated and led by citizen
residents—not private developers—to address collective action problems in their communities.
This contrasts with the multipurpose community development districts (CDDs), which are
created by developers and are prevalent in the unincorporated areas in other states, such as
Florida, Texas, and Colorado, as recently introduced by Carter, Deslatte, and Scott (2018).
The rationale behind the creation of these CSDs closely accords with public choice
theory. When citizen residents are not satisfied with the existing services from their county
authorities, in principle, they can dissolve the county service area in their jurisdiction and create
a new local government. Figure 5-1 shows a series of administrative and procedural steps that
community stakeholders need to take in order to create a CSD under California state law. While
it seems complicated at a glance, citizen residents actually do not need to go through the state
legislature. They can file proposals for both the dissolution of the county service area and the
creation of an CSD to the LAFCO that exists in each county to manage such processes. Once the
review from the commission is complete, the community can then hold an election for CSD
creation and obtain majority approval from resident voters.
141
[Figure 5-1]

Since the creation of CSD is sparked by a clear managerial problem under the county’s
service provision, it is expected to improve services. At the same time, however, communities
also face issues of how to manage new institutions and new governing regimes. Establishing new
plans and policies, for example, and making long-term investments in infrastructure may be
challenging tasks for a new, incoming local governing body and management team to handle.

Creation of Local Governing Institutions and Property Value Change
The creation of new local governing institutions such as CSDs opens an opportunity for scholars
to measure the net benefits of the services they provide compared to those provided by county
authorities.
As Tiebout (1956) suggests, if the service benefits granted by CSDs produce better
amenities to the jurisdictions than those under the county’s service provision, such positive
amenities will not only be valued by current resident customers but also be attractive to potential
homebuyers. This will then translate into an increase in property values (Oates 1969). This
instance suggests that property values might have been suppressed under the county’s
management due to its inefficiency in service provision. Since homebuyers will adjust their
expectations to incorporate changes in both the perceived benefits and costs of public services,
capitalization of property values should capture the benefits for residents of net changes in
service costs as well (i.e., net service benefits).
142
On the other hand, if the quality of services provided by a newly established CSD is poorer than
it was under the county authority, the resultant inefficiencies and disamenities will lead to a
decline in the jurisdiction’s property values. In such cases, the county’s service provision would
be more efficient, while the CSD would suppress property values. The relative changes in home
values therefore should capture the values that the housing markets place on the net service
benefits granted by CSDs compared to those from the county authorities.
With a quasi-experimental research design this essay features (more on this below), using
housing price change as a dependent variable allows us to measure the benefits of sub-local
governance organizations as well as their service costs. This metric overcomes one of the long-
standing methodological limitations in special district research: most of the existing studies use
service expenditure as a dependent variable as a proxy for the cost of services and examine its
association with horizontal and vertical fragmentation and/or concentration of local governments
(Boyne 1992; Burns 1994; Foster 1997; Berry 2008; Hendrick, Jimenez, and Lal 2011; Goodman
2015). This metric fails to capture the benefits of services, and as a result, high levels of
expenditure could mean low effectiveness of special districts (i.e., more costs for service
provision), but it could also mean citizens’ willingness to pay more for quality services (i.e.,
more benefits from service provision). By contrast, the housing price metric allows for
measuring the net benefits of services from special districts compared to their general-purpose
government counterparts, especially with a research design that utilizes the before-and-after
analysis between the two government entities.

Conditional Effect of a Special District by a Community’s Governing
Capacity
143
Few theories exist that indicate the relative performance of special districts to general-purpose
governments. In addition to the positive effect of special districts predicted by a general theory of
public choice, the literature has yet to develop a rich theoretical framework for special districts’
performance. One exception is Mullin (2008), who proposes a conditional theory of a special
district’s responsiveness. She argues that the relative policy responsiveness of special districts to
city and county counterparts should be contingent on the severity of policy problems they
address. The theory posits that when the issue of policy problems is not controversial, special
districts should be more responsive to their constituents due to the specialized services. By
contrast, Mullin claims that when the issues are in more contentious policy domains, general-
purpose local governments should be more responsive than special-purpose districts due to the
greater political pressures that city and county officials face from their constituents.
While Mullin’s theory significantly enhanced our understanding of special districts’
effectiveness, it also makes us question what other local contextual factors would be relevant.
One such factor that this study considers is the community’s governing capacity. Among the
various concepts that have been used for this term across the disciplines, this study adopts the
definition from Ostrom’s institutional capacity (1990; 2005). Namely, how much of collective
action competence a community has developed to take advantage of new institutions they create.
In particular, the literature on collaborative governance illuminates what factors would constitute
this capacity. Studies that focus on environmental management often found that factors such as
entrepreneurial leadership (Heikkila and Gerlak 2005; Thomas 2002; Blomquist 1992), trust and
social capital (Leach and Sabatier 2005; Lubell 2007), scientific knowledge about the problem
(Lubell 2005; Weber 1998), and shared norms and ideas (Weber 2009) determine the success of
a locality’s initiatives on collective action.
144
When these factors are mapped onto the local contexts within which the communities
create their own, self-governing institutions such as CSDs from the grassroots level, we can
reconfigure them into the following three variables: 1) the clarity of the problem statement; 2)
the extent to which local leadership is prepared for new management and governance; 3) the
degree to which the new institution is supported by their citizen residents. Prior research also
suggests that communities need to establish their own capacity for a smooth transition when they
experience institutional change process (Ostrom 1990; Daniels and Walker 2001; Thomson and
Perry 2006).
One of the most glaring managerial and governing challenges associated with such a
grassroots approach is the citizens’ low level of expertise in policy issues. By contrast, a county
system can be more effective because the county government can utilize a high level of technical
expertise, resources, and economies of scale over many of their service areas. If the scale and
characteristics of management are things that community leaders have no prior experience
with—for example, building a large-scale plant and financial planning and management skills—
the county system would be a better institutional choice. Communities therefore need to develop
the essential capacity before they create new institutions as governing authorities.
In the context of CSDs, we would expect that communities with clear problem
statements, prepared civic managerial leadership, and strong public support would manage an
institutional transition process more smoothly and fully take advantage of new representational
structures, which more closely aligns with community boundaries. Using the property value as
an outcome variable that captures the performance of CSDs, this leads to the two following
hypotheses.
145
H1. The overall impact of community services districts on property values will be greater than
that of county authorities.
H2. The impact of community services districts will be conditional on a community’s governing
capacity, which includes the clarity of the problem statement, the preparedness of civic
managerial leadership, and support from citizen residents.

Methods and Data
This study uses a mixed-methods research design. To test the first hypothesis, first, a quantitative
analysis is used to measure the impact of CSD creation on the community’s housing prices,
which serve as a proxy for service quality. Then, the subsample analysis dives deep into each
community that created a CSD and its impact on housing prices. To test the second hypothesis,
the study then uses a qualitative case study approach to compare the communities that showed
divergent trajectories in the impact of CSD creation. The qualitative data were gathered during
the summer of 2017 from the author’s site visits and field interviews with the managers of
special districts. All interviews were semistructured, and each lasted for approximately one hour.
Transcription services were used to accurately interpret the interviewees’ narratives. The list of
interviewees and the questions asked are reported in the Appendices B and C, respectively.
The data for quantitative analysis come from the same sources as reported in the previous
chapter (chapter 4). Namely, the administrative records on housing transactions were obtained
from DataQuick (now CoreLogic), and special district GIS boundary files were collected from a
local formation commission agency in each county. Housing transaction data cover the years
146
between 1988 and 2012, and the analysis considers the event of special district creation. As a
result, seven special districts in two counties constituted the final sample.
Using the same equation (1) as in chapter 4, the analysis measures the difference-in-
difference (DID) impact of CSD creation on home sale prices. The treated group is housing
transactions within the CSD community boundary, and the control group is the transactions from
the surrounding areas within a few miles. As was the case in chapter 4, an alternative control
group is also drawn from the nearest county service areas that are still active and provide similar
services.

Descriptive Data
Table 5-1 shows the mean statistics and t-test results for home characteristics between the paired
treated and control groups before and after the districts were formed. If special districts improved
services, we would expect that their housing price relative to that in the control group would
increase after they were formed compared with their price before formation. This is the case for
Helendale, Phelan-Pinon, San Miguel, and Ground Squirrel Hollow. After the districts were
formed, the treated homes in Helendale show higher prices, and in Phelan-Pinon Hills and San
Miguel, the mean difference in the prices decreases. In Ground Squirrel Hollow, the difference
converges.
Other districts, however, deviate from this pattern. For example, the mean difference in
the home prices in the Chino Valley and Los Osos increased, which suggests that the creation of
the districts may have impacted property values negatively. The treated group in the Los Osos
community shows little qualitative differences in the home structural characteristics compared to
147
those of the control group before the district was established, whereas after the district was
formed, they were sold at a significantly discounted price. Similarly, in Heritage Ranch, there is
no significant difference in home prices before the district was formed, yet the treated group sold
at a significantly discounted price compared with that in the control group after the district was
established. These patterns in Los Osos and Heritage Ranch imply inefficiency under the special
district governance structure.

[Table 5-1]

Quantitative Analysis
As shown in chapter 4, Figure 4-5 presents the DID analyses from the pooled sample across the
seven communities. The results show that the creation of a CSD increases home sale prices by 12
percent compared to prices in the surrounding areas. The effect is not statistically significant in
the year of district formation, but it becomes significant a year later. Additionally, the effect
remains stable for the next three years, the time window the data permit for all seven districts. To
test the robustness of these results, the alternative control group, which draws the home
transactions from the nearest county service areas, was used for the analysis. Figure 4-6 shows
this analysis. The DID estimates do not change under this alternative model, indicating the
robustness of the findings. These data analyses confirm the first hypothesis that, on average,
CSDs generate greater impacts on their places than county authorities do.
[Figure 4-5]
[Figure 4-6]
148

To examine whether the effects were universally positive across the districts, the next
exercise repeats the analysis for each district (i.e., subsample analysis). Figure 5-2 shows that
most special districts in the data had positive impacts on their property values. By contrast, two
districts (Los Osos and Heritage Ranch) had negative impacts. To test the robustness of this
heterogeneity in impacts, the next subsample analysis uses the alternative control group drawn
from the county service areas that are nearest to each special district. Note that two districts were
excluded from this analysis in Figure 5-3 because they lacked a comparable county service area
for their service functions. The results remain robust under this approach: while other districts
show positive impacts, impacts are negative for the Los Osos and Heritage Ranch.

[Figure 5-2]
[Figure 5-3]

Qualitative Case Study Findings
Using the qualitative interview data, this section presents a case study analysis that compares two
communities—Helendale and Los Osos—that are highly similar in their demographic profiles
but exhibited contrasting impacts on their property values.
38
The district-specific analysis
showed that the establishment of the special district increased property values in Helendale but

38
These demographic profiles include population density, average household income and household size,
race and ethnicity, education level, unemployment and poverty rate, homeownership rate, and percent of
the elderly population, among others. For more details, see Table 5A-1 in the Appendix A.
149
decreased property values in Los Osos. These markedly different results were consistently
obtained regardless of which control group was used (see Figure 5-2 and 5-3).
In brief, the case study demonstrates the relevance of the local contexts within which the
special districts are created and operate. Specifically, the indicators of a community’s governing
capacity shape the success of special district management. As hypothesized, the key indicators
identified here are the clarity of the problem statement, the readiness of local managerial
leadership, and public support. Interestingly, the analysis also finds the importance of
intergovernmental coordination between a special district and a county and state agency as one
of the critical determinants of success. This result suggests that while special districts are
independent authorities, they are also embedded within the broader system of federalism, and as
such, even for core services, these districts often need to coordinate with county and state
authorities. Consistent with the quantitative findings, the case study analysis clearly suggests that
the creation of CSDs, despite bringing innovation to the representational structure, it does not
necessarily translate into greater impacts on the places. Rather, communities need to be ready
and develop the capacity to resolve the complex problems that arise from creating and managing
a new governing institution.

Success Story of the Community Services District in Helendale
The Helendale community is located in the Mojave Desert between the City of Victorville and
Barstow in San Bernardino County. As it is located in a rural desert area, water has been the
primary and most significant resource for its 6,000 residents. Until 2006, as one of the county
service areas (CSA 70B and C), Helendale’s water policy was governed and managed by San
Bernardino County. Under the county provision, the residents recalled that they regularly
150
experienced water service system failures. The service lines that carried the water from the main
distribution line to households reached their life span and were rapidly failing. As a result,
leakages often occurred. To make things worse, the county’s responses to the residents’ reports
of leaks took a long time—often between six and eight hours—and the problem was only
temporarily fixed.
From the residents’ perspective, they were paying a great deal for 18 full-time county
employees working in their service area at the time, but the residents did not feel that they were
receiving any preventive maintenance services. This was not acceptable, either from a resource
efficiency standpoint or from a representative standpoint. This problem prompted the
community’s demand for local representation. As the problem statement was easy to define, the
community had an election in November 2006 to form its own CSD. Due to a strong and clear
problem statement, the district had a short transition period and began providing full services in
just four months with 8 full-time employees, fewer than the 18 employees under the county
management system. The general manager of the district recalled: “The community experienced
a leakage almost every day in the first year we took over. The first action we took was investing
$2 million to replace and fix the aged infrastructure, the work that would not have happened
under the county control.”
The Helendale CSD also negotiated its contract terms with a private waste hauler to
better meet the needs in the community. For example, under county control, the hauler used to
provide a household with one waste container and one recycling container a week. After the
district formed, it added one more recycling container for each house to support the community’s
recycling needs. Among the many programs the district negotiated and initiated, two programs
that have added convenience to their residents are a service for collecting bulky items (e.g.,
151
televisions or mattresses) and a green waste drop-off program. Now, the residents do not need to
travel all the way down to the county landfill whenever they cut trees; they simply drop them off
in the backyard of the district office closer to their properties. The district has served its residents
as a collection point, leveraging all of these services from a community-centric perspective,
which was perhaps impossible under county control.
Having a governing body elected from the residents’ own community also meant their
ability to change their water provider. Traditionally, the community purchased water imported by
the Mojave Water Agency for certain amounts allotted to their area. As the yearly available
supply decreased, the cost rose. The district board considered the community’s greater present
and future needs for water from a local source for a more reliable supply of water. Recently, the
district found a local provider at a more affordable price and paid $4 million for 814 acre-feet
water rights. Given that an acre foot is approximately the amount of water that two households
use in a year, this was a very large investment. This action brought the district’s water-use deficit
to within 81 acre-feet of demand (Victorville Daily Press, 2014). To the residents of the
Helendale community, establishing a CSD was meant to maximize their benefits from public
service delivery. The clarity of the problem statement, prepared civic leadership and
management, and continuing support from the public explain the successful trajectory of
Helendale after the establishment of the CSD, as evidenced by the positive effects on property
values in Figure 5-2 and 5-3.

Failure Story of the Community Services District in Los Osos
From the beginning, the Los Osos CSD grew out of a controversy between the community and
the county and state governments. The community used to treat the wastewater with its own
152
septic tanks, but the discharged wastewater contained higher levels of nitrates than allowed by
the regional water quality control board, a state agency. Thus, the discharged wastewaters were
severely polluting the environment (Central Coast Regional Water Quality Control Board, 2004).
Since 1983, the regional water quality control board had prohibited some areas in the community
from using septic tanks (State Water Resources Control Board Resolution 84-13, 1984). To
address this problem, San Luis Obispo County proposed building a wastewater treatment system.
The proposal soon faced strong opposition from community organizations due to the high cost
anticipated to complete the project (San Luis Obispo County, 2011; CMD Group 2018). In 1998,
the community decided to form its own local government—the Los Osos CSD—to build the
sewer system on its own and provide services under community control.
However, even after the formation of the district, public opinion among the residents was
still divided as to whether and where to build the sewer plant. Unfortunately, the agenda of the
district board members at the time failed to manage the public dispute. Consequently, the
residents called for a recall election in 2005 to remove some board members from office even
before their terms expired (Los Osos CSD Recall Committee, 2005). The board nonetheless
adhered to their original plan for the construction of the sewer plant at the designated location. In
the recall election, three of the five board members who had pushed the agenda were eventually
replaced. The new board stopped the ongoing project for sewer plant construction but instead
proposed a new project that would face less opposition. However, there were at least two
problems with the new board’s action. First, they ignored a warning from the regional water
control board that their new plan would not address the environmental concerns in the required
manner. Second, there was a fiscal management problem in the conduct of the new project under
the new board members. The district first defaulted on a state loan, which was the primary source
153
from which the district borrowed the new sewer construction project, and it later filed for
bankruptcy protection (United States Bankruptcy Court, 2006).
39

Consequently, the power to provide sewers was taken from the state and given to the
county government. During the process for the proof of claim in bankruptcy, the district had to
show proof of fiscal solvency and sold its authority to collect solid waste to a private company,
which was later sold and taken over by the county government. As a result, as of 2018, the Los
Osos CSD provides only services for water and drainage, while the other two existing services—
sewers and solid waste collection—are provided by the county government. The historical
incidents and the ultimate bankruptcy and loss of service provision power in the Los Osos
community alert us that, like other municipal governance, the community-control governing
structure also has a potential pitfall of poor management and accountability. While the problem
statement seemed clear initially, the intergovernmental coordination was weak in solving the
complex service problem. Additionally, unorganized civic leadership, subsequently poor
management, and divided public opinion explain the unfortunate trajectory of Los Osos after the
establishment of the CSD, as evidenced by the negative effects on property values in Figure 5-2
and 5-3.

Discussion and Conclusion
This study examines the impacts of CSDs, a class of special districts created as an instrument for
neighborhood governance in unincorporated places in California. Notably, the creation of these
districts is typically sparked by a service quality and infrastructure maintenance problem under

39
The district was not able to pay the contractors it hired to build the failed sewer project. The estimated
debts were $45 million (The Tribune, San Luis Obispo. April 3, 2012).
154
county management. Hence, by the nature of the motivation underlying their choice to opt out of
the county’s service control and the prediction from general public choice theory, the CSDs are
expected to improve neighborhood-level services. To measure such impacts, the analysis utilizes
the CSD creation event in Southern California, and it uses housing price as a metric for
performance measurement. Using DID methods, the analysis finds that the creation of a CSD
enhances residential property values by 12 percent compared to the values in the surrounding
areas still managed by county authorities.
The essay, however, also argues that such an effect would be conditioned on local contexts,
specifically whether the communities have developed the necessary governing capacity, such as
the clarity of the problem statement, managerial local leadership, and support from citizen
residents. Indeed, when the quantitative analysis focuses on each district, the effects varied
across the sample districts. Using qualitative data collected from the field site visits and
interviews with practitioners, the deeper case study analysis then compares two districts that
show markedly contrasting impacts on their property values. The case study finds that the more a
community has a clear problem statement, managerial leadership, and public support, the more
likely it is to benefit from the creation of special districts to improve services.
By contrast, if a community lacks such conditions, the analysis suggests that even if it
creates a special district, the existence of the institution per se would not necessarily guarantee
improved services. Rather, as shown in the case of Los Osos, such a governance model could be
detrimental to the fiscal welfare of a community. Hence, it seems that communities can take
advantage of institutional innovations in special districts if and only if they have developed the
required governing capacity.
155
This study has several implications for public administrators, managers, and citizens who
are interested in community-centered governance and the likely improvement in local service
provision. First, evidence from the aggregate analysis suggests that the design and choice of local
governance structure are relevant to the performance of public organizations. Specifically, a
governance structure where representation is aligned with the scale of the problem benefits
jurisdictions more than a structure where representation is delegated to a larger and higher
authority. The study also reveals that citizens can be at the center of the local governance
restructuring process through the creation of new institutions such as CSDs.
Second, the challenges identified in such community-centered governance suggest that the
process of institutional creation should involve an extensive discussion with other stakeholders
of governmental authorities for a more rigorous evaluation and approval. The current standard
for the formation of CSDs in California is that commissioners of an LAFCO review these
processes with their agency staff. As long as the communities demonstrate their financial
capacity to create and run special-purpose districts, the reviews are generally approved for an
election, which then requires majority support from registered voters in communities. However,
once these governments are created, it is hard to dissolve them even though they reveal problems
in accountability and service performance (Little Hoover Institution, 2017). Elected officials and
public administrators should oversee the formation process more carefully so that only the
communities that demonstrate sufficient capacity and support are allowed to create new special
districts as their governing authorities.
Last, public managers and citizens in special districts should understand the broader
governmental system and the importance of intergovernmental coordination. Such attention and
understanding are particularly urgent when they have to comply with environmental regulations
156
and mandates from the county and state authorities or if they are planning a large infrastructure
project that could generate spillovers to other jurisdictions in the region. Even if the power of
municipal services is delegated to communities for special districts, it does not mean that the
exercise of service power should always be independent of other governmental authorities. In the
end, as the case study illustrates, special-purpose governments are embedded in a broader federal
governmental system in which communities should confer with higher-level local and state
governments to manage their complex service environment.
The case study approach used here constrains the ability to generalize the findings.
However, the fewer cases allow for disaggregate quantitative analysis, which provides evidence
that the impact of special district governance is not unconditional as Mullin (2008) argues. This
finding is important, as it often is overlooked in aggregate analysis. Beyond the quantitative
analysis, the fewer cases also allow us to identify the factors of a community’s governing
capacity and explore whether they determine the success and failure of CSDs.
As Skelcher (2006) notes, one of the promising avenues for future study is a varying type
of governing structures within special districts. While this study focuses on neighborhood
governance in small and medium-sized local jurisdictions that are primarily in unincorporated
areas, some special districts provide services across a metropolitan area with their governing
board members representing each municipality in the jurisdiction. Others have local board
members appointed by the city and county legislature. Examining the benefits and costs of
having special-purpose governments under different governing arrangements and comparing
them to those found in this study may help clarify the consequences of this understudied
governmental entities in a broader sense than what this essay focuses on. As more communities
are choosing special districts as their new institutions to improve service quality, public
157
administrators and government managers need to pay greater attention to them and understand
the implications of this trend.

158
References
Berman, Evan M., and Jonathan P. West. 2012. Public Values in Special Districts: A Survey of
Managerial Commitment. Public Administration Review 72(1): 43-54.
Berry, Christopher. 2008. Piling on: Multilevel Government and the Fiscal Common‐
Pool. American Journal of Political Science 52(4): 802-820.
Blomquist, William. 1992. Dividing the Waters: Governing Groundwater in Southern
California. ICS Press Institute for Contemporary Studies.
Bourdeaux, Carolyn. 2004. A Question of Genesis: An Analysis of the Determinants of Public
Authorities. Journal of Public Administration Research and Theory 15(3): 441-462.
Bovaird, Tony. 2007. Beyond Engagement and Participation: User and Community
Coproduction of Public Services. Public Administration Review 67(5): 846-860.
Boyne, George A. 1992. Local Government Structure and Performance: Lessons from America?
Public Administration 70(3): 333-357.
Box, Richard C. 1997. Citizen Governance: Leading American Communities into the 21st
Century. Thousand Oaks, CA: Sage Publications.
Burns, Nancy. 1994. The Formation of American Local Governments: Private Values in Public
Institutions. Oxford University Press on Demand.
Carter, David P., Aaron Deslatte, and Tyler A. Scott. 2018. The Formation and Administration of
Multipurpose Development Districts: Private Interests Through Public Institutions. Perspectives
on Public Management and Governance 2(1): 57-74.
Central Coast Regional Water Quality Control Board. 2004. Frequently Asked Questions
Regarding Water Quality Issues in Los Osos Community. Retrieved April 16, 2018 from
https://www.waterboards.ca.gov/rwqcb3/water_issues/programs/wmi/docs/los_osos/faq6.pdf
Cooper, Terry L., Thomas A. Bryer, and Jack W. Meek. 2006. Citizen‐centered Collaborative
Public Management. Public Administration Review 66(s1): 76-88.
Chen, Chung-An, Evan M. Berman, Jonathan P. West, and Robert J. Eger III. 2013. Community
Commitment in Special Districts. International Public Management Journal 16(1): 113-140.
CMD Group Website 2018. Municipal Water and Wastewater Facility Project Case Study in
California. Retrieved on April 16, 2018 from http://www.cmdgroup.com/building-types/water-
treatment/california/projects/1000213240/
Craw, Michael. 2017. Institutional Analysis of Neighborhood Collective Action. Public
Administration Review 77(5): 707-717.
Daniels, Steven E., and Gregg B. Walker. 2001. Working through Environmental Conflict: The
Collaborative Learning Approach. Praeger. Westport, CT.
159
Deslatte, Aaron, Tyler A. Scott, and David P. Carter. 2019. Specialized Governance and
Regional Land-use Outcomes: A Spatial Analysis of Florida Community Development Districts.
Land Use Policy 83: 227-239.
Eger III, Robert J. 2005. Casting Light on Shadow Government: A Typological
Approach. Journal of Public Administration Research and Theory 16(1):125-137.
Foster, Kathryn A. 1997. The Political Economy of Special-Purpose Government. Georgetown
University Press.
Goodman, Christopher B. 2015. Local Government Fragmentation and the Local Public Sector:
A Panel Data Analysis. Public Finance Review 43(1): 82-107.
Heikkila, Tanya, and Kimberley Roussin Isett. 2007. Citizen Involvement and Performance
Management in Special‐purpose Governments. Public Administration Review 67(2): 238-248.
Heikkila, Tanya, and Andrea K. Gerlak. 2005. The Formation of Large‐scale Collaborative
Resource Management Institutions: Clarifying the Roles of Stakeholders, Science, and
Institutions. Policy Studies Journal 33(4): 583-612.
Hendrick, Rebecca M., Benedict S. Jimenez, and Kamna Lal. 2011. Does Local Government
Fragmentation Reduce Local Spending? Urban Affairs Review 47(4): 467-510.
Ho, Alfred, and Paul Coates. 2004. Citizen-initiated Performance Assessment: The Initial Iowa
Experience. Public Performance & Management Review 27(3): 29-50.
Hong, Sounman. 2015. Citizen Participation in Budgeting: A Trade‐Off between Knowledge and
Inclusiveness? Public Administration Review 75(4): 572-582
Irvin, Renee A., and John Stansbury. 2004. Citizen Participation in Decision Making: Is It Worth
the Effort? Public Administration Review 64(1): 55-65.
Leach, William D., and Paul A. Sabatier. 2005. Are Trust and Social Capital the Keys to
Success? Watershed Partnerships in California and Washington. Swimming Upstream:
Collaborative Approaches to Watershed Management pp. 233-258. Edited by Sabatier, Paul A.,
Will Focht, Mark Lubell, Zev Trachtenberg, Arnold Vedlitz, and Marty Matlock. Cambridge,
MA: MIT Press.
Little Hoover Institution. 2017. Special Districts: Improving Oversight & Transparency.
Available at http://www.lhc.ca.gov/sites/lhc.ca.gov/files/Reports/239/Report239.pdf
Los Osos Recall Committee. Aug 18, 2015. Retrieved on April 16, 2018 from
http://www.fppc.ca.gov/content/dam/fppc/documents/advice-letters/1995-2015/2005/05145.doc
Lubell, Mark. 2007. Familiarity Breeds Trust: Collective Action in a Policy Domain. Journal of
Politics 69(1): 237-250.
Mullin, Megan. 2008. The Conditional Effect of Specialized Governance on Public Policy.
American Journal of Political Science 52(1): 125-141.
Mullin, Megan. 2009. Governing the Tap: Special District Governance and the New Local
Politics of Water. Cambridge, MA: MIT Press.
160
Ostrom, Elinor. 1990. Governing the Commons. Cambridge University Press.
Ostrom, Elinor. 2009. Understanding Institutional Diversity. Princeton University Press.
San Luis Obispo County. 2011. Public Comment for Item A-10 to the Clerk of the Board of
Supervisors. Retrieved on April 16, 2018 from
http://slocounty.granicus.com/MetaViewer.php?view_id=2&clip_id=1020&meta_id=202436
Skelcher, Chris. 2006. Does Democracy Matter? A Transatlantic Research Design on
Democratic Performance and Special Purpose Governments. Journal of Public Administration
Research and Theory 17(1): 61-76.
State Water Resources Control Board Resolution 84-13. 1984. Retrieved on April 16, 2018 from
https://www.waterboards.ca.gov/centralcoast/water_issues/programs/los_osos/docs/master_docs/
1984_01_19_res_84-13_000.pdf
Thomas, Craig W. 2002. Bureaucratic Landscapes: Interagency Cooperation and the
Preservation of Biodiversity. MIT Press.
Thomson, Ann Marie, and James L. Perry. 2006. Collaboration Processes: Inside the Black
Box. Public Administration Review 66(s1): 20-32.
Tribune, San Luis Obispo. April 3, 2012. “Federal judge upholds CSD bankruptcy plan in Osos.”
Retrieved on April 16, 2018 from
http://www.sanluisobispo.com/news/local/article39199407.html
United States Bankruptcy Court. 2006. Court News “Los Osos Community Services District
Files For Chapter 9 Protection.” Retrieved on April 16, 2018 from
http://www.cacb.uscourts.gov/sites/cacb/files/documents/publications/Sept_Oct_1.pdf
Victorville Daily Press. Sep 9, 2014. CSD purchases water rights for $4.07 million. Available at
http://www.vvdailypress.com/article/20140909/NEWS/140909811
Weber, Edward P. 2009. Explaining Institutional Change in Tough Cases of Collaboration:
“Ideas” in the Blackfoot Watershed." Public Administration Review 69(2): 314-327.

161
Table 5-1. Mean Statistics and T-test of Housing Transaction Characteristics
Group
Before
/After
Log of
sale price
Log of
lot size
Log of
living area
Bed
rooms
Baths
Age
(at sale)
Obs
All Districts

Treated Before 11.47*** 9.65** 7.45*** 3.18*** 2.21*** 9.06*** 26,117
Control Before 11.59 9.61 7.33 3.02 2.12 15.69 23,788
Treated After 12.42*** 8.72*** 7.47*** 3.35*** 2.39*** 16.31*** 70,782
Control After 12.12 9.18 7.31 3.09 2.19 26.67 57,869
Chino Valley

(Control homes
within 1.5 miles)
Treated Before 11.85*** 8.69*** 7.39*** 3.32*** 2.33*** 8.95*** 7,860
Control Before 11.68 9.13 7.30 3.09 2.24 14.49 9,088
Treated After 12.45*** 8.60*** 7.48*** 3.41*** 2.43*** 15.70*** 60,925
Control After 12.10 9.09 7.29 3.10 2.19 27.47 47,993
Phelan-Pinon
Hills
(Control homes
within 2.5 miles)
Treated Before 11.27*** 11.41*** 7.55*** 3.36*** 2.27*** 7.15*** 7,521
Control Before 11.63 10.25 7.47 3.04 2.15 15.13 6,484
Treated After 11.88*** 11.40*** 7.58*** 3.40*** 2.31*** 14.87*** 1,799
Control After 12.05 10.30 7.47 3.07 2.16 26.14 1,239
Helendale

(Control homes
within 10 miles)
Treated Before 11.08 9.09*** 7.52*** 3.07*** 2.17*** 7.04*** 6,654
Control Before 11.07 9.69 7.13 2.94 1.77 18.16 4,210
Treated After 11.91*** 9.11*** 7.53*** 3.10*** 2.20*** 17.57*** 1,947
Control After 11.49 9.48 7.26 3.20 1.98 20.50 2,003
Los Osos

(Control homes
within 5 miles)
Treated Before 11.85*** 8.77*** 7.26*** 2.69*** 1.94*** 18.57*** 2,962
Control Before 12.00 8.37 7.29 2.72 2.09 20.32 2,550
Treated After 12.64*** 8.78*** 7.27*** 2.71*** 1.96*** 29.06* 3,501
Control After 12.82 8.47 7.34 2.81 2.18 29.87 3,110
Heritage
Ranch
(Control homes
within 9 miles)
Treated Before 11.32 9.04*** 7.33*** 2.84*** 2.16*** 4.68*** 290
Control Before 11.21 10.53 7.52 3.11 2.36 9.81 290
Treated After 12.01* 9.13*** 7.39*** 2.91*** 2.20*** 12.57*** 1,752
Control After 12.06 10.22 7.45 3.11 2.29 16.33 2,936
San Miguel

(Control homes
within 5.5 miles)
Treated Before 11.05*** 9.22*** 7.06*** 2.77*** 1.58*** 26.18*** 286
Control Before 11.78 12.79 7.60 3.45 2.50 11.18 574
Treated After 12.38*** 8.87*** 7.23*** 3.15*** 1.89*** 18.14 605
Control After 12.97 12.54 7.66 3.50 2.58 18.42 377
Ground Squirrel
Hollow
(Control homes
within 2 miles)
Treated Before 11.52*** 11.20 7.48 3.22 2.11 3.18*** 544
Control Before 11.71 11.25 7.45 3.22 2.13 9.94 592
Treated After 12.71 11.10** 7.53* 3.41 2.16* 9.16*** 253
Control After 12.71 11.22 7.48 3.42 2.25 17.97 211
Notes: CSD is Community Services District. T-test was conducted for each of the variables in the treated group,
compared to the characteristics of the control group during the periods before and after district formations.
*** p < 0.01, ** p < 0.05, * p < 0.1.

162
Figure 5-1. Administrative Process to Create Community Services District

Figure 5-2. Impact of Special Districts on the Property Values

Notes: The figure shows the results estimating equation (1) in chapter 4 for each special district
community with standard errors clustered by properties in the first four years since the district formation.
Point estimates represent the coefficients on the DID interaction term (SD*Post_4) and 95% confidence
intervals are reported. House transaction sample sizes are in parentheses.

163
Figure 5-3. Robustness Analysis Using County Service Areas as Control Group

Notes: The figure shows the results estimating equation (1) in chapter 4 for each special district
community with standard errors clustered by properties in the first four years since the district formation.
Point estimates represent the coefficients on the DID interaction term (SD*Post_4) and 95% confidence
intervals are reported. House transaction sample sizes are in parentheses.

164
Chapter 5 Appendix A. Tables
Table 5A-1. Demographic Profiles of Communities

Demographics/
Communities
Chino-
Chino
Hills
Phelan-
Pinon
Hills
Helend
ale
County-wide
(San
Bernardino)
Los
Osos
Heritage
Ranch
San
Miguel
County-
wide
(San
Luis
Obispo)
Total Population 159,004 20,688 5,992 2,094,769 15,388 2,910 2,461 276,517
Population Density 2,139 224 1,159 104 1,206 283 1,443 84
Average HH Size 3.33 2.97 2.81 3.33 2.35 2.56 3.35 2.51
Elderly Population (%) 9 15 23 10 20 16 4 17
People with at least
College Education (%)
64 46 68 52 79 64 42 70
Median HH Income ($) 85,865 49,545 69,157 53,433 62,857 71,641 49,274 60,691
Average HH Income ($) 100,367 62,803 72,458 69,064 82,459 92,442 60,393 80,404
Unemployment Rate (%) 9 15 12 13 8 4 13 7
Poverty Rate (%) 9 17 11 19 10 3 19 15
Non-Hispanic White (%) 28 64 81 31 80 81 40 70
Non-Hispanic Black (%) 5 3 1 8 0.24 1 0 2
Non-Hispanic Asian (%) 21 3 1 7 3 1 0.33 4
Non-Hispanic Others (%) 4 2 5 3 5 5 9 3
Hispanic (%) 42 28 12 51 12 12 51 22
Housing Units 46,054 8,036 2,677 705,962 6,916 1,962 752 118,806
Vacancy rate (%) 5 14 20 13 6 42 3 13
Home Owners (%) 75 81 68 60 65 72 44 57
Renters (%) 25 19 32 40 35 28 56 43

Source: American Community Survey 2011-2015 5-year Estimate.
The information was not available for Ground Squirrel Hollow Community.

165
Chapter 5 Appendix B. Individuals Interviewed

San Bernardino County
Local Agency Formation Commission in San Bernardino County
Kathleen Rollings-McDonald, Executive Director
Chino Valley Fire District
Tim Shackelford, Fire Chief
Crestline Sanitation District
Rick Dever, General Manager
Ronald Scriven, Lead Plant Operator
Dawn M. Grantham, Accountant/Bookkeeper
Helendale Community Services District
Kimberly Cox, General Manager

San Luis Obispo County
Local Agency Formation Commission in San Luis Obispo County
David Church, Executive Director
Los Osos Community Services District
Renee Osborne, General Manager
Heritage Ranch Community Services District
Scott B. Duffield, General Manager
San Miguel Community Services District
Rob Roberson, Fire Chief
Ground Squirrel Hollow Community Services District
Dan Gilmore, General Manager

166
Chapter 5 Appendix C. Interview Questionnaire

FOR SPECIAL DISTRICT MANAGERS
Basic background questions
1. Please tell me when you were hired to the current position.
2. Before this position, have you worked for the current special district? If so, please tell me in which
position and from when.
3. Before working for the current special district, have you worked for any other special districts or
municipalities?
4. If yes, please tell me what work experience it was.

Basic special district formation questions
1. Please tell me what the motivations and reasons were in the formation of your special district.
2. Currently, what services does the special district provide? How many parcels/houses or
households/populations are served?
3. Please tell me how the district services are financed
4. Before the special district was formed, by whom the services were provided? Was it the county?
5. I want to understand the various choices the residents in your special district could have had at the time
of the district formation. Was 1) being annexed to a neighboring incorporated city, or 2) having an
agreement to receive service from a neighboring city, one of the choices? If not, please tell me why.
6. Do you think the special district was formed, in part, by expectation to see appreciation in real estate
values in the community?
7. Who initiated the formation efforts? Were they mainly the residents with needs, or were there any other
actors in leading this?

Governance and management questions
1. Is the special district you serve an independent institution? Or is it dependent to county government?
2. Please tell me the major differences between independent and dependent special districts.
3. Please explain the decision-making process in your district. For any major decisions regarding the
service provision (e.g., raising the service fee), how are those decisions made?
4. Please describe the relationship between the general manager and the board members. What is the
hiring process for the manager and what is the decisional boundary of the manager’s authority? How are
responsibilities divided?
5. Please tell me if there is any particular policy or process that the board uses to improve transparency
and responsiveness to the residents’ needs and preference.
6. Please describe any event or issues if the residents did not favor the actions or decisions of the board
since the district was formed.
7. In a situation where the board members think the district needs to increase the service fee but the voters
are resistant, how does it get resolved?

Service delivery questions
1. What do you think the advantages are in the use of special districts in service provision? Why?
2. If there are any disadvantages in the use of special districts for service provision, what do you think
those can be?
3. Do you think that special districts provide services cost effectively and efficiently? If yes, what
evidence do you have on this claim?
4. How do you compare the quality and responsiveness of service provision between the special districts
and county government?

Supplemental questions
167
1. Please tell me which neighboring community or County Service Area may be a good comparison group
in evaluating your district’s specific service performance.

FOR EXECUTIVE DIRECTORS OF LAFCO

Basic background questions
1. Please tell me when you were hired to the current position.
2. Before this position, have you worked for any other LAFCO or municipalities?
3. How many years have worked for San Luis Obispo County LAFCO?
4. Please tell me the role and mission of SLO LAFCO and your role in particular

Special district formation questions
1. Please tell me what the general motivations and reasons are for the formation of special districts.
2. Do you think it is the case that the special districts are formed, in part, by expectation to see real estate
values in the jurisdiction?
3. Who initiates these formation efforts? Are they mainly the residents with needs, and who are other
actors in leading this?
4. Many special districts were formed in the first half of the 20
th
century. Are there any reasons many of
them were established long time ago?

Governance questions
1. Please tell me the major differences between dependent and independent special districts.
2. I understand that SLO County uses special district arrangement for the service provision to
unincorporated areas. Do all unincorporated areas in SLO County have dependent special districts, if they
don’t have their own independent one? If not, who provide the services there?
3. Please describe the governance structure and management in the dependent special districts. Is it
County Board of Supervisors who makes a major decision in these jurisdictions? Is there any
intermediate-level board or agency between Board of Supervisors and residents?

Annexation questions
1. I understand that unincorporated in a county has at least three choices regarding their service provision.
Is this a true statement? If not, why? And what other choices can there be?
1) provided by county 2) form their own special districts 3) consider annexation to a neighboring
incorporated city or receiving the service from the neighboring city under a contract.

Service delivery questions
1. Do independent special districts use contracting-out to private entities for the service delivery?
2. In general, are the services provided to residential properties only or to commercial as well?
3. Many special districts claim that, with their own local control, their service delivery is more cost-
effective and efficient. Do you agree on this?

Supplemental questions
1. Please tell me for the following six special districts, which neighboring community or County Service
Area may be a good comparison group in evaluating your district’s specific service performance.
1) San Miguel CSD 2) Los Osos CSD 3) Heritage Ranch CSD 4) Avila Beach CSD 5) Ground
Squirrel Hollow CSD 6) Independence Ranch CSD
2. If there is any change you might want to see in the future regarding the use of special districts in
service provision and delivery, what those could be?

168
Chapter 6. Conclusion and Policy Implications

Conclusion

The provision of public goods is essentially a realm of market failure. Hence, government and
the public are key players that make investments more productive to enhance the welfare of our
society. Additionally, where these investments are allocated and who reaps the benefits and bears
the costs are not only questions of policy analysis and public management but also political
questions that are relevant to our lives. Further, these questions are directly related to our
residential choice dynamics and quality of life.
Given the multilevel governance structure and the substantive authority that local and
regional governments in the United States have in public goods allocation, it is essential to
understand what factors influence the provision of public goods and the spatial allocation of
public investments. This research sought to contribute to our understanding of the provision of
local public goods. In doing so, I asked the following three related questions: 1) Across and
within metropolitan areas, does the design of institutional rules in regional governments affect
the equitable distribution of public investment between cities? 2) At the school district, city, and
metropolitan area levels, does demographic structure—particularly, the racial structure of income
inequality—dampen the levels of public investment in various urban public goods? 3) At the
community level, does the type of local governance determine the quality of local public goods?
More specifically, does the creation of a community services special district provide citizen
residents with higher-quality services than are available when the services are managed and
provided by county authorities?
169
The empirical findings of this dissertation show the important roles that these governance
and institutional forces play in the local and regional provision of public goods in the United
States. The first essay (chapter 2) demonstrates that the institutional design of regional
governments affects the spatial allocation of public goods across cities in metropolitan areas.
Specifically, the study shows that cities with greater voting power in regional governments
receive more public transportation investments than other cities with less power do. The results
also show that when the power is heavily concentrated in the largest cities, the other cities in the
regions do not have any meaningful influence on the allocation of public investment due to the
dominant influences exercised by the largest cities.
The second essay (chapter 3) illustrates that the local demographic structure also shapes
the overall levels of investment in various urban public goods. In particular, this chapter
demonstrates that the growing income inequality between racial groups shrinks public spending
on an array of goods, including education, health, hospitals, police protection, and parks and
recreation facilities. These results were obtained consistently across different levels of local
jurisdictions, including school districts, cities, and localities in metropolitan areas and even
states. In contrast, other demographic structures, such as racial diversity and income inequality,
were not shown to impact the provision of public goods. My research suggests that inequality in
the representation and political power of racial groups profoundly influences the provision of
public goods, and they are limited to certain geographic and political units.
The third essay (chapter 4) shows that the governance and institutional choice of local
communities markedly affect the quality of public goods provided. The creation of specialized
local governments that citizens in these areas (unincorporated and often rural areas) initiate
produces more net service benefits compared to the benefits of services provided by county
170
authorities. My research design allows me to capture and measure these net service benefits
through changes in housing prices. The quantitative analysis shows that the creation of CSDs
enhances residential property values by 12 percent compared to those in the surrounding areas.
The next chapter then builds on this finding and presents the heterogenous impacts on the
property values in each community in the data. While a majority of them indicate a positive
impact on property values, some districts showed otherwise. The key innovation of the second
essay (chapter 5) is the use of a qualitative case study, which utilizes in-person interview data
with the local special district managers in the sample. Through a qualitative case study approach,
this chapter further identifies a set of community governing capacities that determine the success
and failure in the management and implementation of CSDs. The factors identified as critical
determinants are the clarity of problem statements, local managerial leadership, support from the
public, and intergovernmental coordination (with county and state authorities). Hence, the essay
features a conditional theory of special districts for those created as an instrument for
neighborhood governance.

Policy Implications
Several policy implications are derived from the three essays in this dissertation. Below, I
present some thoughts from each essay and then synthesize them at the end.

What Determines Where Investment Goes? Regional Governance and the Role of Institutional
Rules and Power (Chapter 2)
171
Decisions on the allocation of public goods are made not only at the local level but also at the
regional level. Unraveling regional processes and determining how these processes contribute to
the spatial allocation of public goods across cities further enhance our understanding of the
geography of local public goods distribution. In the U.S. context, one of the ideal venues to
explore this question is regional governments (i.e., regional planning organizations) that allocate
federal, state, and local funds on transportation infrastructure projects at the metropolitan area
level. This chapter shows that the ways in which rules are designed in these intergovernmental
regional organizations—i.e., voting power allocation in the governing policy boards—affect
which cities receive more investments.
The purpose and existence of these regional governments have been for equity-based
cooperation among the municipalities in the region and such policy outcomes. Nonetheless,
the findings of this research show that the allocation of voting power is significant in the
distribution of public funds. For example, the results illustrate that if voting power is skewed
toward a few dominant players (i.e., cities), then the power held by the rest of the less
powerful cities does not influence the allocative decision-making process in organizations,
hence the concentration of public investments to a few cities.
The finding that the public investment that local jurisdictions receive is further
determined by the institutional design of their regional governments contributes to the
academic literature but also has a real policy implication: Policy makers need to consider the
design of institutional rules, specifically their voting power rules, to make them more evenly
distributed if the goal of these regional bodies is to improve equity considerations in the
allocation of public investments that they provide and receive.

172
Racial Inequality and the Local Provision of Public Goods in the United States (Chapter 3)
This research illustrates the role of demographic structure in influencing the local provision of
public goods. Because public goods are provided for and to citizens, scholars have examined
whether and to what extent their demographic characteristics—as an aggregate of citizens’
preferences—influence local and state government spending, including both general public
goods and specific welfare policies. In this line of research, two areas that have received much
attention from scholars are the relationship between income inequality/racial diversity and states’
welfare programs and local spending on public goods.
The innovation of this research is its focus on the intersection of income inequality and
racial diversity, namely, the racial structure of income inequality and the finding that this income
inequality between racial groups has a dampening effect on local governments’ investments in
various urban public goods.
Recently, media and scholars have highlighted a growing trend of income and wealth
inequality as a significant barrier to individuals’ upward mobility and society’s democratic
governance and social goods overall. What has been less understood is the implication of
society’s collective investments in various public goods, particularly from the viewpoint of the
structural composition of inequality between racial groups. The implication of the findings of
this research is straightforward: Policy makers need to close such gaps in racial inequality not
only on the grounds of racial equity or integration, as previously thought, but also from the
perspective of collective action dilemmas and the resultant negative impact on collective public
investments.
173
Further, the effects of budget cuts based on racial inequality that are more salient for
privately substitutable public goods (e.g., education, health, hospitals, police protection, and
parks and recreation) underscore the role of public goods in the era of growing inequality,
distinct from the role of private goods. These privately more substitutable public goods were
those most affected by growing racial inequality because richer households rely more on the
public goods in these policy areas in their consumption. It is thus these policy areas (i.e.,
privately more substitutable goods) that policy makers need to target in the allocative priority of
government subsidies for the poorer households who will be most affected by the dampening
effect of racial inequality.

Community Services Special District as a Local Governing Model (Chapters 4 and 5)
First, even though specialized governments have been the fastest-growing local governments in
the United States, few scholars have examined the consequences. Thus, despite their prevalence
in American local governance, policy communities have lacked evidence to assess their
effectiveness. While my analysis is limited in geographic scope (i.e., eight communities
primarily in unincorporated areas in Southern California), this research provides rigorous
evidence to evaluate the effectiveness of these growing new local governance institutions.
Second, the research question in this essay is directly related to the boundary and
designation of institutional power in the choice of various local governance modes. For example,
should we designate the municipal power of service provision tailored to local community
boundaries when considering the creation of new local institutions? My findings suggest the
answer is “yes” because at the local community level, the creation of these specialized
174
governments that citizens lead from the ground up tends to result in more benefits for the
community than the top-down county control does.
Third, however, designing community control institutions for quality services requires
careful examination in terms of whether the communities are prepared to leverage such
institutional choice. The case studies illustrate that factors such as the clarity of the problem
statement, prepared civic leadership, and supporting public opinion are critical to the success of
these local control institutions’ performance. In other words, without such conditions met,
communities may see lower-quality services for creating and managing these specialized local
governments.

Contribution to the Literature
This dissertation focuses on how the organization of public institutions and political power
affects the distribution of resources and value. Three of the dissertation chapters focus on the
structures of governmental institutions and specifically look at how these structures affect
outcomes on the ground. Two chapters examine special districts, which have become
increasingly prevalent in the United States. Especially, these essays focus on the emergence
of community services districts in California as an instrument for neighborhood governance,
created by citizen residents’ collective action. The other looks at metropolitan planning
organizations, which are important institutional vehicles in the allocation of transportation
resources.
In both cases, the dissertation employs mixed methods research design and finds that
structure matters, such that the distribution of resources and value are affected by how
175
institutions are organized. These papers are interesting and important contributions to the
field for several reasons. First, they represent significant efforts to address longstanding
questions in the field. Regarding the metropolitan resource planning, the question of how the
allocation of power is associated with the spatial allocation of resources is one that many
have highlighted. But it has proven difficult to directly study because of a lack of available
data. This was overcome in the current dissertation through the author’s diligent and tireless
effort. The spatial metropolitan planning organization data on transportation allocation is a
landmark resource for studying how regional governing bodies make decisions and how those
decisions affect localities.
Similarly, the use of special districts has grown dramatically over the past several
decades, yet there has been relatively little research attention given to them. Related to this,
each project resulted in a new dataset that allows for the pursuit of projects that heretofore
could not be undertaken. The high-quality data on the selected special districts will provide
opportunity for study that had not existed previously. While the existing views on special
districts were quite negative, this dissertation introduces a new perspective to the literature,
namely, a class of them that are created as a positive instrument for community governance.
Another unique contribution of special district study is that it features a quantitative analysis,
which shows that special districts can enhance property values in their jurisdictions. This
finding combined with the qualitative case study approach helps broaden our view on the
political economy of local governance.
Third, the findings challenge conventional wisdom and inform practice. With the
essay on metropolitan planning organization in particular, the findings ran counter to what
practitioners expected regarding the importance of institutional rules and political power
176
factors. The special district essay is also interesting in that it raises issues of implementation
and community’s governing capacity as important co-determinants of whether governance
structures will produce expected impacts. These findings and knowledge should inform
practitioners and policymakers on when and how they should take advantage of such
institutional innovation in local governance.
The other chapter of dissertation analyzes a critical issue in urban politics and
policy—the relationship between race/ethnicity of the population and the likely provision of
services by local government. The essay teases out impacts of income inequality as well as
racial diversity by considering whether variation in the between-race component of income
inequality, not just total income inequality, will effect local government expenditures on a
variety of services. The essay focuses on a different dimension of the public goods in
developing hypotheses, where the key factor is not the traditional redistributive/economic
development split, but rather whether the service has private substitutes available for the
public-provided option. An extensive data set was put from multiple sources, including
demographic, economic, political variables as well as government financial information to
test the arguments. The questions addressed are grounded in political theory and important in
urban debates today. Notably, the paper has received the best paper award in urban and local
politics commendation in 2017 by the American Political Science Association and selected as
the best article published in 2018 at Urban Affairs Review and it demonstrates its contribution
to urban governance and politics.

Abstract (if available)

Abstract This dissertation examines how representation and political power in local and regional institutions influence the quality, quantity, and spatial allocation of public goods in the United States. Specifically, this research focuses on the impact of three major governance and institutional forces: institutional rules in regional governance, racial structure of income inequality and type of local government. Each of these factors is examined at different levels of geographic scope including urban, rural, and metropolitan areas and from the communities in unincorporated areas to school districts, cities, metropolitan areas, and states. The findings of the four essays suggest that each of these forces significantly influence the allocation of local public investments. Unraveling what governance and institutional forces and to what extent they impact the local provision of public goods provides important implications for policy makers to manage their performance and to design better institutional arrangements for efficient and equitable allocation of public resources.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A framework for good local governance: achieving prosperity in an increasingly complex environment

PDF

Governing regional collaboratives: institutional design, management and leadership

PDF

Civic associations, local governance and conflict prevention in Indonesia

PDF

Three essays on the causes and consequences of China’s governance reforms

PDF

The effects of interlocal collaboration on local economic performance: investigation of Korean cases

PDF

Foreign-related activities of the Chinese local governments and agents of globalization: a case study of 31 provinces in mainland China

PDF

Why go green? Cities' adoption of local renewable energy policies and urban sustainability certifications

PDF

Goal setting and performance in federal agencies

PDF

Essays on fiscal outcomes of cities in California

PDF

Governance networks: internal governance structure, boundary setting and effectiveness

PDF

The crisis of potable water in Mexico City: institutional factors and water property rights as conditions for creating adequate metropolitan water governance

PDF

Local ground-level ozone policies: Governance structure and institutions In achieving federal attainment

PDF

The functions of the middleman: how intermediary nonprofit organizations support the sector and society

PDF

The risks and rewards of city management: how city managers evaluate the nature of the job and compensation

PDF

Local votes and outside money: campaign contribution geographic origins and their impact on Los Angeles City Council election outcomes

PDF

Community foundations and new governance networks: three studies exploring the role of regionally-networked philanthropic organizations in local problem-solving

PDF

Satisfaction with local "public goods" and services: the effects of household income and privatization in southern California

PDF

Institutional contestation, network legitimacy and organizational heterogeneity: interactions between government and environmental nonprofits in South Korea

PDF

A study of Chinese environmental NGOs: policy advocacy, managerial networking, and leadership succession

PDF

Beyond regulation: special improvement districts, design review, and place-making in New Jersey

Asset Metadata

Creator An, Yeokwang (Brian) (author)

Core Title Governing public goods: how representation and political power in local and regional institutions shape inequalities

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Public Policy and Management

Publication Date 05/10/2019

Defense Date 05/10/2019

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

Tag institutions,local and regional governance,local public goods,OAI-PMH Harvest,Political power,representation

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Tang, Shui-Yan (committee chair), Levy, Morris Eli (committee member), Thom, Michael (committee member)

Creator Email yan5@utk.edu,yeokwana@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-155833

Unique identifier UC11660727

Identifier etd-AnYeokwang-7443.pdf (filename),usctheses-c89-155833 (legacy record id)

Legacy Identifier etd-AnYeokwang-7443.pdf

Dmrecord 155833

Document Type Dissertation

Format application/pdf (imt)

Rights An, Yeokwang (Brian)

Type texts

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

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA