University of Southern California Dissertations and Theses

Who learns where: understanding the equity implications of charter school reform in the District of Columbia

(USC Thesis Other)

Who learns where: understanding the equity implications of charter school reform in the District of Columbia

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content WHO LEARNS WHERE:
UNDERSTANDING THE EQUITY IMPLICATIONS
OF CHARTER SCHOOL REFORM IN THE DISTRICT OF COLUMBIA
by
Andrew Eisenlohr
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
URBAN PLANNING AND DEVELOPMENT
December 2020
Copyright 2020 Andrew Eisenlohr
ii
ACKNOWLEDGMENTS
This dissertation would not have been possible without the time and support of many people. First,
a massive thank you to my advisor, Professor Marlon Boarnet, and my two other committee
members, Professors Julie Marsh and Dowell Myers. Each of you has indelibly shaped my
knowledge, analytical skills, and capacity for critical thinking. I am also deeply grateful for the
research opportunities you and other faculty, including Professors Elizabeth Currid-Halkett and
Annette Kim, have provided me. Second, thank you to my many classmates, who provided countless
feedback and recommendations. I will miss seeing you all in the Gateway research center. Third and
finally, thank you to my fiancée Anna, my family, and my friends. Without a doubt, your
encouragement and enthusiasm helped me to the finish line.
iii
TABLE OF CONTENTS
Acknowledgments.............................................................................................................................................. ii
List of Tables ...................................................................................................................................................... v
List of Figures ................................................................................................................................................... vii
Abstract .............................................................................................................................................................. ix
Introduction ........................................................................................................................................................ 1
Chapter 1: Reviewing the Literature on School Choice Mechanisms ........................................................ 9
Theoretical Justifications for School Choice .................................................................... 10
Arguments against School Choice and the Issue of Interest Convergence ................. 15
Empirical Evidence on the Efficiency Argument ............................................................ 16
Empirical Evidence from Charter Schools on the Equity Argument ........................... 25
Chapter 2: Reform for Whom and for What? Establishing the Policy Context for the
DC School Reform Act of 1995 .................................................................................................................... 35
Section 2. The Policy Process ............................................................................................. 37
Section 3. Implementation and Outcomes ....................................................................... 54
Section 4. Directions for Future Research ........................................................................ 74
Chapter 3: Assessing the Relationship between Charter Schools and Neighborhood
Gentrification in the District of Columbia................................................................................................... 77
Section 2. Literature Review ............................................................................................... 80
Section 3. Data and Definitions ......................................................................................... 85
Section 4. Descriptive Analyses .......................................................................................... 94
Section 5. Regression Analyses ......................................................................................... 110
Section 6. Discussion and Conclusion............................................................................. 117
Chapter 4: Detecting Patterns in Charter School Closure in the District of Columbia ...................... 123
Section 2. Literature Review and Conceptual Framework ........................................... 126
Section 3. Data and Definitions ....................................................................................... 132
Section 4. Descriptive Analyses ........................................................................................ 141
Section 5. Regression Analyses ......................................................................................... 155
Section 6. Discussion and Conclusion............................................................................. 163
iv
Chapter 5: Examining the Locational Decisions of Charter Schools Operating within
the District of Columbia ............................................................................................................................... 168
Section 2. Literature Review and Conceptual Framework ........................................... 173
Section 3. Data and Definitions ....................................................................................... 181
Section 4. Descriptive Analyses........................................................................................ 193
Section 5. Locational Regression Models ....................................................................... 203
Section 6. Discussion ......................................................................................................... 215
Section 7. Conclusion ........................................................................................................ 219
Chapter 6: Concluding Remarks .................................................................................................................. 224
A Final Story of the District ......................................................................................................................... 229
References ....................................................................................................................................................... 231
Appendices
Appendix A: Select Literature on School Choice Mechanisms and Academic
Outcomes ................................................................................................................................... 269
Appendix B: Results of Principal Component Analysis,
Changes in Census Tracts 2000-2017 ..................................................................................... 283
Appendix C: Constructing a Longitudinal Dataset of Charter Campus Locations ................ 284
Appendix D: Constructing a Dataset of Charter School Types ................................................ 293
Appendix E: Map of Landmarks and Neighborhoods in the District of Columbia .............. 300
Appendix F: Patterns in Charter Sector Growth by School Type,
School Year 2000-01 to 2017-18 ............................................................................................. 301
Appendix G: Spatial Proliferation of Charters by Type, School Years
2000-01 through 2017-18 ......................................................................................................... 303
Appendix H: Descriptive Statistics for Measures Used in Main Regression
Analyses, Chapter 3 ................................................................................................................... 309
Appendix I: Results of Sensitivity Analyses, Chapter 3 .............................................................. 310
Appendix J: Enrollment Statistics by Charter School Type, School Years
2004-05 through 2017-18 ......................................................................................................... 320
Appendix K: The Community Eligibility Provision.................................................................... 321
Appendix L: Descriptive Tables for Non-Adult Charter Enrollment,
by Geography, School Year 2010-11 to 2018-19 .................................................................. 322
Appendix M: Summary of Discrete Hazard Modeling ............................................................... 324
Appendix N: Logistic Regression Results, Chapter 4 ................................................................. 329
Appendix O: Supplementary Model Results, School Year 2010-11 to 2018-19,
Chapter 4 .................................................................................................................................... 331
Appendix P: Results of Principal Component Analysis, Changes in Census Tracts
2000-2018 ................................................................................................................................... 332
Appendix Q: Sources Used in Verifying non-DCPS School Buildings ................................... 333
Appendix R: Summary of Nested Logistic Modeling Comparisons......................................... 335
v
LIST OF TABLES
Table 1. Racial composition of the District of Columbia, 1800- 1990 ..................................................... 40
Table 2. Typology of District charter LEAs ................................................................................................ 90
Table 3. Dimensions of District gentrification, 2000-2017 (in 2017 USD where applicable) .............. 95
Table 4a. Enrollment composition of charter LEAs by LEA opening time
(pre- vs. post-PERAA), school year 2017-18............................................................................................. 105
Table 4b. Enrollment composition of charter LEAs by LEA opening time
(pre- vs. post-Recession), school year 2017-18.......................................................................................... 106
Table 5. Enrollment composition of charter school campuses by type,
school year 2017-18 ....................................................................................................................................... 106
Table 6. White alone, non-Hispanic share of charter campuses’ enrollment versus
white alone, non-Hispanic share of occupied census tracts’ school-age population,
by geographic location and charter school type, school year 2017-18 ................................................... 109
Table 7. Baseline regression results ............................................................................................................. 115
Table 8. Racial composition of non-adult charter students, by geography and school year ............... 147
Table 9. Summary statistics on charter LEA closures, across different ranges
of school years ................................................................................................................................................ 151
Table 10. Summary statistics on charter LEAs operating between school years (SYs)
2010-11 and 2018-19, by those that experienced academic closure versus those that
did not close for any reason ......................................................................................................................... 153
Table 11. Impacts of academic closure by student sub-group, school year 2010-11
to 2018-19 ....................................................................................................................................................... 154
Table 12. Complementary log-log regression results, school year 2010-11 to 2018-19 ...................... 158
Table 13. Marginal effects of Regression Model 4 explanatory factors ................................................. 161
Table 14. Supplementary model results, by functional form, dependent variable,
and primary explanatory factor, school year 2010-11 to 2018-19........................................................... 163
Table 15. Summary statistics for certain tract measures, across all census tracts and
for all school years between 2010-11 and 2020-21 ................................................................................... 194
vi
Table 16. Summary statistics for certain tract measures, selected versus unselected
tracts, school year 2012-13 ........................................................................................................................... 195
Table 17. Conditional logistic regression results, school year 2010- 11 to 2020-21 .............................. 209
Table 18. Marginal effects of Regression Model 7 explanatory factors ................................................. 212
vii
LIST OF FIGURES
Figure 1. For Washington Post articles filed under “DC Finances” and “DC School Reform”
between 1990 and 1995, shares of articles published under each search term by calendar year .......... 45
Figure 2. Theory of action for the School Reform Act of 1995 ............................................................... 53
Figure 3a. Locations of operating charter school campuses, school year 1997-98 ................................ 58
Figure 3b. Locations of operating charter school campuses, school year 2007-08 ................................ 58
Figure 3c. Locations of operating charter school campuses, school year 2017-18 ................................ 59
Figure 4a. Locations of operating charter school campuses, school year 2017-18, and
quintile rankings of census tracts per white alone, non-Hispanic share of population, 2017 .............. 60
Figure 4b. Locations of operating charter school campuses, school year 2017-18, and
quintile rankings of census tracts per median household income, 2017 .................................................. 61
Figure 5. Gentrification of District census tracts per Relative Scale measure, 2000- 2017 .................... 97
Figure 6. Gentrification of District census tracts per Absolute Index measure, 2000-2017 ................ 97
Figure 7. White alone, non-Hispanic shares of census tracts’ populations, 2017 .................................. 98
Figure 8. Median household incomes of census tracts, 2017 .................................................................... 98
Figure 9. District of Columbia Public Schools and public charter schools enrollment,
school year 1996-97 to 2017-18 ................................................................................................................... 100
Figure 10. Locations of all charter campuses, school years 2000- 01 through 2017-18,
superimposed on census tracts’ white alone, non-Hispanic population shares, 2017 ......................... 101
Figure 11. Charter-years attributable to each census tract, school years 2000-01
through 2017-18 ............................................................................................................................................. 102
Figure 12. Global Culture charter-years attributable to each census tract,
school years 2000-01 through 2017-18 ....................................................................................................... 103
Figure 13. Learner Centered charter-years attributable to each census tract,
school years 2000-01 through 2017-18 ....................................................................................................... 104
Figure 14. Conceptual framework, with hypotheses depicted ................................................................ 132
viii
Figure 15a. Share of schoolchildren (ages 0-17) considered black American,
by census tract, 2010...................................................................................................................................... 143
Figure 15b. Share of schoolchildren (ages 0-17) considered black American,
by census tract, 2018...................................................................................................................................... 143
Figure 16a. Share of schoolchildren (ages 0-17) in household below
federal poverty line, 2010 .............................................................................................................................. 144
Figure 16b. Share of schoolchildren (ages 0-17) in household below
federal poverty line, 2018 .............................................................................................................................. 144
Figure 17. District Ward boundaries, with census tract boundaries underlaid,
2010 through 2019 ......................................................................................................................................... 146
Figure 18a. Distribution of non-adult charter students considered Hispanic/Latino
or white, by Ward, school year 2018-19 ..................................................................................................... 148
Figure 18b. Share of non-adult charter students considered black American, by Ward,
school year 2018-19 ....................................................................................................................................... 148
Figure 19a. New charter campus locations and all charter campus locations,
school year 2010-11 to 2020-21 ................................................................................................................... 197
Figure 19b. New charter campus locations and all DCPS campus locations,
school year 2010-11 to 2020-21 ................................................................................................................... 197
Figure 20. New charter campus locations and density of children per tract,
school year 2012-13 ....................................................................................................................................... 198
Figure 21. New charter campus locations and gentrification since 2000 per tract,
school year 2012-13 ....................................................................................................................................... 199
Figure 22. New charter campus locations and median gross rent per tract,
school year 2012-13 ....................................................................................................................................... 199
Figure 23. New charter campus locations and facilities available per tract,
school year 2012-13 ....................................................................................................................................... 201
Figure 24a. New charter campus locations and tracts’ nearby charter quality,
school year 2012-13 ....................................................................................................................................... 202
Figure 24b. New charter campus locations and tracts’ nearby District of Columbia
Public Schools quality, school year 2012-13 .............................................................................................. 202
ix
ABSTRACT
In its analysis, this dissertation assesses a specific policy, the District of Columbia School Reform
Act of 1995 (Pub. L. 104-134), which most notably legalized charter schools in the District of
Columbia. The research questions it pursues, however, are broader and relevant to the planning,
public policy, and education fields. First, how do components of a policy intersect with historical
and structural inequalities associated with race and place? And second, how do these interactions
influence a policy’s ability to effect greater equity of access? In pursuing answers to the above
questions, this dissertation identifies: (1) a significant relationship between the presence of charter
schools and the gentrification of District neighborhoods, albeit one that is dependent on type of
charter school (e.g., STEM); (2) regressive and discriminatory trends in charter school closure; and
(3) an inequitable distribution of new charter school campuses based on a model of charter
operators’ locational decisions. Taken together, these findings tell a story of differing results for
three disparate geographic regions in the District. For communities west of Rock Creek, consistently
the whitest and most affluent, charter reform has been largely irrelevant. For communities east of
the Anacostia River, consistently the least white and most impoverished, charter reform has been
largely a backfilling mechanism, whereby charter campuses mainly move in as replacements for
closed or relocated traditional public, private, and charter schools. For communities between Rock
Creek and the Anacostia, which have gentrified sharply since the School Reform Act’s passage,
charter reform has actually supplemented, not replaced, public educational options. In brief, the
District’s most marginalized and impoverished communities do not appear to have benefited most
from charter reform. As a result, this dissertation concludes with several recommendations to help
this policy achieve its stated goal of empowering “families of limited means” .
1
Introduction
Developing consensus around legislation to overhaul education in the District [of Columbia]
hasn’t been easy, and it’s not over yet. No effort on this scale can be unanimous (I have been
telling groups that it is their responsibility to like 80 percent of the package). But if we keep
focused on the children we are all trying to benefit, we can create a system of schools on a
par with the best in the world. And the nation will have a blueprint embodying the principles
of urban education reform that can be applied in any American city.
–U.S. Representative Steve Gunderson, “World Class Schools for Washington”, Washington Post
America is a nation increasingly concerned with equity. America is a nation that is
increasingly inequitable. For the typical American, these two statements are a daily experience. In
fact, it is probably more accurate to say: America is increasingly concerned with equity because it is
increasingly inequitable.
Occupy Wall Street, debates over urban gentrification and displacement, the Black Lives
Matter movement, the Me Too movement, and the turmoil of the coronavirus outbreak are all
instances of national unrest within the last decade. Each has sought to lay bare America’s
socioeconomic inequities. Each has gained traction at the same time that the country’s wealthiest
and most powerful have further consolidated their resources and influence (Pfeffer, Danzier, &
Schoeni, 2013). Stock market indices like the S&P 500, the Russell 3000, and the Dow -Jones
Industrial Average have continually attained new peaks in recent years (e.g., Business Insider, 2020).
Income inequality has reached five-decade highs at the same time (Semega, Kollar, Creamer, &
Mohanty, 2019).
Seeing these juxtaposed realities produces a discomfort – some sense of moral tension. How
has an awareness of inequity grown in parallel with inequity itself, the latter swallowing larger and
2
larger swaths of Americans despite outcries from politicians, scholars, advocates, media members,
and the public?
Unquestionably, power dynamics have played a vital role in this societal trajectory. What is
best for the socioeconomic elite, who dominate our corporate and political systems (Gilens & Page,
2014), is often not best for the typical American. Reducing the capital gains tax is unlikely to help a
family struggling to make ends meet (Oprysko & Kimbel-Sannit, 2019).
Even when the social elite redistribute resources and influence in sporadic moments, their
concessions may be temporary or outright disingenuous (Bell, 1979). After the Brown v. Board of
Education ruling, many southern white families avoided desegregation by transferring to white-only
private schools. In the north, many white families engaged in similar behavior, moving to suburban
school districts far from city centers (Walters, 2001). Even the order for inter -district busing,
intended to address severe race-based segregation between school districts, soon dissipated after
challenges from America’s executive, legislative, and judicial branches (Forman, 2004; Orfield, 1978;
Milliken v. Bradley).
Besides power dynamics, though, the word “equity” provides a separate and equally
important clue about our persistent social schism. At its worst, and in the same way as “diversity”,
equity can become a token of cultural currency, used by individuals and institutions as an instrument
of legitimacy (Ahmed, 2007). At its best, however, it carries a more profound connotation than
“equality”, one that encourages a historical and structural understanding of social disparities. If
America has truly attempted to make itself more equitable, and if those efforts have failed, then the
connotation of equity gives us a reason why: so many of our inequalities are so deeply entrenched
and insidiously intertwined.
As many have documented, events in our collective history appear to support this
conclusion. The permanence of social, demographic, and economic inequalities, or maybe more
3
appropriately their repeated reproduction, is common enough that sociologists have devised the
term “horizontal inequality” to refer to them (Stewart, 2009).
Institutionalized racism against black Americans has persisted, morphing from slavery to Jim
Crow to mass incarceration (Alexander, 2010) and manifesting itself in racially -divided
neighborhoods and schools (Kozol, 1991; Rothstein, 2017). Discrimination against women has
remained too, visible in pay gaps (AAUW, 2018), the unfair distribution of emotional labor
(Wharton, 2009), and unequal access to many upper echelon professional positions (Baker &
Cangemi, 2016). As of 2020, no president or vice president of the United States has been female.
America’s poor remained trapped in poverty. The benefits of globalization and advances in
technology have accrued with the already wealthy owners of capital (Friedman, 2015). Rents levels
have grown faster than real wages over the past 30 years (Collinson, 2011; Myers & Park, 2019).
These inequalities are not only intractable, but also combine with and reinforce one another.
The intersection of racism and sexism is one such example (Browne & Misra, 2003). As another
case, consider the different ways that America’s federal, state, local governments have attempted to
alleviate poverty. Coordinating even just two prospective anti- poverty policies – the development of
affordable housing and the expansion of employment opportunities – could prove exceedingly
difficult. After all, researchers suggests that occupants of affordable housing units tend to be further
from employment opportunities than occupants of market-rate units (Covington, 2009; Holder,
2020).
And still, even if governments could coordinate these two programs effectively, it is unlikely
to be sufficient. The vast majority of new jobs now go to workers with postsecondary degrees
(Goldstein, 2018), and so efforts to reduce poverty must also prioritize expanded pathways to
college. But affordable housing programs, especially public housing projects, can concentrate
poverty in neighborhoods and therefore in schools (Hunt, 2001). Meanwhile, high concentrations of
4
impoverished students predict lower standardized assessment results (Reardon, 2011), and thus
reduced prospects for a college degree. Of course, expanding employment opportunities also loses
its effectiveness when systemic racism and sexism eliminate many job opportunities (Yearby, 2018).
Evenly distributing affordable housing units does not reduce the cost of college tuition either, which
represents another major barrier to education and jobs for America’s poor (Cai & Heathcote, 2018).
The durability and complexity of America’s inequalities may explain why monolithic but
narrowly-focused planning initiatives in the mid-20th century, such as the dramatic expansion of
public housing (Bickford & Massey, 1991), failed to eliminate them (Scott, 1998). Perhaps
responding to these failures, in recent decades American policymakers and planners have
increasingly relied upon market-based interventions to promote equity instead. In the case of
affordable housing, they have shifted towards subsidizing the private development of affordable
housing rather than directly constructing and owning housing themselves.
The country’s system of public education has experienced a similar transition (Labaree,
1997), one that is the overarching inspiration for this dissertation. With attempts to move students
across districts defeated, education policymakers and planners have recalibrated their focus to
ensuring equality in three facets of schooling within districts. These three aspects are: (1) inputs, e.g.,
funding levels per student; (2) outputs, e.g., standardized assessment scores; and (3) processes, e.g.,
enrollment mechanisms like common lotteries (Bulkley, 2013). Implicit in their approach is that
equalizing inputs, outputs, and processes across a district’s schools – in a way, equalizing the
schooling “market” – also equalizes access to high-quality schooling (Forman, 2004).
Policymakers and planners have introduced an array of market-based reforms to drive this
initiative forward, including performance-based teacher pay and an accountability framework based
on standardized assessment results (e.g., No Child Left Behind). Of these efforts, however, none is as
visible as school choice reform, which resulted in the creation and legalization of three new choice
5
mechanisms between 1988 and 1991 : (1) intra-district/inter-district enrollment, i.e., allowing
traditional public students to attend a school other than their neighborhood one (Mikulecky, 2013),
(2) private school vouchers (Laws of Wisconsin, 1990, Act 36), and (3) public charter schools (Laws
of Minnesota, 1991, Chapter 265, H.F. No. 700). These new mechanisms represent an expansion to
choice reform that began with magnet schools, an older school choice mechanism, under the
original Elementary and Secondary Education Act of 1965 (Pub. L. 89- 10).
Since their initial implementation, these three mechanisms have proliferated to varying
degrees. The most current statistics show the number of charter students growing from 0.4 million
in 2000 to 3.1 million in 2017, actually outstripping the 1.3 million increase in traditional public
school students over the same time period (NCES, 2020). While nationwide figures for private
school voucher use are harder to find, one pro-choice organization estimates 192,660 recipients as
of 2019, across 16 states and the District of Columbia (EdChoice, 2020). And although counts of
students utilizing intra-district and inter-district enrollment options are unavailable, we can compare
the number of states with such policies to the number of states with enacted public charter school
laws. As of 2017, 32 states and the District of Columbia had some form of intra -district enrollment
policy, whereas 43 states had some form of inter-district enrollment policy (ECS, 2017). By the same
year, 44 states and the District had legalized public charter schools (Ziebarth & Palmer, 2018).
As these mechanisms have grown in popularity, proponents have variously asserted their
ability to boost academic outcomes and close achievement gaps (i.e., make outcomes more equal) or
to produce less segregated student bodies (i.e., make processes more equal). Yet as the literature
demonstrates, race and income-based segregation continue to plague American school districts with
expanded choice, accompanied by persistent disparities in academic outcomes. Even worse, students
in some districts with choice are more segregated than if they could only attend traditional public
schools in their own neighborhoods (Bifulco & Ladd, 2007). Despite the claims of advocates,
6
“additional schooling options” does not seem to mean “better access to high-quality schooling
options”. But why is this the case?
This is the foundational and motivating question for my research. To answer it, I study the
implementation of charter school reform within the District of Columbia. The District is a salient
case study for a number of reasons. First, it provides an extended period for observing choice,
having first authorized charter schools in 1996. Second, the District has a remarkably competitive
charter school sector relative to its traditional public schools (DCPS). As of school year 2019- 20,
44.1% of its non-adult public school students were enrolled in charter schools and 55.9% in DCPS
(OSSE, 2020).
Third, the District is a semiautonomous territory overseen by Congress, and therefore has
only partial control over the rules and regulations affecting its residents. It is a particularly
appropriate setting for understanding the power dynamics between those who enact policies – in
this case, the federal government – and those who experience them daily – in this case, District
residents.
Finally, and perhaps most importantly, the District and its residents have long suffered stark
residential segregation and socioeconomic inequality, along the highly-correlated lines of race and
income (Asch & Musgrove, 2017; Burner v. Washington). This societal divide is an embedded feature
of the city’s public education system (Washington Lawyers’ Committee for Civil Rights, 2005), and
was in fact a cited factor in the federal government’s decision to allow charter schools in the District
in 1996 (Gunderson, 1995a, 1995b). Achieving greater equity, via more equal access to high-quality
schools, was ostensibly a core tenet of District charter reform.
Nonetheless, I find that the District’s inequalities have effectively reproduced themselves
during charter reform, in ways that are more subtle than the heightened levels of segregation
identified by other scholars (Orfield & Ee, 2017). This dissertation analyzes three such
7
reincarnations along with their implications for equity of access under charter reform: (1) the acute
and severe gentrification of neighborhoods exposed to charter schools, (2) racially-biased and
regressive school closures, and (3) the clustering of new charter schools away from the District’s
poorest and nonwhite residents. All of this despite a uniform per student funding formula, a
universal enrollment system, and local and national scrutiny of the District’s academic outcomes and
achievement gaps.
Before continuing, it is important to consider the fundamental question: why do my results
matter? Of all the growing inequalities in America, why should public education and school choice
garner special attention? For one, gaps in schooling quality predict gaps in many other aspects of
life, including in income and in health (Ross & Wu, 1995). If we want to achieve greater equity in
our society, addressing educational inequalities seems like an important aspect to prioritize (Glomm
& Ravikumar, 2003). Improving school access may also be a rare opportunity for bipartisan action in
our polarized political atmosphere (Fisher, Ury, & Patton, 2011). It is relatively difficult to vote
against the interests of children, who cannot express themselves politically and who represent our
collective future.
This dissertation is not the first piece of research to consider the themes of equity, school
choice, and school access. Many distinguished scholars and widely-read publications have explored
the same terrain (e.g., Orfield, 2014). Furthermore, I note that my findings stem from a single
jurisdiction. Future research must determine whether they generalize to other American cities and
suburban and rural regions with charter schools.
Those caveats aside, I believe this dissertation accomplishes three broad objectives. First, it
provides a new set of evidence for the intractability and intricacy of societal inequalities. Second, it
underscores the need for greater and more nuanced efforts to address them. And third, it illuminates
8
the plight of many long-oppressed and marginalized Washingtonians, who deserve far better from
the local and federal governments supposedly serving them.
By communicating the above, I hope to reframe the concept of “equitable school access” for
educational planners and policymakers. It is clear that addressing in equalities in inputs, outputs, and
processes is not nearly sufficient. We must confront the fundamental inequalities underlying them,
most notably our country’s wide and racially-defined gaps in income and the segregated
neighborhoods those gaps have produced.
As the earlier discussion of equity suggests, this will require addressing a number of
ingrained and interwoven inequalities. More practically, it will mean coordinating educational
interventions with those in affordable housing, public transportation, employment opportunities,
pathways to and financial assistance for college. Already, researchers are beginning to point out the
necessity of doing so; the work of the National Center for Research on Education Access and
Choice is an extraordinary step in this direction. Unquestionably, establishing connections between
government agencies and policies will be incredibly difficult. But until we do so, the warning that
“equity” carries – of inequality’s stubborn reappearance – will likely ring true once again.
The remainder of this dissertation is structured as follows. Chapter 1 summarizes the
literature on school choice mechanisms in the United States, paying particular attention to findings
on charter schools and equity of access. Chapter 2 then reviews the policy context for charter
reform in the District. In doing so, it sets the stage for the three core analytical chapters that follow.
Chapter 3 assesses the relationship between charter schools and neighborhood gentrification within
the District. Chapter 4 considers patterns in the closures of District charter schools. Chapter 5
examines the locational decisions of charter school operators across District neighborhoods. Finally,
Chapter 6 provides some concluding remarks. It is accompanied by a final and salient story.
9
CHAPTER 1
Reviewing the Literature on School Choice Mechanisms
School choice has always been present in America’s elementary and secondary education
systems, a fact that garners insufficient recognition from many who cover the topic in contemporary
research. Aside from the more-recently established “choice mechanisms” that I describe and
distinguish below, American households have historically had a right to exercise choice in two ways:
(1) changing school districts or attendance zones by physically moving (Forman., 2007, p. 844), or
(2) opting out of a traditional public school to attend a private school (Lankford & Wyckoff, 1992).
Scholars typically consider the first traditional method – choosing schools by physical moving across
districts – as a manifestation of Tiebout sorting (e.g., Hoxby, 2000a; Tiebout, 1956). Their
understandings of private schools are more varied because such schools also provide households
with ways of maintaining religious programming in their children’s curricula (Forman, 2006).
Regardless, these two traditional methods of choice share a few problematic traits. First,
each imposes a significant cost on households. Physically moving or paying private school tuition are
expensive undertakings. Second, each method leaves the institutional structure of traditional public
school systems unchanged (Chubb & Moe, 1990), including the spatial link between residential
location and schooling location. Third, the literature suggests an inverse relationship between the
two methods: the more school districts around households in a metropolitan area, the lower the
share of district students enrolled in private schools (Hoxby, 2000a ; Urquiola, 2005). This last
phenomenon is especially noteworthy, as it reveals a consistent preference for choice among
households traditionally able to exercise that choice. If households who can choose do choose,
expanding the choice capabilities of disadvantaged households may have significant ethical appeal
(Rawls, 2001; Sen, 1999).
10
Over the past several decades, the dynamics outlined above have indeed inspired broad calls
for augmented school choice. As I describe below, the theoretical justifications underlying these calls
are not mutually exclusive. In fact, they have frequently converged, helping facilitate the creation of
four school choice mechanisms in the 20th century: magnet schools, private school vouchers,
intra/inter-district enrollment, and charter schools.
In the sections that follow, I review the theoretical arguments in favor of school choice
mechanisms and the arguments against them. I then summarize the empirical evidence on these
mechanisms’ implementation. The majority of the literature examines choice mechanisms’ effects on
schooling efficiency. Their results are decidedly mixed in this regard. Moreover, multiple papers
indicate a strong connection between choice mechanism’ potential to increase efficiency and their
potential to improve equity of access. It seems imperative to more thoroughly understand the
potential for the latter.
Accordingly, a section of the literature has developed a consistent body of evidence that
school districts with charter schools, in particular, experience greater segregation in their charter
schools than in their traditional public schools. Given the topic of this dissertation, I emphasize the
findings for charter schools, but similar findings, albeit more limited in number, are available for
other choice mechanisms too (e.g., for magnet schools, Saporito, 2003; for voucher systems,
Brunner, Imazeki, & Ross, 2010). Besides heightened segregation, the literature fails to sufficiently
articulate the other ways that choice mechanisms may complicate issues of equity in access rather
than resolving them. A growing strand does, however, contemplate the prospective association
between charter schools and urban gentrification.
Theoretical Justifications for School Choice
The academic literature reveals three distinct justifications for introducing choice
mechanisms: (1) they will increase academic outcomes without requiring an increase in spending, i.e.,
11
they will increase schooling efficiency (see Hoxby, 1996, 2003); (2) they will increase poor or
historically marginalized households’ levels of access to schools, i.e., they will increase schooling
equity; and (3) they will establish school choice as a fundamental right of households. To these three
justifications, I add two other factors that may be helping drive the popularity of school choice
mechanisms: (4) the opportunity for households to move into cities without enrolling their children
in traditional public school districts, and the benefits this dynamic confers to said cities; and (5) the
performance pressures that accountability frameworks place upon traditional public schools.
The Efficiency Case for School Choice
Proponents have articulated three ways in which expanding school choice may boost the
efficiency of education systems. First, expanding choice may allow for better matching between
households’ schooling preferences and the types of schools their children attend (Friedman, 1962).
If this matching improves student outcomes, then it suggests an increase in schools’ efficiency.
Second, expanding choice may enhance competition among schools in beneficial ways. For
example, greater threats of exit may make traditional public schools more sensitive to household
needs or force them to allocate resources more responsibly (Hirschman, 1970). Alternatively,
competition may induce innovations in the delivery of education, i.e., it may act as a “learning tool”
for policymakers and other schools (Schneider & Ingram, 1990). Given that schools are highly
visible entities and that most of their innovations should be easily replicable, such innovations may
diffuse effectively across schools and widely raise student outcomes (Rogers, 2010).
Third, providing households with choices other than traditional public schools may allow
them to attend schools with more outcome-oriented institutional logics (Thornton & Ocasio, 1999).
Chubb and Moe (1990) frame the harmful and bureaucratic institutional logics of traditional public
schools, as well as their misallocation of resources due to influences like teacher unions, as a
principle reason for expanding school choice.
12
The efficiency case, then, actually defines two distinct pathways by which school choice
mechanisms may boost student outcomes. First, the mechanisms themselves may have a direct
treatment effect on those students utilizing them. Second, school choice mechanisms, by
introducing competition or inducing innovation, may also have a “spillover” effect on students
continuing to attend traditional public schools. As I survey the empirical evidence regarding school
choice mechanisms’ efficiency, I consider the potential for each of these effects. Without doing so, it
is impossible to determine whether expanding school choice can truly establish a general equilibrium
of higher efficiency in the delivery of elementary and secondary education.
The Equity Case for School Choice
The equity case for school choice mechanisms is more streamlined than the efficiency case.
The singular argument made by its proponents is that no child should be consigned to attending a
particular school based on residential location. American states and districts have historically
assigned children to public schools via residential attendance zoning. Because of this and because
traditional methods of choice are costly, poorer and historically marginalized households, who may
also be residentially segregated, inherently have fewer schooling options than their counterparts.
Further, because traditional public school districts receive a large amount of funding from local
taxes, wealthier households are also likely to attend traditional public schools that have far more
resources than those attended by poorer households (Kozol, 1991).
By weakening the traditional neighborhood-school link and schooling location and
expanding the choice capabilities of poor and marginalized households (Forman, 2006, 2007), school
choice mechanisms may reduce the race-based or income-based segregation of students between a
district’s schools or across districts themselves (Fiss, 1965; Forman, 2004). Such integration may
even boost the outcomes of children from poor and marginalized households by granting them
access to better-resourced or more effective schools (Forman, 2007, pp. 853- 854; Reardon, 2011).
13
This represents a potential connection between the equity and efficiency cases, which I revisit when
considering the empirical evidence on school choice mechanisms. Moreover, if having diverse
student bodies in schools – or at least ones that are representative of local populations – represents a
social norm, such integration may have intrinsic value. This intrinsic value may be reflected in visible
commitments to diversity in student body, which traditional, attendance-zoned public schools would
be unable to act upon (Potter & Quick, 2018).
Charter schools, in particular, may also improve equity of access in a separate way. Because
they are distinct educational institutions with a high degree of curricular autonomy at the campus
level, charter schools can provide communities with culturally-specific and culturally-sensitive
learning spaces, which Wilson (2016) refers to as “counterpublics”. These spaces may be especially
vital for communities receiving insufficient attention and resources or facing stigmatization in
existing private and traditional public schools; such as immigrant populations (Wilson, 2016) or
special needs students.
The Deontological Case for Choice
Although less popular than the efficiency or equity cases, a deontological argument for
school choice mechanisms also exists. Policymakers and voters may feel a moral obligation to grant
households a “right to choose” (Stone, 2012). As Ball (1993) argues, “‘choice might serve as a
powerful means of political legitimation… cloaking the system in the powerful American Ideology
of consumer sovereignty’” (p. 16).
An array of legal scholars have considered this fundamental right to school choice in the
years since private school vouchers, intra/inter-district enrollment, and charter schools first
appeared in American education systems (e.g., Rinas, 1996; Ryan & Heise, 2002; Smith, 1995).
Contrary to Ball’s (1993) theory, their conclusions indicate that legislators and voters do not
generally feel an obligation to expand choice for households, consistently pointing to historical
14
restrictions on and legislative failures of inter-district enrollment policies, which appear most
attributable to suburban households’ desires to exclude their schools from the choice sets of urban
residents (Ryan & Heise, 2002).
Two Additional Undercurrents Advancing School Choice
Although they are not theoretical justifications for choice mechanisms, I find it conceptually
helpful to mention two other dynamics that may be helping drive school choice reform. First, each
of the mechanisms I consider (i.e., magnet schools, private school vouchers, intra/inter-district
enrollment, and public charter schools) enables households to enter school districts and cheaply
enroll their children outside of the traditional public school within their neighborhood (Renzull i &
Evans, 2005). Thus, the presence of school choice mechanisms may be quite valuable to gentrifying
households who want to avoid traditional public schools, perhaps concerned by those schools’
quality, their relatively poor or nonwhite student bodies (Saporito, 2003), or their institutional
identities. In one case study, Hankins (2007) documents the establishment of a charter school by
gentrifiers of an Atlanta neighborhood, who expressly wished to separate their children from the
neighborhood’s traditional schools and longtime residents. Consequently, cities may improve their
financial standing by introducing choice mechanisms that incentivize wealthier households to move
inside their borders (Johnson, 2015).
Second, the No Child Left Behind reauthorization of the Elementary and Secondary
Education Act (Pub. L. 107-110) acutely and drastically increased the accountability pressures on
public schools from the federal and state governments (Wirt & Kirst, 2005). Some executives within
traditional public school systems may support school choice mechanisms as a necessary way of
redefining the principal-agent relationship with governments, diluting those governments’ oversight
activities by spreading them across multiple agents, i.e., multiple education institutions (Heinrich &
Marshcke, 2010).
15
Arguments against School Choice and the Issue of Interest Convergence
So far, I have presented three distinct theoretical justifications for the introduction of school
choice mechanisms above the traditional choice methods of Tiebout sorting and private schooling. I
have also noted two contemporary trends that may help explain the growth of the 20th century
mechanisms. Do any scholars, though, argue against the provision of school choice? If so, on what
grounds?
In fact, arguments against school choice are consistently made on the basis that systems of
choice reproduce or worsen the equity issues within America’s existing elementary and secondary
systems. Ball (1993) argues that the “market [of school choice] provides a mechanism for the
reinvention and legitimation of hierarchy and differentiation via the ideology of diversity,
competition, and choice” (p. 16). Henig (1995) similarly states: “Faced with disparate groups… it is
tempting to beat a retreat from the idea of collective purpose, and to settle more comfortably into
bite-size communities of like minded folks…. It is this urge to retreat that ultimately accounts for
much of the popular appeal of both the broad privatization movement and the specific segment of
the education-reform movement that draws on market models for inspiration” (p. 9). As a third
example, Orfield (2014) claims “[Policymakers] already knew that unrestricted school choice would
leave segregation virtually intact because it had been tried in hundreds of districts and that much
more was needed” (p. 275).
According to the above scholars, the expansion of school choice is a path-dependent
exercise. It allows issues of segregation by race and income, already inequitable when considering the
traditional choice methods of physically moving or private schooling (Fiss, 1965; Ryan & Heise,
2002), to replicate. This could stem from power dynamics affecting who shapes policies (Stone,
2012). It could originate from the desires of white or wealthy households to avoid schools with
different peers (Lankford & Wyckoff, 1992; Saporito, 2003). Or it could be that diff erent
16
sociodemographic groups engage with systems of choice in different ways (Ball & Vincent, 1998;
Reay & Ball, 1998). If experiencing segregation significantly affects the outcomes of poor, nonwhite,
or other marginalized groups, then this reduces the efficiency argument for school choice as well
(Henig, 1995; Reardon, 2011).
Despite these counterarguments, the implementation of school choice mechanisms
continues to spread. Why is this the case? For one, the theoretical justifications I describe above
have a high level of interest convergence (Bell, 1979). School choice proposals have received strong
support from both the left and right sides of the political spectrum, especially when implemented in
historically under-resourced urban districts (Orfield, 1978, 2014). Suburban households, who do not
favor inter-district school choice mechanisms (Ryan & Heise, 2002), may especially support intra-district
choice mechanisms in urban districts (e.g., magnet schools, private school vouchers, intra-district
enrollment, and public charter schools) precisely because they reduce public calls for inter-district
choice. If this is the case, it only reaffirms the interest convergence at work.
Having summarized the theoretical debate over the 20th century school choice mechanisms,
I now turn to considering the empirical findings on these mechanisms. In order, I focus on whether
they indicate improvements in schooling efficiency or equity. For the latter, I restrict my attention to
charter schools, given this dissertation’s focus on the equity implications of District charter reform.
Empirical Evidence on the Efficiency Argument
What do we know about the effects of school choice mechanisms on student outcomes in a
given school district, and how do we know it? Within this section, I survey the literature to address
the question. For intra-district and inter-district schemes, I focus on the former as inter -district
enrollment has received less research. I outline in order:
1. The methods researchers have used to identify causal effects of school choice mechanisms
on student outcomes;
17
2. The wide variety in findings on whether these mechanisms have a significant causal effect;
3. Reasons for the wide variety in findings, as well as issues in the methodological tools relied
upon by researchers; and
4. Future items for researchers’ consideration when assessing the relationship.
As I articulate earlier, introducing school choice mechanisms can generate efficiency gains
via two distinct channels. One, the mechanisms may directly improve the outcomes of students
attending schools of choice. Two, the mechanisms may have spillover effects on the outcomes of
students remaining at the traditional public schools in their neighborhoods. I consider the evidence
for each type of causal effect below.
Methods for Identifying the Causal Effects of School Choice Mechanisms
In an ideal world, researchers would be able to observe a true counterfactual when
estimating the effect of exercising school choice on a student’s future outcomes. Put another way, it
would be helpful to compare (a) a student’s outcomes when he/she exercises school choice, versus
(b) the same student’s outcomes when he/she does not exercise school choice. In the same vein, it
would be helpful to compare (a) the outcomes of a student attending a traditional public school
when school choice mechanisms are available for use, versus (b) outcomes of the same student
when no choice mechanisms are competing with the traditional public system.
In reality, observing these true counterfactuals is not possible (Lamarche, 2008). Instead,
researchers must compare the outcomes of students who receive a treatment (i.e., utilize a school
choice mechanism, or attend a traditional public school where choice mechanisms are available) with
those who do not receive that treatment (i.e., do not utilize a school choice mechanism, or attend a
traditional public school where choice mechanisms are unavailable or compete with the traditional
public system to a lesser extent). Appendix A tabulates some academic papers exploring the effects
18
of school choice mechanisms on student outcomes; see its “Analytical Framework” column for
detailed examples of the methodological approach I describe above.
In addition to the lack of a true counterfactual, estimating the causal effects of school choice
mechanisms must overcome endogeneity concerns (e.g., Ballou, Goldring, & Liu, 2006; Imberman,
2011a, 2011b). The most intuitive example is the self -sorting of students across a district’s schools
based upon some hard-to-measure or unmeasurable characteristic, like the value they place on
education. For instance, students who utilize intra-district enrollment options may have greater
academic motivation than students who elect to remain at neighborhood public schools (Cullen,
Jacob, & Levitt, 2005). Alternatively, students who elect to use private school vouchers may tend to
do so when their performance is suffering in traditional public schools. In both cases, attributes of
students or their households are associated with both the use of school choice mechanisms and
academic outcomes. In turn, this increases the potential for biased estimates of choice mechanisms’
treatment effects.
To overcome issues of endogeneity, the extant literature adopts three disparate quasi-
experimental methodologies: (1) manipulating discontinuities in admission decisions; (2) using fixed -
effects panel datasets; and (3) developing propensity-score matching systems.
For the first methodology, researchers tend to estimate treatment effects by comparing the
outcomes of: (a) students who win lotteries to attend schools of choice, versus (b) students who lose
those same lotteries (e.g., Cullen, Jacob, & Levitt, 2006). In the case of magnet schools specifically,
when those schools require a certain level of academic performance for admission, regression
discontinuity at the admission cut-off point generates similar effect estimates (e.g., Abdulkadiroğlu,
Angrist, & Pathak, 2014). Because these regressions depend upon lottery victories or offers of
admission, they typically use a two-stage least square or reduced functional form where lottery
victory or offer of admission is an instrumental variable.
19
For the second methodology, researchers use fixed-effects panel datasets to estimate the
effect of utilizing school choice relative to attending a traditional public school ( e.g., Bifulco & Ladd,
2007). Effects are almost always fixed at the student level, frequently fixed at the school level, and
sometimes at the joint student-grade or student-school level. Because of their fixed-effects nature,
these panel regressions produce effect estimates for only those students who both attend traditional
public schools and utilize a school choice mechanism in their academic careers, i.e., for only students
with longitudinal variation in the type of schooling they receive.
For the third methodology, researchers develop a system of propensity-score matching
system to pair a “control” student in traditional public schools with each student in a school of
choice. Match criteria can include sociodemographic characteristics and prior academic outcomes. It
is sometimes further stipulated that the control student must come from the same traditional public
school the “treated” student was most likely to attend (e.g., CREDO, 2013, 2015). The difference in
outcomes across the paired students then represents the estimated treatment effect.
I discuss the validity of these three methodologies in detail below. First, though, I note that
the first two methodologies present appealing ways to control for unobservable student
characteristics that may be endogenous to regression analyses. Within a given district and school
year, the distribution of the potentially endogenous traits of students who enter a lottery or change
from traditional public schools to schools of choice (or vice versa) may, in fact, be randomly
distributed. However, the third method is concerning in that it may not present a valid control if the
“control” student continuing to attend a traditional public school is differentially exposed to school-
level characteristics that affect academic performance (e.g., the traditional public school has a higher
concentration of impoverished students).
Meanwhile, in estimating the spillover effects that choice m echanisms may have on
traditional public school students, empirical researchers have developed various proxy measures for
20
the increased “competition” that traditional public schools may experience (e.g., Booker, Gilpatric,
Gronberg, & Jansen, 2008). While these measures may be the best available, they do not appear to
sufficiently address inherent endogeneity concerns. In particular, trends that correlate with the
implementation of school choice mechanisms, such as heightened school accountability schemes,
may also relate to changes in the outcomes of traditional public school students. For examples of
proxy measures for competition, see Appendix A.
Finally, I find it important to note that the near-unanimous dependent variable of interest,
regardless of analysis, is standardized assessment scores. This makes a good deal of sense. States had
to record standardized assessment scores under No Child Left Behind, and standardized assessment
scores represent already-scaled measures of student skill levels. A few pieces of research do
incorporate alternative measures of student outcomes as their dependent variables, including: Cullen
et al. (2005), who use completion of high school grades as well as high school graduation; Cullen et
al. (2006), who use attendance rates, high school graduation, and non-academic outcomes like arrest
and disciplinary incidences; Deming, Hastings, Kane, and Staiger (2014), who use measures of
postsecondary matriculation and persistence; and the aforementioned Imberman (2011a, 2011b),
who uses measures of attendance and disciplinary incidences. I further discuss the literature’s
reliance on standardized assessment scores below.
Findings on the Causal Effects of School Choice Mechanisms
I now consider the distribution of findings in the literature on school choice mechanisms
and schooling efficiency. Given the text above, this section largely consists of evidence on whether
choice mechanisms significant alter the standardized assessment scores of students exercising choice
or the scores of students at nearby traditional public schools. Within Appendix A, I tabulate relevant
literature across the following dimensions: study geography, time period of study, school grades
observed, the analytical framework employed (e.g., comparing lottery winners to lottery losers),
21
whether the paper tests for the direct effect of choice mechanisms on students exercising choice or a
spillover effect, and the empirical findings.
I emphasize Appendix A’s right-most column, which categorizes empirical findings as
“Negative”, “Mixed”, or “Positive”. I classify a paper’s empirical findings as “Negative” if it: (1)
reports a significant and negative effect of a choice mechanism on students utilizing it or students at
nearby traditional public schools; or (2) reports both significant and negative as well as insignificant
effects. I classify a paper’s findings as “Mixed” if it reports both significant and negative as well as
significant and positive effects, again on the students utilizing it or on nearby traditional public
students. Finally, I classify a paper’s empirical findings as “Positive” if it: (1) reports a significant and
positive effect of a choice mechanism, or (2) reports both significant and positive as well as
insignificant effects. None of the papers I survey reports wholly insignificant findings.
Per the above categorization scheme, the literature on the relationship between choice
mechanisms and schooling efficiency is decidedly mixed. This is true for magnet schools (two papers
considered “Mixed”), private school vouchers (two papers considered “Mixed”, two papers
considered “Positive”), intra-district enrollment programs (two “Negative”, one “Mixed”, one
“Positive”), and charter schools (three “Negative”, six “Mixed”, four “Positive”). The mixed nature
of the literature also holds true across methodologies. Regarding direct versus spillover effects of
choice mechanisms, the literature is more in favor of the latter. Three of four papers estimating
charter schools’ competitive effects report “Positive” findings, and both papers estimating vouchers’
competitive effects report “Positive” findings too. Again though, the literature does not sufficiently
address potential endogeneity concerns in these causal estimations. I detail my criticisms below.
Reasons for Mixed Findings
I propose a number of reasons why the literature on choice mechanisms and schooling
efficiency is so mixed. First, the literature estimates the effects of choice mechanisms via
22
standardized assessment scores. However, it may be the case that some parents select schools of
choice for reasons unrelated to assessment scores, such as specific academic programming, e.g., arts-
infused learning (Bifulco & Ladd, 2007), or factors like safety and extracurricular activities
(Imberman, 2011a).
Second, it is possible not all schools of choice in a study area are over-enrolled; parents may
overridingly select those with high reported assessment scores (Abdulkadiroğlu, Pathak,
Schellenberg, & Walters, 2017). As a result, studies that focus on lottery winners and lowers may
produce upwardly biased estimates of treatment effects relative to analyses of students who switch
between neighborhood public schools and schools of choice. In turn, this may explain the results
that Hoxby, Murarka, and Kang (2009) and Abdulkadiroğlu, Angrist, Dynarski, Kane, and Pathak
(2011) report for New York City and Boston charter schools, respectively.
Third and relatedly, econometric methods and definitions of treatment and control groups
vary widely across the papers surveyed. As one example, Figlio and Hart (2010) use a 5-mile radius
to estimate the competitive effect of voucher-receiving private schools on nearby traditional public
schools, whereas Imberman (2011b) employs 1-mile and 1.5-mile radii to estimate the competitive
effect of charter schools on nearby traditional schools.
Fourth, existing papers have studied a wide range of places in terms of geography: single
urban districts, all districts in a state, districts in multiple states, etc. Variations in local
sociodemographic contexts, such as the extent to which residents are segregated by race or income
(Ainsworth, 2002), probably help drive the variable findings I report. It is probable that variations in
political contexts and choice policy structures do too.
Fifth, studies tend to focus on third grade and afterwards in a student’s academic career,
because standardized assessments typically begin no earlier than third grade. Yet what if changes in
students’ proficiency levels are greater in magnitude or more frequent in prior grades? Sixth and
23
relatedly, studies rely on standardized assessments in elementary and secondary grades as the
outcome of interest. But what if measures of later outcomes like household income produce
different or stronger results, as the human capital or signaling benefits of education may materialize
much later in life (Chetty, Friedman, & Rockoff, 2014)?
Seventh, district or state policies to close ineffective charter schools may upwardly bias
estimates of charter school effects, a phenomenon acknowledged in the literature itself (e.g.,
CREDO, 2013). Eighth, the maturity of the charter sector being studied, i.e. , how many years since
charter schools began operating in a district, appears to matter both in terms of effects on their own
students (Clark, Gleason, Tuttle, & Silverberg, 2015; Ladd, Clotfelter, & Holbein, 2017) as well as
nearby traditional public school students (Imberman, 2011b).
Ninth, all of the “spillover” studies I have reviewed include data for school years 2002- 2003
or afterward. This is unfortunate timing given the massive expansion of federal and state oversight
activities in 2002 via No Child Left Behind, including added pressure to improve standardized
assessment scores. If schools of choice tend to locate in more urban areas (a sound strategy given
their need to compete for existing students, which likely requires a dense population) and urban
districts tended to respond more to accountability frameworks (feasible given their greater resources
available), this raises the prospect of significant endogeneity in estimates of spillover effects.
Tenth and finally, the literature often fails to contemplate how changes in school-level
characteristics experienced by children moving from neighborhood public schools to schools of
choice may shape changes their academic outcomes. For example, Hanushek, Kain, and Rivkin
(2009) estimate significant and negative effects of racial segregation on black American students’
standardized assessment scores in traditional public school systems. Similarly, Reardon (2016) has
shown that black American students’ greater exposure to peers in poverty relative to white students
is a significant predictor of the black-white achievement gap observed between schools in the same
24
district. These findings suggest that outcomes of school choice mechanisms in terms of efficiency
are inextricably associated with their outcomes in terms of equity.
Future Considerations for Scholars Analyzing Choice Mechanisms and Efficiency
Within the prior section, I have noted a diversity of gaps in the existing literature on school
choice mechanisms and schooling efficiency. While each of these imply future areas for academic
research, I reemphasize my final two points. First, understanding school choice mechanisms’ effects
on traditional public school students is crucial, because some economic theory supports these
mechanisms most strongly in terms of how they impact traditional public schools (e.g., Hirschman,
1970). Scholars of this phenomenon should consider ways to avoid the endogeneity of changes in
accountability frameworks; this will be difficult given the newest reauthorization of the Elementary
and Secondary Act, the Every Student Succeeds Act (Pub. L. 114-95), which shrank the federal and
state accountability apparatus.
Second, the fact that segregation of students by race or income across schools appears to
shape their academic outcomes adds weight to the idea that all scholars of school choice should keep
equity of access in mind. In doing so, they should draw comparisons between sectors (e.g., traditional
public schools versus charter schools) as well as between schools in a given sector (e.g., majority
white charter schools versus majority nonwhite charter schools).
Each of the above points hints at a larger gap in the literature on choice mechanisms and
efficiency. Of the papers I have reviewed, only one considers the general equilibrium effect that
introducing choice mechanisms has on student outcomes, i.e., the effects on both students exercising
choice and students not exercising choice (see Sass, 2004). All others demonstrate a partial
equilibrium effect, either on students exercising choice or on students remaining in neighborhood
public schools. Although solving for general equilibria is incredibly difficult, in this context it seems
to be a necessary endeavor.
25
Empirical Evidence from Charter Schools on the Equity Argument
In addition to examining the efficiency argument, a separate strand of the literature concerns
itself with the equity implications of choice mechanisms. Although incomplete, this strand has
developed a consistent body of evidence that school districts with charter schools, in particular,
experience greater segregation in their charter schools than in their traditional public schools. These
findings therefore indicate a new way for segregation to occur in school districts, one that is distinct
from the Tiebout sorting of households across attendance zones within a district or across districts
themselves (Black, 1999).
The literature fails to definitively identify which mechanism(s) are most responsible for such
segregation. Nevertheless, it does present evidence that both avoidance of “outgroups” by white or
wealthy households (Saporito, 2003) and differing choice behaviors across sociodemographic groups
contribute to this outcome. Notably, the literature does not adequately consider how the designs of
systems of choice themselves may undermine efforts to improve equity of access. Because the body
of evidence regarding equity is smaller than t hat regarding efficiency, I summarize relevant pieces of
literature individually and within-this chapter’s body rather than tabulating them. In general, works
that assess the equity implications of choice mechanisms have examined outcomes in two ways:
1. At the macroscopic level, is there evidence that systems of choice are associated with greater
segregation by race or income than traditional public schools, either by comparing schools of
choice to neighborhood schools or comparing schools of choice against each other?
2. At the household level, is there evidence that households differentially act upon systems of
choice in ways that may lead to segregation? If so, does this evidence uncover outgroup
avoidance by some households, disparate tastes for educational programming, or gaps in
households’ knowledge and use of choice along sociodemographic characteristics?
I summarize the following studies accordingly.
26
Bifulco and Ladd (2007)
Among the studies I have reviewed regarding equity of access and charter schools, two of
the most salient are those by Bifulco and Ladd (2007) and Ladd et al. (2017). Both analyses compare
the charter school sector with the traditional public school sector in North Carolina.
The former piece explores the potential for segregation via both of the channels I delineate
above. From a descriptive vantagepoint, the authors note that North Carolina’s charter schools
generally exhibit increased racial segregation relative to traditional public schools within the same
district. They then use a conditional logistic framework to understand how black American and
white households select particular charter schools for their children. Controlling for types of
educational programming, types of assessments given, enrollment size, residential distance to school,
and student-teacher ratios, they conclude that black American students are significantly most likely
to attend a charter school with a racially balanced student body (defined as 40% - 60% black
American students). Controlling for the same factors, they find that white students are significantly
most likely to attend a charter school with minimal black American students (defined as 0% - 20%
black American students).
These findings do not prove outgroup avoidance by white households attending charter
schools in North Carolina, as residential segregation along racial lines may also play an outsized role.
Still, the authors argue that the majority of black American and white students could choose to
attend charter schools across the spectrum of racial composition (p. 49). The authors acknowledge
that correlations between race, educational attainment, and income further frustrate attempts to
identify racially-motivated decision-making by households in North Carolina’s charter sector.
Of additional note is the divergent preferences by household race related to school type and
educational experience. Bifulco and Ladd report that black American households significantly prefer
schools geared towards gifted students and one that are community-oriented, while they significantly
27
avoid schools that emphasize character or moral development. In contrast, the authors find that
white households significantly prefer schools that are designed for at-risk students, are community-
oriented, focus on experiential instruction, or provide alternative assessments. These differing
educational preferences may also drive the state’s segregated charter schools.
In terms of the effects of racial segregation on academic outcomes, Bifulco and Ladd also
study whether segregation is significantly associated with changes in standardized assessment scores.
Using a student-level fixed-effects panel regression, they isolate black American students who
transfer from traditional public schools to charter schools where: (a) the share of students
considered black American is at least 60%, and (b) that share is at least 10% higher than the
traditional public school left by a particular student (in percentage point terms). Via this somewhat
arbitrary protocol, the authors find that such students tend to experience significantly less growth in
math assessment scores relative to black American students who transfer to less segregated charter
schools, as well as nonblack students in general (see Bifulco & Ladd, 2007, Table 5, p. 46). They find
no such relationship when relying on reading assessment scores.
Because of the student-level fixed-effects framework of their regression, the authors are able
to estimate the effects of segregation for only students who switch from traditional public schools to
charter schools. They cannot say anything about the academic outcomes of students who begin their
educations in the charter sector.
Ladd et al. (2017)
Follow-up work by Ladd et al. (2017) unveils another dynamic in North Carolina’s cha rter
sector. According to this more recent study, the segregation across charter schools observed by
Bifulco and Ladd (2007) actually worsened in the years leading up to 2012. This subsequent research
also reports that the share of charter students considered black American consistently decreased
28
from 1999 to 2012, to the point where a larger share of traditional public school students were black
American than charter school students. The opposite trend occurred for white students.
Analyzing the rate at which children reenroll at their charter schools year-over-year, the
authors also find that children attending majority white charter schools were more likely to return to
that school the following year than children attending majority nonwhite charter schools. They
interpret these findings as evidence that charter schools fit white families’ preferences for
segregation better than they fit nonwhite families’ preferences regarding schools.
Unlike Bifulco and Ladd (2007), Ladd et al. (2017) do not estimate th e relationship between
a charter school’s level of segregation and its students’ academic outcomes. Nevertheless, they
conclude that the growth of white students in North Carolina’s charter sector likely explains at least
part of the increase in charter students’ average performance on standardized assessments, given
that these white students tend to have higher previous standardized assessment scores.
Renzulli and Evans (2005)
In this study, the authors analyze 688 charter schools throughout the United States. They use
an ordinary least squares regression framework to predict the white student enrollment shares at
these charter schools. They specify two explanatory factors of interest: (1) the level of racial
integration in the school district containing each charter school, i.e., the extent to which each charter
school’s racial composition aligns with its district’s overall racial composition, and (2) the level of
“contact” between white and nonwhite students in the school district containing each charter
school, a factor that scales the previous integration .measure by a district’s share of nonwhite
enrollment.
Notably, the authors find that a traditional school district’s level of integration is a significant
and positive predictor of the white student enrollment shares at its charter schools. The authors also
find that a traditional district’s level of contact is a significant and negative predictor of the white
29
student shares at its charter schools. As higher contact measures imply greater portions of a district’s
students are nonwhite, it is possible that white households view charter schools in “high contact”
districts as an imperfect method of avoiding nonwhite households.
Kleitz, Weiher, Tedin, and Matland (2000), and Weiher and Tedin (2002)
Both Kleitz et al. (2000) and Weiher and Tedin (2002) look at the relationship between a
household’s racial status and its decision regarding charter school enrollment. Kleitz et al. ( 2000)
survey over 1,000 households whose children attended charter schools in 1997 or 1998 to determine
which factors influenced their selection of a charter school. Within this survey, they do not ask
parents whether the sociodemographic composition of classmates is an important factor. Of all
sociodemographic sub-groups, the proximity of a charter school to a household’s residence is most
important to low-income households. I stress this finding when discussing gaps in the literature on
charter schools and equity of access below.
Weiher and Tedin (2002) build on Kleitz et al.’s (2000) work by considering the composition
of classmates as a factor shaping households’ selection of charter schools. The authors report that
“no [sociodemographic sub-group] of respondents says that it is important to them that their
children attend schools with children who are predominantly of the same racial or ethnic group” (p.
82). Yet when looking at students transferring form traditional public schools to charter schools, the
authors find that a student’s race and ethnicity are statistically significant predictors of enrolling in a
charter school with a greater share of students falling in the same racial and ethnic grouping than the
departed traditional public school.
Summary of Findings and Gaps in the Literature
What does this selection of studies tell us about the potential for charter schools to mitigate
or exacerbate segregation in schools districts? First, within the school districts analyzed, it appears
that charter schools tend to be more segregated by race and ethnicity than traditional public schools.
30
Bifulco and Ladd (2007) further explore this phenomenon by estimating the relationship between
such segregation and academic outcomes. Their findings are consistent with the main result from
Hanushek et al.’s (2009) analysis of intra-district enrollment policies in Texas: that racial segregation
has a statistically significant and negative effect on black students’ standardized scores. They also
align with work by Reardon (2011, 2016), who finds that the segre gation of students by race and
income predicts achievement gaps across a district’s schools, although he concludes that racial
achievement gaps are actually proxies for income-based achievement gaps.
In aggregate, this literature strand suggests that the concerns lodged by critics of school
choice have validity. White households may indeed use systems of choice as a way of avoiding
nonwhite households while still residing in the same traditional school district (Renzulli & Evans,
2004). At the same time, households may engage in systems of choice in disparate ways that are
correlated with sociodemographic background. For instance, households may have differing
preferences for educational programming at schools, which may result in widened segregation in
charter schools due to the various “types” of programming they can offer (Bifulco & Ladd, 2007).
Alternatively, resource and information constraints may prevent some households from participating
in systems of choice as fully as other households; poor households may rely on more proximate
charter schools due to transportation or other costs, and thus systems of choice may reinforce
residential segregation by income and race (Weiher & Tedin, 2000). Ultimately, however, these
potential pathways and estimated correlations fall far short of causal proof, and the literature on
charter schools and equity of access suffers from multiple gaps.
First, given their descriptive and associational approaches, existing studies of the
relationships between charter schools and segregation cannot adequately control for endogenous
factors shaping household enrollment patterns. As Bifulco and Ladd (2007) acknowledge, white
families may attend charter schools with majority white children because they value the academic
31
performance levels of other students, the income levels of other students’ households, and/or the
educational attainment levels of other students’ parents. To uncover a racially-motivated mechanism
likely requires qualitative examination, survey work, or ethnography; b ut given the ways in which
white households code their language to be race-neutral when discussing schooling decisions
(Holme, 2002), even these approaches are likely to be arduous undertakings.
Second, the existing literature on charter schools and their segregating effects has largely
concerned itself with segregation by race, yet segregation by income may be just as or even more
important. Again, Reardon (2016) has shown that black American students’ greater average exposure
to peers in poverty is a significant predictor of the black-white achievement gap between schools in
the same district. In work analyzing the enrollment decisions of households in the United Kingdom,
Ball and Vincent (1998) and Reay and Ball (1998) identify a household’s income level as a significant
determinant of how that household interacts with systems of choice, with low-income households
tending to rely on proximity and children’s friend groups when selecting a school (Reay & Ball,
1998), and households discussing educational decisions with only households in the same income
group (Ball & Vincent, 1998).
Third, and much as I argue with the literature on choice mechanisms and schooling
efficiency, findings regarding charter schools’ potential to further segregation are likely dependent
upon local context. To reiterate, residential segregation by race and income is likely to play a crucial
role.
Fourth and finally, scholars of choice mechanisms and equity of access have focused their
energies on the extent to which the presence of these mechanisms – particularly charter schools – is
associated with racial segregation. They have produced a very compelling body of evidence for this
specific relationship. Nonetheless, the presence of choice mechanisms – again, particularly charter
schools – may complicate equity of access in ways unrelated to student segregation.
32
For example, and as I mention earlier in this chapter, school choice mechanisms may be
especially attractive to families that gentrify relatively nonwhite or impoverished school districts. If
this dynamic exists, it has notable implications for equity of access, and a small group of researchers
have realized as much. One of the most influential works in this vein, by the geographer Hankins
(2007), demonstrates that gentrifiers with children have taken advantage of the “neoliberalization”
of public services to create charter schools and develop a new, place-based urban space for their
community. But econometric analyses of this question paint an unclear picture. An analysis of the
presence of charter schools in Atlanta, Chicago, and Philadelphia does not find clear evidence of
gentrification leading to charter schools (Davis & Oakley, 2013). In contrast, Burdick- Will, Keels,
and Schuble (2013) demonstrate that gentrifying neighborhoods in Chicago experience decreases in
the number of children attending local traditional public schools, resulting in lower levels of funding,
drops in quality, and potentially an increased market for local charter schools. Related work finds
that neighborhood-based city revitalization efforts predict the opening of charter schools in those
neighborhoods (Keels, Burdick-Will, & Keene, 2013).
More recently, academics have asked whether the presence of charter schools predicts
neighborhood change. From a theoretical standpoint, it is feasible that the presence of charter
schools causes neighborhood change for several reasons: (1) separating school enrollment from
residence could make lower-income and higher-minority neighborhoods more appealing to higher-
income and white households; (2) the phenomenon described by Hankins (2007) could work in
reverse – school choice could attract gentrifiers by facilitating construction of new place-based
communities, akin to Johnson’s (2015) claim; (3) charter schools can give admission preference to
nearby residents; and (4) the physical construction of charter schools on distressed properties may
itself spur further redevelopment of the area. Although this literature strand is quite nascent, at least
three pieces of research have found a significant and positive relationship between the presence of
33
charter schools and subsequent gentrification or increases in property values (Beracha & Hardin,
2019; Hicks, 2017; Pearman & Swain, 2017).
If public charter schools are in fact linked with gentrification, then their very presence may
make it more difficult for poor and nonwhite households to access them. But significant holes in
this part of the literature still remain. As research has found that gentrification predicts the presence
of public charter schools and vice versa, future research should consider whether these two
phenomena fall within a larger feedback loop. Further work should also address the ways in which
the operators of public charter schools may selectively market to gentrifiers, as well as the other
gentrification-related costs that new charter schools may impose on existing communities, such as a
loss in their sense of place (Crisman & Kim, 2019; Shaw & Hagemans, 2015) .
Separately, while researchers recognize the role that school closure can play in shaping
estimates of charters’ effectiveness (CREDO, 2013), the literature on charter schools and equity of
access has barely considered how charter closures interact with existing inequalities in school
districts. In fact, only one published article has modeled charter school closure based on student-
level characteristics (Paino, Boylan, & Renzulli, 2017).
Similarly, scholars have failed to consider the locational behaviors of charter school
operators, which like gentrification and school closure, may interact with districts’ existing
inequalities and therefore amplify them. For example, charter schools may systematically locate away
from a district’s poorest residents, perhaps scared away by historically low academic outcomes and a
fear of subsequent closure. As with charter closure, only one published article has considered the
equity of access implications of charter operators’ locational decisions (Henig & MacDonald, 2002).
This does not mean that no instances of improved access exist in the literature on charter
schools. For example, a recent and promising study details the establishment of “diverse-by-design”
charter schools, which intentionally enroll students from various sociodemographic backgrounds
34
(Potter & Quick, 2018). Meanwhile, Wilson (2016) describes the benefits that segregated charter
schools may offer to historically marginalized student groups by giving them time and safe spaces to
adapt to new places, cultures, and languages. Still, much like scholars have identified increased
segregation in districts with charter schools, it seems probable that additional research will unveil
other challenges to the notion that choice mechanisms inherently improve equity of access.
To help grow the field in this regard, I now turn to my research on charter reform in the
District of Columbia. Chapter 2 begins this work by establishing the policy context for initial reform.
In doing so, it motivates my analyses on charter schools and gentrification (Chapter 3), patterns in
charter school closure (Chapter 4), and biases in charter operators’ locational decisions (Chapter 5).
35
CHAPTER 2
Reform for Whom and for What?
Establishing the Policy Context for the DC School Reform Act of 1995
We keep hearing, Children First, and it seems like it’s children last.
–Parent responding to the first District charter school closure in 1998, Washington Post
When President Bill Clinton signed the District of Columbia School Reform Act of 1995
into law on April 26, 1996 (P.L. 104-134), he authorized a dramatic shift in governance for the
District of Columbia’s elementary and secondary public education system. The School Reform Act
(SRA or the Act) is an incredibly rare instance of the federal government directly influencing a local
school district. In its scope, it imposed federal reporting and spending requirements on the District’s
traditional public schools. It also provided grant funding for curricular overhaul in the traditional
system as well. It is most renowned, though, for legalizing public charter schools within the District.
In the twenty-four years since the SRA’s authorization, the District’s institution of public
education has markedly evolved. Yet to what extent is this change attributable to the Act itself?
While some effects of the SRA are obvious, such as the 44.1% of non-adult District students
currently enrolled in charter schools, its role in catalyzing other phenomena (e.g., increases in
standardized assessment scores) remains unclear (OSSE, 2020). Accordingly, I seek to better
understand the motivating factors for the Act’s becoming law and whether it has fulfilled its lofty
ambition to create within the District a “a world-class education system that prepares students for
lifetime learning in the 21st century” (P.L. 104-34, Sec. 2101(b)).
My investigation identifies a complex set of circumstances that converged to produce the
necessary environment for passage of the SRA, not all of which directly related to the District’s
36
schools. For one, the SRA strongly symbolized a comprehensive system change to avert the
District’s growing public infrastructure crisis (McDonnell & Elmore, 1987; Stone, 2012). At the
same time, the SRA’s framing was advantageously amorphous; it grew from multiple policy
alternatives and appealed to multiple political principles, including notions of efficiency, eq uity, and
liberty (Stone, 2012). Third, the SRA’s issue area – governance and performance of the District’s
public education system – was located in a remarkably one-sided power dynamic between the federal
and District governments.
Using my findings on the political process, I develop a theory of action for the Act. Based
on the empirical evidence available, I then assess whether the Act’s multifaceted theory of action has
materialized. Although I find clear evidence of charter schools’ successful diffusion throughout the
District, I fail to confirm that the SRA – specifically via the presence of charters – has substantively
improved the efficiency or equity of the District’s public education system, or the satisfaction levels
of households using the system. As another complication, I document scholarly work that identifies
heightened patterns of segregation in charter schools relative to traditional public schools.
A major result of my investigation is the need for additional academic research. The School
Reform Act, and its legalization of charters in particular, has been the subject of many descriptive
analyses (e.g., Osborne, 2015) but very few causal ones. Complicating matters is the significant
number of other system-changing policies affecting the District’s public schools since the SRA, such
as the transition to mayoral control in 2007 (D.C. Law 17- 9). Such policies represent a “feedback
loop” of responses to the Act (Wirt & Kirst, 2005). They also make it more difficult for researchers
to isolate the Act’s effects from the other policies’.
Nevertheless, my analysis of the SRA reveals many opportunities for impactful and timely
research, as well as policy interventions. This is especially true in developing a more thorough
37
literature on charter schools’ implications for equitable access to schools, which motivates the core
analyses of this dissertation in Chapters 3, 4, and 5.
I structure this paper as follows. Section 2 outlines the political process that produced the
SRA. Due to the Act’s origins, I apply Kingdon’s (2011) seminal work on federal policymaking; I
also utilize theories from Stone’s (2012) on the “polis” and from Schneider and Ingram (1993) on
target populations and their constructions. Section 3, drawing on the theory of action developed in
section 2, describes the implementation of the SRA as authorized; it also weighs the extant literature
and data regarding its outcomes. Within this section, I discuss the Act’s unintended consequences as
well as the feedback loop of subsequent policies imposed on the District’s schools. Finally, I
conclude in Section 4 with recommendations for future research and policy.
2. The Policy Process
2.1. The District Context
2.1.1. The History of Black Americans and Whites in the District, 1800 to 1990
For many Americans, the District of Columbia is known as the federal district housing the
U.S. government. But for most of its existence, the District has also been a sizeable city home to
many non-federal employees. It is a place with a distinct and local culture (Austermuhle, 2016), such
as its mid-to-late 20th century identity as “Chocolate City” (Mock, 2019).
Seen through the lenses of agglomeration and urban economics, this local “side” of the
District seems easy to explain. In the same way as in other American cities, industries and workers
have located in the District due to a competitive advantage. In the District’s case, this advantage is
the presence of the federal government and its required functions. Yet this argument is insufficient.
Unlike other American cities, which fall within county and state governments, the District is directly
overseen by the U.S. government. Federal actions have defined its context and development in
unique and consistent ways. Nowhere is this truer than in the city’s racia l dynamics. Given the
38
District’s role as symbol for the nation and America’s extensive record of racial injustice, this is an
unfortunate and fitting exercise.
Even since the District’s inception, its racial composition and the federal government have
been intimately linked. At least in part, Thomas Jefferson, James Madison, and Alexander Hamilton
even selected its geographic location in 1790 – between the Southern states of Maryland and
Virginia – to indicate federal support for slavery and the plantation economy (Mann, n.d). Slave
trading and slave owning were initially permissible activities in the District , although they were
eventually banned in 1850 and 1862, respectively (see the Compromise of 1850, ch. 63, 9 Stat. 467,
and the District of Columbia Compensated Emancipation Act, Pub. L. 37-50).
Over the same post-Revolutionary period, the District was home to a large and growing
population of manumitted slaves. While Southern states typically barred freed black Americans from
remaining in their jurisdictions, the federal government, and therefore the District, had no such laws
(Library of Congress, n.d). In 1800, almost one-third of all Washingtonians were black American, of
whom only 16% were considered “free” (see Table 1). By 1860, i.e., two years before the
compensated emancipation of District slaves, the share of black Americans considered “free” had
risen to 78% (see Table 1).
This clustering of freed black Americans in the District foreshadows even longer-term
trends in the city’s overall racial composition. In 1800, 69.6% of Washingtonians were considered
white and 30.4% were considered black American. By 1990, these shares had inverted, with only
29.6% of residents considered white and 65.8% considered black American (see Table 1).
Five events are associated with accelerations in this general trend. The first is the
aforementioned freeing of black American slaves during the Civil War and their subsequent
relocation to the District. The second is the outbreak of World War II, the desegregation of
government-related jobs in areas like the defense industry, and the accompanying increase in black
39
American migration to the District’s metropolitan area (Austermuhle, 2016). Notably, while the
white share of Washingtonians decreased during these first two events, the total number of white
Washingtonians increased at the same time.
The third event is the desegregation of District schools in 1954, with white flight to
suburban school districts reinforced by existing and discriminatory New Deal housing policies and
rising car ownership (Knoll, 1959; Simons, 2018). Fourth is the closing of the Naval Gun Factory
and the reduction in jobs at Join Base Anacostia-Bolling in 1962, both of which encouraged
formerly-employed white households living in the southeastern portion of the District to relocate to
segregated suburbs in Maryland (Asch & Musgrove, 2018). Fifth and finally is the outbreak of civic
unrest, particularly among black American communities, in response to the assassination of Martin
Luther King Jr. in 1968. Further white flight appeared to soon follow (Brooks & Valadez, 2017).
Unlike the first two events, both shares and counts of white Washingtonians decreased
substantially during these latter three events. This was especially the case during court-ordered
school desegregation. White Washingtonians decreased by 172,602 between 1950 and 1960,
equivalent to 22% of the city’s total population in 1950. The primacy of public school access and
white concerns over school integration (Simons, 2018) are especially salient given this paper’s focus
on District charter reform.
2.1.2. The History of Home Rule in the District, 1800 to 1990
The relationship between the District’s racial composition and the federal government’s
activities is not one directional. At the same time that the District’s racial composition was inverting,
the federal government seems to have adjusted its governance of the District in response.
Table 1. Racial composition of the District of Columbia, 1800-1990 (recreated per Gibson & Young, 2002)
Year
Total
residents
White
residents*
Black American
residents*
White
share (%)
Black American
share (%)
Free Black
Americans
1800 8,144 5,672 2,472 69.6 30.4 400
1810 15,471 10,345 5,126 66.9 33.1 1,572
1820 23,336 16,058 7,278 68.8 31.2 2,758
1830 30,261 21,152 9,109 69.9 30.1 4,604
1840 33,745 23,926 9,819 70.9 29.1 6,499
1850 The Compromise of 1850 bans slave trading in the District (ch. 63, 9 Stat. 467)
1850 51,687 37,941 13,746 73.4 26.6 10,059
1860 75,080 60,763 14,316 80.9 19.1 11,131
1862 The District of Columbia Compensated Emancipation Act frees remaining slaves in the District (Pub. L. 37-50)
1863 The Emancipation Proclamation frees slaves in Confederate states
1870 131,700 88,278 43,404 67.0 33.0 -
1880 177,624 118,006 59,596 66.4 33.6 -
1890 230,392 154,695 75,572 67.1 32.8 -
1900 278,718 191,532 86,702 68.7 31.1 -
1910 331,069 236,128 94,446 71.3 28.5 -
1920 437,571 326,860 109,966 74.7 25.1 -
1930 486,869 353,981 132,068 72.7 27.1 -
1933 Various New Deal housing initiatives begin to take effect, encouraging suburbanization by urban whites (Capps, 2015)
1940 663,091 474,326 187,266 71.5 28.2 -
1941-45 World War II; the defense industry, primarily in the District’s metropolitan area, desegregates (Austermuhle, 2016)
1950 802,178 517,865 280,803 64.6 35.0 -
1954 In Bolling v. Sharpe, the Supreme Court finds the racial segregation of District public schools unconstitutional
1960 763,956 345,263 411,737 45.2 53.9 -
1962 Naval Gun Factory closes and Joint Base Anacostia-Bolling reduces capacity (U.S. Navy, n.d.a; n.d.b)
1968 Martin Luther King Jr. is assassinated; the District experiences civic unrest in the following days (Taylor, 2018)
1970 756,492 209,272 537,712 27.7 71.1 -
1980 638,333 171,768 448,906 26.9 70.3 -
1990 606,900 179,667 399,604 29.6 65.8 -
Note: Counts of residents by race include both Hispanic and non- Hispanic individuals
40
41
In the District’s earliest days, from 1802 to 1870, it functioned as a true municipality, with a
locally-elected mayor and city council responsible for its day-to-day operations (DC Council, n.d;
Austermuhle, 2013). By 1848, all white male residents were a ble to participate in these local
elections, and by 1867, all black American male residents were similarly enfranchised (Masur, 2011).
As black Americans comprised one-third of the District population in 1870, the suffrage of black
American men had significant implications for the District’s local politics, including the election of
black American officials (Masur, 2011). White District residents and businessmen concerned with
these racial changes lobbied Congress to restructure its governance of the District (Masur, 2011). In
support of these groups, in 1870 Congress asserted its ultimate governing authority over the District,
replacing the locally-elected mayor and city council members with Presidentially-appointed
individuals (DC Council, n.d). Four years later, Congress further corroded the District’s municipal
status, delegating District management to a Board of Commissioners comprised of three Presidential
appointees (Austermuhle, 2013).
For the next 100 years, all District operations continued to fall under the exclusive control of
the federal government, despite repeated calls for renewed home rule. In the late 1960s and early
1970s, leveraging the momentum of the national civil rights movement, key District residents again
lobbied Congress for some semblance of local control. A notable leader of this effort was Walter
Fauntroy, Martin Luther King Jr.’s representative in the District, who in 1971 became the District’s
first-ever non-voting Congressional delegate (Austermuhle, 2013).
In 1973, Congress passed the District of Columbia Home Rule Act (P.L. 93-198), finally
allowing District residents to once again elect a mayor and city council to manage the city’s day -to-
day operations. Yet in hindsight, this acquiescence seems poorly planned at best and outright
duplicitous at worst. Within the same law that permitted local governance in the District, Congress
preserved its ultimately authority over the District’s functions and residents, restricted local revenue
42
generation, and assigned crushing debt obligations formerly managed by the federal government
(GAO, 2003a; P.L. 93-198). As the next sub-section shows, the latter two actions acutely and
severely impacted the District’s newly-formed local government. In some ways, they may have even
created the SRA itself.
Before moving on to the District context immediately surrounding the SRA’s passage, I
again stress the relevant racial dynamics connected to the Home Rule Act’s passage. By 1970, over
70% of Washingtonians were considered black American. Many of the vocal proponents for local
District governance were black American too, including Walter Fauntroy. Further, they were
leveraging a civil rights movement most tangibly connected to black Americans. In contrast, only 16
of the 93
rd
Congress’ 533 voting members identified as black American (U.S. Government Printing
Office, 1973; U.S. House of Representatives, n.d). It is conceivable, albeit difficult to prove, that the
93
rd
Congress intentionally designed the Home Rule Act as a punitive and racially -motivated piece of
legislation, a Pyrrhic victory for District residents rather than true and lasting change (Masur, 2011).
Given welfare reforms legalized by the same 93rd Congress, this is not an outlandish theory
(Neubeck & Cazenave, 2002).
2.1.3. The State of the District in the 1990s
Regardless, by the early 1990s, the District’s local government was experiencing a crisis. It
was nearing insolvency and losing thousands of residents across all racial groups to outlying suburbs,
but especially middle and upper class households (Cohn, 1991; Cohn & Sanchez, 1994; Cunningham
& Remy, 1993; Forstall, 1996). The exit of these households, who were leaving due to the District’s
worsening financial standing and failing public service infrastructure, inexorably exacerbated the
city’s fiscal shortfall by shrinking the tax base (Abramowitz, 1990a; Bouker, 2008). As
aforementioned, these conditions were clearly associated with the budgetary implications of the
Home Rule Act (GAO, 2003a).
43
The District’s traditional public school system was not immune to this turbulence. Media
coverage from 1990 through 1995 paint a picture of a system breaking down: intentional inflation of
enrollment figures to cover operating costs; a dropout rate of 19.1% among 16 to 19 year-olds in
1990; delays to schools opening in 1994 due to fire code violations; and schools’ lacking basic
resources like soap, toilet paper, and textbooks (Abramowitz, 1990b; Brown, 1995a; Horwitz, 1995a;
Horwitz & Jordan, 1992; Washington Post, 1994). Schools across the District’s neighborhoods,
which were highly segregated by race and income, told a tale of a public education system riven in
two (REACH Project, 2019).
Superintendent Franklin Smith of the District’s public schools sought to redress these
problems by cutting school administrator positions, nominating schools for closure, and lengthening
the instructional day (Horwitz, 1993; Richardson & Ragland, 1991; Washington Post, 1993). He even
went a step further, proposing an intra-district enrollment policy for the system and privatization of
school services (Harris, 1994; Smith, 1992). He suffered defeat on both of these proposals. In
particular, his defeat on privatization highlighted two political factors: (1) the formidable strength of
District politicians and the Washington Teachers Union on local policy disputes; and (2) the racial
tensions associated with “outsider” privatization companies changing an institution long seen as
serving the black American residents of the majority black American District (Horwitz, 1994; King,
1995).
Despite reform attempts by local government officials like Smith (Sanchez, 1993), a
Republican Congress led by Speaker Newt Gingrich decided federal intervention was necessary
(Schneider & Vise, 1995b). In 1995, President Clinton approved the creation of the District of
Columbia Financial Responsibility and Management Assistance Authority or “Control Board” (P.L.
104-8). Its explicit purpose was to address the District’s twin crises of insolvency and population
44
decline (Bouker, 2008). In the name of reform, the federally appointed Control Board enjoyed broad
power over District governmental bodies and set the precedent for the SRA.
2.2 The Evolution of the SRA: Applying Theories of Policymaking
2.2.1. Kingdon’s Problem, Policy, and Political Streams, and Their Couplings
In Kingdon’s (2011) treatise on federal policymaking, a window for policy enactment
successfully opens when problem, solution, and political “streams” are successfully “coupled”. This
is conceptually similar to a convergence of interests. In the case of the SRA, each of the above
streams helped catalyze the Act’s formation. The District’s overwhelming public infrastructure and
financial issues, which bled into its public education system, presented a stark problem for federal
lawmakers, who were responsible for allocating federal tax dollars to the District (Vise, 1995; Vise &
Horwitz, 1995). The District’s issues in this regard had been accumulating for years (Bouker, 2008).
Nonetheless, the problem appears to have reached a sudden crisis status with the news of insolvency
and recordation of the largest annual deficit since Home Rule arriving during budget formulation for
the 1995 fiscal year (GAO, 1995).
This genesis of policymaking as a response to crisis is central to Kingdon’s (2011) findings,
and it suggests that the issue of school reform quickly progressed from irrelevance to significance
(Ch. 5). A simple analysis tracking media coverage of the District’s financial issues with its coverage
of District school reform corroborates this assumption (see Figure 1). It indicates that the media
played a large role in linking these problems and bringing them to the fore. Headlines of articles at
the time include: “Students, Teachers Tell Barry of a Crisis in D.C. Schools”; “D.C. Schools: ‘Do
Not Enter’?”; “Ten Ways to Turn D.C. Around”; “D.C. Dropout Rate Among Worst in U.S.: Other
Area Cities Far Above Average”; and “Barry Says D.C. Deficit Now $722 Million: Mayor To Ask
Congress to Cut City Wages” (Horwitz & Jordan, 1992; Loeb, 1995; Osborne, 1990; Schneider &
Vise, 1995a; Washington Post, 1994). Of course, the most obvious sign of this problem coupling at
45
the federal level is the fact that Congress actually made the SRA part of the District’s general
appropriations bill for the 1996 fiscal year. It is also worth noting that traditional school systems are
one of the bedrock bureaucratic institutions of American jurisdictions (Chubb & Moe, 1990). A
threat to their viability may be particularly able to provoke widespread concern among lawmakers.
Figure 1. For Washington Post articles filed under “DC Finances” and “DC School Reform” between
1990 and 1995, shares of articles published under each search term by calendar year
Note: Searches conducted within Proquest’s Washington Post archives by search term. Search period limited to
calendar years 1990, 1991, 1992, 1993, 1994, and 1995. Gross counts of published articles found for each
search term were divided, by calendar year, into the total count of published articles for their respective search
term over the 1990-1995 period to determine shares.
Turning to the policy stream, the advocacy work of charter school entrepreneurs like Albert
Shanker, who was president of the American Federation of Teachers, may have “softened up”
Congressional legislators to radical school reform via choice (Kahlenberg & Potter, 2014).
0%
5%
10%
15%
20%
25%
30%
35%
40%
1990 1991 1992 1993 1994 1995
"DC Finances" "DC School Reform"
46
Neoliberal reforms to education likely satisfied the majority of Republicans in Congress, many of
whom had been influenced by the Reagan administration’s agenda (Clark & Amiot, 1981).
Additionally, school choice should have been attractive to federal policymakers given its laissez faire
implementation and budget neutrality in a per pupil funded system. In aggregate, these facts help
explain why school choice, as a policy, was able to advance from its nascent stages rather easily
(Kingdon, 2011). Finally and separately, the SRA’s strong federal oversight tool was surely attractive
as well, since it symbolized a simple reversion to Congress’s pre -Home Rule management practices
for the District.
The swing in Congressional majority from Democrats to Republicans in 1994 abruptly
shifted the federal government’s ideology and created the requisite opportunity in the political stream.
This shift in ideology accelerated the return of federal intervention in District affairs . Republican
Representative Newt Gingrich, the new Speaker of the House, articulated a desire for Congressional
action to “move beyond the District’s financial crisis and rejuvenate the city’s schools,
neighborhoods and economy” (Schneider & Vise, 1995b). As alluded to by Gingrich, this political
energy built off the precedence-setting decision to install a federal Control Board. Importantly, when
Speaker Gingrich and Representative Steve Gunderson of Wisconsin began authoring the SRA in
1995, they and fellow lawmakers perceived a national mood in support of school choice reform
(Vise & Schneider, 1995a). Various states had recently enacted both charter school and voucher
policies, including Minnesota in 1991 (charters), California in 1992 (charters), and Gu nderson’s
home state of Wisconsin (charters in 1993 and vouchers in 1990).
1

Even if they were not truly associated, the District’s schooling and general financial issues
were coupled within the problem stream by Republican lawmakers and the media. This coupling
1
See the National Council for State Legislature’s Interactive Guide to School Choice Laws
(http://www.ncsl.org/research/education/interactive-guide-to-school-choice.aspx)
47
gave school rehabilitation a high symbolic status, making it necessary to fix the city through its
schools (Stone, 2012). It is evident that the three different streams were closely coupled to each
other at a higher level as well. By its very definition, the policy stream of school choice reform was
coupled to the problem stream of a school system in need of intervention. Similarly, the political
stream of a change in Congressional leadership was coupled with the problem stream; after all, the
new House and Senate were responsible for making federal payments to the District on an annual
basis (Fauntroy, 2003). Third, the ideological implications of the school choice policy stream meshed
with the neoliberal principles of the Republican-defined political stream, which gave the support of
“choice” strong deontological value for federal lawmakers (Stone, 2012). To add even more fuel, the
spillover effects of other states’ school choice laws had rippled over to Congress. A window for the
SRA was wide open.
2.2.2. Kingdon’s Actors Shape the SRA and Move It Towards the Policy Window
As Kingdon (2011) describes it, both agenda-setters and alternative-setters mold federal policies as
they move towards the window of enactment. Typical agenda-setters include the President,
politically appointed staff, and high-ranking members of Congress, such as heads of committees and
the Speaker of the House. Agenda-setters help move policy closer to the window. Typical alternative-
setters include special interest groups (e.g., “outsider” public officials in affected governments,
lobbying groups, and organized labor) and technical experts (e.g., consultants and academics). They
shape the policy as it moves toward the window. Kingdon suggests that some groups, such as interest
groups and outsider government officials, also enjoy some negative blocking powers for policy.
Drawing on the analysis of streams above, it is clear Speaker Gingrich was the primary
federal agenda-setter for the SRA. Republican Representative Steve Gunderson, whom Gingrich
tasked with identifying educational reform strategies for the District, became the second major
48
agenda-setter (Gunderson, 1995a). And as Gingrich and Gunderson began designing the Act in
August 1995, alternative-setters began to mold its shape and blockers moved into position.
As its name suggests, the SRA’s authors intended to change the system of the District’s
public schools. To accomplish this, they contemplated two mechanisms for shifting authority away
from the traditional public system: private school vouchers and public charter schools. Alternative-
setters responded to these two mechanisms disparately, with private school vouchers receiving
disproportionate attention. As the Republican House passed the first version of the SRA in the fall
of 1995, national interest groups converged on the voucher issue, with lobbying groups such as the
Christian Coalition supporting vouchers while the National Education Association (a national
teachers’ union) stood in firm opposition (Schneider & Sanchez, 1995). Some federal politicians also
voiced their concern, such as the District’s non-voting Congressional delegate Eleanor Holmes
Norton and other Democratic lawmakers (Strauss, 1995a). The true death knell for vouchers as a
policy alternative, though, came when two groups exercised their blocking capabilities: Democrats in
the Senate filibustered the SRA over its inclusion of vouchers, and President Clinton threatened a
veto (Schneider & Yang, 1996). Speaker Gingrich and the House had no choice but to remove
vouchers from the SRA.
The District’s local government actors had also been divided over the alternative o f
vouchers. Some, including Superintendent Smith and Mayor Marion Barry, voiced cautious support
for the alternative; others, including local school board officials, city council members, and the
Washington Teachers Union, rejected it (Vise & Horwitz, 1995; Washington Post, 1995). District
residents, caught in the crossfire, seemed at least somewhat supportive (Washington Post, 1995). But
it is unclear whether these contentions at the District level had any effect on the federal policy arena,
a microcosm of the power imbalance I describe in the next sub-section.
49

Unlike vouchers, the alternative of public charter schools received relatively little attention
by federal officials, large special interest groups, or the majority of District policymakers and
residents (Lacireno-Paquet & Holyoke, 2007). Some tension between the city’s councilmembers,
who eventually supported charters, and the local school board, who remained opposed, was present
(Loeb, 1995). However, this tension was overshadowed by the presence of a new local interest
group named Friends of Choice in Urban Schools (FOCUS), formed in 1995 to advocate for charter
schools in the District (Lacireno-Paquet & Holyoke, 2007). FOCUS enjoyed astounding invisibility
at the time – a thorough review of the Washington Post archives reveals no coverage of the group. The
group also enjoyed an “insider” relationship with Republicans in Congress, even helping draft the
charter school portion of the School Reform Act (FOCUS, 2018). This insider status with Congress
almost certainly stemmed from the social networks of FOCUS’ founder, Malcolm E. Peabody, who
had served as a Deputy Assistant Secretary within the Department Housing and Urban
Development (HUD) under President Nixon and worked under his brother and Massachusetts
Governor Endicott Peabody (Washington Monthly, 2016).
Ultimately, and perhaps distracted by the prospect of school vouchers, no parties attempted
to block the inclusion of charter schools in the law; with the support of FOCUS, they survived and
became the policy’s main reform mechanism.
2.2.3. Stone’s Theories on the Polis
Stone’s (2012) conceptions of policymaking in the polis help explain why the SRA successfully made
it through Kingdon’s policy window. First, the SRA possessed effectively nebulous framing as a
solution. Its goal of school reform was quite broad, and therefore could be interpreted and
supported by parties in a number of ways (Brown, 1995b; Stone, 2012, Ch. 7). The guiding
principles publicized by Gunderson also appealed to different conceptions of political organization:
providing households with choice conveyed a respect for liberty (Stone, 2012, Ch.11); empowering
50
poor households evinced a goal of equity; and creating a “world-class” school system, while a vague
statement, seems to connote improved efficiency (Gunderson, 1995b; Stone, 2012). I assert this
third point on efficiency (getting more out of what is being spent) is reasonable given the
outstanding concerns over the District’s and its schools’ financial viability (Stone, 2012). Second,
and as described above (see sub-sections 2.1 and 2.2), Republicans in Congress effectively made the
failing school system a synecdoche for the District’s financial collapse; reform of the District could
not be achieved without reform of the schools (Stone, 2012).
Fourth and most critically, a tremendous disparity in p ower existed between the federal
policymakers and their “beneficiaries”, i.e. the local District government and its residents. Even in
the era of Home Rule, Mayor Barry admitted Congress’s domination of District government (Vise &
Schneider, 1995b). The District’s financial crisis and Congress’s status as a funder of District
operations inevitably calcified this one-sided relationship (Schneider & Vise, 1995c). Through this
lens, some District officials’ reluctant acceptance of the SRA appears less motivated by aligned
interests and more motivated by perceptions that “‘Congress was going to it anyway’” (Brown,
1995c) and District opposition would provoke “sterner legislation” (Strauss, 1995b). While local
residents and District policymakers were able to bloc k each other, as evidenced in Superintendent
Smith’s failure to privatize schools, they had no such power over federal policymaking and were
frequently “left on the sidelines” (Vise & Schneider, 1995c). The stature of the federal Control
Board made disputing the SRA even more difficult. Infighting between the District’s Superintendent
Smith, the Teachers Union, the School Board, and the city council made just a coordinated District
response impossible (e.g., Horwitz, 1995b).
2.2.4. Targeted Populations and Policy Tools
The federal government’s construction of its target population reaffirms this power
imbalance. In short, the federal government viewed the District government as a deviant
51
bureaucracy – a “mess” of a city whose officials “fail[e d] to cooperate” (Schneider, 1995; Schneider
& Ingram, 1993; Vise & Schneider, 1995d). Critical race theorists might add that the largely white
Congress viewed the District as deviant due to its majority black American status ( e.g., Masur, 2011).
Contemporary media coverage describes District residents’ and policymakers’ suspicions of a racist
reform agenda conducted by “outsiders” (Grove, 1995; Vise & Schneider, 1995b; Woodlee & Morin,
1995). As noted in sub-section 2.1, these very suspicions had doomed Superintendent Smith’s earlier
call to privatize some school services.
McDonnell and Elmore’s (1987) and Schneider and Ingram’s (1993) research suggests that
policymakers dealing with a deviant bureaucracy may use policy tools that remove authority and/or
punish. The SRA’s design tools confirm this expectation. The implementation of school choice via
charter schools was a clear system-changing tool that shifted educational authority away from the
bureaucracy, including the locally-elected school board, to local non-profits and to District residents,
the latter of whom could now choose where to attend school.
2.2.5. The Act’s Final Design
In addition to the system-changing tool of public charter schools, the Act’s architects
included two other tools to effect change. Consistent with the federal government’s constitutional
right to District oversight, which includes the ability to review all District laws, budgets, and
financial activities, Congress required the District’s traditional school system to submit annual
corrective action reports and aligned budgets as well as enter capital spending agreements intended
to rehabilitate the traditional schools’ physical infrastructure (P.L. 104- 134, Sec. 2101(a), Sec. 2552).
On top of these mandates, Congress authorized a $2 million capacity-building curricular overhaul
grant for the creation of a “World Class Schools Task Force” (P.L. 104-34, Sec. 2311-2315)
52
2.3. A Theory of Action
The findings of sub-section 2.2 allow me to develop a theory of action for the SR A (see
Figure 2). In its design, the Act had two explicit goals: (1) to raise the performance of the District’s
public education system (as I argue, probably via heightened efficiency), and (2) to make the system
more equitable.
The SRA’s primary architects, Speaker Gingrich and Representative Gunderson, did not
specify how the Act would enhance efficiency. Presumably, this would arise from two distinct
phenomena that could reshape the “education production function” (e.g., Krueger, 1999). The
system-changing expansion of the education market via choice could: increase competition and
associated innovation (Friedman, 1962); allow charter schools to more effectively allocate resources
given their separation from the traditional schools’ bureaucracy (Chubb & Moe, 1990); and/or
encourage more efficient resource allocation and effort expended within the traditional system via a
new threat of exit (Hirschman, 1970). Second, the federal government’s oversight of the traditional
schools’ corrective action plan, annual budget, and facilities investment plan could force a more
efficient allocation of funds, one more oriented to maximizing the education production function.
For this second pathway to hold true, the federal government would need to have a more accurate
comprehension of and/or greater interest in specifying the correct input-output relationship than
the District.
The explicit goal of equity had a clearer mechanism. Namely, decoupling residential location
from school of attendance would allow poorer households to access higher-performing or preferred
schools.
Figure 2. Theory of action for the School Reform Act of 1995
Note: Explicitly stated goals in solid line boxes, implicit ones in dashed line boxes, and ultimate goals on bottom row
53
54
On top of these explicit goals, I believe it is important to mention three potential, implicit
goals as well. First, the need for school reform was ostensibly associated with the need for general
reform of the District (Vise & Horwitz, 1995). As such, it is quite feasible that the SRA’s authors
intended for it to revitalize the District by attracting new residents and expand the tax base. Second,
the SRA’s explicit goals may have masked an attempt by whites to reassert control over the majority
black American District and black American-dominated bureaucracy and/or to re-segregate the
public education system (Fauntroy, 2003). I explore the potential for this second implicit goal in
greater detail in section 3, where I consider the segregating effects of charter schools. Third, the
SRA’s authors may have shaped the Act per prevailing institutional logics of the capitalist market
and an innate cultural value of “choice” (Thornton & Ocasio, 1999).
2.4 The SRA Becomes Law
Due to all of the factors delineated in this paper, the SRA quickly moved up the federal
agenda and to the policy window in the latter half of 1995, in parallel with the District’s growing
fiscal crisis. Although the alternative of school vouchers was blocked in early 1996, the alternative of
charter schools persisted. Packaging the SRA within the annual appropriations bill for the District,
the Gingrich-led House of Representatives narrowly passed the modified Act (H.R. 3019) on March
7, 1996 (209-206). The Senate followed suit less than two weeks later on March 19 (79- 21). After a
month-long committee conference, the revised appropriations bill – and the SRA within it – arrived
at President Clinton’s desk on April 25. The next day, he signed Public Law No. 104- 134.
3. Implementation and Outcomes
With my analysis of the Act’s political process complete, I now turn to its implementation
and outcomes. To do so, I explore the effects of the SRA on the District’s public education system
and on the city writ large. I focus on the implementation of the Act’s main legislated mechanism –
public charter schools – for three reasons. First, the introduction of public charter schools was a
55

dramatic structural change to the District’s public education system and how households access it;
they are the reason the SRA is a “system-changing” piece of legislation (McDonnell & Elmore,
1987). Second, no evidence suggests the District’s traditional public schools ever engaged in the
curricular overhaul incentivized by another SRA mechanism, the one-time $2 million grant program
for a “World Class Schools Task Force”.
2
Third, the Act’s mechanism of mandated reporting and
capital spending for the traditional public school system is uninteresting to investigate due to its
binary structure of outcomes (i.e., did the District’s traditional public schools submit a report and
spend their capital funds or not?). More importantly, it was unlikely to have any strong effect as an
individual piece of legislation, given Congress’ already existing authority to oversee the District in all
respects.
3.1. Implementation of the SRA’s Charter Reform
To effect systemic change in the District’s public schools, the Act’s Congressional architects
reduced the influence of the District’s traditional school system in three ways. First, they provided
District households with greater agency in their choice of schooling by creating charter schools
accessible to all residents. Second, they granted chartering authority to both the District’s existing
Board of Education and a new independent Public Charter School Board (PCSB) (P.L. 104-34, Sec.
2214). Third, they made charter schools autonomous from the traditional public schools’ oversight
and accountability framework (P.L. 104-34, Title II-B).
The federal government thereby legislated a framework for bureaucrats and households in
the District to alter the city’s public education system. Although as the last sentence suggests, this
was only a framework. As a result, it seems natural to ask questions about whether the system
changes actually occurred. Did charter schools open? If so, how did households respond? How did

2
Future research should verify this finding, which was made by reviewing available online documentation, including
contemporary media coverage and District of Columbia Public Schools reports, where available. Verification could
involve interviews with District of Columbia Public School leaders who were working at the time.
56
the two chartering bodies manage implementation of the SRA’s charter mechanism? And can we
observe any other consequences? The following sub-sections attempt to answer these questions,
grounding descriptive data in theories of policy diffusion and street level bureaucrats.
3.1.1. Diffusion Theory
In the first chapter of his foundational piece on policy diffusion, Rogers (2010) reviews the
relative success of various innovations. From this exercise, he proposes five dimensions of an
innovation that determine the extent of its successful diffusion throughout a population: “relative
advantage”, “compatibility”, “complexity”, “trialability”, and “observability” (pp. 15-16).
In applying Rogers’ work to the District’s charter school reform, charters fit squarely within
at least three of the dimensions on a theoretical level. Charter schools are an innovation that is easy
for households to understand; in many ways, they even imitate the structure of the older magnet
school choice mechanism. Charter schools are also “trialable” for both households and the
bureaucrats authorizing them. If households do not like a particular charter school, they may exit
and enroll at another one. If bureaucrats deem a charter school to be sufficiently underperforming,
they may revoke its charter. Finally, charter schools are a highly observable innovation. They are a
physical, local, and public good. Further, they are part of an institution, namely elementary and
secondary education, that imposes compulsory attendance on the vast majority of youth.
It is unclear whether District residents would view charter schools as relatively advantageous
or compatible with their values. This is pa rticularly true given the history of the SRA’s charter
reform mechanism. As I described in section 2, the vast majority of District residents did not
strongly advocate for charter schools. Their inclusion in the SRA appears due to the support of
federal policymakers and a single local lobbying group. Regardless, Rogers’ (2010) theory suggests
the diffusion of charter schools was likely, given their trialability, their lack of complexity, and their
observability as an innovation.
57

3.1.2. The Diffusion of the District’s Charter Schools
In April 1996, the School Reform Act of 1995 became law. By the end of the calendar year,
five different organizations had received charters from the District’s Board of Education. Of these
five, Options Public Charter School and Marcus Garvey Public Charter School were the first to
open their doors (Wilgoren, 1997). Three years later, nearly 10% of the District’s public school
students were attending one of 29 different charter school campuses (Wilgoren & Strauss, 1999). In
the school year 2019-20, 44.1% of District students attended public charter schools across over 60
charter operators and over 100 distinct campuses (OSSE, 2020).
The statistics provided above demonstrate an extensive diffusion of public charter schools in
the District’s public education system. This aligns with their predicted success via Rogers’ five
dimensions of diffusion. When I delineated these five dimensions, I claimed that it is unclear on a
theoretical level whether charter schools’ would appear to be a “relative advantage” for or
“compatible” with District households’ preferences. Charters’ documented proliferation suggests
they may have appeared so for many households.
A more nuanced investigation of charter schools’ diffusion over time helps identify the
households for which charters were an appealing innovation. Figures 3.1-3.3 show the geographic
locations of charter school campuses in the District at three points in time: school years 1997-98,
2007-08, and 2017-18, respectively. The evolution of the campuses’ spatial distribution is quite
notable. In 1997-98, charter schools were rather centrally located (see Figure 3a). In 2007-08 and
2017-18, we can see that charter schools aggressively expanded to the city’s northeastern and
southeastern quadrants; however, essentially no charter schools located in the city’s northwestern
quadrant (see Figures 3b and 3c).

58
Figure 3a. Locations of operating charter school campuses, school year 1997-98
Figure 3b. Locations of operating charter school campuses, school year 2007-08
59
Figure 3c. Locations of operating charter school campuses, school year 2017-18
Why this is the case becomes clearer when we consider the distribution of households by
race and income over the city. In particular, Figures 4a and 4b depict an absence of charter school
campuses in the neighborhoods with the whitest and highest-income households in school year
2017-18, respectively.
It is difficult to believe that no households in the city’s westernmost portion, i.e., in the area
west of Rock Creek, have preferred schooling options other than the District’s traditional public
schools. In fact, recent work by the Brookings I nstitution demonstrates that the majority of private
schools within the District are located within this region (Perry, 2018), indicating that wealthy
households in the District continue to rely on private schools for school choice rather than public
charter schools. Such behavior could reflect a mentality of sunk costs or loyalty tied to investments
made (Hirschman, 1970). Alternatively, these households could prefer that their children attend
60
schools with other children of similar sociodemographic class (Holme, 2002), which is easier to
ensure in a private system with frequently high costs of entry.
Figure 4a. Locations of operating charter school campuses, school year 2017-18, and quintile
rankings of census tracts per white alone, non-Hispanic share of population, 2017 5-year American
Community Survey (ACS)
Note: 2017 total population comes from 2017 5-year ACS Table B01003; 2017 white alone, non- Hispanic total
population comes from 2017 5-year ACS Table B01001H
This does not mean the traditional public schools west of Rock Creek Park are performing
poorly. Both of the traditional high schools located in this area appear in U.S. News’ “Best High
Schools” portal for the District (U.S. News & World Report, n.d). Given the District’s traditional
schools give first right of enrollment to households within each school’s attendance zone (D.C. Law
17-9), it is unlikely many households outside of these white and wealthy neighborhoods have access
to their high-performing schools. Households within these neighborhoods may prefer that lack of
access to outsiders (Saporito, 2003).
61
Figure 4b. Locations of operating charter school campuses, school year 2017-18, and quintile
rankings of census tracts per median household income, 2017 5-year ACS
Note: 2017 median household income comes from 2017 5-year ACS Table S1901; adjustments for inflation
made using the U.S. Bureau of Labor Statistics’ Consumer Price Index (CPI) for All Urban Consumers in the
Washington-Arlington-Alexandria Metropolitan Statistical Area
Returning to Rogers’ (2010) dimensions of diffusion, the correlation between charter
schools’ locations and neighborhoods’ income levels makes sense as well. High-income households
have the ability to exit their local public schools by physically moving to another attendance zone or
paying for private school, while low-income households cannot afford to exit local public schools
similarly (Hirschman, 1970). For the latter households, the innovation of public charter schools
would be compatible with a preference for a low-cost alternative to their local public schools. If
these households are dissatisfied with their local public schools as contemporary media accounts
indicate (Loeb, 1995), charter schools may represent a relative advantage as well. To the ext ent
proximity to their residence plays a role in households selecting a particular school, charter schools
should tend to locate near the households they believe are responsible for demand (Jacobs, 2013).
62
In an example of charters’ potential to improve equity, the first charters to open in the
District did serve as exits or “counterpublics”, i.e., dedicated spaces for historically oppressed and
underserved groups, who were previously stuck in traditional public schools (Henig, Holyoke,
Lacireno-Paquet, & Moser, 2001; Wilson, 2016). Marcus Garvey and Options, the first two charter
schools to open in the District, are salient examples. Marcus Garvey provided students with an
Afrocentric curriculum not available in the traditional public schools. Options exclusively served
students with special education needs. Even within the first several years of their legalization,
however, public charter schools opening in the District appeared less likely to be such
counterpublics (Henig et al., 2001). I hypothesize why th is shift occurred in sub-section 3.4.
In the preceding paragraph, I have used the past tense in describing Marcus Garvey and
Options. This is because both organizations no longer exist, a fact that bridges to a final empirical
application of Rogers’ diffusion theory: the high level of trialability of charter schools. Of the 36
organizations that opened public charter schools in the first five school years after legalization ( i.e.,
between school years 1996-97 and 2000-01), almost two-thirds (23) no longer operate any schools in
the District (PCSB, 2019a). Of the 23 such organizations that closed, reasons for closure vary. What
is most notable, though, is that only 11 closures were associated with low academic performance. In
contrast, a majority of closures (15) related to financial or management deficiencies.
These numbers suggest a fundamental tradeoff in the autonomy charter schools are granted.
They receive independence by agreeing to be highly accountable to chartering authorizers and
families; but this accountability is often on an annual basis, when the District completes annual
reviews and families enter the school lottery. Decisions of charter renewals are even more
infrequent, typically over five-year intervals (see Chapter 4). In intervening periods, it may be quite
difficult for charter authorizers to recognize poor performance and intervene on a timely basis.
63
Studying this phenomenon forms the crux of the next sub- section, where I evaluate charter reform
as implemented by street level bureaucrats.
3.1.3. Street Level Bureaucrats and the Principal-Agent Problem
Over the past several decades, scholars have considered how policies as implemented differ
from their initial designs (McLaughlin, 1987). Naturally, the “street level bureaucrats” responsible
for implementation have become a focal point for scholars (Weatherley & Lipsky, 1977). Academics
from numerous fields have identified ways that these individuals shape policies on the ground,
including via schema and institutional logics (Marsh, 2012; Spillane, Reiser, & Gomez, 2006). In this
case, I apply a framing of bureaucrats as agents within a principal -agent relationship (Loeb &
McEwan, 2006).
Principal-agent theory concerns the behaviors of actors (i.e., agents) and the groups
responsible for measuring the performance of these actors across stipulated criteria (i.e., principals).
Its conceptual framework suggests problems such as: (1) principals lack the ability to collect full
information on agents’ performance across one or more criteria; (2) agents are often incentivized to
align their effort with only those areas that principals are most able to assess; and (3) agents may
have incentives to behave in ways that contradict what principals desire (Heinrich & Marschke ,
2010).
Looking at the implementation of the SRA’s charter reforms, two sets of such agents are
apparent. The first set are the Board of Education and the PCSB, which were responsible for
approving charter applications and overseeing charter schools’ performance. The Board of
Education’s chartering authority fell within an array of existing responsibilities to the District’s
traditional public schools. In contrast, the new Public Charter School Board established by the SRA
had no responsibilities other than charter authorization and oversight. While both bodies had diffuse
principals overseeing their performance in the form of District residents and government officials,
64

the Board of Education was also answerable to the District’s traditional public school system and
the households enrolled in that system.
Principal-agent theory posits that agents will allocate their efforts to the areas upon which
principals assess their performance (Heinrich & Marschke, 2010). In this respect, it is clear that the
two chartering agents faced differing incentive structures. The Board of Education had a wide set of
responsibilities, many of which did not relate to public charter schools. One such responsibility was
the management of the DCPS system until Mayor Adrian Fenty assumed control in 2007. The
Public Charter School Board’s sole responsibility was the approval of charters and oversight of
chartered schools. Principal-agent theory suggests the Charter School Board would expend greater
effort on reviewing charter applications and monitoring charters’ performance. Further, it is
plausible that the Board of Education, with a vested interest in the traditional DCPS system, feared
the threat of a competitive charter sector and had an incentive to approve lower quality charter
applications, unlike the Public Charter School Board.
The second set of agents engaged in the implementation of the District’s charter reform are
the charter leaders themselves. These charter leaders have more narrowly defined principals than the
chartering bodies, specifically the body that chartered their organization and the households their
organization serves. Of course, charter leaders’ principals do not include the traditional public
schools’ bureaucratic administration. Given their oversight by both the households they serve and
the government body who chartered their organization, charter leaders may act differently as agents
depending on to whom they feel most accountable. Who chooses to become a charter leader – who
chooses to become a street level bureaucrat – may have changed over time as the District’s
demographics or the behavior of chartering bodies shifted.

65
3.1.4. The Chartering Bodies and Charter Leaders as Agents
Again, the disparate responsibilities of the Board of Education and the PCSB suggest
differences in their behavior as agents. The history of District charter reform confirms this
prediction. Soon after the SRA’s enactment, the Board of Education became viewed as less
competent in its management of charter schools than the PCSB. A Washington Post article from 1998
lauds the PCSB’s comprehensive accountability scheme relative to the Board, commending its
“approving an accountability plan for each [charter school]…. [and] providing consulting assistance
to schools as they develop their plans” (Finn, Manno, & Vanourek, 1998).
This stands in stark contrast to media coverage of the Board’s approval of Marcus Garvey’s
charter in 1996, which described it as “born of confusion” and the Board as “[misunderstanding] its
role from the beginning” (Wilgoren, 1996). A 2005 assessment on the returns of the District’s
charter school reform noted that “charter schools authorized by the Board of Education are, on
average, not performing as well as those authorized by the PCSB” (Mead, 2005). As supporting
evidence, the report pointed to the Board of Education’s revocation of seven charters since 1996,
compared to the PCSB’s revocation of only one. Perhaps given these results, the District removed
chartering authority from the Board of Education in 2007 (D.C. Law 17- 9). Since that time, the
Public Charter School Board has been the sole body approving and overseeing charters in the
District.
The composition of charter organizations acting as agents within the District has changed
over time too. In sub-section 3.1.2, I noted that the first charter schools in the District tended to be
havens for certain student sub-groups; the organizations running such schools were locally based
and typically operated a single campus. More recently, the organizations operating charter schools in
the District are increasingly likely to operate multiple campuses, including in other cities and states.
66
In school year 2019-20, I estimate that 62 of 94 non-adult charter campuses, i.e., 66%, were helmed
by an organization that operated multiple campuses.
I posit this change has occurred due to the more stringent oversight exerted by the PCSB as
principal, especially since it assumed the Board of Education’s oversight authority in 2007. Because
they operate multiple schools, multi-campus operators tend to have centralized staff in charge of
operations, budgeting, financing, etc. At the same time, because they operate multiple schools, I
hypothesize multi-campus operators are less likely than one-campus charter organizations to offer
very specialized curricula for specific types of households; doing so would make it more difficult to
meet necessary enrollment figures. Due to their more sophisticated management schemes, it is likely
that multi-campus operators are better able to fulfill the PCSB’s accountability requirements. Again,
sub-section 3.1.2 revealed that the majority of early charter school closures were due to “financial
and management deficiencies”. Yet the distance between PCSB as principal and charter school
leaders as agents continues to be wide, as evidenced by the continued management issues in the
District’s charter schools today (e.g., Stein, 2018a). Thus, the autonomy of charters in the District
may have negative effects due to the way it has distanced them from principals. Charters may
overcome these negative consequences of autonomy if they have positive effects on the efficiency or
equity of public education. The next sub-section weights the evidence for this case.
3.2. Did Charter Schools Improve Efficiency or Equity?
3.2.1. An Examination of Efficiency via Standardized Assessment Scores
In sub-section 2.3, I argued the Act’s goal of “world class” schools had implications tied to
reshaping – or making more efficient – the “education production function” of the District’s public
education system. To measure whether this has occurred, I rely on measures of District students’
standardized assessment scores. As my theory of action articulates (see Figure 2), and per the
theoretical justifications I summarize in this dissertation’s literature review ( see Chapter 1), improving
67
students’ assessment scores while keeping investment levels constant could occur for at least three
reasons.
First, charter schools are institutions unaccountable to the traditional public school system
and its bureaucracy. Such a redefinition of the principal-agent relationship may allow charter schools
to more effectively allocate resources and implement innovative practices in a competitive
environment (Chubb & Moe, 1990, Ch. 1-2). Second, the threat of a credible exit by households to
charter schools may incentivize the traditional public school system to improve its services
(Hirschman, 1970). Third, the presence of many enrollment options for families could result a more
competitive marketplace, which could boost outcomes by incentivizing greater effort and innovation
and/or by matching household preferences with actual schools attended (Friedman, 1962).
Of course, the District’s foundational per pupil funding level almost doubled between fiscal
years 2002 and 2019, from $5,907 to $10,658 (OSSE, 2008; OCFO, 2018). If the SRA were going to
significantly improve the efficiency of the District’s public education system via higher standardized
assessment scores, the increase in funds invested means a large increase in standardized assessment
scores should occur.
Nonetheless, using students’ standardized test scores as a barometer does not indicate a
dramatic increase in efficiency attributable to charter reform. The standardized test scores of District
students in both traditional public schools and charter schools, particularly on NAEP, have
undoubtedly increased over the past two decades (ORA, 2015). Analysis conducted by the Urban
Institute finds that: (1) changes in District students’ demographics due to gentrification explain only
some of the increase in NAEP scores; (2) charter schools do not appear to systematically
outperform traditional public schools; and (3) it is difficult to determine which of the District’s
policy interventions are responsible for this increase in scores (Blagg & Chingos, 2016). Like the
Urban Institute’s work, which simply compared the magnitudes of changes in sub-populations’
68
average scores over time relative to the District’s demographic composition per these sub-
populations, the vast majority of analyses conducted regarding changes in District students’ test
scores have been descriptive in nature. Not only are these studies mixed on how the District’s
traditional public schools have performed relative to public charter schools (e.g., ODCA, 2014;
Osborne, 2015), but their descriptive nature renders them unable to identify a causal effect of
charter reform on changes in academic performances.
Yet a couple of econometric analyses do exist that purport to show a causal effect of District
charter schools on student performance in said charter schools. I argue each has notable flaws.
Nichols and Özek (2010) employ an instrumental variable framework to estimate that students
opting to enroll in charter schools are expected to experience increases in District standardized
assessment scores of approximately 15% of a standard deviation, relative to “similar” students who
remain in neighborhood public schools. The authors’ instrument used – the student’s residential
proximity to a charter school – seems a poor choice. Studies have shown that households’
preferences for academic quality trump proximity in cities with choice (e.g., Teske, Fitzpatrick, &
O’Brien, 2009). To the extent households’ choices to enter the choice market relate to their
children’s academic outcomes, such endogeneity weakens the instrument.
Another analysis by CREDO (2015) estimates that the average marginal effects of a District
public charter school on the growth of a student’s reading and math scores on District standardized
assessments were approximately 0.10 and 0.13 standard deviations, respectively, relative to
traditional public schools. However, given the majority of non-adult charter students in the District
are black American (77.5% as of the 2018-19 school year, PCSB, 2020a) and that the proficiency
rates of black American students on the District’s standardized assessment fall so far below white
students (Weeden & Jacobson, 2017), this average marginal effect likely translates to a very low
effect size. Although it is hard to compare effect sizes to achievement gaps, 10% of a standard
69
deviation does not seem to overcome black-white profiency rate gaps of 50% and 48% for English
language arts and math (in percentage point terms) on the District’s PARCC assessment (PCSB,
2019b), respectively. Regarding the potential for charter schools to have a spillover effect on
traditional public schools, no empirical analysis is available. Given this is one of the theoretical
justifications for charter schools (see Chapter 1), this represents a significant gap in the literature.
3.2.2. An Examination of Equity
It is unclear whether the SRA’s charter reform has improved the efficiency of the public
education system, at least as measured by standardized assessment scores. Such scores have
increased modestly, but it is currently impossible to ascribe these increases to the SRA; at the same
time, District taxpayers’ investments in public schools have risen tremendously.
The evidence for the SRA’s effects on equity is similarly ambiguous. In theory, charter
reform could improve equity by improving access to educational opportunities for disadvantaged
households, and indeed, the opening of charter schools has potentially provided historically
underserved and marginalized households with a valuable exit from the traditional public school
system (Frey, 2002). Unfortunately, the very types of charter schools that may be most valuable to
these households – the counterpublics described by Wilson (2016) – have frequently closed (e.g.,
Haynes, 2005; Stein, 2018b; Strauss & Slevin, 1998; Turque, 2009). Such school closures present a
pressing potential equity issue if they disproportionately affect nonwhite, poor, or otherwise
vulnerable students, one that contradicts Representative Gunderson’s original goals for the SRA
(Gunderson, 1995a). Academic research is needed to understand whether this is the case or not.
Separately, and despite the fact that the District requires each charter operator to determine
its own operating locations, only one academic study assesses patterns in District charter operators’
locational decisions (Henig & MacDonald, 2002). Among other things, the authors report that
charter campuses are significantly more likely to be located in tracts with low rates of student
70
proficiency on the Stanford 9 math exam, high levels of homeownership, high shares of the
population considered black American or Latina/o, higher numbers of vacant public school
buildings, and lower numbers of private schools. They also find charter campuses are significantly
more likely to be located in tracts that contain at least one rail transit station, are close to the
District’s downtown area, and whose residents participate in local elections at high rates. However,
these findings may no longer be relevant, as the study assesses locational decisions near the
beginning of charter reform, in school year 1999-2000. An updated analysis of locational decisions is
vital, as such decisions may have significant equity of access implications, e.g., if new charter
campuses tend to cluster away from nonwhite or poor residential areas.
My review of charters’ efficiency also highlights another grave equity concern: the large
achievement gap between white and black American students that continues to yawn, and even
grow, despite the SRA’s charter reform. The share of children in the District’s affluent Ward 3 who
are considered “college and career ready” via PARCC examination scores is more than 50
percentage points higher than the share in the District’s Ward 7, which is over 95% black American
(Balingit & McLaren, 2017; see Chapter 4).
3.2.3. Subsequent System-changing Reforms
Perhaps because of these unclear efficiency and equity outcomes, federal and District
politicians have acted to impose additional system -changing reforms. In 2004, the federal
government enacted legislation providing private-school vouchers to 2,000 District residents, adding
a second choice mechanism for households (P.L. 108-199). In 2007, the District passed the Public
Education Reform Amendment Act (PERAA) (D.C. Law 17 -9), which established mayoral control
over all public schools, made PCSB the sole chartering authority, and permitted open enrollment in
the traditional public school system. Most notoriously, education reformist Michelle Rhee became
chancellor of the traditional public school system in 2007. She quickly set in motion many new
71
policies, including the mass closure of schools (Brown, 2012), the transition to a merit-based teacher
compensation and retention regime known as IMPACT (Haynes, 2008), and even the co -
management of traditional public schools by CMOs (Chandler, 2015 b). The presence of these
additional policies makes it extremely difficult to isolate charter reform’s effects on the efficiency or
equity of the District’s public education system. I re-examine these reforms in sub-section 3.4.
3.3. Other Effects of Charter Schools, Particularly Gentrification and Segregation
I have argued that the federal government’s enactment of the SRA was inextricably tied to
efforts to revitalize the District, which was nearing fiscal insolvency and hemorrhaging higher
income households. The District has significantly gentrified since the passage of the SRA. From
2000 to 2018, the District’s median household income increased from $60,472 to $82,604; in the
same period, the share of District residents aged 25 and older with at least a bachelor’s degree has
increased from 39.1% to 57.6%.
3
The fiscal condition of the District has improved as well, likely in
part due to this gentrification. Its bonds, which infamously received “junk” status in the 1990s (New
York Times News Service, 1995), are now rated up to AAA by S&P (Washington, DC Investor
Relations, n.d).
Are charter schools partially responsible for the District’s gentrification? This is a question of
vital importance given the SRA’s virtually explicit link between improving schools to improve the
city. Until recently, though, this a question that researchers have not widely identified (Jordan &
Gallagher, 2015). The only piece to study this question using District data employs a spatial analysis
correlating tracts undergoing gentrification with tracts that contain clusters of public charter schools;
but it falls short of estimating a causal relationship (Kerr, 2012). As with the spatial and
sociodemographic patterns of charter school closures and locational decisions for new campuses,
3
2000 data come from the U.S. Census 2000 Summary File 3. 201 8 income data come from the 2018 5-year ACS Table
S1901, and 2018 educational attainment data come from the 2018 5 -year ACS Table S1501.
72
this unanswered question of whether District charters cause or stem from gentrification is a glaring
hole in the literature on the SRA.
I have further stressed we must also situate understandings of the SRA – and charter reform
generally – within the broader neoliberal movement for public service provision based on choice
(Wolfe, 2009). The presence of choice may be beneficial by better matching households with their
preferred schools (Friedman, 1962). The exercise of choice may also have intrinsic value, which
aligns with Stone’s (2012) political conception of liberty. If this is the case, it is possible that
households exercising their right to school choice in the District are more satisfied than households
who do not. The few studies conducted to date show that District parents’ satisfaction levels with
schooling increase directly after selecting a charter school but attenuate over time. After five years,
satisfaction levels of parents whose children attend charters are largely indistinguishable from those
of parents whose children attend traditional public schools (Buckley & Schneider, 2006; Schneider &
Buckley, 2003). However, such studies are more than ten years old, which suggests value in
reapplying them to new data.
Much like the state of evidence on efficiency and equity, it is unclear whether the SRA’s
charter reform has led to the gentrification its authors hoped for or if it increased household
satisfaction with schools. However, one consequence of charter reform does stand out: the
augmented segregation of the District’s public education system, both between charter schools and
between the charter and traditional sectors. Black American, Latino/a, and English learners all have
higher exposure rates
4
in charter schools relative to traditional public schools (Jacobs, 2013). Over
85% of the District’s charter enrollment continues to be black American and Latino/a (see Appendix
4
An exposure rate describes how likely a student with particular sociodemographic characteristics is to encounter
students with the same sociodemographic characteristics in his/her school. For instance, an exposure rate of 90% for
black American students means that the typical black American student goes to a school where the 90% of enrolled
students are black American.
73
L). Recent work by the Civil Rights Project has concluded that gentrification of the District has
improved the traditional public schools’ racial diversity more than charter schools’ (Orfield & Ee,
2017, p. 36). In this light, charter schools’ choice -structured system may be fulfilling a latent
preference of white households for “status” segregation by race and disability (Holme, 2002). Given
correlations of race/ethnicity and sociodemographic class in the District, these segregating effects
may directly translate into the achievement gaps on standardized assessments I have presented
(Reardon, 2016). It therefore seems reasonable to hypothesize that the SRA’s architects actively used
charter reform as a way to appease disadvantaged households while maintaining the support of
white and wealthy households. By preserving attendance zoning for the District’s traditional public
school system, Congress guaranteed that white and wealthy District households, ensconced in their
own neighborhoods, would enjoy continued access to exclusively white and wealthy traditional
public schools (Chapman & Donor, 2015). From an empirical standpoint, however, such motivation
is difficult to prove.
3.4. The Policy Feedback Loop
The School Reform Act was a federally-dictated, system-changing law intended to broadly
reshape the District’s public education system. As I describe in sub-section 3.2.3, I believe the SRA’s
unclear effects have inspired the majority of its “feedback loop”, evident in subsequent system-
changing reforms (Wirt & Kirst, 2005). Republicans in Congress, initially rebuffed during the SRA’s
drafting, finally succeeded in legalizing private school vouchers for the District in 2004 (P.L. 108-
199). While political ideology probably played an outsized role, this action may have been an attempt
to aid the SRA in effecting systemic change as well. District officials, continuing to face negative
media depictions of their public education system well after the passage of the SRA (e.g., Keating &
Haynes, 2007), introduced system-changing legislation too. In 2007, Mayor Adrian Fenty initiated
mayoral control over the public education system and introduced an open enrollment policy for the
74

traditional schools (D.C. Law 17-9). His appointment of Michelle Rhee as chancellor of the
traditional schools, made possible only by the authority he assumed from the District’s Board of
Education, was a harbinger of significant changes to the traditional schools’ structure, which I
summarize in sub-section 3.2.3. The fact that the District took dramatic action in 2007 may have
also stemmed, at least in part, from a desire to control reforms begun at the federal level in the
1990s.
One visible issue in the SRA’s implementation did arise and inform a direct policy response.
Due to the poor performance of charter schools authorized by the District’s Board of Education,
Mayor Fenty stripped the Board of Education of its chartering powers when he assumed control in
2007.
4. Directions for Future Research

Over the course of this paper, I have noted the many unanswered questions regarding
charter schools’ efficacy in creating a “world-class” school system for the District. It is undecided
whether charter schools caused the gentrification the Act’s authors hoped for; it is similarly
undecided whether they continue to multiply due to gentrification. While it seems charter schools
may not significantly improve citizens’ satisfaction with schools through an enhanced sense of
liberty, the literature is incomplete. Despite this lack of clarity, charter schools have become a
District institution, and racial segregation across schools appears to have worsened as a result. Taken
together, these facts suggest a convergence of interests undergirding charters’ growth. Charters’
institutional logic of access, even if it does not play out in reality, seems attractive to those on both
the “left” and “right” sides of the ideological spectrum; charters may be exits and safe havens, or
they may be the paragon of neoliberal reform. And while charters’ may not satisfy households’
preferences for liberty, they may satisfy white households’ preferences for racial segregation. In
short, it does not seem that charter schools are going away anytime soon, in the District and
75
elsewhere, regardless of their actual performance records. Thus, examining policies like the SRA is a
paramount exercise.
My theory of action identifies several large goals of the SRA’s Congressional authors. As is
often the case with education policy work, understanding how these goals have been – or have not
been – met requires the contributions of multiple academic fields, such as economics, planning,
sociology, and geography. In short, many areas to expand the literature on the SRA are quite evident
and span multiples schools of thought.
Still, after reviewing the extant literature, what is most clear is the differing levels of attention
paid to the various goals of the SRA. In particular, questions about the SRA’s outcomes in terms of
performance (and probably efficiency) appear to outweigh questions about the SRA’s outcomes in
terms of equity. Many researchers have attempted to connect District students’ standardized
assessment scores to charter reform. Fewer researchers have raised concerns over segregation in the
District’s traditional public and charter schools. Almost none have explored other questions of
households’ access to schools within systems of choice, which again is the motivation for my work.
I believe this last point is vital, and it presents an opportun ity to re-key the analysis of choice
mechanisms, from focusing on performance to one focusing on equity of access. Currently, we
cannot say whether charter schools in the District cause neighborhoods to gentrify or whether they
choose to locate in particular neighborhoods, potentially those that are already gentrifying. Yet this
has clear implications of which households are most likely to access these schools. Gentrifying
neighborhoods may displace residents and thereby force them to attend different schools; they may
also “crowd out” existing residents from accessing charter schools due to greater demand for a
limited number of seats. Furthermore, if charter schools do cause gentrification, this may
disproportionately affect certain households, a clear detriment to equity of access. More generally,
we have not identified the factors that most greatly influence the locational decisions of charter
76
operators. And we do not understand patterns in charter school closure, and whether closures have
disproportionate effects on certain student populations. All of the above require concerted academic
investigation.
Finally, I am gravely concerned by the number of articles (both academic and otherwise) that
purport to demonstrate a significant and positive effect of District charter schools on charter
students’ academic performance levels. My concern primarily relates to the translation of these
findings to pragmatic context. A substantial body of evidence shows continuing and vast gaps in
students’ performances along sociodemographic lines, with white and/or wealthy students
significantly outperforming others. Whatever effects are rightfully attributable District charters, it is
becoming clear they are not enough to overcome this gap.
In closing, I return to this paper’s epigraph. Public school systems like the District’s are
incredibly large and complex entities. They connect to virtually all aspects of the city, operating on
local tax dollars, providing employment opportunities for residents, potentially even shaping urban
development, and, of course, educating children. Too often children come last in the considerations
of education policymakers, when they are the very reason that schools exist. We must never lose
sight of this fact.
77
CHAPTER 3
Assessing the Relationship between
Charter Schools and Neighborhood Gentrification in the District of Columbia
Tired of seeing money siphoned from neighborhood schools into the uncertain hands of
charter operators, a group of public school parents filed a lawsuit in 2004, accusing city and
federal officials of “creating a two-track system” of education that favors charters and
impoverishes children who remain in the D.C. school system. The lawsuit accused the city of
promoting the Two Rivers Public Charter School east of Capitol Hill so white and middle-
class parents could escape neighborhood schools that are “too black”.
–“The Future of D.C. Public Schools: Traditional or Charter Education”, Washington Post, 2006
For several decades, various academic fields have examined the connections between
neighborhoods and their public schools. Such research has typically identified significant relationships
between neighborhoods’ sociodemographic compositions ( e.g., race and income), financial
characteristics (e.g., home values), and measures of school quality ( e.g., standardized assessment scores).
Due to the traditional organization of schooling along geographic boundaries, these examinations
have often drawn comparisons across districts or their attendance areas (e.g., Black, 1999).
Charter school reform, however, has made more intricate the potential relationship between
neighborhoods and schools by allowing diverse public schooling options within districts and their
attendance areas. And because urban charter schools have proliferated at the same time as the
substantial gentrification of cities, an emergent strand of literature now attempts to connect the two
phenomena. While this strand is underdeveloped, it suggests charter schools may both result from
and result in the gentrification of certain neighborhoods within a given city.
Still, more work is necessary to refine the prospective connection between charter schools
and urban gentrification. I therefore seek to expand the field in several ways. First, I develop and
78
utilize an 18-year longitudinal dataset that contains the location of each charter school campus in
each school year within the District of Columbia, studying a longer period of charter school
presence than most existing papers. Second, I consider multiple and multi-dimensional definitions of
gentrification that account for different aspects and stages of the phenomenon. Third, I consider
both overall charter school presence as well as the presence of different “types” of charter schools
(e.g., adult-serving schools, Montessori or learner-centered schools, etc.). Fourth, I explicitly control
for the presence of rail transit, which is known to influence neighborhood characteristics and may
also influence where charter schools choose to operate, the latter for both neighborhood zoning and
enrollment reasons. In accordance with the literature, I hypothesize that: (1) neighborhoods’
exposure to charter school campuses is significantly associated with their levels of residential
gentrification; and (2) the charter school-gentrification relationship significantly varies by type(s) of
charter schools present.
The District of Columbia is a fitting place to test my hypotheses for multiple reasons. First,
it is currently home to over 60 unique charter local educational agen cies (LEAs), enrolls
approximately half of all public school students, and has been operating since school year 1996-97
(Gaines & Ly, 1996; also see section 4 and Appendices C and D). Second, improving District
households’ equity of access was an explicit goal of Congressional lawmakers when they drafted the
District’s charter school law in 1995-96 (Gunderson, 1995a, 1995b; Pub. L. No. 104-134). Third, and
potentially in tension with making access to schools more equitable, historical evidence sugg ests that
charter school reform was part of a larger Congressional agenda to revitalize and gentrify the
District (Bouker, 2008; Vise & Horwitz, 1995).
I develop a variety of descriptive analyses showing that: (1) the District has gentrified in a
spatially uneven way, i.e., in certain neighborhoods; (2) the expansion of charter school campuses has
occurred in a similarly uneven way; (3) within the physical distribution of charter campuses across
79
the District, significant variation exists in proliferation by type of school; and (4) sociodemographic
enrollment within District charter schools has been uneven across multiple dimensions as well,
including: the time at which a charter LEA began operating in the District, the type of school a
student attends, and the neighborhood in which a charter school campus operates.
I then present several regression analyses that empirically estimate the relationship between
charter school presence and neighborhood gentrification within the District, again considering
multiple measures of gentrification. In doing so, I identify a significant link between the overall
presence of charter schools and neighborhood gentrification, as well as significant links between
certain types of charter schools and levels of neighborhood gentrification. Thus, I generate
supporting evidence for each of my hypotheses.
My findings have important implications for both policymakers, scholars, and the general
public. First, they highlight a significant relationship between charter school reform and urban
gentrification at the neighborhood level, a key finding at tension with equity-based justifications for
charter schools (Forman, 2004). Second, they reveal variation in the charter school-neighborhood
relationship that depends upon type of charter school – a complexity that this thread of literature
has not yet considered. Third, they underscore a new way in which inequitable access to urban
schools may occur: the gentrification of neighborhoods surrounding charter schools, and the
subsequent “crowding out” of long-residing, displaced, or otherwise disadvantaged households from
those charter schools due to limited enrollment space and physical inaccessibility.
This paper is structured as follows. Section 2 provides a review of the literature on urban
gentrification and the connection between charter schools and neighborhood-level gentrification in
cities. Section 3 describes the data used in my analyses. Section 4 provides descriptive analyses
regarding the gentrification of District neighborhoods, the proliferation of charter school campuses
in the District, and enrollment trends in charter schools both temporally and spatially. Section 5
80
presents results from regression analyses that empirically estimate the relationship between a District
neighborhood’s exposure to charter schools and its level of gentrification. Section 6 discusses my
findings and concludes.
2. Literature Review
2.1. Defining, Measuring, and Understanding Urban Gentrification
Over the past several decades, the phenomenon of urban gentrification has received
attention from a variety of academic fields, including urban geography (e.g., Hankins, 2007),
economics (e.g., Guerrieri, Hartley, & Hurst, 2013), planning ( e.g., Freeman & Braconi, 2004), and
urban studies (e.g., Marcuse, 1985).
In their definitions of urban gentrification, academics are consistent in at least three ways.
First, they characterize urban gentrification as a phenomenon occurring at the neighborhood scale
(e.g., Marcuse, 1985). Second, they agree that a necessary component of urban gentrification is an
increase in a neighborhood’s income or wealth, accompanied by an increase in its housing prices
(Newman & Wyly, 2006; Smith, 1979). Third, and frequently implicitly, they identify shifts in a
neighborhood’s income or wealth as gentrification only if such changes are attributable to in-moving
households with higher incomes or greater wealth than existing residents (Vigdor, 2002).
Aside from the above three facets, though, scholars’ definitions of urban gentrification
differ. Some consider an upward shift in residents’ educational attainments as an integral component
of gentrification’s definition (Brummet & Reed, 2019; Freeman, 2005). Others include changes in
neighborhoods’ racial or ethnic compositions, typically looking for an increase in the share of white
alone, non-Hispanic residents (Ellen, Horn, & Reed, 2019). Still others associate gentrification with
younger household heads or families without children (Hwang & Lin, 2016, p. 14).
Studies also differ in their measurement periods for gentrification and hence how quickly it
can manifest. Hackworth (2002) measures neighborhood changes over a 30-year period, from 1960
81
to 1990. In contrast, Guerrieri et al. (2013) analyze changes in housing price data o ver an
approximately 7-year span.
Complicating the question of measurement periods is that gentrification may happen over
multiple stages (Zuk, Bierbaum, Chapple, Gorska, & Loukaitou-Sideris, 2018). To provide a
hypothetical example, a neighborhood just beginning to experience gentrification pressures may see
an increase in the share of adult residents who are young, white professionals. Years later, these
early-stage “gentrifiers” may form families and have children in the same neighborhood. Depending
on how this happens, the share of children who are white may increase in that latter period, despite
the share of adult residents who are white remaining constant during that timeframe. Hankins (2007)
observes this dynamic in her study of gentrification and charter schools in Atlanta.
Researchers additionally diverge on whether gentrification is an absolute or relative process.
Some assume that whole cities may gentrify via gentrification in each of their neighborhoods
(Newman & Wyly, 2006). Conversely, others assume that gentrification is a process occurring
relative to citywide dynamics (Freeman, 2005).
Scholars have also identified associations between certain physical amenities and the
gentrification of nearby neighborhoods. One such amenity is rail transit (Baker & Lee, 2019; Lin,
2002). Another notable one is charter schools (Burdick-Will, Keels, & Schuble, 2013). Both the
academic literature and contemporary media coverage have noted that other place-specific amenities,
like restaurants and other retail spaces, tend to change as a neighborhood’s composition alters in
terms of income, race, or ethnicity (Shaw & Hagemans, 2015; Toth, 2015; Zukin et al., 2009).
Because many amenities have value or use that are specific to certain populations, such as a
Vietnamese grocery store, an African Methodist Episcopal church, a hipster dive bar, or a school
designed to serve an immigrant community, it seems valuable for investigations of the amenity -
gentrification connection to include racial/ethnic change in their measures of gentrification.
82
2.2. The Relationship between Charter Schools and Urban Gentrification
Across various geographies, researchers have observed race-based and income-based
segregation of students enrolled in charter schools (Chapman & Donnor, 2015; Frankenberg, Siegel-
Hawley, & Wang, 2010; Ladd, Clotfelter, & Holbein, 2017; Urquiola, 2005), sometimes to an even
greater extent than adjacent traditional public schools (Bifulco & Ladd, 2007, Orfield & Ee, 2017).
These outcomes could be attributable to divergent household preferences for schools or the traits of
their students (Bifulco & Ladd, 2007; Saporito, 2003), residential segregation and a proximity bias in
the schools that households choose to attend (Kleitz, Weiher, Tedin, & Matland, 2000; Lubienski,
Gulosino, & Weitzel, 2009), asymmetries in households’ access to information on schooling options
(Ball & Vincent, 1998), or other factors entirely.
Yet observing levels of sociodemographic integration is only one potential way of
determining whether charter schools have addressed issues of inequitable access. Another strand of
literature modifies the analyses of traditional public schools and their neighborhoods referenced in
section 1 (e.g., Black, 1999), seeking to understand whether charter schools are responding to or
shaping neighborhoods in ways that favor some households more than others. This strand assesses
whether the presence of charter schools is associated with the gentrification of nearby
neighborhoods in urban settings. Although incomplete (Jordan & Gallagher, 2015), it suggests an
association does exist and raises the need for more sophisticated study of the subject.
For instance, Hankins (2007) collects and reviews qualitative data regarding the opening of a
charter school campus in a gentrifying area of Atlanta, Georgia. She concludes that middle -class
gentrifiers with children have taken advantage of the “neoliberalization” of public services, creating
charter schools to develop a new, place-based urban space for their own community. Separate and
compelling descriptive research by Kerr (2012) concludes that clusters of charter schools do spatially
correlate with gentrifying census tracts in the District of Columbia and Brooklyn, New York. And
83
consistent with the notion that urban charter schools are an outgrowth of neighborhood
gentrification, research based in Chicago finds that charter schools are more likely to open in
neighborhoods targeted by city revitalization programs (Burdick-Will, Keels, & Schuble, 2013).
However, contemporary work by Davis and Oakley (2013) does not find clear evidence of
neighborhood gentrification leading to the creation of charter schools in Atlanta, Chicago, or
Philadelphia; associations that did appear in Chicago and Philadelphia seem attributable to
endogenous coordination efforts by local governments.
Other studies assert that the presence of charter schools may lead to the future gentrification
of nearby neighborhoods. Hicks (2017) reports that the presence of a charter school campus in a
Los Angeles neighborhood is a significant and positive predictor of increases in: neighborhood
home values; the share of white residents in that neighborhood; and the share of white, married
households with children. Similarly, a survey of urban neighborhoods nationwide notes that
expanding school choice options in a neighborhood is a significant and positive predictor of:
increases in the neighborhood’s share of white, college-educated residents above the citywide
average; and increases in that neighborhood’s home values (Pearman & Swain, 2017). Beracha and
Hardin (2019) also identify a home value premium attached to nearby charter schools, albeit only for
multiple-bedroom housing units, in Florida neighborhoods that contain “high-quality” charter
schools relative to nearby traditional public schools. This work qualifies the results of other studies
that found minimal or no relationship between the presence or availability of charter schools and
home values (Brehm, Imberman, & Naretta, 2017; Horowitz, Keil, & Spector, 2009).
In addition to the literature’s thinness, I note several limitations of the literature on charter
schools’ association with urban gentrification. First, the majority of charter-gentrification papers
measure gentrification over an approximate 10-year period (Beracha & Hardin, 2019; Burdick-Will et
al., 2013; Hankins, 2007; Pearman & Swain, 2017), which is on the lower side when compared to
84
research focused on urban gentrification (e.g., Hackworth, 2002). Second, many studies do not
consider changes in neighborhoods’ school-age populations when defining gentrification, even
though charter schools primarily serve the school-age population and gentrification may occur over
multiple phases or generations. Relatedly, the majority do not test the use of both relative and
absolute measures of gentrification in their analyses. Exceptions to the prior two points include
Hicks (2017) and Pearman and Swain (2017), respectively. Third, no research in this vein has
controlled for types of charter school, despite intriguing findings from Bifulco and Ladd (2007) and
the literature on charter schools and equity of access (Wilson, 2016). Beracha and Hardin (2019)
come closest by distinguishing charter schools per grades served, e.g., elementary versus high
schools. Fourth, no work has accounted for the presence of rail transit, an amenity that the literature
has found related to urban gentrification, and one whose location may be correlated with charter
schools (see sub-section 3.7).
2.3. Research Questions and Hypotheses.
Building from the literature, I propose the following two research questions and hypotheses:
Research question 1: Is the gentrification of District neighborhoods significantly associated with
the presence of nearby charter schools?
Hypothesis 1: The gentrification of District neighborhoods is significantly associated with
the presence of nearby charter schools. Furthermore, the extent of neighborhoods’
gentrification is significantly associated with the total number of charter schools that have
operated close to the neighborhood over time.
Research question 2: Does the relationship between a District neighborhood’s level of
gentrification and the presence of nearby charter schools vary by the type(s) of charter schools that
have operated nearby?
85
Hypothesis 2: The relationship between a District neighborhood’s level of gentrification
and the presence of nearby charter schools does vary by the type(s) of charter schools that
have operated nearby. Drawing from Bifulco and Ladd’s (2007) work, I further hypothesize
that: (1) the presence of “Global Culture” schools is significantly and negatively associated
with gentrification in nearby District neighborhoods; and (2) the presence of “Specific
Population” and “Learner Centered” schools is significantly and positively associated with
gentrification in nearby District neighborhoods.
Having summarized the extant literature and presented my research questions and hypotheses, I turn
to describing my data, definitions, methods, and findings.
3. Data and Definitions
3.1. Summary of the District of Columbia School Reform Act of 1995 (Pub. L. No. 104-134)
President Clinton signed the District of Columbia School Reform Act of 1995 (SRA) into
law on April 26, 1996, authorizing a dramatic shift in governance for the District of Columbia’s
elementary and secondary public education system. Most notably, the SRA immediately legalized
the establishment of public charter schools within the District. Per the law, and unlike the District’s
traditional public schools, charter schools in the District do not have attendance zones. Rather, all
District residents have equal claim to each charter schools’ seats regardless of residential location. If
demand for a charter school’s seats exceeds supply, a randomized lottery is held.
The federal law’s Congressional authors intentionally minimized the number of
accountability tools available to local District of Columbia public officials in the planning and
oversight of public charter schools. In particular, other than managing closed traditional public
school facilities and deciding whether to allow another school to operate within such facilities,
District officials play no formal role in charter schools’ acquisition of facilities or locational
decisions. Charter school operators must typically participate in the District’s private real estate
86
market and select where they will participate. Because of this legislated autonomy, the District
provides an especially salient context for assessing how charter schools’ locational decisions are
associated with neighborhood-level gentrification in an urban setting.
3.2. Using Census Tract-level Data
To understand the relationship between District neighborhoods’ level of gentrification and
their exposure to charter schools, I use tract-level data from the U.S. Census Bureau to develop
measures of gentrification for each District neighborhood, defining neighborhood as tracts.
Per the 2010 decennial census boundaries, there are 179 census tracts in the District (U.S.
Census Bureau, 2019). I exclude four of these tracts from my analysis because their land usage is
entirely or almost entirely institutional: Tract 2.01, containing Georgetown University; Tract 23.02,
containing the Washington Hospital Complex, Federal Armed Forces Retirement Home, and
Lincoln’s Cottage; Tract 62.02, containing Federal Parklands including the National Mall; and Tract
73.01, coterminous with the Anacostia-Bolling Joint Military Base.
I use decennial census data from 2000 as my baseline for measuring gentrification in the
remaining 175 tracts given that District charter schools began operating in August 1996 (Gaines &
Ly, 1996). I use the 2017 5-year American Community Survey (ACS) as my endpoint. I use relational
files from Brown University’s Longitudinal Tract Data Base to translate 2000 data into the same
tract geography as 2017 data (Brown University, n.d). I note th at the 17-year period I employ for
measuring gentrification is consistent with the time periods used by others studying the subject (see
sub-section 2.1).
3.3. Defining and Measuring Gentrification for District Neighborhoods (i.e., Census Tracts)
As my review of the literature indicates, urban gentrification is a nebulously defined
phenomenon. Scholars have measured it via disparate criteria, likely reflecting both its multi- faceted
nature and the various stages over which it may occur. Accordingly, I develop two distinct, multi-
87
dimensional measures of gentrification to test my hypotheses and thereby generate more robust
results; the individual dimensions included in each measure are well supported by the literature
summarized in section 2.
My first measure of gentrification is similar to those employed by Freeman (2005) and Ellen
et al. (2019). It compares changes in each census tract from 2000-2017 with changes in the entire
District during the same period. It does so across four dimensions:
5

1. share of the population considered white alone, non-Hispanic;
2. share of the school-age population (0-17 years) considered white alone, non-Hispanic;
3. share of the population with at least a bachelor’s degree (25 years and older); and
4. growth in median household income, inflation-adjusted to 2017 dollars.
Using these four dimensions, I construct an ordinal scale that evaluates a census tract’s level
of gentrification relative to that of the overall District. I award a census tract 1 point for each
instance where change in a dimension exceeded District-level change in the same dimension. This
results in an integer scale ranging from 0 points to 4 points, where a score of “0” indicates a lack of
gentrification relative to the overall District and a score of “4” indicates a high level of gentrification
relative to the overall District. As the District has experienced substantial and positive change in
each of the four dimensions listed above (see section 4), this ordinal scale is a conservative identifier
of which census tracts have gentrified from 2000-2017. Throughout the remainder of this paper, I
refer to this first measure as my “Relative Scale”.
5
2000 total population comes from 2000 Census Table SF1/P001; 2000 white alone, non- Hispanic total and school-age
population come from 2000 Census Table SF1/P012I; 2000 school-age population comes from 2000 Census Table
SF1/DP1; 2000 share of population with at least a bachelor’s degree comes from 2000 Census Table SF3/DP2; and
2000 median household income comes from 2000 Census Table SF3/DP3. 2017 total population comes from 2017 5 -
year ACS Table B01003; 2017 white alone, non- Hispanic total and school-age population come from 2017 5-year ACS
Table B01001H; 2017 school-age population comes from 2017 5-year ACS Table B01001; 2017 share of population with
at least a bachelor’s degree comes from 2017 5-year ACS Table S1501; and 2017 median household income comes from
2017 5-year ACS Table S1901. Adjustm ents for inflation are made using the U.S. Bureau of Labor Statistics’ Consumer
Price Index (CPI) for All Urban Consumers in the Baltimore -DC Metropolitan Statistical Area.
88
My second measure of gentrification is a continuous index, like that utilized by the
University of Illinois at Chicago (UIC, 2014). Unlike my first measure, this second measure does not
compare census tracts to the overall District. As a result, it is a more aggressive identifier of which
census tracts have gentrified from 2000-2017, producing a less spatially-concentrated plot of
gentrification. It considers absolute changes in each census tract across the following five
dimensions (the first four the same as my first measure):
1. share of the population considered white alone, non-Hispanic;
2. share of the school-age population (0-17 years) considered white alone, non-Hispanic;
3. share of the population with at least a bachelor’s degree (25 years and older);
4. growth in median household income, inflation-adjusted to 2017 dollars; and
5. percentile rank of the census tract per its median household income, compared to all
other District census tracts.
I employ principal component analysis (PCA) to transform each census tract’s changes
across these five dimensions into a single, indexed measure of its gentrification. I utilize the first
component from the PCA to accomplish this. I select only the first component as it explains
approximately 70% of total variance, meeting the indexing criteria outlined in the literature (Jolliffe
& Cadima, 2016). The results of this principal component analysis are available in Appendix B.
Throughout the remainder of this paper, I refer to this second measure as my “Absolute Index”.
3.4. Constructing Locational Data for Charter School Campuses
I define a charter school campus as a specific charter LEA operating in a specific physical
facility. If a single charter LEA operates multiple schools in a given facility ( e.g., KIPP DC PCS often
co-locates multiple academies), I consider those schools to be a single campus. However, if two or
more charter LEAs operate in one facility, I consider each as a distinct campus. Distinguishing
multiple charter LEAs that occupy the same facility is important given my second hypothesis: that
89
the charter school-gentrification relationship varies by the type(s) of charter schools proximate to a
neighborhood.
I cross-reference a wide array of resources to construct a longitudinal dataset that locates
District charter school campuses by academic year since the first year of their operation (school year
1996-97) I provide an exhaustive list of references for this dataset in Appendix C.
I focus on campus locations for school years 2000-01 through 2017-18, which mirrors my
2000-2017 period for measuring gentrification. I include school year 2017- 18 data since new campus
locations are typically announced to the public about one year in advance. For each of these school
years, I assign each charter school campus a location equivalent to the center of the physical facility
in which it operated.
3.5. Developing a Charter School Typology
Like the A+ Colorado report (2018), I define charter school type as a charter LEA’s
educational emphasis, its core educational philosophy, and its main curricular programming. I
develop this typology at the LEA-level, rather than campus-level, because my research indicates a
consistency in educational experience across the campuses a charter LEA operates, at least within
the District. The only exception to my classifying at the LEA level is Maya Angelou PCS; I classify
its Young Adult Learning Center as an Adult charter school and its other campuses as Specific
Population charter schools. As with my locational data for charter schools, I cross-reference a wide
array of resources to develop this typology and classify LEAs accordingly (see Appendix D). I define
eleven types in Table 2 below.
90
Table 2. Typology of District charter LEAs
Type of charter school Definition
Adult Charter schools that focus on serving students at least 16 years of
age. Educational emphasis is often on graduate equivalent degree
(GED) attainment or credit recovery for a high school diploma.
Blended Learning Charter schools whose core philosophy is the regular incorporation
of technology to promote improved learning and educational
outcomes.
Comprehensive Charter schools whose main identity is the offering of a rigorous,
robust, and cross-disciplinary learning experience. Such charter
schools do not brand themselves as emphasizing a certain curricular
topic or discipline (e.g., STEM), do serve students in grades
kindergarten-12 (K-12), and do not focus on serving a specific
population of students.
Creative Arts Charter schools that regularly incorporate creative arts into their
curricula and emphasize creative arts as a crucial element of thei r
core philosophy of learning.
Dual Language Charter schools that orient themselves around dual
language/language immersion programming.
Early Childhood Charter schools that restrict their services to early childhood
education. Such charter schools serve students in pre-kindergarten
grades (age 3 or age 4) and often serve students through the
beginning or middle of elementary school.
Global Culture Charter schools who organize their curriculum around teaching
students through a culturally-specific or global/international lens
(e.g., schools offering an Afrocentric education).
Learner Centered Charter schools whose core philosophy is student-driven learning
(e.g., schools employing the Montessori educational model).
Postsecondary/Vocational Charter schools that focus on students’ success in college and
careers. These charter schools often brand themselves as having
“college preparatory” programming and offering vocational or
technical training.
Specific Population Charter schools that advertise themselves as serving a specific
population of students in grades kindergarten-12 (K-12). Examples
of specific populations include: neglected, abused, or delinquent
students; alternative-needs students; special education students;
English learner students (EL); foster and homeless students;
students who are teenage mothers; low-income students; and
students of a certain race, ethnicity, or gender.
STEM Charter schools that organize their curriculum around the teaching
of science, technology, engineering, or mathematics (STEM).
Sources: See Appendix D for an exhaustive list of references used in developing this typology.

91

In some cases, charter LEAs appear to fall within multiple categories of the typology. For
example, Carlos Rosario International PCS is both an adult charter school and a charter school
oriented to serving English learners and immigrant communities (PCSB, 2018a). Therefore, and
when necessary, I assign charter LEAs a primary type as well as a secondary type, again consistent
with the approach of A+ Colorado (2018). When presenting the results of my regression analyses, I
include a sensitivity analysis that considers secondary types in place of primary types (see section 5).
3.6. Measuring a Neighborhood’s Exposure to Charter Schools
For each of the 175 census tracts in my study, I use geographic information software (GIS)
to locate its centroid. I then construct a circular catchment areas with 1-mile radius around each
centroid. I employ a 1-mile radius given median travel distances to school reported for students
attending District of Columbia Public Schools (Chandler, 2015a), along with findings from the
literature on traveling to school (e.g., Su et al., 2013).
I count how many charter school campuses fall within each catchment area. I conduct these
counts by school year, from school year 2000-01 through school year 2017-18. Because some
District census tracts have a surface area that is less than 3.14 square miles (i.e., the area of a circle
with 1-mile radius), a single charter school campus can fall within multiple catc hment areas. I use the
term “charter-year” to refer to the unique combination of (i) a specific charter school campus that
(ii) is operating in a specific school year and (iii) falls within a specific census tract’s catchment area.
Based on these criteria, I consider a census tract’s exposure to charter schools to be the
charter-years attributable to that tract between school years 2000-01 and 2017-18. To provide a
hypothetical example: if a census tract’s catchment area contained three charter campuses in each of
the eighteen school years I study, I would attribute a total of 54 charter-years to that census tract.
Because my second hypothesis concerns individual charter school types, I also disaggregate a census
tract’s total charter-years down to charter-years by type. When presenting the results of my
92
regression analyses (see section 5), I include sensitivity analyses that define a tract’s charter -years
using a catchment area with 0.5-mile radius instead of a 1-mile radius. I use a 0.5-mile radius to
match my baseline catchment area for rail transit exposure (see sub-section 3.7 below).
It is also possible that the relationship between a tract’s level of gentrification and its
exposure to charter schools depends on the enrollment size of those charter schools. Large charter
schools may influence the characteristics of proximate neighborhoods more than small charter
schools, since more households can attend them. Using official enrollment counts from the District
of Columbia Office of the State Superintendent of Education, I conduct a sensitivity analysis that
weights each charter-year by the number of students enrolled at that charter school campus relative
to the total population of charter school students in each applicable school year (see section 5).
3.7. Measuring a Neighborhood’s Exposure to Rail Transit
The literature indicates that the presence of rail transit is directly associated with
neighborhood gentrification, at least in some cities (see sub-section 2.1). Still, this is not the only
reason why controlling for rail transit may be important in the context of this study.
First, the District does not actively provide transportation services for students attending
traditional public or charter schools, other than those with transportation services in their heir
individualized education plans (IEPs) (OSSE, n.d.a). Instead, it fully covers the cost of using public
transit for students through a program called Kids Ride Free (DDOT, 2018). Since District charter
schools do not have any guaranteed, attendance zone-based enrollment (Pub. L. 104-134, § 2206(a)),
it appears advantageous for charter schools to locate near rail transit stations (along with bus transit
hubs).
A recent survey also finds that over 75% of District land within 0.5 miles of rail transit
stations (Metrorail) is “upzoned” to allow construction of properties taller or denser than detached
single-family homes, including commercial properties (Loh, 2019). This compares to 58% of all
93
District land being upzoned (Loh, 2018). Since District charter schools require large spaces in which
to operate, historically either commercial spaces or vacated school buildings (GAO, 2003b ), they
may also tend to locate near rail transit stations because of favorable zoning for commercial
properties.
Each of the Metrorail’s six lines provides access to the District’s central business district ( see
sub-section 3.8 below), and by the end of 2000, all Metrorail stations within 1.5 miles of the
District’s boundary were operational or under construction (Broadway, 2000; WMATA, 2014).
Because of these two facts, I treat each Metrorail system equally in terms of its prospective
association with District census tracts’ gentrification from 2000- 2017.
I use geographic information software (GIS) to locate the centroids of the 175 census tracts
in my study. I then construct a circular catchment area with 0.5-mile radius around each centroid,
per the literature on rail transit access and catchment areas (e.g., Guerra, Cervero, & Tischler, 2012).
The District provides locational data for all Metrorail station entrances.
6
I use that data to
assign each rail station a location equivalent to the mid -point between that station’s physical
entrances, not including separate elevator entrances. I then identify whether a District census tract is
proximate to rail based on whether its 0.5-mile radius catchment area contains one or more
Metrorail stations, including stations in Maryland or Virginia. In my sensitivity analyses, I consider
an alternate catchment area size with 1-mile radius, consistent with my catchment area for charter
school exposure and some literature on rail transit (e.g., Ewing et al., 2015).
3.8. Calculating a Neighborhood’s Distance to the Central Business District
The literature also identifies proximity to a central business district as a predictor of
gentrification (Baum-Snow & Hartley, 2017; Brummet & Reed, 2019). Using the District
6
This information is available for public use in ArcGIS via the following link:
https://services.arcgis.com/neT9SoYxizqTHZPH/arcgis/rest/services/Metro_Station_Entrances_Status/FeatureServe
r
94
Department of Transportation’s definition of the city’s central business district, I calculate the linear
distance between each tract’s centroid and the centroid of the central business district (DDOT,
n.d).
7
3.9. Including Other Control Variables
In addition to the data I describe above, I use three additional control variables in my
regression analyses (see section 5). They measure: (i) how close a census tract’s demographic
composition was to 75% not white alone, non-Hispanic in 2000 (i.e., 25% white alone, non-
Hispanic)
8
; (ii) a tract’s density of housing units in 2000 (housing units per square mile of land); and
(iii) the share of a tract’s occupied housing units that were rental units in 2000.
9
As with other census
tract-level data, I use relational files from Brown University’s Longitudinal Tract Data Base to
translate 2000 data into the same tract geography as 2017 data (Brown University, n.d).
4. Descriptive Analyses
In the following sub-sections, I describe gentrification within the District, the proliferation
of charter schools in the District, and patterns in charter schools’ enrollment compositions. I
observe spatially clustered gentrification in the District, spatially uneven growth of charter school
campuses (and within that spatially uneven growth, further imbalances by type of charter school),
and patterns in students enrolled at public charter schools over time and by type. These observations
provide compelling evidence that the gentrification of District neighborhoods is related to the
presence of charter schools.
7
The central business is approximately bounded: on the west by 23
rd
Street, NW; on the north by Massachusetts
Avenue, NW; on the east by 2
nd
Street, NE and 2
nd
Street, SE; on the south by Constitution Avenue, NW between 23
rd

Street, NW and 14
th
Street, NW; and also on the south by D Street, SW and D Street, SE between 14
th
Street, NW and
2
nd
Street, SE.
8
Calculated as: 1 / abs (0.75 – census tract share of population not white alone, non-Hispanic in 2000)
9
2000 total population comes from Census Table SF1/P001; 2000 white alone, non-Hispanic population comes from
Census Table SF1/P012I; 2000 total housing units comes from Census Table SF1/H001; land surface areas of census
tracts come from 2019 GIS shapefiles (U.S. Census Bureau, 2019); 2000 number of occupied units and number of
occupied units rented in 2000 come from Census Table SF1/DP1.
95
4.1. The Overall District Sharply Gentrified from 2000 to 2017
The District of Columbia gentrified across many dimensions since charter schools began
operating in school year 1996-97. As Table 3 shows, over 80% of District population growth from
2000 to 2017 was attributable to new white alone, non -Hispanic residents. Even in 2017, that group
accounted for only 36% of the total population. Change in the school-age population was even more
dramatic: the increase in the white alone, non-Hispanic school-age population actually exceeded the
increase in the total school-age population, implying a net decrease in its nonwhite or Latino
children. The share of residents with at least a bachelor’s degree increased by 18 percentage points,
median household income grew at a 27% rate, and the median home value more than doubled.
Table 3. Dimensions of District gentrification, 2000-2017 (in 2017 USD where applicable)
Dimension of change
2000-2017
gross change
2000-2017
growth rate
2017 level 2000 level
total population 100,332 0.18 672,391 572,059
...white alone, non-Hispanic 82,723 0.52 241,901 159,178
...white alone, non-Hispanic share 8% n/a 36% 28%
...black alone -22,250 -0.06 321,062 343,312
...black alone share -12% n/a 48% 60%
...Hispanic or Latino 26,876 0.60 71,829 44,953
...Hispanic or Latino share 3% n/a 11% 8%
school-age population 3,066 0.03 118,058 114,992
...white alone, non-Hispanic 10,536 0.77 24,231 13,695
...white alone, non-Hispanic share 9% n/a 21% 12%
Population share with at least bachelor's degree 18% n/a 57% 39%
median household income $16,309 0.27 $77,649 $61,340
median home value (owner-occupied) $297,098 1.24 $537,400 $240,302
Sources:
1. 2000 total population comes from 2000 Census Table SF1/P001; 2000 white alone, non- Hispanic total and
school-age population come from 2000 Census Table SF1/P012I; 2000 black alone population comes from 2000
Census Table SF1/P012B; 2000 Hispanic or Latino population comes from 2000 Census Table SF1/P004; 2000
school-age population comes from 2000 Census Table SF1/DP1; 2000 share of population with at least a
bachelor’s degree comes from 2000 Census Table SF3/DP2; 2000 median household income comes from 2000
Census Table SF3/DP3; and 2000 median home value comes from 2000 Census Table SF3/DP4.
2. 2017 total population comes from 2017 5 -year ACS Table B01003; 2017 white alone, non-Hispanic total and
school-age population come from 2017 5-year ACS Table B01001H; 2017 black alone population comes from
2017 5-year ACS Table B01001B; 2017 Hispanic or Latino population comes from 2017 5 -year ACS Table
B03002; 2017 school-age population comes from 2017 5-year ACS Table B01001; 2017 share of population with
96
at least a bachelor’s degree comes from 2017 5-year ACS Table S1501; 2017 median household income comes
from 2017 5-year ACS Table S1901; and 2017 median home value comes from 2017 5-year ACS Table B25077.
Note: Adjustments for inflation made using the U.S. Bureau of Labor Statistics’ Consumer Price Index (CPI) for All
Urban Consumers in the Baltimore-DC Metropolitan Statistical Area.
4.2. The District’s Gentrification Occurred in a Spatially Uneven Way
According to my Relative Scale and Absolute Index measures, the District’s gentrification
was concentrated in certain areas. Figures 5 and 6 show heat maps of gentrification for each of the
175 census tracts I study, per my Relative Scale and Absolute Index measures respectively.
As anticipated, the Relative Scale measure produces a more spatially concentrated pattern of
gentrification than the Absolute Index measure. Nevertheless, my two measures are largely
consistent in identifying which areas of the District experienced the most gentrification versus the
least gentrification from 2000 to 2017. My identification patterns also align with contemporary
media and academic accounts, including neighborhoods like Columbia Heights, the U Street
corridor, Catholic University/Brookland, Gallaudet University/NoMa, and Waterfront (Richardson,
Mitchell, & Franco, 2019; Stein, 2017).
The most gentrified census tracts tend to fall between the two large geographic dividers of
the District: Rock Creek Park and the Anacostia River ( see Appendix E for a map of District
landmarks and neighborhoods). Over the last few decades, the District’s most affluent, and typically
white, residents have tended to live west of Rock Creek Park, in neighborhoods such as Georgetown
and Cleveland Park. Conversely, the District’s poorest, and typically black American, residents have
tended to live east of the Anacostia, in neighborhoods such as Congress Heights and Douglass. See
Figures 7 and 8 for graphical depictions.
97
Figure 5. Gentrification of District census tracts per Relative Scale measure, 2000-2017
Sources: My own calculations from U.S. Census data, as described in section 3.
Figure 6. Gentrification of District census tracts per Absolute Index measure, 2000-2017
Sources: My own calculations from U.S. Census data, as described in section 3.
98
Figure 7. White alone, non-Hispanic shares of census tracts’ populations, 2017
Sources: Total population from 5-year ACS Table B01003; white alone, non-Hispanic count from 5-year Table
B01001H.
Figure 8. Median household incomes of census tracts, 2017
Source: 2017 median household income comes from 2017 5-year ACS Table S1901.
99
Thus, it appears the District has gentrified in a way consistent with the framework of
Guerrieri et al. (2013): gentrifying households have tended to locate in areas next to, but not within,
the District’s most affluent and white neighborhoods. While the authors posit that desirable
consumption opportunities in high-income neighborhoods drive this behavior, it seems plausible
that gentrifying households are also avoiding overly nonwhite or poor neighborhoods in the District.
4.3. Charter School Campuses Proliferated Substantially from School Year 2000-01 to 2017-18
Concurrent with this substantial gentrification, enrollment at charter schools (collectively,
the “charter sector”) grew markedly – both in level and the share of total public students. Figure 9
plots historical enrollment figures for the District’s traditional public schools (DCPS) versus its
charter sector. As of school year 2000-01, a clear majority of public students were enrolled in DCPS
schools: 68,925 in DCPS versus 9,881 in the charter sector. By school year 2017-18, public student
enrollment was split almost evenly between DCPS and the charter sector: 48,144 and 43,393,
respectively. During the same timeframe, total public school enrollment grew from 78,806 to 93,016.
Unsurprisingly, the number of charter LEAs and campuses in the District also multiplied
between 2000 and 2017. Appendix F contains two tables, which list the number of charter LEAs
and charter campuses operating in each school year since 1996-97. In school year 2000-01, 34
charter LEAs operated 39 campuses throughout the District – an average of 1.15 campuses per
charter LEA. By school year 2017-18, 66 charter LEAs operated 112 campuses throughout the
District – an average of 1.70 campuses per charter LEA. The tables in Appendix F further
disaggregate the number of operating charter LEAs and charter campuses by the eleven types of
charter schools I identify (see sub-section 3.5 and Table 2). For instance, the number of Adult LEAs
quintupled between 2000-01 and 2017-18, going from two to 10. The number of Dual Language
LEAs increased by a factor of nine, from one to nine. Three types were not operating at all in school
year 2000-01: Blended Learning, Early Childhood, and Learner Centered.
Figure 9. DCPS and PCS enrollment, school year 1996-97 to 2017-18
Sources:
1. DCPS and PCS enrollment counts for school years 2007-08 through 2017-18 come from the District of Columbia Office of the State Superintendent of
Education’s (OSSE) annual enrollment audit reports.
2. DCPS and PCS enrollment counts for school years 1996-97 through 2006-07 come from Appendix C of OSSE’s Quality Schools Report, other than PCS
enrollment for school years 1996-97 and 1997-98.
3. PCS enrollment for school year 1996-97 is a conservative estimate based on planned enrollment levels for Marcus Garvey PCS and Options PCS (Gaines & Ly,
1996).
4. PCS enrollment for school year 1997-98 comes from the 21
st
Century School Fund and Brookings Greater Washington Research Program joint report titled DC
Public School and Public Charter School Capital Budgeting: Task 3 Report (April 4, 2005).
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
80,000
90,000
Enrolled Students
DCPS Charter Sector
100
101
4.4. The Proliferation of Charter School Campuses Occurred in a Spatially Uneven Way
Like the District’s pattern of gentrification, the geographical expansion of its charter sector
occurred unevenly. Figure 10 plots the locations of charter school campuses in every school year
from 2000-01 to 2017-18, superimposed over census tracts’ 2017 white-alone, non-Hispanic
population shares. Figure 11 employs my definition of a census tract’s exposure to charter schools,
depicting the total charter-years attributable to each census tract (see sub-section 3.6). Both Figures
10 and 11 demonstrate that almost no charter campuses have ever operated in the area west of Rock
Creek, its whitest and most affluent region (see Figures 7 and 8). This is a striking geographic divide,
one hinting that residents of this area have not felt the need for charter schools despite that option.
Figure 10. Locations of all charter campuses, school years 2000-01 through 2017-18, superimposed
on census tracts’ white alone, non-Hispanic population shares, 2017
Sources:
1. See sub-section 3.4 for information on charter school campus locations.
2. 2017 total population comes from 2017 5 -year ACS Table B01003; 2017 white alone, non-Hispanic population
comes from 2017 5-year ACS Table B01001H.
102
Figure 11. Charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-section 3.4 for information on charter school campus locations.
Yet perhaps this is unsurprising. The School Reform Act layered charter schools on top of
attendance zoning for its traditional public schools. As a result, households continue to have a
geographic “school of right” based on where they live, resulting in segregation across the District’s
traditional public schools that reflects its residential segregation (Orfield & Ee, 2017). Households
may value this attendance zoning mechanism, which grants them some control over their children’s
peers at school. Such hypothetical behavior would be consistent with the literature (Lankford &
Wyckoff, 1992; Lankford & Wyckoff, 2006; Saporito, 2003). Alternatively, given their typically high
incomes, residents west of Rock Creek can rely on private schools for options separate from the
District’s traditional public schools.
103
Figure 11 provides an additional and valuable insight. Specifically, the pattern in total
charter-years by census tract seems to correspond to the District’s pattern of gentrification (Figures
3 and 4).
4.5. Within their General Pattern of Proliferation, Charter School Campuses Display
Disparate Patterns of Expansion Depending on Charter School Type
Disaggregating counts of total charter-years by type reveals further variation within the
charter sector’s proliferation. The growth patterns of some charter types appear positively correlated
with the District’s pattern of gentrification, while others do not. Figure 12 shows the Global Culture
charter-years attributable to each census tract; the census tracts most exposed to them do not appear
to be the census tracts that significantly gentrified. In contrast, Figure 13 shows the Learner
Centered charter-years attributable to each census tract; while the sample size of these schools is
small, the census tracts most exposed do appear to be census tracts that significantly gentrified.
Figure 12. Global Culture charter-years attributable to each census tract, school years 2000-01
through 2017-18
Sources: See sub-sections 3.4 and 3.5 for information on charter campus locations and their types.
104
Figure 13. Learner Centered charter-years attributable to each census tract, school years 2000-01
through 2017-18
Sources: See sub-sections 3.4 and 3.5 for information on charter campus locations and their types.
Please see Appendix G for the growth patterns of all charter school types. As with Figures 12 and
13, the other types display significant variations in the spatial patterns of their growth.
4.6. Student Enrollment at District Charter Schools Varies Temporally and by Type
I conclude my descriptive analyses by noting variations in District charter enrollment both
temporally and by charter school type. If the District’s charter schools in general are positively
associated with gentrification, it seems plausible that recently -opened charter schools may have
whiter and wealthier students than charter schools that have been operating for longer periods of
time. More specifically, I expect that initial groups of students attending recently-opened charter
schools tend to be whiter and wealthier than the students that initially attended older charter schools
when they first opened. I hypothesize that the outgroup avoidance and divergent household
preference mechanisms observed by Saporito (2003) and Bifulco and Ladd (2007) produce positive
feedback loops related to these initial compositions of enrolled students over time.
105
I do, in fact, find evidence supporting this dynamic. Using school year 2017-18 enrollment
data from the District’s Public Charter School Board (PCSB), I compare the shares of white, non-
Hispanic students and “at-risk” students
10
(a proxy for low-income students) at newer versus older
charter LEAs. For robustness, I consider two temporal cut-points for bifurcating 2017-18 charter
LEAs into newer versus older. First, I use the year 2008, the first full year of the District’s Public
Education Reform Amendment Act (PERAA; DC Law 17-9) being active. I do so because PERAA
was a highly visible symbol of education reform in the District and may have been a synecdoche of
overall District change for gentrifying households (Stone, 2012). Second, I use the year 2010, the
first full year after the Great Recession (NBER, 2010); I do so because of the literature exploring
linkages between the Great Recession and urban gentrification (e.g., Hyra & Rugh, 2016).
Using each cut-point, I find that students enrolled at “newer” charter LEAs (those that
began operating in the District in 2008 or later, and 2010 or later, for the cut -points respectively) are
significantly more likely to be white, non-Hispanic and are significantly less likely to be at -risk (see
Tables 4a and 4b below). Using the same enrollment data, I also assess the sociodemographic
compositions of charter school campuses by type in school year 2017-18. Table 5 presents the
results of this analysis.
Table 4a. Enrollment composition of charter LEAs by LEA opening time (pre- vs. post-PERAA),
school year 2017-18
** for p < 0.01
Charter LEA opening time
Number of
charter LEAs
Total
enrollment
White alone,
non-Hispanic
share of enrollment
At-Risk
Share of Enrollment
pre-PERAA (before 2008) 31 26,630 4.4% 52.2%
post-PERAA (2008 and after) 26 11,768 11.7% 46.7%
7.4%** -5.5%**
10
The District’s Public Charter School Board defines at -risk students as “non-adult students who receive Temporary Assistance for
Needy Families or Supplemental Nutrition Assistance Program benefits, are homeless or in the foster system, or are… a year or more
behind in high school” (PCSB, 2018b).
106
Table 4b. Enrollment composition of charter LEAs by LEA opening time (pre- vs. post-
Recession), school year 2017-18
** for p < 0.01
Charter LEA opening time
Number of
charter LEAs
Total
enrollment
White alone,
non-Hispanic
share of enrollment
At-Risk
Share of Enrollment
pre-Recession (before 2010) 36 30,560 4.4% 51.9%
post-Recession (2010 and after) 21 7,838 15.1% 45.1%
10.7%** -6.9%**
Sources:
1. Total enrollment figures for each charter LEA (and its campuses) come from OSSE’s official 2017-18 enrollment
audit report and documentation (OSSE, 2018).
2. Demographic data for charter students are interpolated using PCSB’s 2018 Quality Reports in for each charter
campus; these data are provided in share (percentage) format (PCSB, 2018c).
Notes:
1. Excludes Adult campuses as PCSB does not provide at-risk data for those schools.
2. Because shares were interpolated using figures reported in share format, rounding errors are present.
Table 5. Enrollment composition of charter school campuses by type, school year 2017-18
Highlighted cells represent some – not all – significant departures in a sociodemographic group’s representation in a
given charter school type versus its representation in total charter enrollment, p < 0.01; columns do not sum to 100%
Charter school type
White alone,
non-Hispanic
share of
type’s enrollment
Black alone,
non-Hispanic
share of
type’s enrollment
English Learner
share of
type’s enrollment
At-risk
share of
type’s enrollment
adult 0.9% 44.3% unavailable unavailable
blended learning 0.2% 96.7% 1.7% 81.2%
comprehensive 4.3% 84.0% 5.9% 52.1%
creative arts 13.1% 74.1% 4.8% 43.7%
dual language 25.2% 37.9% 15.1% 18.8%
early childhood 5.1% 86.7% 8.4% 59.8%
global culture 0.0% 100.0% 5.9% 45.8%
learner centered 35.6% 41.5% 8.4% 15.2%
postsecondary/vocational 1.2% 85.0% 6.2% 58.3%
specific population 0.8% 95.6% 1.6% 69.1%
STEM 0.3% 96.3% 3.1% 59.9%
Total 6.0% 74.9% 6.1% 44.7%
Sources:
1. Total enrollment figures for each charter LEA (and its campuses) come from OSSE’s official 2017-18 enrollment
audit report and documentation (OSSE, 2018).
2. Demographic data for charter students are interpolated using PCSB’s 2018 Quality Reports in for each charter
campus; these data are provided in share (percentage) format (PCSB, 2018c).
Notes: Because shares were interpolated using figures reported in share format, rounding errors are present; color table
can be viewed at wileyonlinelibrary.com.
107
I find that white alone, non-Hispanic students are significantly overrepresented in Creative Arts
and Dual Language charter schools and, to an even greater extent, in Learner Centered charter
schools. Unsurprisingly, black alone, non-Hispanic students are significantly underrepresented in the
latter two charter types, as are at-risk students.
In addition, I find that black alone, non-Hispanic students are significantly overrepresented
in Global Culture charter schools, consistent with the fact that each such campus provides an
Afrocentric curricula (see table in Appendix D). English learners are overrepresented at Dual
Language charter schools, consistent with the idea that these schools are working to establish
students’ fluency in multiple languages. Also, at-risk students are significantly overrepresented at
Specific Population charter schools, consistent with the fact that many such schools aim to serve the
at-risk population. In closing, I note that the enrollment composition of Comprehensive schools
approximates the enrollment composition of the charter sector overall.
Vitally, my descriptive findings appear to align with those of Bifulco and Ladd (2007), who
conclude that black households prefer community-oriented schools (i.e., potentially the same types
of schools that I consider to be Global Culture schools) while white households prefer experiential
learning and alternative-assessment schools (i.e., the Learner Centered schools I consider).
Furthermore, these type-based variations in enrollment appear correlated with type-specific spatial
patterns of proliferation between school year 2000-01 and 2017-18 (see Appendix G), as well as the
distribution of white-alone, non-Hispanic households in the District (see Figures 7 and 10).
It is possible that the figures depicted in Table 5 simply result from certain types of charter
schools tending to open in neighborhoods with certain sociodemographic characteristics. For
example, Creative Arts charter schools may enroll white alone, non-Hispanic students at a higher
rate than the charter sector overall because: (a) they tend to locate in neighborhoods with greater
108
concentrations of white alone, non-Hispanic children; and (b) households tend to attend schools
that are physically proximate.
Yet the statistics I report in Table 6 cast doubt on such a clean spatial correlation. Within
Table 6, I compare: (a) the white alone, non-Hispanic share of a charter school campus’ student
populations to (b) the white alone, non-Hispanic share of the school-age population in the census
tracts that contains the campus. I disaggregate this comparison first by whether a specific charter
school campus is located between Rock Creek Park and the Anacostia River versus east of the
Anacostia River, and then by charter school type. I do so as white alone, non-Hispanic households
are much less likely to reside east of the Anacostia River (see Figure 7).
Table 6 demonstrates that the majority of charter school types, and therefore the majority of
charter schools, enroll fewer white alone, non-Hispanic students than expected based on the
sociodemographic compositions of the census tracts that contain them. This is especially
pronounced for schools located between Rock Creek Park and the Anacostia. Nonetheless, I
observe two significant exceptions to this trend: Dual Language and Learner Centered charter
schools. Not only do these school types enroll white alone, non- Hispanic students at a rate
significantly higher than the charter sector overall (see Table 5), they also enroll white alone, non-
Hispanic students at a rate greater than expected based on the sociodemographic compositions of
the census tracts where they are located.
Table 6. White alone, non-Hispanic share of charter campuses’ enrollment versus white alone, non-Hispanic share of occupied census
tracts’ school-age population, by geographic location and charter school type, school year 2017-18
Blue cells indicate the two sub-groups of campuses for which white-alone, non-Hispanic enrollment shares exceed the white-alone, non-Hispanic shares of the school-
age populations in the census tracts occupied by such campuses
Charter school's white alone, non-Hispanic
share of SY17-18 enrollment minus census
tract's 2017 white alone, non-Hispanic
share of the school-age population
Geographic location
of charter campus
Charter school type
SY17-18
campuses
operating
SY17-18
enrolled
students
Average Median
Between Rock Creek and Anacostia Adult 11 4,020 -22.8% -22.1%
Between Rock Creek and Anacostia Comprehensive 26 11,644 -11.9% -7.6%
Between Rock Creek and Anacostia Creative arts 2 781 -23.7% -23.7%
Between Rock Creek and Anacostia Dual language 12 4,571 11.9% 12.1%
Between Rock Creek and Anacostia Early childhood 5 895 -12.9% -10.6%
Between Rock Creek and Anacostia Global culture 2 118 -26.1% -26.1%
Between Rock Creek and Anacostia Learner centered 3 586 15.0% 20.3%
Between Rock Creek and Anacostia Postsecondary/vocational 8 3,014 -7.9% -9.2%
Between Rock Creek and Anacostia Specific population 2 368 -40.1% -40.1%
Between Rock Creek and Anacostia STEM 3 600 -14.9% -6.3%
East of Anacostia Adult 4 930 -1.7% -1.7%
East of Anacostia Blended learning 2 633 -1.0% -1.0%
East of Anacostia Comprehensive 9 6,424 -3.3% -2.1%
East of Anacostia Early childhood 7 1,721 -2.4% 0.0%
East of Anacostia Postsecondary/vocational 10 5,179 -1.7% -2.0%
East of Anacostia Specific population 4 1,423 -7.1% -1.7%
Total 110 42,906 -7.50% -3.44%
Sources:
1. Total enrollment figures for charter LEAs (and campuses) come from OSSE’s official 2017-18 enrollment audit report and documentation (OSSE, 2018).
2. Demographic data for charter students are interpolated using PCSB’s 2018 Quality Reports in for each charter campus; these data are provided in share
(percentage) format (PCSB, 2018c).
Notes: Creative Minds International PCS excluded from analysis as it is located on federal government property, i.e., a non-residential census tract; Sustainable Futures
PCS excluded from analysis as sociodemographic data on enrolled students unavailable for SY17 -18.
109
110
5. Regression Analyses
In the previous section 4, I describe that:
1. the overall District sharply gentrified from 2000 to 2017;
2. the District’s gentrification occurred in a spatially uneven way;
3. charter school campuses proliferated substantially from school year 2000-01 to 2017-18;
4. charter school campuses’ proliferation also occurred in a spatially uneven way;
a. the pattern in charter campuses’ proliferation appears to correspond to the pattern in
the District’s gentrification;
b. within the pattern of campuses’ proliferation are significant variations in locations of
expansion by school type;
5. students who attend more recently-opened charter LEAs are significantly more likely to
be white alone, non-Hispanic and significantly less likely to be low -income;
6. the sociodemographic composition of charter school enrollment varies significantly by
school type, and in ways that appear to correspond to the District’s pattern of
gentrification; however,
7. these variations in enrollment composition by type do not cleanly correlate on a spatial
basis with variations in the compositions of neighborhoods’ school-age populations.
Taken together, my descriptive analyses offer compelling evidence in support of my two hypotheses.
Nevertheless, a more controlled statistical approach is needed to determine the validity of this
evidence. In the remainder of section 5, I develop six regression analyses and present their results. In
general, these results generate additional support for my two hypotheses.
5.1. Regression Modeling Framework
As mentioned above, I conduct six separate regression analyses to estimate the relationship
between District census tracts’ levels of gentrification and the levels of their exposure to charter
111
schools, both overall and by type. The first two regression analyses (Regressions 1 and 2) use my
Relative Scale measure as the dependent variable and predict its tract-level values by accounting for
the total charter-years and type-based charter-years attributable to those tracts, respectively. The next
two regression analyses (Regressions 3 and 4) use my Absolute Index measure as the dependent
variable, also predicting its tract-level values with the total charter-years and type-based charter-years
attributable to those tracts, respectively. Finally, I include two additional regression analyses
(Regressions 5 and 6) that use each tract’s change in the white alone, non-Hispanic share of its
school-age population as the dependent variable; as with the first two sets of regression analyses,
these analyses predict tract-level values with the total charter-years and type-based charter-years
attributable to those tracts. I consider this dependent variable for two reasons. First, of all groups,
the school-age population may be most likely to change based on shifts in nearby educational
opportunities. Second, I seek to test the extent to which this single measure captures the multi-
dimensional and multi-stage nature of urban gentrification I conceptualize.
I use an ordinal probit model for Regressions 1 and 2 because my Relative Scale is an ordinal
and discrete measure. Use of an ordinal probit model is appropriate when attempting to predict an
outcome that is measured via ordered and discrete numbers (i.e., my Relative Scale) but that
theoretically occurs on a continuous spectrum (i.e., urban gentrification) (McKelvey & Zavoina,
1975). In econometric terms, the model assumes that this true, latent measure (y*) is a linear
combination of a set of explanatory factors and an error term that follows the standard normal
distribution:
𝑦𝑦 𝑖𝑖 ∗
= 𝒙𝒙 𝒊𝒊 𝜷𝜷 + 𝑒𝑒 𝑖𝑖 , 𝑒𝑒 𝑖𝑖 ~ 𝑁𝑁 (0,1), ∀ 𝑖𝑖
The model further assumes that the observable version of the dependent variable falls across a set of
discrete, ordered values 0 through m:
112
𝑦𝑦 𝑖𝑖 = 𝑗𝑗 ↔ 𝜇𝜇 𝑗𝑗 − 1
< 𝑦𝑦 𝑖𝑖 ∗
≤ 𝜇𝜇 𝑗𝑗
where j = 0,…,m
Thus, the probability of each discrete value occurring is:
𝑃𝑃 [ 𝑦𝑦 𝑖𝑖 = 𝑗𝑗 ] = 𝜑𝜑 � 𝜇𝜇 𝑗𝑗 − 𝒙𝒙 𝒊𝒊 𝜷𝜷 � − 𝜑𝜑 � 𝜇𝜇 𝑗𝑗 − 1
− 𝒙𝒙 𝒊𝒊 𝜷𝜷 � where
𝜑𝜑 � 𝜇𝜇 𝑗𝑗 − 𝒙𝒙 𝒊𝒊 𝜷𝜷 � = 𝑃𝑃 [ 𝑒𝑒 𝑖𝑖 ≤ 𝜇𝜇 𝑗𝑗 − 𝒙𝒙 𝒊𝒊 𝜷𝜷 ]
Maximum likelihood estimation is the appropriate method of estimating the parameters in this
model, and the log-likelihood function is:
ln[ 𝐿𝐿 ] = ∑
𝑁𝑁 𝑖𝑖 = 1
∑
𝑚𝑚 𝑗𝑗 = 0
𝑧𝑧 𝑖𝑖 𝑗𝑗 ln [ 𝜑𝜑 𝑖𝑖 𝑗𝑗 − 𝜑𝜑 𝑖𝑖 𝑗𝑗 − 1
] where
𝑧𝑧 𝑖𝑖 𝑗𝑗 equals 1 if 𝑦𝑦 𝑖𝑖 = 𝑗𝑗 and 0 otherwise;
𝜑𝜑 𝑖𝑖 𝑗𝑗 = 𝜑𝜑 � 𝜇𝜇 𝑗𝑗 − 𝒙𝒙 𝒊𝒊 𝜷𝜷 � ;
𝜑𝜑 𝑖𝑖 𝑗𝑗 − 1
= 𝜑𝜑 � 𝜇𝜇 𝑗𝑗 − 1
− 𝒙𝒙 𝒊𝒊 𝜷𝜷 � ;
Via this model, I specify the following latent relationship for Regression 1:
Equation 1:
𝑦𝑦 𝑖𝑖 ∗
= 𝑥𝑥 𝑖𝑖 𝛽𝛽 + 𝑤𝑤 𝑖𝑖 𝛾𝛾 + 𝑧𝑧 𝑖𝑖 𝛿𝛿 + 𝑣𝑣 𝑖𝑖 𝜌𝜌 + 𝑡𝑡 𝑖𝑖 𝜃𝜃 + 𝒛𝒛 𝒊𝒊 𝝎𝝎 + 𝑒𝑒 𝑖𝑖 , 𝑒𝑒 𝑖𝑖 ~ 𝑁𝑁 (0,1) where
• y i* is the latent level of gentrification for census tract i, which I measure via my
Relative Scale;
• x i is the total number of charter-years attributable to census tract i;
• w i is the distance from census tract i to the District’s central business district;
• z i is an indicator that equals 1 when census tract i is located between Rock
Creek Park and the Anacostia River, and 0 otherwise;
• v i is an indicator that equals 1 when census tract i is located east of the
Anacostia River;
• t i is a indicator that equals 1 when the centroid of census tract i is within 0.5
miles of a Metrorail station; and
• z i is a set of additional controls for the state of census tract i in 2000, namely:
the proximity of census tract i’s 2000 population being 25% white alone, non-
Hispanic (i.e., 75% not white alone, non-Hispanic); the density of housing units
in census tract i in 2000; and the share of occupied units that were rented
(versus owned) in census tract i in 2000.
113
I specify the same latent relationship for Regression 2, except instead of x i (i.e., the total number of
charter-years attributable to census tract i), I specify x i, which is a vector representing the eleven
distinct, type-based charter-years attributable to census tract i.
I employ an Ordinary Least Squares model (OLS) in Regressions 3 and 4 as the dependent
variable is a continuous and linear measure (i.e., my Absolute Index is a linear combination of
individual dimensions of gentrification). I also utilize OLS in Regressions 5 and 6 since the
dependent variable is a continuous and relatively linear measure. I linearly transform my dependent
variable in Regressions 5 and 6 by subtracting the District -level change in the white alone, non-
Hispanic share of the school-age population. As a result, positive coefficients listed in the results for
Regressions 5 and 6 are associations predicting an increase in a census tract’s white alone, non-
Hispanic share of the school-age population beyond the increase for the entire District. Below,
Table 7 provides the results of my six baseline regression analyses. I provide descriptive statistics for
the Table 7 variables in Appendix H.
5.2. Regression Results
Because I utilize different statistical models across Regressions 1-6, and because in each
Regression a higher dependent variable value represents a greater level of gentrification, I focus on
the level of significance and sign (i.e., positive or negative) of the estimated coefficients listed in
Table 7. In general, a positive coefficient implies a higher level of gentrification associated with a
given variable. I discuss Table 7’s results within section 6.
I perform several sensitivity analyses to test the stability of the estimated coefficients in
Table 7. In order, these sensitivity analyses account for the following:
1. I adjust my methodology for attributing charter -years to census tracts, by shrinking the
circular catchment areas around tracts’ centroids from 1-mile radii to 0.5-mile radii;
114
2. I adjust my methodology for measuring census tracts’ exposure to rail transit, expanding
the circular catchment areas around tracts’ centroids from 0.5-mile radii to 1-mile radii;
3. I make both circular catchment area adjustments, as described above, simultaneously;
4. I substitute secondary types for charter school campuses, when applicable; and
5. I assign a weight to each charter-year (i.e., each charter school location in a given school
year), where the weight is equivalent to the charter school’s share of total charter
enrollment in a given school year.
11
I present the results of these sensitivity analyses in Appendix I. The results of these sensitivity
analyses support the general findings I discuss in section 6.
11
Because campus-specific enrollment figures are not available prior to school year 2004 -05, in this fifth sensitivity analysis I consider
only charter-years between school years 2004-05 and 2017-18.
Table 7. Baseline regression results
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(1) (2) (3) (4) (5) (6)
Gentrification measure
Relative
Scale
Relative
Scale
Absolute
Index
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used
Ordinal
probit
Ordinal
probit
OLS OLS OLS OLS
TOTAL charter-years 0.0047** 0.0018** -0.0004+
...Adult charter-years 0.0105 0.0036 0.0011
...Blended Learning charter-years 0.3443 0.0713 -0.0177
...Comprehensive charter-years 0.0163* 0.0073** 0.0010
...Creative Arts charter-years -0.0199 0.0028 -0.0017
...Dual Language charter-years -0.0152 -0.0091* -0.0033*
...Early Childhood charter-years 0.0062 0.0037 -0.0008
...Global Culture charter-years -0.0432** -0.0119** -0.0033*
...Learner Centered charter-years 0.2746* 0.0213 0.0171
...Postsecondary/Vocational charter-years -0.0040 0.0001 -0.0009
...Specific Population charter-years 0.0243* 0.0063 0.0015
...STEM charter-years -0.0189 0.0113 -0.0090**
Distance to central business district -23.5686** -14.2750+ -7.9344** -1.7736 -2.1544** -1.8188*
Between Rock Creek and Anacostia 0.8842* 0.8966* 0.1850+ 0.1496 0.2783** 0.2731**
East of Anacostia -0.6541 -1.2596* -0.1222 -0.3758 0.2017** 0.1823**
Within 0.5 miles of Metrorail station 0.2486 0.4176+ 0.2157** 0.2395** 0.0190 0.0252
Proximity of population to 75% non-WANH in 2000 0.0120 0.0112 0.0010 -0.0001 0.0006* 0.0004
Housing units per square mile in 2000 -0.0000 -0.0000 -0.0000* -0.0000* 0.0000 0.0000
Share of occupied units that were rented in 2000 -0.7822 -1.0654+ 0.3277 0.2917 -0.1442* -0.1539*
115
cut1 -1.2170 -0.7707
cut2 -0.4249 0.0866
cut3 0.0213 0.5986
cut4 0.7815 1.4837
constant 0.3853 0.0134 -0.0222 -0.0284
n (# of census tracts) 175 175 175 175 175 175
Wald statistic or F-statistic 150.54 176.46 25.58 16.35 15.44 8.34
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.2908 0.3409 0.3894 0.4488 0.3433 0.3630
Notes: Standard errors in Regressions 1 and 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 3, 4, 5, and 6 are
adjusted for heteroskedasticity using the Huber-White sandwich estimator.
116
117
6. Discussion and Conclusion
Within this paper, I examine the relationship between the District’s charter schools and the
gentrification of its neighborhoods. To test my hypotheses in a more robust manner, I conduct my
examination in multiple ways. I use both descriptive and regression analyse s (sections 4 and 5),
include multiple dimensions in my measures of gentrification (section 3), consider both relative and
absolute measures of gentrification (section 3), and study changes in charter schools’ enrollment
compositions on top of neighborhood changes (section 4). I also include multiple sensitivity analyses
to test the stability of my baseline regression estimates (Appendix I).
From a descriptive perspective, I uncover compelling spatial patterns in both the District’s
gentrification and the proliferation of its public charter school sector over time. I also find variations
in charter schools’ enrollment over time and by type, which suggest both a linkage between charter
schools and gentrification as well as a distinct preference among gentrifying households for certain
types of charter schools. The case for divergent preferences by type is buoyed by my regression
analyses, which indicate that neither Creative Arts nor Dual Language charter schools are
systematically located in the District’s most gentrifying neighborhoods (Table 7). Therefore, a spatial
proximity bias in the schools that households attend does not explain why a disproportionately high
share of white households enroll at such charter schools (Table 5). It is possible that the relatively
small enrollment size of such schools disincentivizes many gentrifying households from moving
closer to reduce transportation costs, as few households have the ability to attend such schools
relative to Comprehensive schools (see Appendix J for enrollment size statistics by charter type).
Using both my Relative Scale and Absolute Index measures, I estimate a significant and
positive coefficient between a District neighborhood’s level of gentrification ( i.e., a census tract’s
gentrification) and its level of exposure to charter schools overall; and I identify a similar
relationship using a neighborhood’s exposure to Comprehensive charter schools only (Table 7).

118

Furthermore, using both measures, I estimate a significant and negative relationship between a
District neighborhood’s level of gentrification and its level of exposure to Global Culture charter
schools (Table 7). Finally, my Relative Scale measure results in significant and positive estimated
relationships between a District neighborhood’s level of gentrification and the extent of its exposure
to Learner Centered and Specific Population charter schools, while my Absolute Index measure
results in a significant and negative estimated relationship between gentrification and exposure to
Dual Language charter schools. Use of the Relative Scale measure results in more types being
identified as significantly related to neighborhoods’ gentrification levels than use of the Absolute
Index measure: four types versus three types, respectively (Table 7, Regressions 2 and 4).
I find the significant and negative coefficients associated with Dual Language and Global
Culture charter schools in Regressions 2 and 4 especially notable (Table 7). Charter schools tailored
to serving English learners, as well as Afrocentric schools, are manifestations of the
“counterpublics” that Wilson (2016) documents.
12
Both types appear oriented toward serving
longtime residents of the District and their communities. However, while Afrocentric schools seem
highly unattractive to inmoving gentrifiers, Dual Language schools seem highly attractive. In school
year 2017-18, for instance, Dual Language schools enrolled white alone, non-Hispanic students at a
rate over four times that of the overall charter sector; they also enrolled at-risk students at a rate less
than half that of the overall charter sector (Table 5). Meanwhile, every student enrolled at an
Afrocentric school in year 2017-18 was black alone, non-Hispanic (Table 5). Thus, a unique tension
is apparent in these Dual Language schools, which appear to be serving both long-residing
immigrant communities from Latin America as well as new, gentrifying households. This unique
tension warrants further, and likely qualitative, study.

12
Per Appendix D, all of the District’s Global Culture sc hools have offered – and if still open, continue to offer – Afrocentric
curricula.
119
The results of Regressions 5 and 6 are also quite illustrative (Table 7). Unlike the regressions
using my multi-dimensional measures of gentrification, Regression 5 estimates a significant and
negative relationship between the change in a District neighborhood’s white alone, non-Hispanic
share of its school-age population and the level of its exposure to charter schools. Similarly,
Regression 6 estimates a significant and negative relationship between the dependent variable and
the level of its exposure to Dual Language, Global Culture, and STEM charter schools, with the first
two types likewise identified in Regression 4. I highlight the indicator variable for a census trac t
being east of the Anacostia River, which has a positive and significant estimated coefficient in
Regressions 5 and 6 but a negative – and sometimes significant – estimated coefficient in Regression
1 through 4. This reveals a divergent spatial pattern in the school-age population’s compositional
change over time compared to the other dimensions of gentrification I consider (e.g., growth in
household income). This divergence, along with the divergent coefficient estimate for total charter-
years in Regression 5, reinforces my argument that gentrification is a complex and multi-dimensional
phenomenon. Understanding this dimensionality and developing a robust conceptual framework for
gentrification is an area of the field that requires extensive additional stu dy.
The results I present in Table 6 supplement my relatively consistent findings from
Regressions 2,4, and 6 (i.e., the regressions assessing charter school presence by type). Per Table 6, it
appears that certain households’ preferences for particular types of charter schools may outweigh
the convenience of a school’s physical proximity. Table 6 also demonstrates that even charter
schools located in significantly gentrified areas of the District tend to have enrollment compositions
that are significantly less white than the school-age populations that reside around them. Thus, it
may be the case that gentrifying households are attracted to neighborhoods with charter schools (or,
charter schools are attracted to gentrifying neighborhoods) more than these households are seeking
for their children to actually attend said charter schools. This could stem from the type-based
120
preferences I mention above, from outgroup avoidance on the part of gentrifying households who
want their children to attend other schools (e.g., private schools, traditional public schools, etc.) with
peers who are socio-demographically similar, or from entirely other mechanisms.
Regarding other mechanisms, it is possible that certain charter types intentionally locate in
gentrifying areas because they seek to enroll the types of households moving there, while other
charter types intentionally locate outside of gentrifying areas but are still highly attractive to
gentrifying households. Such a phenomenon would help explain my consistent findings regarding
the enrollment compositions of Dual Language and Learner Centered schools (see Tables 5 and 6)
as well as my divergent findings regarding their spatial locations relative to gentrifying areas (see
Regressions 2 and 4). It would also align wit h evidence that leaders of District charter schools are
designing their schools with specific neighborhoods and communities in mind (Langhorne, 2019).
Without question, additional research is necessary to articulate a cohesive framework.
Finally, and separately, I note that the estimated coefficient for a District census tract’s
exposure to rail transit is positive and statistically significant in three of my first four regressions.
This finding is consistent with the literature ( see sub-section 2.1). It also stresses the need for
researchers to adequately control for other neighborhood-based amenities when exploring the
relationship between neighborhood gentrification and a specific amenity, including charter schools.
Prior studies on the link between urban gentrification and charter schools have failed to include such
controls (e.g., Pearman & Swain, 2017).
In sum, a clear majority of my results are consistent with my two hypotheses and build upon
the literature I review in section 2 ( e.g., Bifulco & Ladd, 2007; Hankins, 2007; Kerr, 2012). I stress,
though, that my results are purely associational in nature. They do not reveal the mechanistic
framework by which charter schools cause or result from the gentrification of urban neighborhoods.
121
Understanding that framework remains a critical area for future research, since it is likely that
multiple and interrelated mechanisms are responsible. It may be the case that gentrifying households
demand charter schools because systems of choice allow gentrifying households to practice
outgroup avoidance more effectively (e.g., Hankins, 2007; Saporito, 2003). Or it may be the case that
institutional logics associated with charter schools allow a greater level of diversification than in
traditional public school systems (Thornton & Ocasio, 1999), and that gentrifying households prefer
different types of charter schools than other households. I generate some evidence in support of the
latter.
It is also possible that the opening of charter schools may increase nearby home values if
those charter schools are perceived as academically superior to proximate, traditional public school
options (Beracha & Hardin, 2019). Alternatively, because they are neighborhood-based and highly
visible amenities, new charter schools may be potent symbols communicating that a neighborhood is
changing or is intended for a population different from existing neighborhood residents (Stone,
2012). They may act as signals to gentrifiers moving into urban areas. Finally, and similarly, the
establishment of charter schools may be a signal to public investors, helping attract redevelopment
activity as bellwethers of community change. Whatever the case, it is likely that both quantitative and
qualitative methods are necessary to disentangle them.
In addition to my findings’ associational nature, I note two other limitations. First, my results
come from a single city, the District of Columbia. It is possible that the relationship between charter
schools and urban gentrification varies significantly across other cities. Second, I have taken pains to
define and measure my data in as clear and supportable a way as possible. Still, I am aware these
methods are my own and do not perfectly match those of other researchers.
Yet despite these limitations, I believe my results have significant implications for scholars,
policymakers, and the general public. First, they highlight a significant relationship between charter
122
school reform and urban gentrification at the neighborhood level, a finding at tension with general
equity goals associated with charter reform. Second, they indicate an important variation in the
charter school and neighborhood relationship that depends upon type of charter school – a nuance
only separate areas of education literature have considered. Third, they underscore a new way in
which inequitable access to urban schools may occur: the gentrification of neighborhoods
surrounding charter schools, and the subsequent crowding out of long-residing, displaced, or
otherwise disadvantaged households from those charter schools due to limited enrollment space and
physical inaccessibility. And specific to policy analysis of the SRA, they suggest that charter reform
may indeed have been an effective component of the District revitalization sought by federal
lawmakers in the 1990s.
At a minimum, this paper raises a legitimate challenge to the argument that instituting
charter reform de facto makes households’ access to schools more equitable, much as findings on
exacerbated segregation in the charter sector have done. One of the defining characteristics of
charter schools as a reform mechanism is their greater flexibility and autonomy relative to traditional
public schools (Chubb & Moe, 1991). But this fundamental trait may be highly problematic as well.
Charter schools have the potential to mean different things to different people. They may help
gentrifying households establish more robust communities separate from existing neighborhoods or
residents, much as they may divide student enrollment rather than diversifying it. They may also
tend to operate relatively far from the families who most desperately need public schooling options.
Given the increasing popularity of charter schools and ongoing gentrification of American cities, it is
crucial that we understand whether this is the case.
123
CHAPTER 4
Detecting Patterns in
Charter School Closure in the District of Columbia
But now, in an unexpected turn of events, National Collegiate Prep has launched a new
campaign to fight its closure, and to fight the [Public Charter School Board], which
organizers say is unfairly targeting black-led charter schools in the District. Under the banner
of “Save Our Students, Save Our School, Save Ourselves” and backed by local clergy,
community organizations, and Ward 8 City Councilmember Trayon White, National
Collegiate Prep staff and students have taken the unusual step of refusing to accept the
[Board’s] decision.
–This Ward 8 High School Had Its Charter Revoked. Now It’s Fighting Back, DCist, 2019
Even after the Brown v. Board of Education ruling, severe segregation remains entrenched in
America’s public education system (e.g., Fiss, 1965; Green & Gooden, 2016; Kozol, 1991; Walters,
2001). States have responded over the past 30 years by legislating new choice mechanisms for
households, such as intra-district enrollment, private school vouchers, and public charter schools
(Forman, 2004; Smith, 1995).
Researchers have established, however, that these choice mechanisms can actually replicate
or worsen de facto segregation (Frankenberg, Siegel-Hawley, & Wang, 2010) by:
1. facilitating “outgroup avoidance” by white and wealthy households (Saporito, 2003);
2. reinforcing informational gaps across types of households (Ball & Vincent, 1998); and
3. diversifying the “types” of education available (e.g., dual language schools), some of which
may attract certain households (Bifulco & Ladd, 2007).
Yet for the reasons below, charter schools may especially complicate issues of equitable access rather
than resolving them.

124

First, although new charter schools require physical facilities, many jurisdictions do not plan
where charter operators should locate (e.g., the District of Columbia). This is particularly concerning
given evidence that charter schools systematically locate away from districts’ most impoverished
communities (Kerr, 2012), producing clear inequities in households’ physical access to schooling
options (Valant & Lincove, 2018).
Second, and serving as the focus of this paper, policymakers and scholars have advocated for
aggressively closing those not meeting academic performance goals (CREDO, 2013; Orfield, 2014;
Osborne, 2012), along with those exhibiting deficiencies in other areas, e.g., chronic under-
enrollment.Like operators’ locational decisions, it is possible this closure instrument exacerbates
inequities in school access. For example, closure policies may disproportionately target impoverished
students, given findings that income-based segregation is a powerful predictor of achievement gaps
in a district (Reardon, 2011). Insofar as race and income are correlated, closure policies may ta rget
students of particular races as well. But despite these potential outcomes, virtually no work examines
the factors predicting charter closures. As a result, I aim to help develop a new strand of literature by
studying the phenomenon in the District of Columbia.
In accordance with the literature on achievement gaps (see sub-section 2.2), I hypothesize
that charter operators with high concentrations of impoverished students – and in the District’s
context, also high concentrations of black American stude nts – are more likely to experience closure
due to poor academic performance (i.e., “academic closure”). Separately, based upon my conceptual
framework (see sub-section 2.4), I hypothesize that charter operators enrolling low shares of all
charter students in the District, as well as charter operators not belonging to a charter management
organization (CMO), are more likely to close for academic reasons too.
The District is a salient study area for several reasons. First, charter schools have operated in
the District for an extended period of time, since school year 1996-97, and now enroll about half of

125

all public students in the District (OSSE, n.d.b). Second, of the 104 charter local education agencies
(LEAs) that have operated in the District, about 25% have closed for academic reasons, and over
40% for any reason (see sub-section 4.3). Third, and unlike many other jurisdictions ( e.g., the City of
Los Angeles), a single entity has made all charter granting and revocation decisions since school year
2007-08. Fourth, an array of information on charter LEAs and closure decisions is publicly available,
including enrollment demographics and reasons for closure. And fifth, District residents have long
suffered stark residential segregation, along both racial and income lines (see sub-section 4.1). This
divide has produced markedly unequal levels of access to high-quality District schools (Orfield &
Ee, 2017), and was in fact a cited factor in the federal government’s decision to allow charter schools
in the District in 1996 (Gunderson, 1995a, 1995b).
To test my hypotheses, I hand-construct a longitudinal dataset of District charter LEAs. I
then conduct descriptive and regression analyses to identify the factors associated with academic
closures between school years 2010-11 and 2018-19. I select that nine-year period as the Public
Charter School Board began implementing its hallmark accountability tool for academic
performance, the Performance Management Framework, in school year 2010-11. I focus on
academic closures as they are a direct consequence of the Framework’s implementation, comprising
74% of all closures during the nine-year period (see sub-section 4.3). This focus also helps me draw
from the well-developed and adjacent literature on gaps in student achievement.
I find that that a charter LEA is significantly more likely to experience academic closure if:
(1) it educates high concentrations of black American (and probably impoverished) students, or it
operates in neighborhoods with high concentrations of impoverished and black American children;
or (2) it enrolls a low share – less than 1% – of all District charter students. I do not identify a
significant relationship between CMO membership and academic closure.
126
From a policy and planning perspective, why do these results on matter? For one, the
literature finds that school closure may permanently reduce the academic outcomes of affected
students. If so, then a crucial policy within the District’s choice landscape (i.e., the Performance
Management Framework), is dealing disproportionate damage to its most vulnerable and historically
oppressed communities. In addition, even if closures are justifiable per uniformly -applied criteria,
their tendency to occur in certain communities exposes a non-uniform landscape of high-quality
schooling options. And finally, school closure is a visible, disruptive, and stigmatizing event ( e.g., see
section 6). The symbolic message it sends to enrolled students, affected communities, and those
observing from afar may exacerbate social biases and stereotypes regarding who is expected to
receive a high-quality education. To convey these points collectively: rather than redressing a system
of inequitable access, the District’s charter reform may be actively reinforcing it.
This paper is structured as follows. Section 2 reviews the literature on factors predicting
charter closure, factors predicting achievement gaps, and the effects of school closure, and also
outlines my conceptual framework. Section 3 details the data used in all analyses. Section 4 describes
the segregation of District neighborhoods, the segregation of students across charter campuses, and
patterns in and impacts of charter closure. Section 5 presents results from my primary regression
analysis, which empirically estimates the predictors of academic closure, as well as supplementary
regression models. Section 6 discusses my findings and concludes.
2. Literature Review and Conceptual Framework
2.1. Predicting Charter School Closure
To-date, only one published article has modeled charter school closure from student-level
data. Using a national dataset for years 1992 through 2005, Paino, Boylan, and Renzulli (2017) find
that a charter school’s probability of closure is significantly and positively associated with its share of
students considered black American, and significantly and negatively associated with its enrollment

127

size. The authors observe diverging probabilities of closure for two groups of schools as they
operate over time. Probabilities are roughly constant over time for charter schools that enroll high
shares of black American students, while they diminish over time for those that do not.
I note four limitations with Paino et al.’s (2017) research, which I hope to build on and
address with this paper. First, the authors’ models do not distinguish between different types of
closure, e.g., poor academic performance or low enrollment, despite the fact that disparate
mechanisms may drive different types of closure. Second, the authors use relatively dated
information, from a period when charter schools were only beginning to gain footholds in many
districts. Given the substantial growth and maturation of charter schools across districts since then,
it is possible that the documented relationships no longer hold. For instance, the authors report that
financial or operational issues was the predominant reason for closure between 1992 and 2005; and
in fact, my own descriptive analysis generate an identical result for the District between 1996 and
2007 (see sub-section 4.3). But my descriptive analysis also indicates that these closures have steadily
decreased in the District since then, while academic closures have steadily increased.
Third, the authors use closure data collected across many jurisdictions, acknowledging that
each jurisdiction has specific rules and regulations for the oversight of charter schools, and thus each
may approach charter closure in distinct ways. Fourth, they do not distinguish between: (1) the
closure of a single charter campus, but the continued existence of that campus’ operator; and (2) the
closure of an entire charter operator. Conceptually and procedurally, these are distinct outcomes and
may have differing impacts on a district and its students.

128
2.2. Predicting Achievement Gaps Across Schools
In contrast to the above lack of research, a large body of work weighs the factors correlated
with student achievement gaps. Within this literature, scholars consistently conclude that greater
levels of student segregation predict wider gaps in student achievement.
Undergirding many of their studies is a similar conceptual framework: (1) student
segregation may occur across districts (a result of residential segregation) or within districts (not
necessarily a result of only residential segregation when choice mechanisms are present) (e.g.,
Rumberger & Willms, 1992); (2) due to correlations between race/ethnicity and income, it is feasible
that race-based achievement gaps proxy income-based achievement gaps (e.g., Fahle, Reardon,
Kalogrides, Weathers, & Jang, 2020; Reardon, 2011; Reardon 2016); and (3) students’ differential
exposures to peers of a particular race, ethnicity, or household income level ( i.e., “peer effects”) may
be the mechanism through which segregation affects academic performance (Hanushek, Kain, &
Rivkin, 2009; Hoxby, 2000b). A notable portion of the literature contends that discrepancies in
educational quality across segregated schools, i.e., “school effects”, are actually the true mechanism
driving achievement gaps (e.g., Fryer & Levitt, 2004).
In addition to measuring gaps as differences in mean achievement levels, aggregated across
whole schools, researchers report gaps within certain achievement levels, school grades, and time
periods. For example, studies have found that the greatest gaps exist between high-achieving white
students and high-achieving black American students, i.e., segregation has the greatest negative effect
on high-achieving black American students (Clotfelter, Ladd, & Vigdor, 2009; Hanushek et al.,
2009). Other work reports that racial achievement gaps grow most during early schooling grades,
particularly between kindergarten and third grade (Fryer & Levitt, 2006), and that gaps expand most
quickly during summer months, due to inequalities in children’s access to educational materials and
programming during that time (McEachin & Atteberry, 2017).

129

2.3. The Effects of School Closure on Students and Communities
A separate strand of literature considers the effects of school closure on students and their
communities. Because closure policies typically target schools with “poor academic performance”, a
nebulous criterion (e.g., Bross, Harris, & Liu, 2016), most research e stimates impacts on academic
outcomes. And while findings are mixed in this regard, evidence for neutral or negative effects
decidedly outweighs the evidence for positive effects (Sunderman, Coghlan, & Mintrop, 2017).
Looking at traditional high schools in New York City, Kemple (2015) does not find a
significant effect of closure on ninth graders’ outcomes, e.g., attendance rates, graduation rates,
probability of earning a Regents diploma. In contrast, Engberg, Gill, Zamarro, and Zimmer (2012)
conclude that school closure has a significant and negative effect on students’ math and reading
scores if students do not relocate to higher-performing schools. Gordon et al. (2018) also note
negative effects, observing pronounced and persistent gaps in math scores between displaced
students and similar, non-displaced students, plus temporary gaps in reading scores. Further, they
generate evidence that closures decrease the scores of students at receiving schools.
Meanwhile, an assessment of closures in Baton Rouge and New Orleans identifies negative
effects for high school students and positive effects in younger grades (Bross et al., 2016). Similar to
Engberg et al. (2012), those attending a higher-quality new school tended to experience significant
improvements in outcomes, while those attending a low-quality new school tended to experience no
change or reduced outcomes.
Critically, Bross et al. (2016) find that greater planning and oversight can mitigate negative
effects. Immediate school closures, whereby all students depart the same year that a closure decision
is made, appear more detrimental than phased-out closures. And permanent facility closures seem
worse than charter “takeovers” that allow students to re-enroll at the same facility. The generally
positive effects associated with New Orleans closures, compared to the generally negative effects
130
associated with Baton Rouge closure, may stem from the former’s emphasis on moving high-quality
charter operators into closing schools (Bross et al., 2016).
Numerous pieces also contemplate the broader, community-based effects of school closure.
Consistent with findings on student achievement gaps (see sub-section 2.2), most closures are located
in non-white and impoverished neighborhoods (e.g., Gordon et al., 2018). Frequently, these closures
contradict the wishes of and psychologically damage the vulnerable communities they target, who
would rather see greater investments in public services and the continuation of existing schools
(Barnum, 2019; Flynn & Arguelles, 2001; Sunderman et al., 2017). In this regard, the closure
decision-making process may be especially problematic, as it tends to eschew community and
student participation in favor of technical analyses (Sunderman et al., 2017). And finally, school
closures are costly decisions (Lytton, 2011); by definition, they absorb public funds that otherwise
could provide the expansions in public services that historically oppressed communities desperately
seek.
In closing, the literature suggests that school closure is not an intrinsically defensible
strategy: student outcomes may deteriorate; closures may target certain communities; and
underserved, disadvantaged families may continue to be underserved and disadvantaged. These
outcomes may be highly probable in America’s segregated urban school districts.
2.4. Conceptual Framework
Informed by the above literature, I construct my research question, hypotheses, and
conceptual framework (see Figure 14).
Research Question: What charter LEA-specific factors predict academic closure under the
Performance Management Framework, as administered by the Public Charter School Board?
131
Hypothesis 1 (H1): Having a high concentration of impoverished students – and in the
District’s context, black American students – significantly increases the likelihood of
academic closure.
Hypothesis 2 (H2): Operating on a small scale, i.e., enrolling a low share of all District
charter students and occupying few facilities, significantly increases the likelihood of
academic closure.
Hypothesis 3 (H3): Belonging to a CMO significantly decreases the likelihood of academic
closure.
Regarding operator size, it seems feasible that small operators are more prone to academic
closure. For instance, authorizers may prefer to close small operators since such closures will affect
few students and few campuses. Alternatively, because charter expansion often requires authorizer
approval (e.g., in the District, PCSB, 2020b), and because academic outcomes may comprise a
portion of such review (e.g., in the District, PCSB, 2019c, pp. 7-8), operators that continuously
occupy few facilities or enroll few students may do so precisely because they are deemed “poor
performers”. In other words, small operator size may be merely a correlate of poor academic
performance.
Separately, operators who belong to a CMO may benefit from their CMO’s accrued
experience and knowledge, achieving superior outcomes across their campuses. Alternatively, due to
centralized and sophisticated management, CMO membership may augment an operator’s
enrollment and number of facilities occupied. If authorizers are less likely to close larger operators,
then CMOs may indirectly reduce the likelihood of closure via this size mechanism.
132
Figure 14. Conceptual framework, with hypotheses depicted (H1, H2, and H3)
3. Data and Definitions
3.1. Summary of the District of Columbia School Reform Act of 1995 (Pub. L. No. 104-134)
When President Clinton signed the District of Columbia School Reform Act of 1995 (SRA)
on April 26, 1996, he authorized a dramatic shift in the District’s public education system. Among
other provisions, the SRA immediately legalized the chartering of local education agencies (charter
LEAs) and the opening of campuses within the District.
The Act’s Congressional authors intentionally minimized District officials’ capacity for the
planning of charter LEAs. The few oversight instruments provided are: (1) granting charters to
tentative operators; (2) approving charter LEAs to operate in traditional public school facilities or
other publicly-owned properties; and (3) revoking charters from LEAs during: a standard 5-year
charter review cycle (from first year of operation), a standard 15-year charter renewal cycle (from
first year of operation), or an ad-hoc charter revocation proceeding.
H1(+)
Mechanism 1:
High concentration of
impoverished students
Mechanism 2:
Small operating size
“Poor” academic
outcomes for enrolled
students
Small operating size
Mechanism 3:
CMO membership
Large operating size
Outcome:
Academic closure
H2(+)
H3(-)
H2(-)
133
The SRA initially delegated charter granting and revocation decisions to two District entities:
the Board of Education (DC BOE), which already managed the District’s traditional school system,
and a newly-created Public Charter School Board (PCSB). Between 1996 and 2006, DC BOE and
PCSB acted in parallel, granting charters to and revoking charters from the operators they oversaw.
In 2007, the District passed the Public Education Reform Amendment Act (D.C. Law 17- 9), which
established mayoral control over traditional public schools and replaced DC BOE with a new State
Board lacking charter oversight responsibilities. PCSB has acted as the only District charter
authorizer since 2007.
3.2. The Public Charter School Board’s Performance Management Framework
Since its inception, PCSB has become known as a rigorous and exemplary charter authorizer
(Cohen, 2017), possibly because of its accessible and comprehensive policies (PCSB, 2020b). This
paper examines outcomes (i.e., academic closures) attributable to one such policy, the Performance
Management Framework (PMF).
Since school year 2010-11, PCSB has applied the PMF to calculate composite “scores” for
most charter campuses, and to thereby rank their academic performances (PCSB, 2010, 2020a). The
scale for scores ranges from 0% to 100%, with a higher score indicating superior academic
performance. PCSB calculates each score via measurable factors, each intended to capture a
dimension of academic quality. In the current iteration of the PMF (PCSB, 2019c), factors for
campuses educating pre-kindergarten through eighth grade students include: students’ growth in
math and English, per prior and current year standardized assessments; students’ achievement levels
in math and English, per current year standardized assessments; and attendance rates, re-enrollment
rates, and other criteria reflecting the “school environment”.
134
For high school campuses (i.e., those teaching ninth through twelfth grades), PCSB currently
relies on the same factors, along with a fourth category of factors intended to capture students’
readiness for college and careers (e.g., graduation rates, college matriculation rates).
Because they educate a distinct and non-assessed population, PCSB relies on other methods
to score adult-serving campuses, via a separate “Adult Education” framework. PCSB does not score
campuses designed for high-needs special education students, at-risk students, or ungraded students
on diploma or individualized education plan (IEP) certification pathways; these campuses fall under
an “Alternative Accountability” framework (PCSB, 2019c).
After determining composite scores, PCSB converts them to a tiered grading system: Tier 1
for “high performing” campuses (composite score of at least 65%); Tier 2 for “mid performing”
campuses (composite score between 35% and 64.9%); and Tier 3 for “low performing” campuses
(composite score below 35%). PCSB then consults these Tier rankings to determine subsequent
oversight activities for each campus.
Due to their high scores, PCSB does not subject Tier 1 or Tier 2 campuses to additional
oversight. Moreover, PCSB actively encourages LEAs to replicate, expand, or grow enrollment when
all of their campuses consistently receive Tier 1 rankings (PCSB, 2019c , pp. 7-8).
Unlike Tier 1 and Tier 2 campuses, PCSB subjects Tier 3 campuses to Qualitative Site
Reviews in the next school year (for additional information on Site Reviews, see PCSB, 2020c). PCSB
also may begin an ad-hoc charter revocation proceeding for any Tier 3 campus that: (1) has a
composite score below 20%; (2) was also Tier 3 in the prior year and h ad its composite score
decrease by at least 5 percentage points compared to that prior year; or (3) received a Tier 3 ranking
in three of the previous five school years.
PCSB completes the above steps within six months of a school year’s conclusion; thereafter,
the agency publishes a report for each campus containing descriptive statistics from that most recent
135
school year, historical and current composite scores, and historical and current Tier rankings (e.g.,
PCSB, 2017). Because reports for a given scho ol year become public during the following year (e.g.,
reports for school year 2016-17 became public during 2017-18), it is likely that household responses
to score or Tier ranking changes happen one school year later (i.e., score or Tier shifts in school year
2016-17 may predict enrollment changes in 2018-19).
Although the PMF’s structure has been relatively constant since its first application (PCSB,
2019c; PCSB, 2013), it is worth mentioning the following adjustments. First, PCSB has handled both
“Early Childhood” charter campuses (those educating students no older than third grade) and
“Adult” charter campuses (those educating adult students) in disparate ways over different school
years. PCSB did not score and tier Adult campuses prior to school year 2014-15, Early Childhood
campuses teaching kindergarten or older grades prior to 2015-16, and Early Childhood campuses
teaching only pre-kindergarten students prior to 2016-17 (PCSB, 2014, 2015, 2016). This is likely
because the District has historically assessed students between third and twelfth grades (e.g., OSSE,
2011, 2019), and much of the PMF relies on these assessments’ scores. As such, it is clear that even
the current iteration of the PMF scores the performance of Early Childhood campuses differently
than elementary or middle campuses.
As a second note, in school year 2014-15, the District changed its standardized assessment
vendor, moving from the DC Comprehensive Assessment System to the state consortium-designed
Partnership for Assessment of Readiness for College and Careers (OSSE, n.d.c). No charter
campuses – other than Adult campuses – received an official score for school year 2014- 15,
although three charter LEAs did close immediately afterward.
136
3.3. Constructing a Longitudinal Dataset for Charter Campuses and Charter LEAs
3.3.1. Locational Data for Each Charter Campus in Each School Year
For my purposes, I define a charter campus the same way as PCSB. Typically, this is: (1) a
specific LEA, (2) operating in a particular physical facility, and (3) educating grades subject to
identical PMF scoring factors. By this definition, Paul PCS’ Middle School and International High
School are separate campuses, even though they share the same physical facility.
I cross-reference a variety of resources to construct a longitudinal dataset locating District
charter campuses by school year, from 1996-97 to 2018-19 (see Appendix C).
3.3.2. Characteristics of Schoolchildren in Census Tracts
In multiple analyses, I consider the historical distributions of District schoolchildren across
neighborhoods. Many of these schoolchildren did not attend charter campuses, enrolling in
traditional public or private school campuses instead.
To measure schoolchildren’s characteristics, I access the U.S. Census Bureau’s 5-year
American Community Surveys (ACS) administered between 2010 and 2018, plus TIGER/Line
shapefiles at the tract level (U.S. Census Bureau, 2019). I treat each District census tract as a
neighborhood, and I define schoolchildren as those aged 0-17 years, based on the granularity of ACS
data. I use ACS Tables B01001, B01001B, and B17024 to calculate the following measures for each
census tract in each school year: (1) the share of schoolchildren considered black American alone;
and (2) the share of schoolchildren whose households earned below the federal poverty line, which I
refer to as “impoverished children”.
Per the 2010 decennial census boundaries, there are 179 census tracts in the District (U.S.
Census Bureau, 2019). I exclude four of these tracts from all analysis because their land usage is
entirely or almost entirely institutional: Tracts 2.01, 23.02, 62.02, and 73.01.
137
Within my primary regression analysis (see sub-section 5.2), I also account for the
characteristics of schoolchildren who live near each charter LEA in each school year. To do so, I: (1)
identify all census tracts whose centroids are within 1 mile of a certain LEA’s campuses (based on
Chandler, 2015a); (2) sum all schoolchildren, black American schoolchildren, and imp overished
schoolchildren across these census tracts; and (3) calculate the pooled shares of schoolchildren near
the LEA that are black American or impoverished. I then compare these pooled shares to the shares
of actual tracts, determining the pooled shares’ percentile ranks relative to actual tracts’ shares.
3.3.3. Closure Data for Each Charter Campus and Each Charter LEA in Each School Year
PCSB provides data on the closures of all charter campuses and LEAs between school years
1996-97 and 2018-19 (PCSB, 2019a). Data elements include: (1) status of LEA’s charter (i.e., closure
of campus only, with no charter status change; forcible revocation of LEA’s charter by relevant
authorizer; or voluntary relinquishment of charter by the LEA); and (2) reason(s) for closure, which
can fall across three categories: poor academic performance, financial or operational
mismanagement or fraud, and low student enrollment. While National Collegiate Prep PCS was
scheduled to close at the end of the 2019-20 school year, its closure decision was made before the
2018-19 school year; I therefore treat it as a “closed” LEA within this paper. Historically, both
relinquishment and revocation decisions have occurred right before or during the final year of
operation.
To reiterate, this paper focuses on charter LEAs experiencing academic closure. I do not
distinguish between charter relinquishment and charter revocation for academic reasons, since PMF
data indicate similar performance levels across the two closure actions (see sub-section 3.3.4 for
details on PMF data). In fact, of the LEAs that relinquished their charters since school year 2010- 11
and whose data are available, all but one received a Tier 3 ranking in its penultimate year. The lone
138
exception, Septima Clark PCS, received a PMF score of 37.1% before relinquishing its charter – only
2.2 percentage points above a Tier 3 ranking.
3.3.4. PMF Scores and Tier Rankings for Each Charter Campus in Each School Year
Within six months after the end of each school year, PCSB publishes reports with PMF
scores and Tier rankings for each charter campus that: (a) operated in that school year, and (b) did
not close during or after that school year. Campus reports for all school years since 2011-12 are
available on PCSB’s website (PCSB, 2020a). Reports prior to 2014-15 are referred to as “School
Performance Reports”, while later reports are termed “School Quality Reports”.
Although the 2010-11 School Performance Reports are not publicly available, PMF scores
and Tier rankings for campuses operating in that year are accessible in their 2011- 12 and 2012-13
reports. This is because each Report lists a campus’ PMF scores and Tier rankings from the three
most recently completed school years (e.g., a campus’ 2012-13 School Performance Report lists its
2010-11, 2011-12, and 2012-13 PMF scores and Tier rankings). Unfortunately, data from school year
2010-11 are irretrievable for campuses that operated in school year 2010- 11 and then closed in the
following year.
3.3.5. Original Charter Authorizer for Each Charter LEA
Via its website, PCSB lists the chartering authorizer for each charter LEA that has ever
operated in the District (PCSB, 2019d). Again, only PCSB or the now-defunct DC BOE have ever
authorized District charter LEAs, and only PCSB sin ce school year 2007-08.
3.3.6. Charter Management Organization (CMO) Membership for Each Charter LEA
PCSB disseminates annual financial summary reports for charter LEAs. Beginning with
school year 2012-13, these reports list the LEAs that fall under a broa der management organization
(PCSB, 2018d). Within this paper, I define a CMO as an entity that either: (a) oversees the
operations of at least two charter LEAs in the District; or (b) oversees the operations of a District
139
charter LEA and an organization chartered in another jurisdiction. I do so as I hypothesize that the
experience and knowledge accrued from managing multiple networks help CMOs achieve superior
academic outcomes across constituent campuses (see sub-section 2.4).
In some cases, a District charter LEA has permitted certain schools or grades to fall under
the management of a separate organization, typically for early learning or virtual learning (e.g., Perry
Street Preparatory PCS deferred management of pre-kindergarten grades to the AppleTree Institute).
In others, a District charter LEA has reported to a non-profit organization that managed only that
single charter LEA in the District (e.g., St. Coletta Special Education PCS reported to Saint Coletta
of Greater Washington, Inc.). In either instance, I do not label that charter LEA as belonging to a
CMO.
3.3.7. Total Student Enrollment Counts for Each Charter Campus and Each Charter LEA in
Each School Year
I assign each campus-and-school year combination in my locational dataset a total student
enrollment count, as published in audits reports from the District’s Office of the State
Superintendent of Education (OSSE). These reports are available for school years since 2004-05 and
are accessible via OSSE’s official website (OSSE, n.d.b). For all analyses that occur at the LEA level,
I simply aggregate enrollment counts across LEAs’ campuses.
OSSE’s audit reports tabulate enrollment data by LEA, campus, grade, and some student
educational characteristics, such as special education needs level and English learner status.
However, they do not include any tabulations by race, ethnicity, or household income.
3.3.8. Student Enrollment Counts by Race and Ethnicity for Each Charter Campus and
Each Charter LEA in Each School Year
Within each campus-level School Performance or Quality Report, PCSB provides total
enrollment counts and enrollment composition statistics from the most recently completed school
140
year (PCSB, 2020a). These include the shares of students considered black American,
Hispanic/Latino, and white. Importantly, the prior three “race” categories are mutually exclusive: all
black American and white students listed in the reports are non-Hispanic/non-Latino students. I
emphasize these three racial groups as they comprised over 95% of non-adult charter students
between school years 2010-11 and 2018-19 (see sub-section 4.2 below).
For each campus-and-school year combination with an available PCSB Report, I use the
listed shares of students by racial group. I use Reports’ listed shares alone, as raw enrollment figures
do not always match OSSE’s official enrollment audit counts.
I utilize the National Center for Education Statistics’ (NCES) Common Core of Data for
school year 2010-11, given the unavailability of PCSB Reports (NCES, n.d). For each charter
campus operating in school year 2010-11, I use the NCES data to calculate its share of black
American, Hispanic/Latino, and white students, dividing each racial category’s enrollment count by
total enrollment. As with the PCSB Reports, I rely on calculated shares, rather than raw counts, as
NCES’ enrollment figures do not always match OSSE’s official audit reports. I use calculated shares
from other years in the NCES database to corroborate the PCSB data. I also use them as primary
data for LEAs’ final school years, as PCSB Reports omit closing LEAs.
To calculate the aggregate enrollment compositions of each LEA by school year, I: (1)
multiply each of its campus’ racial shares (per NCES and PCSB data) by that campus’ total
enrollment (per OSSE data); (2) by each racial group, add the resulting counts across all campuses;
and (3) divide each racial group’s aggregated count by the LEA’s total enrollment.
3.3.9. A Lack of Reliable Student Poverty Data, and Using Student Race as a Proxy
Although an imperfect measure (Snyder & Musu-Gillette, 2015), researchers have
traditionally relied upon a student’s eligibility for free and reduced price lunch (FRPL) as a binary
indicator of that student’s poverty status (Greenberg, 2018). But because the District adopted the
141
Community Eligibility Provision in school year 2012-13, reliable FRPL counts are unavailable for
the majority of my study period (GAO, 2017). (For more information on the Community Eligibility
Provision and how it affects estimations of students in poverty, see Appendix K.)
Yet race and income are unfortunately and inextricably intertwined within the District.
Numerous studies have found that the average black American Washingtonian has significantly
lower income and markedly less wealth than the average wh ite or Hispanic/Latino counterpart
(Kijakazi et al., 2016; Hendey, 2017). And my own descriptive analyses have similar results,
demonstrating a strong spatial correlation between census tracts’ concentrations of black American
schoolchildren and schoolchildren whose families are under the federal poverty line (see sub-section
4.1). Therefore, a charter campus’ or District neighborhood’s concentration of black American
schoolchildren should also provide significant information on its concentration of impoverished
schoolchildren.
To further validate this prospective relationship, I calculate the pairwise correlation between
charter LEAs’ concentrations of black American students and of FRPL-eligible students. I do so for
school years 2008-09 to 2011-12, using the Common Core of Data for years 2008- 09 to 2010-11 and
PCSB Performance Reports for 2011-12. I exclude NCES information prior to school year 2008- 09
given significant irregularities, while the Community Eligibility Provision prevents me fr om
including any data after 2011-12. I calculate a correlation of 61.21% for this four-year period, which
is significant at the 0.1% level. Therefore, it is probable that racial data decently proxy poverty data,
at least in the District context.
4. Descriptive Analyses
Within this section, I describe: (1) the segregation of the District’s schoolchildren across
their residential neighborhoods; (2) the segregation of the District’s charter students across the
142
campuses they attend; (3) patterns in the closure of District charter LEAs, spatial and otherwise; and
(4) the differential impacts of academic closure on sub-groups of District charter students.
Because many of my analyses are spatial in nature, I generate a number of maps plotting the
District’s geography. These maps demarcate two key geographic features, Rock Creek and the
Anacostia River, which divide the District into three discrete regions. F or reference, I provide a map
of District landmarks and neighborhoods in Appendix E.
I note that within sub-section 4.1, my definition of schoolchildren is those individuals aged
0-17 years, due to the granularity of Census data (see sub-section 3.3.2). Within sub-sections 4.2 and
4.4, I utilize enrollment data naturally restricted to ages 3- 17 years (i.e., students enrolled in pre-
kindergarten through twelfth grade). I exclude Adult charter students from all analysis.
4.1. The Segregation of District Schoolchildren Across Neighborhoods, by Race and Income
Between school years 2010-11 and 2018-19, the District’s black American schoolchildren
experienced high – and worsening – levels of residential segregation (see Figures 15a and 15b). While
census tracts east of the Anacostia continuously housed very high concentrations of black American
schoolchildren, and census tracts west of Rock Creek continuously housed very low concentrations
of black American schoolchildren, census tracts between the two experienced decreasing
concentrations of the same group. Figures 16a and 16b display a subtler but identical, and also
intensifying, segregation of schoolchildren whose households earned below the federal poverty line.
143
Figure 15a. Share of schoolchildren (ages 0-17) considered black American, by census tract, 2010
Sources: 2010 5-year ACS Tables B01001, B01001B, and B17024.
Figure 15b. Share of schoolchildren (ages 0-17) considered black American, by census tract, 2018
Sources: 2018 5-year ACS Tables B01001, B01001B, and B17024.
144
Figure 16a. Share of schoolchildren (ages 0-17) in household below federal poverty line, 2010
Sources: 2010 5-year ACS Tables B01001, B01001B, and B17024.
Figure 16b. Share of schoolchildren (ages 0-17) in household below federal poverty line, 2018
Sources: 2018 5-year ACS Tables B01001, B01001B, and B17024.
145
These results are consistent with existing literature. For one, researchers identify the most
aggressive District gentrification occurring between Rock Creek and the Anacostia (Richardson,
Mitchell, & Franco, 2019). Second, the contrast between schoolchildren west of Rock Creek and
those east of the Anacostia is extensively documented (Asch & Musgrove, 2017; Burner v.
Washington), and helped justify the District’s charter reform in the first place (REACH Project, 2019).
Third, the correlated segregations of black American and impoverished schoolchildren confirm the
strong race-and-income relationship found previously in the District (e.g., Hendey, 2017).
4.2. The Segregation of District Charter Students Across Campuses, by Race
Within the prior sub-section, I present schoolchildren’s distributions by census tract – a
natural consequence of using Census data. In this and the following two sub-sections, I present
findings by the District’s eight political Wards, as well as groupings of Wards. Each District Ward is
represented by a particular city councilmember, and each Ward receives specific consideration in
District policymaking, planning, and legislation (DC Planning, n.d).
These Wards’ boundaries can be helpful in aggregating spatial statistics. Ward 3 contains the
majority of neighborhoods west of Rock Creek; Wards 7 and 8 contain the neighborhoods east of
the Anacostia; and the other five Wards cover neighborhoods between Rock Creek and the
Anacostia (see Figure 17). A small portion of Ward 7 does span the Anacostia, and it housed four
charter campuses between school year 2010-11 and 2018-19. Since no charter LEAs operated in
Ward 3 during that period, I tend to compare Wards west of the Anacostia (i.e., Wards 1, 2, 4, 5, and
6) with those east of the Anacostia (i.e., Wards 7 and 8).
Within the area east of the Anacostia, I also distinguish between Wards 7 and 8. This is
because Ward 8 neighborhoods tended to house higher concentrations of impoverished
schoolchildren than Ward 7 neighborhoods, despite near -identical concentrations of black American
schoolchildren (see Figures 15a, 15b, 16a, and 16b). Such a disparity represents a potentially key
146
variance given the lack of data on students’ household income levels (see sub-section 3.3.9). And it is
one I leverage in my primary regression analysis ( see sub-section 5.2).
Figure 17. District Ward boundaries, with census tract boundaries underlaid, 2010 through 2019
Source: U.S. Census Bureau TIGER/Line Shapefiles.
Turning to my results, I find that the residential segregation of District schoolchildren clearly
translated to a campus-based segregation of students, an expected result per literature on school
choice and residential proximity (e.g., Kleitz, Weiher, Tedin, & Matland, 2000; Lubienski, Gulosino,
& Weitzel, 2009) and a District study by Orfield and Ee (2017).
As Table 8 depicts, the share of students west of the Anacostia who were considered black
American dropped significantly between school years 2010-11 and 2018-19: from 78.9% to 64.9%.
In contrast, the share of students east of the Anacostia who were considered black American
remained roughly constant, between 96% and 99%, across both Wards 7 and 8. These diverging
trends are attributable to significant increases in Hispanic/Latino and white charter students, almost
all of whom enrolled at campuses west of the Anacostia.
147
Table 8. Racial composition of non-adult charter students, by geography and school year (SY)
Share of students considered black American (in percentage points), all campuses
Geographic location SY10-11 SY14-15 SY18-19
West of Anacostia River 78.9 70.0 64.9
East of Anacostia River 98.7 97.9 97.0
… Ward 7 98.0 96.7 96.1
… Ward 8 99.6 98.8 97.8
86.7 80.6 77.5
Share of students considered Hispanic/Latino (in percentage points), all campuses
Geographic location SY10-11 SY14-15 SY18-19
West of Anacostia River 14.5 17.4 17.9
East of Anacostia River 0.8 1.2 1.9
… Ward 7 1.3 2.1 2.6
… Ward 8 0.2 0.5 1.3
9.1 11.3 11.6
Share of students considered white (in percentage points), all campuses
Geographic location SY10-11 SY14-15 SY18-19
West of Anacostia River 4.7 8.5 11.5
East of Anacostia River 0.3 0.3 0.2
… Ward 7 0.4 0.4 0.4
… Ward 8 0.1 0.2 0.1
3.0 5.4 7.1
This divergence aside, it was consistently the case that: (1) almost all Hispanic/Latino charter
students and white charter students were enrolled at campuses west of the Anacostia (see Figure 18a);
and (2) almost all students that were enrolled at campuses east of the Anacostia were considered
black American (see Figure 18b). And in conformance with the literature, this segregation of students
by race corresponded to wide achievement gaps by race (Meghjani, 2020). Additional tabulations of
student demographics, covering all school years between 2010-11 and 2018-19, are available in
Appendix L.
148
Figure 18a. Distribution of non-adult charter students considered Hispanic/Latino or white, by
Ward, school year 2018-19
Source: See Appendix L for additional data on charter students.
Figure 18b. Share of non-adult charter students considered black American, by Ward, school year
2018-19
Source: See Appendix L for additional data on charter students.
149
4.3. Patterns in Charter LEA Closures, Especially Academic Closures
Having described the significant segregation of the District’s schoolchildren, across both the
neighborhoods where they live and the charter campuses where they learn, I now describe the
closures of District charter LEAs. All statistics come from my hand -constructed dataset.
Table 9 presents statistics for all closures over three, mutually exclusive time periods: (1)
between school years 1996-97 and 2006-07, i.e., the first eleven years of charter LEAs in the District,
when both PCSB and DC BOE acted as authorizers; (2) between school years 2007-08 and 2009-10,
when PCSB acted as the sole charter authorizer but did not employ the PMF, as well as a period
capturing much of the Great Recession (NBER, 2010); and (3) between school years 2010-11 and
2018-19, when PCSB acted as the sole charter authorizer and employed the PMF. I also report
aggregate figures over all time, i.e., school year 1996-97 to 2018-19.
In total, 99 charter LEAs operated within the District for at least two years (my analyses
exclude charter LEAs that operated for a single year, as such LEAs may have closed for reasons
related to planning or start-up). Of these LEAs, 41 either voluntarily relinquis hed their charter or
had their charter forcibly revoked, equivalent to a “closure share” of 41.4%. Measured temporally,
closure share increased over the three time periods; i.e., the highest closure share occurred when
PCSB alone acted as a charter authorizer and also employed the PMF. The number of charter LEAs
closed per school year – the “closure rate” – crested in the middle period, at a rate of 3.7 charter
LEAs per year. Even the closure rate of the most recent period was twice as high as the closure rate
of the earliest period, based on rates of 2.1 and 1.0, respectively.
Regarding closure type, the number of academic closures and the share of closures deemed
academic in nature increased substantially over time. Whereas only four academic closures occurred
in the earliest time period, equivalent to 36.4% of all contemporary closures, 14 academic closures
took place in the most recent time period, or 73.7% of contemporary closures. An inverse trend is
150
apparent in financial/operational closures, which decreased in number and in their share of all
closures over the three time periods. The latter trend becomes more dramatic when excluding the
three LEAs closed after investigations into misuse of funds: Community Academy PCS (Tucker,
2015), Options PCS (Brown, 2016), and Village Learning Center PCS (DC Auditor, 2004). I note
that the prevalence of financial/operational closures during the initial years of District charter LEAs
aligns with the existing literature (Paglin, 2001; Paino et al., 2017), and its increasing rarity aligns with
statements from PCSB regarding improvements to its charter granting process (REACH Project,
2019).
The data also indicate an increasing closure share for PCSB-authorized LEAs over time, but
a higher and relatively constant closure share for DC BOE-authorized LEAs. The substantially
higher closure share for DC BOE-authorized LEAs is largely inevitable. Given its dissolution, DC
BOE could not authorize any new charter LEAs after school year 2006-07; the LEAs it had already
authorized, however, could later relinquish their charters or have their charters revoked by PCSB.
Finally, Table 9 suggests a lower probability of closure for LEAs belonging to a CMO prior to the
PMF, but a higher probability during its implementation.
Because my research concerns academic closures during the PMF’s application, I use Table
10 to narrow my focus to two sub-groups of charter LEAs operating between school years 2010-11
and 2018-19: (1) those charter LEAs that experienced academic closure, and (2) those that did not
close for any reason. From each of these sub-groups, I exclude charter LEAs with campuses that do
not report assessment scores, i.e., Adult campuses, Early Childhood campuses, or campuses subject
to the Alternative Accountability system, as such LEAs do not experience the same PMF-based
oversight as LEAs reporting assessment scores. As with Table 9, Table 10 also excludes charter
LEAs that operated for a single school year between 2010-11 and 2018-19; in this case, these were
all organizations that began educating students in school year 2018-19.
Table 9. Summary statistics on charter LEA closures, across different ranges of school years (SYs)
SY96-97 to 06-07
(PCSB & DC BOE)
SY07-08 to 09-10
(PCSB, no PMF)
SY10-11 to 18-19
(PCSB, PMF)
All SYs
Total charter LEAs operating*
66 63 77 99
… Number of charter LEAs closed
11 11 19 41
… Charter LEAs closed per school year
1.0 3.7 2.1 1.8
… Share of charter LEAs closed
16.7% 17.5% 24.7% 41.4%
Charter LEAs closed for academic reasons 4 6 14 24
Charter LEAs closed for financial/operational reasons
10 5 4 19
Charter LEAs closed for low enrollment 0 3 1 4
… Share of closed LEAs closed for academic reasons 36.4% 54.5% 73.7% 58.5%
LEAs authorized by PCSB 40 44 64 73
… PCSB-authorized charter LEAs closed 4 5 15 24
… Share of PCSB-authorized LEAs closed
10.0% 11.4% 23.4% 32.9%
LEAs authorized by DC BOE 26 19 13 26
… DC BOE-authorized charter LEAs closed 7 6 4 17
… Share of DC BOE-authorized LEAs closed
26.9% 31.6% 30.8% 65.4%
LEAs part of CMO network 6 7 12 12
… CMO charter LEAs closed 0 0 4 4
… Share of CMO charter LEAs closed
0.0% 0.0% 33.3% 33.3%
LEAs not part of CMO network 60 56 65 87
… Non-CMO charter LEAs closed 11 11 15 37
… Share of non-CMO charter LEAs closed
18.3% 19.6% 23.1% 42.5%
Note: All calculations exclude the 5 charter LEAs that operated for only 1 school year during period of study: Digital Pioneers Academy PCS, Statesmen Prep Academy
for Boys PCS, Sustainable Futures PCS, The Family Place PCS, and Young Technocrats PCS. Of these 5 excluded charter LEAs, two closed during the period of
study: Young Technocrats PCS (1999) and Sustainable Futures PCS (2018).
151
152
Table 10 reveals key differences between charter LEAs that experienced academic closure
and charter LEAs that did not close for any reason. For one, LEAs experiencing academic closure
tended to be smaller while operating, in both the number of students enrolled and the number of
physical facilities occupied. The average academically-closed charter LEA educated approximately
1% of all charter students and operated in a single facility, whereas the average unclosed LEA
educated approximately 2% and operated across two facilities. Moreover, while the largest
academically-closed LEA educated approximately 2% of all charter students, the largest unclosed
LEA educated approximately 16%.
Still, the two groups of LEAs diverge more dramatically when it comes to where they
operated. On average, academically-closed LEAs educated more than half of their students in
facilities east of the Anacostia (i.e., 52%), while unclosed LEAs educated only a quarter of their
students east of the river. Schools in Ward 8 drive this discrepancy. The average academically-closed
LEA educated 45% of its students in Ward 8; the average unclosed LEA educated about one-tenth
of its students – 11% – in the same neighborhoods.
Given these data, it is unsurprising that the two groups had disparate enrollment
compositions. The typical academically-closed LEA taught a student body that was 98% black
American, 2% Hispanic/Latino, and 0% white. The typical unclosed charter LEA taught a student
body that was 70% black American, 15% Hispanic/Latino, and 11% white. Meanwhile, the lowest
concentration of black American students for an academically-closed LEA was 91%, compared to
10% for an unclosed LEA.
Nonetheless, the two groups are similar across two dimensions as well. Near-equal shares of
academically-closed and unclosed LEAs were authorized by PCSB: 86% and 83%, respectively. Also,
an academically-closed LEA was slightly more likely to be part of a CMO than an unclosed LEA,
with shares of 29% and 20%, respectively.

153

Table 10. Summary statistics on charter LEAs operating between school years (SYs) 2010-11 and
2018-19, by those that experienced academic closure versus those that did not close for any reason

SY10-11 to SY18-19*
LEAs
Closed for
Academics
LEAs
Continuously
Operating
Number of charter LEAs* 14 41
Maximum number of facilities for an LEA 2 8
Minimum number of facilities for an LEA 1 1
Average number of facilities per LEA 1 2
Maximum share of all charter students enrolled at LEA 2% (759 students) 16% (6,264)
Minimum share of all charter students enrolled at LEA 0.5% (154) 0.2% (53)
Average share of all charter students enrolled at LEA 1% (388) 2% (745)
Average share of students east of the Anacostia 52% 25%
… Average share of students in Ward 7 6% 13%
… Average share of students in Ward 8 45% 11%
Maximum share of students considered black American 100% 100%
Minimum share of students considered black American 91% 10%
Average share of students considered black American 98% 70%
Maximum share of students considered Hispanic/Latino 8% 85%
Minimum share of students considered Hispanic/Latino 0% 0%
Average share of students considered Hispanic/Latino 2% 15%
Maximum share of students considered white 4% 46%
Minimum share of students considered white 0% 0%
Average share of students considered white 0% 11%
LEAs authorized by PCSB 12 34
LEAs authorized by DC BOE 2 7
… Share of LEAs authorized by PCSB 86% 83%
LEAs part of CMO network 4 8
LEAs not part of CMO network 10 33
… Share of LEAs part of CMO network 29% 20%

Note: All maximum, minimum, and average calculations are made over the 389 unique charter LEA and school year
combinations observed in the dataset. All other calculations are made over the 55 unique charter LEAs observed. All
calculations exclude charter LEAs with Adult, Early Childhood, and Alternative Accountability campuses. All
calculations also exclude 3 charter LEAs that operated for only one school year, due to beginning operations in 2018 -19:
Digital Pioneers Academy PCS, Statesmen Prep Academy for Boys PCS, and The Family Place PCS.
154
4.4. The Differential Impacts of Academic Closure on Student Sub-Groups
Table 10 implies that academic closures under the PMF affected certain types of students,
and certain District neighborhoods, more than others. And this is borne out in the data. From
school year 2010-11 to 2018-19, there were 262,212 unique combinations of a student, a charter
LEA, and a school year. Based on these combinations, Ward 8 students were three times more likely
to experience academic closure than students in any other Ward. And students east of the Anacostia,
i.e., in Wards 7 and 8, were 3.2 times more likely than students west of the Anacostia. Even more
drastically, black American students were 7.3 times more likely to experience academic closure than
Hispanic/Latino students, and an incredible 61.8 times more likely than white students. Given the
correlation I find between race and income (see sub-section 3.3.9), it is probable that impoverished
students were much more likely to experience academic closure, too. Below, Table 11 lists the
impacts of academic closure by Ward and by student racial group.
Table 11. Impacts of academic closure by student sub-group, school year 2010-11 to 2018-19
Student sub-group
Total student
records
Records observed in
academic closures
% of records observed
in academic closures
Ward 1 students 20,713 359 1.7
Ward 2 students 10,724 0 0.0
Ward 4 students 43,385 249 0.6
Ward 5 students 63,001 1,140 1.8
Ward 6 students 22,299 0 0.0
… Students west of Anacostia 160,122 1,748 1.1
Ward 7 students 50,482 820 1.6
Ward 8 students 51,608 2,774 5.4
… Students east of Anacostia 102,090 3,594 3.5
Black American students 210,948 5,205 2.47
Hispanic/Latino students 29,616 102 0.34
White students 14,530 6 0.04
All students 262,212 5,342 2.0
155
5. Regression Analyses
The descriptive statistics I present in section 4 are quite compelling. They illuminate uneven
and potentially regressive effects of academic closure within the District. But they cannot quantify
the extent to which a given factor predicts academic closure. To address this deficiency, I present my
primary regression analysis, which utilizes the same data summarized in Table 10.
5.1. Functional Form and Specification of Primary Regression Analysis
Similar to Paino et al. (2017), I use a hazard modeling framework to predict the academic
closure of charter LEAs between school years 2010-11 and 2018-19. My sample of “survivors”
consists of the 41 LEAs that did not close for any reason and did not have any Adult, Early
Childhood, or Alternative Accountability campuses; my sample of those experiencing the “hazard
event” consists of the 14 academically-closed LEAs, which also had no Adult, Early Childhood, or
Alternative Accountability campuses.
Because academic closure is a discrete event that can occur once per school year, each
observation is a unique combination of LEA and school year of operation (the same as in Table 10).
From this perspective, academic closure is a rare event, with only 14 occurrences out of 389 LEA-
school year combinations – i.e., 3.6% of all observations.
A review of the methodological literature suggests that complementary log-log functional
forms are especially well-suited for discrete hazard modeling (Martuzzi & Elliott, 1998), particularly
when the hazard event occurs rarely (Kitali, Kidando, Sando, Moses, & Ozguven, 2017). I therefore
specify the below regression framework, using the complementary log-log functional form.
Equation 1:
ln( −ln[1 − 𝑝𝑝 𝑡𝑡 ( 𝒙𝒙 𝒊𝒊 )]) = 𝒙𝒙 𝒊𝒊 𝜷𝜷 + 𝑤𝑤 𝑖𝑖 𝛾𝛾 + 𝑧𝑧 𝑖𝑖 𝛿𝛿 + 𝑣𝑣 𝑖𝑖 𝜌𝜌 + 𝒕𝒕 𝒊𝒊 𝜽𝜽 + 𝛼𝛼 1
𝐷𝐷 𝑗𝑗 1
+ 𝛼𝛼 2
𝐷𝐷 𝑗𝑗 2
+ 𝛼𝛼 3
𝐷𝐷 𝑗𝑗 3
where
• p t(x i) is the probability of academic closure for charter LEA i in school year t;
156
• x i is either a single explanatory factor or a set of explanatory factors, capturing either the
sociodemographic composition of the students enrolled in LEA i or the neighborhoods
surrounding LEA i, respectively (as detailed below);
• w i is an indicator that equals 1 when the enrollment of LEA i, expressed as a share of
total enrollment across all charter LEAs in the dataset, is less than 1%, and 0 otherwise;
• z i is an indicator that equals 1 when LEA i falls under the management of a CMO, and 0
otherwise;
• v i is an indicator that equals 1 when LEA i was authorized by DC BOE, and 0 otherwise;
• t i is a set of indicators for how long LEA i has operated within the District, with a
reference category of 1 to 5 school years; the indicator categories are: 6 to 10 school
years; 11 to 15 school years; and 16 to 20 years; and
• each D j is an indicator used to capture the baseline hazard rates for three discrete time
periods: D j1 for school years 2010-11 to 2012-13, D j2 for school years 2013-14 to 2015-
16, and D j3 for school years 2016-17 to 2018-19.
Within this model, the subscript i represents a particular charter LEA (and technically a
particular school year t of operation, given the pooled nature of the dataset); it identifies the factors
that capture LEA-specific risks of academic closure. LEA-specific risks denoted by i are separate from
baseline hazard rates, which capture the risks of academic closure shared by all charter LEAs during
particular school years. In Equation 1, these baseline hazard rates are denoted by subscript j. For more
information on hazard modeling over discrete periods, see Appendix M.
5.2. Primary Regression Analysis Results
I run a number of regression “Models” using Equation 1, varying only the explanatory
factor(s) captured by 𝒙𝒙 𝒊𝒊 . In Model 1, I use a single explanatory factor intended to capture the
sociodemographic composition of students enrolled in LEA i : an indicator that equals 1 if at least
95% of the student body is considered black American, and 0 otherwise. Although this measure
captures the LEA’s actual concentration of black American students, it is likely not the best proxy
for that LEA’s concentration of impoverished students based on data from Wards 7 and 8 (see
section 4).
157
For Models 2 – 4, I include indicators for whether the majority of a LEA’s students are
located east of the Anacostia (i.e., within Wards 7 or 8), within Ward 7, or within Ward 8. These
indicators should capture variations in the race-income relationship more effectively than Model 1’s
measure (see sub-section 4.2).
In a similar way, for Models 5 – 10, I instead include indicators for whether the population
residing in census tracts proximate to a LEA is at least the 80th percentile in: its concentrations of
black American schoolchildren, impoverished schoolchildren, or both ( see sub-section 3.3.2). These
latter indicators are more granular than the prior, Ward-based indicators. And per the literature on
school choice and residential proximity (see sub-section 4.2), I expect them to even more effectively
capture variations between high concentrations of black American students and high concentrations
of impoverished students.
Table 12 contains the results of my ten Models, with all coefficients presented as odds ratios.
Per Model 1, having an exceedingly high concentration of black American students (i.e., at least 95%
of all students) significantly increases a charter LEA’s likelihood of academic closure. Yet per
Models 2 – 4, the spatial location of a particular LEA’s students is a more powerful predictor. While
educating a majority of students east of the Anacostia significantly increases the likelihood of
academic closure (see Model 2), Model 3 demonstrates that educating a majority of students in Ward
8 drives the more generalized relationship. Model 4 provides reinforcement, showing that when
indicators for the high concentration of black American students, majority of students in Ward 7,
and majority of students in Ward 8 are all considered, only the indicator for Ward 8 is estimated to
be statistically significant (at the 5% level).
Table 12. Complementary log-log regression results (dependent variable: academic closure), school year 2010-11 to 2018-19
Notes: + for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001
All coefficients reported as odds ratios
Charter LEAs closed for non-academic reasons excluded Observations: 389
Charter LEAs with Adult, Early Childhood, or Alternative Accountability campuses excluded Zero outcomes: 375
Charter LEAs that operated in the District for only one school year excluded Nonzero outcomes: 14
Student-based
regression
Ward-based regressions
Explanatory factor (1) (2) (3) (4)
Student population >= 95% Black American 3.050+ - - 0.986
Majority of students east of Anacostia River (Wards 7 or 8) - 3.641* - -
… Majority of students in Ward 7 - - 2.500 2.517
… Majority of students in Ward 8 - - 6.896** 6.965*
Proximate tracts >= 80
th
percentile (black American schoolchildren concentration) - - - -
Proximate tracts >= 80
th
percentile (impoverished schoolchildren concentration) - - - -
… Proximate census >= 80
th
percentile (BOTH concentrations) - - - -
Enrollment share < 1% 2.626 3.820* 3.997* 4.008*
Member of charter management organization (CMO) 1.911 2.056 3.132 3.143
Authorized by District’s Board of Education 0.692 0.939 1.007 1.008
Operated 6 to 10 years 2.340 2.875 3.115 3.121
Operated 11 to 15 years 2.840 3.240 5.629* 5.646*
Operated 16 to 20 years 0.495 0.409 0.683 0.685
Period 1 (SY10-11 to SY12-13) 0.003*** 0.002*** 0.001*** 0.001***
Period 2 (SY13-14 to SY15-16) 0.008*** 0.007*** 0.004*** 0.004***
Period 3 (SY16-17 to SY18-19) 0.011*** 0.008*** 0.005*** 0.005***
Converged Log Likelihood -54.190 -53.436 -51.560 -51.560
Wald chi2 126.45 123.85 116.37 116.38
Prob > chi2 0.0000 0.0000 0.0000 0.0000
158
Table 12 cont’d. Complementary log-log regression results (dependent variable: academic closure), school year 2010-11 to 2018-19
Proximate census tract-based regressions
Explanatory Factor (5) (6) (7) (8) (9) (10)
Student population >= 95% Black American - - - - 1.478 1.240
Majority of students east of Anacostia River (Wards 7 or 8) - - - - - -
… Majority of students in Ward 7 - - - - - -
… Majority of students in Ward 8 - - - - - -
Proximate tracts >= 80
th
percentile (black American schoolchildren conc.) 1.558 4.452** - - 3.524+ -
Proximate tracts >= 80
th
percentile (impoverished schoolchildren conc.) 3.347 - 4.883** - - -
… Proximate census >= 80
th
percentile (BOTH concs.) - - - 5.649** - 4.983*
Enrollment share < 1% 3.483* 3.359+ 3.565* 3.425* 3.151+ 3.319+
Member of charter management organization (CMO) 2.662 2.236 2.758 2.790 2.071 2.655
Authorized by District’s Board of Education 0.729 0.745 0.726 0.701 0.736 0.700
Operated 6 to 10 years 3.045 2.967 3.039 3.133 2.803 3.030
Operated 11 to 15 years 4.857* 4.583+ 4.768+ 5.073* 4.297+ 4.906*
Operated 16 to 20 years 0.645 0.499 0.673 0.729 0.487 0.703
Period 1 (SY10-11 to SY12-13) 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002***
Period 2 (SY13-14 to SY15-16) 0.006*** 0.007*** 0.006*** 0.005*** 0.006*** 0.005***
Period 3 (SY16-17 to SY18-19) 0.007*** 0.008*** 0.007*** 0.007*** 0.007*** 0.006***
Converged Log Likelihood -52.072 -52.675 -52.150 -51.508 -52.555 -51.471
Wald chi2 119.00 120.86 119.10 117.29 120.24 117.00
Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
159
160
Indicators for the sociodemographic characteristics of proximate neighborhoods (Models 5
– 10) act very similarly to the indicators for students’ spatial locations (Models 2 – 4). Nevertheless,
it is striking that the attributes of the school-age populations surrounding LEAs’ campuses are
estimated to be more significant predictors of academic closure than the characteristics of the LEAs’
student bodies themselves (see Models 9 and 10). Again, it appears feasible that these non -student
measures are picking up key variations in household income levels that the student race data cannot.
Turning to other LEA attributes, besides Model 1, enrolling less than 1% of all charter
students is a significant predictor of academic closure. Conversely, and across all Models, indicators
for CMO membership and chartering authorizer are estimated to be insignificant predictors of
academic closure. Finally, in a majority of specifications ( i.e., Models 3 – 10), LEAs operating in their
11
th
through 15
th
years are significantly more likely to close than LEAs in th e reference category, i.e.,
those operating in their first through fifth years. My Supplementary Model results below provide one
explanation for why this is the case.
To test the robustness of my analysis, I re -run my ten Models using the logistic functional
form, which is appropriate alternative form for discrete hazard modeling (see Appendix M). The test
results are highly stable, and are provided in Appendix N.
As a final note, the odds ratios I present in Table 12 communicate critical aspects of the
relationships between my explanatory factors and the probability of academic closure. However, it
can also be quite useful to refer to these factors’ marginal effects, i.e., the extent to which a change in
a charter LEA characteristic changes the likelihood of that LEA closing. Accordingly, Table 13 lists
marginal effect sizes for Model 4’s explanatory factors and estimated odds ratios.
161
Table 13. Marginal effects of Regression Model 4 explanatory factors
+ for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001
Explanatory Factor
Change in Probability of
Academic Closure (i.e.,
Marginal Effect Size)
Student population >= 95% Black American -0.03%
Majority of students in Ward 7 1.74%
Majority of students in Ward 8 3.65%*
Enrollment share < 1% 2.61%*
Member of charter management organization (CMO) 2.16%
Authorized by District’s Board of Education 0.02%
Operated 6 to 10 years 2.14%
Operated 11 to 15 years 3.26%*
Operated 16 to 20 years -0.71%
Note: All explanatory factors are indicator variables. Marginal effects calculated at means.
5.3. Supplementary Regression Analysis Results
Before discussing and concluding, I present the results of supplementary regressions, which
reinforce my conceptual framework (see sub-section 2.4) and relate to writings on school choice and
exacerbated segregation (see section 1). These results outline how systems of choice, intersected with
accountability frameworks like the PMF, may lead to a feedback loop of inequitable access: de facto
segregation producing achievement gaps, achievement gaps producing de facto segregation, and
increasingly concentrated academic closures.
I run a total of eight “Supplementary Models” (see Table 14). For these Supplementary
Models, I include only campuses located between Rock Creek and the Anacostia. I do so in light of
the fact that: (1) neighborhood change was largely confined to this area (see sub-section 4.1), and (2)
households residing in this area exercised greater inter-Ward school choice than households east of
the Anacostia (Gallagher, 2019).
In Supplementary Models 1 and 2, the dependent variable indicates whether the PMF Tier
ranking for a charter campus increased between the prior and current school year. For Model 1, the
162
primary explanatory factor is the change in the campus’ share of students considered black
American, also between the prior and current school year; for Model 2, the primary factor is the
change in the campus’ share of students considered white. Models 3 and 4 are identical to Models 1
and 2, respectively, except their dependent variable indicates whether the Tier ranking decreased
between the prior and current school year. Because the dependent variable for Models 1 – 4 is an
indicator variable rarely equal to one (i.e., Tier increases and decreases are rare events), I again
employ the complementary log-log functional form.
In Supplementary Models 5 and 6, the dependent variable measures the change in a campus’
share of students considered black American between the prior and current school year. For Model
5, the primary explanatory factor indicates whether the campus’ Tier ranking increased between the
second prior and prior year; for Model 6, the primary factor instead indicates whether the Tier
ranking decreased. I account for a one school year lag in households’ responses to Tier ranking
changes given PCSB’s timeline for publishing its Reports (see sub-section 3.2). Models 7 and 8 are
identical to Models 5 and 6, respectively, except their dependent variables measures the change in a
campus’ share of students considered white between the prior and current school year. Because the
dependent variable for Models 5 – 8 is a continuous and linear measure, I employ the ordinary least
squares (OLS) functional form for this sub-set.
Building on the literature, my descriptive analyses, and my primary regression results, I
expect that changes in shares of black American students are negatively associated with Tier ranking
increases and positively associated with Tier ranking decreases. Conversely, I expect that changes in
shares of white students are positively associated with Tier ranking increases and negatively
associated with Tier ranking decreases. And in fact, this is precisely what I find. While not all of the
coefficients are deemed statistically significant, all but one’s sign corresponds to expectations, and
163
that coefficient is deemed statistically insignificant (see Supplementary Model 5 in Table 14). For full
results, please see Appendix O.
Table 14. Supplementary model results, by functional form, dependent variable, and primary
explanatory factor, school year 2010-11 to 2018-19
+ for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001
Model
#
Functional
Form
Dependent Variable Primary Explanatory Factor
Factor
z-value or
t-value
1
Complementary
log-log
Tier Rise
(prior to current year)
Black American share change
(prior to current year)
-2.55*
2
Complementary
log-log
Tier Rise
(prior to current year)
White share change
(prior to current year)
2.32*
3
Complementary
log-log
Tier Fall
(prior to current year)
Black American share change
(prior to current year)
-0.01
4
Complementary
log-log
Tier Fall
(prior to current year)
White share change
(prior to current year)
-0.39
5 OLS
Black American share change
(prior to current year)
Tier Rise
(2
nd
prior to prior year)
0.43
6 OLS
Black American share change
(prior to current year)
Tier Fall
(2
nd
prior to prior year)
1.69+
7 OLS
White share change
(prior to current year)
Tier Rise
(2
nd
prior to prior year)
0.42
8 OLS
White share change
(prior to current year)
Tier Fall
(2
nd
prior to prior year)
-0.13
6. Discussion and Conclusion
At its core, this paper asks: which charter LEA-specific factors are associated with academic
closure? This is an important question for the literature on school choice mechanisms and equity of
access. An array of research suggests that school closure can be a regressive policy, depending on
context – its implementation can actually widen the achievement gaps that justified intervention in
the first place. In addition, many studies indicate that school closure is a discriminatory policy,
targeting non-white and impoverished students and their communities, which makes sense given the
literature on achievement gaps. Yet the extant body of work has insufficiently explored what
happens to school closures when systems of choice and accountability frameworks intersect. Due to
164
documented discrepancies in how households interact with choice systems (see section 1), charter
LEA closures within a jurisdiction like the District may be especially regressive and discriminatory,
and in fact, increasingly so over time. They may worsen inequities of access rather than reducing
them.
Considered collectively, my findings offer compelling evidence that this is the case. Between
school years 2010-11 and 2018-19, when PCSB implemented the PMF, the District’s schoolchildren
were highly segregated across their neighborhoods by race and household income (see sub-section
4.1). The District’s charter sector reinforced this residential divide, with children equally segregated
across the charter campuses they attended (see sub-section 4.2).
The increase in academic closures under the PMF (see sub-section 4.3) was distributed in a
similarly uneven way. A majority of students experiencing closure attended campuses in Ward 8,
which contained the District’s most black American and impoverished neighborhoods (see Table 11).
And while 80% of all yearly enrollment records, across all operating charter campuses, were for
black American students, this share jumps to 97% when isolating records for students that
experienced academic closures (see Table 11), a clearly discriminatory impact.
My primary regression Models corroborate these descriptive results, thereby providing even
more support for my first hypothesis. Other than in Model 5, very high concentrations of black
American and/or impoverished children are consistently, significantly, and positively related to the
likelihood of academic closure (see Table 12). The District’s striking correlation between race and
income, combined with the unavailability of student poverty data, makes it difficult to conclude
definitively whether the race-based or income-based segregation of students is the true mechanism
predicting patterns in academic closure.
Nevertheless, I generate results hinting that the concentration of poverty plays a more vital
role. First, my Ward 8 indicator is more significant than my Ward 7 indicator ( see Models 3 and 4).

165

Second, my indicator for a high concentration of black American students is always insignificant,
and my Ward or proximate neighborhood indicators are always significant, when I pair the two types
of indicators (see Models 4, 9, and 10). In other words, the measures appearing to best capture
variations in poverty concentration are the most significant sociodemographic measures in my
Models.
My descriptive and regression analyses also strongly support my second hypothesis. Charter
LEAs enrolling few students and operating in few facilities are systematically more likely to
experience academic closure. Admittedly, this paper is unable to expose the true mechanism
underlying this relationship. Authorizers may prefer to close small LEAs to minimize closures’
impact, or authorizers may force LEAs to remain small due to poor academic performance.
Undoubtedly, additional work is necessary to identify the true mechanisms at play.
Finally and conversely, this paper produces no results in favor of my third hypothesis. On
the one hand, perhaps the true benefit of belonging to a CMO is greater enrollment size or more
facilities occupied, which I already control for by measuring operational scale directly. Alternatively,
perhaps CMO membership does not confer legitimacy or knowledge that protects LEAs from
academic closure. As with above, it is clear more studies are required.
This paper is subject to several limitations, on top of those I have already noted. First, it
relies on data from a single jurisdiction and during particular school years; it is possible other
contexts and periods unveil other closure phenomena. If reliable District data are available for earlier
school years, future studies could examine how variations in academic closures are related to
changes in authorizers’ accountability frameworks versus demographic shifts in the District.
Second, my outcome of interest is academic closure, and so my work does not extend to all
charter closures. This is not to say that other types of closure, such as financial/operational, do not
matter. For example, charter LEAs that stay intentionally small, perhaps because they serve as
166
“counterpublics” for certain groups of students like immigrants and English learners (Wilson, 2016),
may be especially prone to financial/operational closure because they lack the centralized
management structure that CMOs provide. Such a relationship would run counter to the very
equitable access claims that frequently motivate charter reform (Forman, 2004).
As a third limitation, this paper does not look at students’ academic outcomes after closure,
and so it is hard to definitively state that the District’s approach to closure is regressive in addition to
being discriminatory. Still, my spatial analyses demonstrate lesser access to higher-quality charter
campuses (i.e., unclosed campuses) for black American and impoverished students, which the
literature indicates is crucial in preventing negative effects from closure (Bross et al., 2016).
Furthermore, my Supplementary Models imply a positive feedback loop defined by two phenomena:
(1) different types of households respond to Tier rankings (i.e., exercise school choice) in distinct
ways; and (2) Tier rankings increase or decrease due to the presence of different types of students
(see sub-section 5.3). These two phenomena have the potential to work in tandem, inexorably
restricting closures to and further segregating the District’s most vulnerable students. If this does
occur, it is a stirring indictment of the claim that charter reform intrinsically improves access for
schoolchildren and their families. It may also explain why older LEAs are more likely to experience
academic closure (see Table 12).
Fourth and finally, I note that this paper’s analyses are quantitative in nature. Qualitative
research has an equally important role to play in this literature, and it may be especially helpful in
explaining my findings on small charter LEAs and LEAs belonging to CMOs. Such work can also
be invaluable in understanding closure’s psychological impacts on communities, such as a loss in
sense of place (Barnum, 2019).
When the federal government brought charter reform to the District of Columbia, it did so
paternalistically, relying on its own conception of what was best for the District (Gunderson, 1995a,

167

1995b). For all of District officials’ resentment towards this mentality ( e.g., Brown, 1995), they have
recreated it in their management of charter reform. The Performance Management Framework, and
the academic closures associated with it, do not seem best for historically oppressed students and
families. Instead, they appear to actively maintain a segregated and unequal system of school access
for the District’s residents – one that mirrors longstanding inequalities based on race and income.
Until we recognize that replacing communities’ resources is not equivalent to expanding them, we
will fail to admit what inspired the Brown ruling in the first place.

168
CHAPTER 5
Examining the Locational Decisions of
Charter Schools Operating within the District of Columbia
But we have a good example of one right now. Lee Montessori is a charter that does
Montessori pre-K through fifth grade and whatever their grade bands are, but you know pre-
K through fifth generally. They’re in Brookland, northeast DC, and 95% of their children, I
think, are white children…. They came to me a couple of years ago and said, “We want more
diversity in our school. Would you be willing to do a preference for at-risk students to come
into our school building?” They wanted reserve seats for at-risk students, and we’ve debated
that kind of internally whether or not that would be something that we would do. I find it a
little bit offensive to say that you want to ship in at -risk kids. Why can’t you create (a) more
open space that they want to be there, and (b) go somewhere else and set up your
headquarters if that’s what you need to do…. People are always asking about how they can
have a more diverse school. They don’t usually ask what they need to do to make that
happen. They just want to make sure that they keep shipping in kids from out-of-boundary
into their school. It’s kind of a one-way street in that regard. It should be more a two-way
street in my opinion.
–Anonymous DC Council employee, DC REACH Project Interview, 2019
At the end of the 20th century, a trio of novel choice mechanisms began appearing in
American school districts. In 1988, the State of Minnesota introduced the option of “open
enrollment”, permitting students to attend traditional public schools outside the district attendance
area where they resided (Mikulecky, 2013). In 1990, the State of Wisconsin first legalized private
school vouchers for low-income residents and began testing their efficacy in the pilot Milwaukee
Parental Choice Program (Laws of Wisconsin, 1990, Act 36). And in 1991, the State of Minnesota
enacted a third new choice mechanism in the form of public charter schools (Laws of Minnesota,
1991, Chapter 265, H.F. No. 700).
Of these three mechanisms, public charter schools have the potential to alter school districts
in ways that open enrollment and school voucher policies cannot. In particular, while implementing
169
the latter two mechanisms leverages existing educational institutions and their facilities – traditional
public schools and private schools, respectively – introducing public charter schools requires new
institutions and new physical locations for their operation, at least when not converting existing
traditional public schools.
State and local governments plan and oversee these two dimensions of charter reform to
differing degrees. Although they all necessarily create rules and regulations for chartering new
institutions, fewer of them centrally plan where those institutions actually locate their schools
(REACH Project, 2019). Yet where new charter schools locate, and the reasons they select those
sites, may be critical for determining the legitimacy of charter reform, especially if it is intended to
make access to high-quality schools more equitable (Forman, 2004). I now present two hypothetical
examples to demonstrate this.
For one, researches have established that residential proximity plays a major factor in where
children learn, with families preferring schools around where they live (Kleitz, Weiher, Tedin, &
Matland, 2000). Along with outgroup avoidance by white and wealthy families (Saporito, 2003), this
proximity preference helps explain why districts with residential segregation by race or income
exhibit similar segregation within their schools, even when a good degree of school choice is
available (Bifulco & Ladd, 2007; Orfield & Ee, 2017). Significant segregation across schools,
certainly concerning in and of itself, also appears to drive gaps in scholastic achievement across the
same schools (Reardon, 2016). And given the underlying residential segregation at hand, these
achievement gaps may occur across space as well as race and income.
If new charter schools tend to open in relatively wealthy – and probably white –
neighborhoods, perhaps attracted by their high levels of academic achievement, this would represent
a formidable challenge to equitable access. By definition, these schools would be geographically
farther from a district’s poorest residents and therefore less physically accessible. But such long

170

distances also have a disparately high impact on low-income students, who are most likely to rely on
public transit to travel to and from school and therefore least likely to travel long distances for
school (e.g., Gallagher, 2019). The clustering of new charter schools in relatively affluent areas would
be especially regressive in districts that do not provide transportation services for their students
(Valant & Lincove, 2018).
As a second example, new charter campuses may prefer to open in neighborhoods that have
recently lost existing schools, perhaps because of decreased competition. Though as noted above,
residential segregation by race and income appears to drive achievement gaps (Reardon, 2016); and
unsurprisingly, scholars have observed the concentration of closures in non-white and impoverished
neighborhoods (Gordon et al., 2018). If opening where existing schools have closed is indeed a
locational behavior of new charter schools, it does not seem particularly helpful in terms of equity of
access. Rather than having their existing school options supplemented, historically oppressed or
marginalized communities would have them replaced, albeit potentially with better -run or worse-run
schools.
Again, these two examples are merely hypothetical. It is certainly possible that new charter
schools tend to locate in impoverished neighborhoods whose schools have not closed, or
neighborhoods near transit, or neighborhoods that are gentrifying. Still, my two examples
underscore the potential equity implications of new charter school locations. It seems vital for
jurisdictions not centrally planning charter school sites to understand how operators are making
their own locational decisions. Yet unfortunately, the literature on charter operators’ locational
decisions within a district is virtually nonexistent. I therefore aim to address this literature gap by
examining charter campus openings in the District of Columbia between school years 2010-11 and
2020-21.
171
The District of Columbia is a salient study area for several reasons. First, the District does
not centrally plan campus locations for its charter operators (see sub-section 3.1 below). Second,
charter campuses have operated in the District for an extended period of time, since school year
1996-97, and now enroll about half of all public students in the District (OSSE, n.d.b), allowing me
to separate the impacts of traditional public and charter schools. Third, approximately eight new
charter campuses have opened per school year since 2010-11, providing good variation in my
outcome of interest (see sub-section 3.2 below). Fourth, District residents have long been segregated
by both race and income, in their neighborhoods as well as their traditional schools (Asch &
Musgrove, 2017; Burner v. Washington; Orfield & Ee, 2017; REACH Project, 2019). Fifth and finally,
when legalizing charter schools in the District of Columbia in 1996, one of the federal government’s
specified goals was improving school access for households of limited means (Gunderson, 1995a,
1995b). It feels imperative to assess the locational behavior of District charters accordingly.
Based on the limited literature and other applicable research, I develop a conceptual
framework with four types of factors that may influe nce how charter operators select locations for
new campuses (see sub-section 2.2 below):
1. non-schooling characteristics of neighborhoods, e.g., sociodemographic traits of residents;
2. the availability of adequate facilities across neighborhoods;
3. the number and quality of public schooling opportunities close to neighborhoods, both
traditional public and charter; and
4. the locations of other campuses managed by the same charter operator and, if the new
campus is the relocation of an existing campus, the prior location of said campus.
I conduct a variety of conditional logistic regressions to estimate how these four types of
factors predict the actual District neighborhoods chosen for new campuses between school years
172
2010-11 and 2020-21. I generate strong evidence that type 2 and 4 factors play the largest roles in
locational decision-making, and that type 3 factors play a vital role as well. I also produce limited
support for proximity to rail transit, a type 1 factor. Perhaps most interesting, I report that charter
operators are significantly more likely to select neighborhoods already containing high-performing
charter campuses, while they are significantly less likely to select neighborhoods containing high-
performing traditional public campuses.
When assessed collectively, my findings cast doubt on the notion that the District’s charter
reform has equitably enhanced schooling options for its neighborhoods. As I articulate above, if
charter operators prioritize neighborhoods where traditional public and charter campuses close or
otherwise move away from, then charter reform foremost has replaced neighborhoods’ schooling
options rather than supplementing them. Viewed via this lens, although charter operators have chosen
the District’s most impoverished and marginalized communities for many new campuses, it is
because so many traditional public and charter campuses in the same communities have closed as
well (see section 6).
Second, my results indicate that charter campus locations are highly path-dependent, and
that operators prefer neighborhoods already containing other high-performing charter campuses.
This behavior aligns with my first hypothetical example, and validates t he equity concerns it raises;
District charter reform, when it has substantially supplemented a neighborhood’s public schooling
options rather than replacing them, has almost certainly not done so in the city’s most impoverished
and marginalized communities.
When Congressmembers designed and enacted the District’s charter reform, they framed it
as an instrument of equity – one that would resolve historical gaps in access to high-quality public
schools. At the same time, they intentionally minimized the planning and oversight of District

173

charter operators, including locational decisions. This paper highlights a palpable tension in this
theory of action and in the legacy of charter reform and school access in the District.
If equity is a true motivating factor for the District’s – or any jurisdiction’s – charter reform,
managing the locations of charter campuses appears to be a necessary and beneficial function of the
public sector. With that in mind, this paper concludes by identifying ways that state and local
governments can centralize the planning of charter school locations.
2. Literature Review and Conceptual Framework

2.1. The Existing Literature
Few pieces of research explore how charter operators select jurisdictions for their campuses
(e.g., Glomm, Harris, & Lo, 2005). To my knowledge, only one examines where charter operators
locate their campuses inside a given jurisdiction.
Using the locations of charter campuses operating in the District of Columbia in school year
1999-2000, along with contemporary attributes of District census tracts, Henig and MacDonald
(2002) predict the number of charter campuses present in each census tract. Among other things,
the authors report that charter campuses are significantly more likely to be located in tracts with low
rates of student proficiency on the Stanford 9 math exam, high levels of homeownership, high
shares of the population considered black American or Latina/o, higher numbers of vacant public
school buildings, and lower numbers of private schools. They also find charter campuses are
significantly more likely to be located in tracts that contain at least one rail transit station, are close
to the District’s downtown area, and whose residents participate in local elections at high rates.
While Henig and MacDonald’s (2002) paper is quite valuable, I note the following
limitations, which I hope to address with my own paper.
First, the authors use data close to the inception of charter operators in the District. Since
their study year of 1999-2000, the District’s educational context has changed in at least three notable
174
ways. One, the number of charter campuses in the District has more than tripled. Because the
District has become saturated by charters in recent years, it is likely that charter operators’ locational
selections, along with their actual decision-making procedures, have shifted in response.
Two, due to the No Child Left Behind Act’s passage in 2002 (Pub. L. No. 107-110), it is
probable that existing campuses’ standardized assessment scores and performance ratings play a
much more prominent role in more recent charter campus openings relative to the Stanford 9 scores
utilized by the authors. And three, in 2008 and 2013, the District’s traditional public school system
closed a host of its campuses under the leadership of Chancellors Michelle Rhee and Kaya
Henderson, respectively (Brown, 2012, 2013). The data make clear that these mass closings have
influenced new charter campus locations (see sub-sections 4.3 and 5.2), likely altering the estimated
relationship between locations and vacant school building availability.
As a second limitation, and related to the first, Henig and MacDonald (2002) do not control
for the presence of traditional public school and charter campuses separately. Doing so is especially
important when a jurisdiction contains many charter campuses; in such a context, new charter
campus locations should be especially sensitive to the locations of existing ones.
Third, the authors do not include the number of children residing in a census tract relative to
the enrollment levels of nearby traditional public and charter campuses. From a conceptual
standpoint, new charter campuses should target these tracts given that children are more likely to
attend schools close to their residence (Lubienski, Gulosino, & Weitzel, 2009).
Fourth, Henig and MacDonald (2002) do not account for how many campuses fall under
each charter operator. Conceivably, new charter campus locations may be significantly correlated
with the locations of other campuses managed by the same operator (see sub-section 2.2.4).
175
2.2. Conceptual Framework: Four Types of Explanatory Factors
I build on Henig and MacDonald’s (2002) article to construct my own conceptual
framework. To assist in this endeavor, I repeatedly refer to interview data from the REACH Project
(2019), in which I participated as a researcher. The Project, a multi- state initiative funded by the U.S.
Institute of Education Sciences, has conducted a number of interviews with District public
education officials and other relevant parties, oriented around school choice policies and equity
within the District. In addition to these sources, I draw from findings in adjacent literature related to
school choice and charter schools. Aggregating this information, I identify four “types” of
explanatory factors for new charter campus locations.
2.2.1. Non-schooling Characteristics of Neighborhoods
Various studies conclude that charter campuses are significantly more likely to open in
neighborhoods that exhibit signs of gentrification (e.g., Burdick-Will, Keels, & Schuble, 2013;
Hankins, 2007; Kerr, 2012). In fact, Hankins (2007) provides robust qualitative evidence that
gentrifiers directly cause the opening of charter campuses in their neighborhoods, as a way of
redefining the geography they occupy.
Henig and MacDonald (2002) also report that charter campuses are significantly more likely
to open in census tracts that contain at least one rail transit station. This is an intuitive result for two
reasons. One, charter campuses do not have guaranteed enrollment through traditional attendance
zones. Instead, they enroll households across an entire district, and must typically entice households
to “opt out” of their neighborhoods’ traditional public schools. In jurisdictions without busing
services, where many students rely on public transit to travel to and from school, siting a new
campus it near rail transit is a highly sensible strategy.
Two, charter campuses require large spaces in which to operate, historically either vacant
school buildings or other non-residential properties (i.e., commercial or institutional) (GAO, 2003b).

176

Because many jurisdictions, including the District, “upzone” areas near rail transit to allow the
construction of commercial properties (Loh, 2018, 2019), charter campuses may be significantly
more likely to locate near rail transit due to the greater availability of suitable space.
Besides the above two factors, interview data indicate that property rent levels may play an
important role as well. More specifically, one District policymaker has speculated that the absence of
charter campuses in the District’s most affluent neighborhoods is attributable to high rents in those
areas (REACH Project, 2019). Finally, while not discussed in the extant litera ture, controlling for
neighborhoods’ densities of children may be an important consideration as well. Intuitively, charter
campuses may prefer to open in areas dense with children.
2.2.2. Availability of Adequate Facilities
The REACH Project (2019) reveals that acquiring adequate facilities is the largest concern
for District charter operators hoping to open new campuses. Moreover, charter operators in the
District have repeatedly expressed their preference for education-oriented buildings, i.e., “school
buildings” (Gathright, 2019; Sturdivant-Sani, 2018). Given the prohibitive cost of constructing a new
school building, it is unsurprising that Henig and MacDonald (2002) find charter campuses are
significantly more likely to open in census tracts with vacant ones.
On top of vacant school buildings, it is feasible that charter campuses are significantly more
likely to open in tracts with commercial properties formerly occupied by another charter operator.
For one, occupation by another charter operator may be a filtering mechanism, signaling to a new
charter campus that the property is appropriate for use. Second, charter operators may modify
commercial spaces to meet specific needs (e.g., a gymnasium or cafeteria), and new campuses may
prefer to occupy spaces with those needs already satisfied.

177
2.2.3. Nearby Public Schooling Opportunities
In outlining the District’s school choice environment, government officials have described
opening new traditional public schools in neighborhoods where the number of children is increasing
relative to nearby, existing public school seats (REACH Project, 2019). It is reasonable to assume
that charter operators engage in a similar calculus regarding these “enrollment gaps”.
Separately, Henig and MacDonald (2002) report that District charter campuses are
significantly more likely to open in tracts with low proficiency rates on the Stanford 9 math exam.
Again, it is probable that more recent charter campus locations are even more sensitive to nearby
campuses’ levels of “quality”, due to augmented accountability systems. Furthermore, these
locations may be more or less sensitive to the quality of charter versus traditional public schools.
2.2.4. Existing Campuses and Relocation Consideration s
While not explored in any literature, it seems important to account for the locations of a
charter operator’s existing campuses when predicting the location of its new campus. The nature of
this relationship, however, is not intuitively obvious. Charter operators may prefer a diffuse network
of campuses, especially if these campuses educate students in similar grades. Alternatively, charter
operators may prefer to cluster their campuses, even if they educate students in similar grades,
perhaps to better centralize their day-to-day operations and administration.
Similarly, it appears vital to recognize that some new charter campuses are actually the
relocation of an existing campus. In such cases, charter operators may prefer to open relocated
campuses near their prior sites, particularly so that enrolled students do not face changes in their
travel times that cause them to unenroll.
2.3. Regarding Private Schooling Opportunities
Before specifying my research question and hypotheses, I note that I do not account for the
presence of private schools within my locational models. This is a deliberate decision. It is easy to

178

envision families substituting between traditional public schools and public charter schools. Both
types of school are free to attend and secular in their instruction. Yet it is far more intricate to model
how families choose between private schools and public charter schools.
First, families typically spend thousands – or tens of thousands – of dollars per year to enroll
their children in private schools; and to many of them, this is a fair price to pay. After all, on top of
the actual education services they provide, private schools may confer an elite status (Persell &
Cookson, 1985) or institutional “brand” (Cheng, Trivitt, & Wolf, 2016), helping families and
children access prestigious colleges, develop lucrative careers, or maintain social positioning.
Second, many families enroll their children in non -secular private schools so that they
receive instruction according to certain religious practices or values. Third, private school tuition
levels change over time (Dolan, 2018), and it is possible these changes distort families’ willingness to
substitute between private schools and traditional public or public charter schools.
Accounting for these mechanisms, and how they interact with the locational decisions of
public charter schools, is an incredibly complex endeavor and outside the scope of this paper. Still,
scholars have attempted to do so, e.g., Chakrabarti and Roy (2010), who estimate only a small and
negative impact of public charter schools on nearby private schools’ enrollment levels.
2.4. Challenges to Centrally Planning the Locations of Charter Campuses
Before continuing, it is worth understanding how state and local governments actually plan
and oversee their charter operators. Doing so can reveal political contexts that prevent the
centralized planning of charter campus locations. Such contexts’ presence may also explain the lack
of literature on charter operators’ locational decisions. After all, understanding locational behavior is
not pressing if jurisdictions have no way of altering it.
The aforementioned REACH Project (2019) findings are especially useful in this regard. The
Project reviews charter authorization procedures in five states: Colorado, Florida, Louisiana,
179
Michigan, and Oregon. Across these five, the Project identifies two common aspects of their
political contexts that make central planning difficult.
First, each state permits authorization by an array of entities. Each allows local school
districts to act as authorizers, and each allows at least one other type of organization to authorize as
well, such as the state board of education or public postsecondary institutions. As a result, centrally
planning charter locations within a given state would require coordinating both: (a) authorizer
approvals, and (b) the ultimate locations of approved operators (REACH Project, 2019). Certainly,
this is an arduous and politically complicated task that would require a large amount of public
resources.
Second, all states reviewed by the REACH Project require individual charter operators to
secure their own facilities, although some states do require that operators have access to unused
traditional public school space. In addition, charter operators often do not receive public funding to
acquire these facilities. As a result, operators frequently rely on private donations or their public
operating funds to cover their occupancy costs and must also navigate the real estate market.
This is a sharp departure from traditional public schools, whose occupancy costs are usually
covered centrally by local and state governments and whose campus locations are determined via
public planning hearings. And it is possible that states and local school districts have intentionally
designed this facilities discrepancy. They may view inferior facilities environments as the “price” that
operators pay for greater autonomy. I note that this is merely conjecture and requires future
corroboration. Regardless, and like the presence of many different authorizers, it is clear that the
delegation of facility acquisition to individual operators and the lack of public facility funding make
centralized planning more difficult (REACH Project, 2019; National Charter School Resource
Center, 2018).

180

2.5. Research Question and Hypotheses
Drawing from the above, I specify the following research question and hypotheses.
Research Question: Which tract-specific attributes predict the locational decisions of District
charter operators?
Hypotheses for non-schooling characteristics of neighborhoods
Hypothesis 1: District charter operators are significantly more likely to select
gentrified neighborhoods.
Hypothesis 2: District charter operators are significantly more likely to select
neighborhoods containing a rail transit station.
Hypothesis for facility availability
Hypothesis 3: District charter operators are significantly more likely to select
neighborhoods containing either: (a) vacant traditional public school buildings, or (b)
non-school properties recently vacated by another charter operator.
Hypotheses for nearby schooling opportunities
Hypothesis 4: District charter operators are significantly more likely to select
neighborhoods with a high number of resident children compared to the number of
students enrolled at nearby charter or traditional public campuses, i.e.,
neighborhoods with a high “enrollment gap”.
Hypothesis 5: District charter operators are significantly less likely to select
neighborhoods with high-performing traditional public or charter campuses nearby.
Hypotheses for existing campuses and campus relocations
Hypothesis 6: If the District charter operators is opening a new campus and already
operates other campuses, it is significantly more likely to locate its new campus: (a)
181
far away from campuses educating a similar range of grades as its new campus, and
(b) near campuses educating a dissimilar range of grades.
Hypothesis 7: If the locational decision concerns the relocation of an existing
campus, the District charter operators is significantly more likely to select a census
tract near the previous place of operation.
3. Data and Definitions
3.1. Summary of the District of Columbia School Reform Act of 1995 (Pub. L. No. 104-134)
When it became federal law on April 26, 1996, t he District of Columbia School Reform Act
dramatically restructured the District’s public education environment. Among a few other
provisions, the SRA immediately legalized the chartering of local education agencies (“charter
LEAs”) and the opening of campuses within the District.
The Act’s Congressional authors intentionally minimized District officials’ planning and
policy levers regarding charter LEAs. The few oversight instruments provided are: (1) granting
charters to tentative operators; (2) approving charter LEAs to operate in traditional public school
facilities or other publicly-owned properties; and (3) revoking charters from LEAs during: a standard
5-year charter review cycle (from first year of operation), a standard 15-year charter renewal cycle
(from first year of operation), or an ad-hoc charter revocation proceeding.
Salient to this paper, the SRA does not require the District to make vacated traditional public
schools available for use by charter LEAs. When the District has made one available to charter
LEAs, the data indicate no standard timespan between vacating the building and accepting charter
proposals for use.
Besides the release of publicly-owned facilities, the SRA grants the District no control or
oversight regarding the locational decisions made by charter LEAs. These LEAs are free to navigate
the commercial and residential real estate markets and to negotiate with various types of institutions
182
(e.g., post-secondary and religious institutions, as well as the federal government) in order to secure
appropriate spaces for operations.
Since 1998, the District has used a per -pupil funding formula that distributes local tax dollars
to charter LEAs to cover both operating and facilities costs (D.C. Law 12-207). Between school
years 1999-00 and 2007-8, the District determined charter LEAs’ per-pupil facilities allotment by
dividing: (a) the facilities costs of its traditional public schools, by (b) total students enrolled in its
traditional public schools. In the following years, the District has simply stipulated a flat dollar
amount in facility funding per student. In school year 2019-20, the per pupil facilities allotment was
$3,335 (D.C. Code § 38–2908(b-2)(2C)). That the District even has a separate per-pupil allotment
for charter facilities contrasts with the policies of many other stat es and school districts (REACH
Project, 2019).
3.2. Identifying Newly-Opened Charter Campuses in Each Applicable School Year
For the purposes of this paper, I define a charter campus as: (1) a specific charter LEA that
(2) operates in a particular physical facility. By this definition, I consider KIPP DC PCS’ Connect,
Northeast, and Spring Academies to comprise a single charter campus. In contrast, I treat Two
Rivers PCS’ Elementary and Middle Schools as two separate campuses because they occupy distinct
facilities.
I cross-reference a variety of primary resources to construct a longitudinal dataset that
locates all District charter campuses by school year since their legalization in 1996 ( see Appendix C).
I utilize this dataset to identify newly-opened charter campuses in school years 2010-11 through
2020-21. I focus my analysis on this eleven -year period given limitations of census tract-level
measures for the District’s residents (see sub-section 3.3.1 below).
I exclude four types of openings from my analyses. First, I exclude 17 adult charter campus
openings. I do so because adult charter LEAs educate a distinct population with greater mobility

183

than the typical student (e.g., the ability to drive a car); they also typically operate in the evening and
enroll fewer full-time students than other charter LEAs. Therefore, the locational decision-making
process for adult charter LEAs may markedly differ from other charter LEAs. Second, I exclude
seven facilitated campus takeovers. I do so because the agreed transition of a campus’ operation
from one LEA to another does not represent a locational choice by the latter LEA. Third, I exclude
two special education campus openings, as special education charter LEAs serve a specific
population and may make locational decisions in a disparate way relative to other charter LEAs.
Fourth, I exclude a single campus opening that occurred in an entirely institutional census tract,
which makes it impossible to properly pair tract -level data with that opening. Removing these 27
observations leaves 87 campus openings for inclusion in my locational models.
3.3. Measuring a Neighborhood’s Non-Schooling Characteristics
3.3.1. Defining a Neighborhood
Within this paper, I define each District census tract as a neighborhood. Per the 2010
decennial census boundaries, there are 179 census tracts in the District (U.S. Census Bureau, 2019). I
exclude five of these tracts from all analysis because their land usage was entirely or almost entirely
institutional for the majority of my study period. These are: Tract 2.01 (Georgetown University),
23.02 (Washington Hospital Complex, Federal Armed Forces Retirement Home, and Lincoln’s
Cottage), 62.02 (Federal Parklands including the National Mall), 68.04 (Anacostia Park,
Congressional Cemetery, DC Armory, DC Jail, JFK Stadium, and formerly DC General Hospital),
and 73.01 (Anacostia-Bolling Joint Military Base). As mentioned in the previous sub -section, a single
charter campus opened in Tract 23.02 during my study period and is excluded from all analysis.

184
3.3.2. Sociodemographic Characteristics of Neighborhoods
For each school year between 2010-11 and 2020-21, my locational models include two
sociodemographic measures for each census tract: (1) its density of children and, (2) the extent to
which its population has gentrified since the year 2000.
To measure a tract’s density of all children, I divide: (a) its number of residents aged 0- 17
years per the U.S. Census Bureau’s 5-year American Community Survey (ACS) Table B01001, by (b)
its land area per U.S. Census Bureau TIGER/Line shapefiles (U.S. Census Bureau, 2019). In some
locational models, I consider the density of particular age ranges of children, namely those aged 0-4
years, those aged 5-9 years, those aged 10-14 years, and those aged 15-17 years. These ranges roughly
correspond to the ages of students in pre-kindergarten, elementary, middle, and high schools,
respectively (see sub-section 3.5 below). Population counts at this granular level also come from the
5-year ACS Table B01001.
To measure the extent to which a tract’s population has gentrified since 2000, I develop a
continuous index like that utilized by the University of Illinois at Chicago (UIC, 2014). This index
blends absolute changes in each census tract since 2000 across four dimensions:
13

1. share of the population considered white alone, non-Hispanic;
2. share of the population with at least a bachelor’s degree (25 years and older);
3. growth in median household income, adjusted for inflation;
13
2000 total population comes from 2000 Census Table SF1/P001; 2000 white alone, non- Hispanic total population
comes from 2000 Census Table SF1/P012I; 2000 share of population with at least a bachelor’s degree comes from 2000
Census Table SF3/DP2; and 2000 median household income comes from 2000 Census Table SF3/DP3. Later years’
total population counts come from various vintages of the 5 -year ACS Table B01003; later years’ counts of white alone,
non-Hispanic persons come from various vintages of the 5 -year ACS Table B01001H; later years’ shares of the
population with at least a bachelor’s degree come from various vintages of the 5 -year ACS Table S1501; and later years’
median household income levels come from various vintages of the 5 -year ACS Table S1901. Adjustments for inflation
are made using the U.S. Bureau of Labor Statistics’ Consumer Price Index (CPI) for All Urban Consumers in the
Washington-Arlington-Alexandria Metropolitan Statistical Area.

185

4. percentile rank of the census tract per its median household income, compared to all other
District census tracts.

I use relational files from Brown University’s Longitudinal Tract Data Base to translate 2000
decennial census tracts into the same tract geography as 2010 decennial census tracts (Brown
University, n.d). Next, I employ principal component analysis (PCA) to blend each census tract’s
changes across these four dimensions between 2000 and 2018 (i.e., the most current 5-year ACS
data). I select only the first component from the PCA as it explains approximately 80 percent of
total variance in the data, meeting the indexing criteria outlined in the literature (Jolliffe & Cadima,
2016). The results of this principal component analysis are available in Appendix P.
Because a charter LEA must establish the location of a new campus before the campus
begins operating, I establish a one-year lag between census tracts’ sociodemographic measures and
my observed charter campus openings. For instance, I link charter campus openings in school year
2011-12 with: (1) tracts’ density of children levels per 2010 5- year ACS data, and (2) tracts’ indexed
gentrification levels per changes between the 2000 decennial census and 2010 5- year ACS data.
Introducing this one-year lag necessarily restricts the beginning of my study period to school
year 2010-11, as 2009 is the first year that 5- year, census-tract level ACS data is available. Because
my study period runs until school year 2020-21, but the most current 5-year ACS data is from 2018,
I link 2018 5-year ACS data to school year 2019-20 as well as school year 2020-21.
Finally, for all school years in my study period, I use the same first PCA component from
2000 – 2018 changes to blend dimensional changes. And in fact, the starting point to calculate
dimensional changes, i.e., the 2000 decennial census, is the same as well. For a given census tract, the
only input that changes when calculating its indexed gentrification measure in one school year versus
another is the end point data used, e.g., for school year 2011-12, 2010 5-year ACS data, and for
school year 2012-13, 2011 5-year ACS data.

186

3.3.3. Other Non-Schooling Characteristics of Neighborhoods
In addition to census tracts’ sociodemographic traits, I control for two other non-schooling
attributes of each census tract: (1) median gross rent levels in the tract, and (2) whether the centroid
of the census tract lies within 0.5 miles of a rail transit station.
I use median gross rent levels as a proxy for the affordability of school space within the
census tract. For each school year in my study period, these data come from a vintage of the 5-year
ACS Table B25064. As with each tract’s sociodemographic data, I establish a one-year lag between
tracts’ median rent levels and observed charter campus openings, e.g., 2009 ACS rents are associated
with school year 2010-11 charter openings. As the data show (see sub-section 4.3), a high number of
charter campuses have opened within vacant traditional public schools or institutionally-owned
properties. Any measure of residential, or even commercial rents, is unlikely to capture the
idiosyncratic lease rates for such arrangements.
Regarding proximity to rail transit, I use geographic information software (GIS) to locate the
centroids of the 174 census tracts included in my locational models. I then construct a circular
catchment area with 0.5-mile radius around each centroid, per the literature on rail transit access and
catchment areas (e.g., Guerra, Cervero, & Tischler, 2012).
The District provides location data for all of its Metrorail station entrances.
14
I use that data
to assign each rail station a location equivalent to the mid -point between that station’s physical
entrances, not including separate elevator entrances. I then identify whether a District census tract is
proximate to rail based on whether its 0.5-mile radius catchment area contains one or more
Metrorail stations, including stations in Maryland or Virginia.

14
This information is available for public use in ArcGIS via the following link:
https://services.arcgis.com/neT9SoYxizqTHZPH/arcgis/rest/services/Metro_Station_Entrances_Status/FeatureServe
r
187
3.4. Measuring a Neighborhood’s Available and Adequate Facilities
3.4.1. Availability of School Buildings
For each school year in my study period, I determine whether a census tract contains vacant
school space – traditional public or otherwise – that is available for use by a charter campus. To do
so, I first categorize the type of building occupied by each of the 87 new campuses in my study. I
verify whether a building is a vacant traditional public school using a facilities master list published
by the District’s Office of the Deputy Mayor for Education (DME, 2019).
Additionally, I verify whether a campus-occupied property was initially constructed as a
private elementary or secondary school, or as part of a post-secondary institution, by reviewing the
property history and announcements regarding charter campus openings. Sources for this task are
available in Appendix Q.
As I mention in sub-section 3.1, it is impossible to predict when vacant school buildings
become available for charter use. For example, the District of Columbia Public Schools (DCPS) J.F.
Cook campus closed in 2008; a charter campus first occupied the building in 2014, representing a
six-year lag between closure and re-use by a charter LEA. In contrast, DCPS’ Webb campus closed
in 2009, and a charter campus first occupied the building in 2013, equivalent to a four -year lag
between closure and availability for re-use.
Due to such inconsistencies, I assume that a DCPS building is available in a given census
tract and in a certain school year only if a new charter campus opened in that building in the same
school year. Vacant DCPS buildings appear to be the optimal property type for charter LEAs, as
evidenced by recent media campaigns calling for the District to make more vacant DCPS buildings
available for charter use (e.g., Gathright, 2019). Thus, it seems reasonable to assume that a vacant
DCPS building, when available for charter use, is occupied by a charter campus as soon as possible.
188
Similar to DCPS buildings, I assume that a non-DCPS school building (e.g., a private school
building or a post-secondary institution building, such as the Marist School Seminary) is available in
a given census tract and in a certain school year only if a new charter campus opened in that building
in the same school year. I make one exception to this rule, though. Because Potomac Preparatory
PCS (formerly Potomac Lighthouse PCS) occupied the Catholic Sisters College building from 2008
to 2016, and because Mundo Verde Bilingual PCS later opened its Calle Ocho Campus in the
building in 2019, I assume that it was available for charter use in 2016, 2017, 2018, and 2019.
3.4.2. Availability of Non-School Space Previously Occupied by Other Charter Campuses
Besides vacant school space, I also consider the availability of non-school buildings
previously occupied by charter campuses. Such properties are typically commercial or institutional
(e.g., a church annex) in nature.
For each school year, I assess whether a census tract contains a non-school building that was
formerly occupied by a charter campus and is subsequently occupied by a charter campus in that
school year. Like the Catholic Sisters College case I describe directly above, I assume that non-
school buildings are available for charter use during the school years between a charter campus
vacating the space and a new charter campus occupying it, including the year a new charter campus
moves in.
I do not consider non-school buildings that were once inhabited by a charter campus and,
after that campus departed, never housed another charter campus. From a theoretical standpoint, it
is problematic to assume such buildings are available for charter use, as their owners may have
decided against future charter tenants. Nevertheless, I recognize the limitations inherent in defining
the availability of non-school buildings per this sub-section (see sub-section 6.4).

189

3.5. Measuring a Neighborhood’s Nearby Public Schooling Opportunities
I use geographic information software (GIS) to locate the centroids of the 174 census tracts
included in my locational models. I then construct a circular catchment area with 1-mile radius
around each centroid. I employ a 1-mile radius given median travel distances to school reported for
students attending DCPS campuses (Chandler, 2015a), along with findings from the literature on
traveling to school (e.g., Su et al., 2013).
For each school year, I identify the charter campuses and DCPS campuses that are located
within each census tract’s 1-mile catchment area, not including adult and special education
campuses. Longitudinal location data for charter campuses come from my hand-constructed charter
dataset (see sub-section 3.2 and Appendix C). Longitudinal location data for DCPS campuses in
school years 2019-20 and 2020-21 come from the District’s universal enrollment system (My School
DC, 2019, 2020); location data for DCPS campuses in prior school years come from archived
records made available on the DCPS website (DCPS, n.d. a).
3.5.1. Enrollment Gaps Based on Nearby Non -Adult Charter and DCPS Campuses
For school funding and residency verification purposes, the District’s Office of the State
Superintendent of Education (OSSE) is responsible for counting the number of enrolled students in
each charter and DCPS campus at the beginning of each school year.
After an appeals and data review process, OSSE releases finalized enrollment documents
during the same school year, typically around January or February. These documents tabulate
enrollment data by LEA, campus, grade, and some student educational characteristics, such as
special education needs level and English learner status. However, they do not include any
tabulations by race, ethnicity, or household income. Reports are publicly available for all school years
since 2004-05 and are accessible via OSSE’s official website ( OSSE, n.d.b).
190
As I articulate above (see sub-section 2.5), I hypothesize that charter LEAs are significantly
more likely to open new campuses in census tracts that have a high population of children relative to
enrollment levels at nearby charter and DCPS campuses, i.e., census tracts with high “enrollment
gaps”. To calculate enrollment gaps for each census tract and school year, I subtract: (a) the
enrollment count for nearby charter and DCPS campuses, from (b) that census tract’s number of
children. For each school year and census tract, I calculate enrollment gaps in two ways:
1. for all children, by subtracting nearby enrollment counts for pre-kindergarten through
twelfth grades from a tract’s ACS count for the population aged 0-17 years; and
2. for children in specific grade ranges aligning with traditional schooling structures: pre-
kindergarten grades (early learning schools), kindergarten through fifth grades (elementary
schools), sixth through eighth grades (middle schools), and ninth through twelfth grades
(high schools); e.g., by subtracting the number of nearby kindergarten through fifth grade
students from a tract’s ACS count for the population aged 5-9 years old. The applicable age
ranges from ACS population counts are listed in sub-section 3.3.2.
I match each opening charter campus with a grade-specific enrollment gap figure per the
grades that the new campus will educate. For instance, if the opening charter campus will educate
children in third through fifth grades, I match the opening campus with the grade-specific
enrollment gap for nearby “elementary schools”, i.e., across kindergarten through fifth grades.
As with my non-schooling neighborhood measures (see sub-sections 3.3.2 and 3.3.3), I
assume a one-year lag between observed enrollment gaps and new charter campus openings. For
example, I link school year 2015-16 charter campus openings with enrollment gaps calculated from
school year 2014-15 enrollment figures and 2014 5-year ACS data. When linking these enrollment
gap measures to charter campus openings, I exclude all gaps associated with campuses – both
charter and DCPS – that closed in the prior school year.

191

During my study period, I observe 11 new DCPS campus openings. Because DCPS plans
and announces its new campuses many months in advance, I associate enrollment levels at opening
DCPS campuses with charter campus openings in the same school year. For the 10 new DCPS
campuses that opened prior to school year 2020-21, I use actual enrollment audit data from each
new campus’ first school year. For the new DCPS campus opening in school year 2020-21 – Stevens
Early Learning – I rely on enrollment projections published by DCPS (DCPS, n.d.b).
3.5.2. The Quality of Nearby Non-Adult Charter and DCPS Campuses
As a proxy for school quality, I use nearby campuses’ math proficiency rates on standardized
assessments. Between school years 2005-06 and 2013-14, the District administered an annual
standardized assessment for charter and DCPS students in third through tenth grades called the DC
Comprehensive Assessment System (DC CAS). Since school year 2014-15, the District has instead
administered the state consortium-designed Partnership for Assessment of Readiness for College
and Careers (PARCC) for similar grades.
Proficiency rates on DC CAS and PARCC assessments for all school years since 2007-08 are
made available by OSSE (OSSE, n.d.c). Proficiency rates are presented by subject (i.e., math and
reading) and campus. OSSE’s reports sometimes break out separate proficiency rates for traditional
schooling grade ranges at a single campus (e.g., for Columbia Heights Education Campus’ middle
and high school grades separately). In such cases, I blend the intra-campus figures via an enrollment-
weighted average, so that each of my campuses has a single math proficiency rate for each school
year.
Because the District switched from the DC CAS to PARCC assessment halfway through my
study period, I convert each campus’ math proficiency rates to percentile rankings. For each school
year, I calculate three types of percentile rankings: (1) for each charter campus compared to only
other charter campuses; (2) for each DCPS campus compared to only other DCPS campuses ; and
192
(3) for each charter campus and DCPS campus compared to all other charter and all other DCPS
campuses simultaneously.
For each census tract and school year, I then blend these three types of percentile rankings
via an enrollment-weighted average, so that each census tract has three measures of nearby school
quality: (1) the quality of nearby charter campuses compared to all District charter campuses; (2) the
quality of nearby DCPS campuses compared to all DCPS campuses; and (3) the quality of nearby
charter and DCPS campuses compared to all charter and DCPS campuses.
Unlike my neighborhood measures and nearby school enrollment measures, I assume a two-
year lag between school quality measures and new charter campus openings. This is because
standardized assessment results from the prior school year do not become available until the current
school year has started, whereas enrollment counts from the prior year become available in the same
year. When linking these lagged school quality measures with charter campus openings, I exclude all
measures associated with campuses – both charter and DCPS – that closed in the prior school year.
3.6. Measuring a Neighborhood’s Proximity to LEA-specific Campuses
3.6.1. Proximity to Other Campuses within the Same LEA
I use GIS to measure the distances between: (a) the centroid of each census tract, and (b) all
other campuses operated by the same charter LEA. I calculate these distances in the school year that
a new campus opens. I then develop three measures for each tract’s distance to existing campuses:
1. average distance from the tract’s centroid to existing campuses with students in similar
grades as the new campus, again based upon traditional school grade spans;
2. average distance from the tract’s centroid to existing campuses with students in dissimilar
grades; and
3. the average distance from the tract’s centroid to all existing campuses.

193

3.6.2 Proximity to Previous Iteration for Relocating Campuses
I isolate all instances where a charter LEA closed a campus location i n one school year and
moved the campus to a new location the following year. The data reveal that these “relocations” are
almost exclusively associated with single-campus charter LEAs. For each such case, I use GIS to
measure the distance between: (a) the centroid of each census tract, and (b) the prior year location of
the newly-relocated campus.
4. Descriptive Analyses

Within this sub-section, I summarize where new charter campuses opened in the District,
and then proceed to depicting relationships between these locations and my hypothesized
explanatory factors.
Below, Table 15 provides summary statistics for a selection of tract measures. It does so
across all 174 census tracts and across all years in my study period. Most of these measures are
hypothesized explanatory factors, but several are not and are included for informational purposes
only. I note that Table 15 cannot provide summary statistics for LEA-specific or campus-specific
measures, such as the average distance of tracts to a LEA’s existing campuses.
Unfortunately, it is not highly informative to generate statistics across all tracts and school
years, or even across (a) tracts selected by campuses, versus (b) tracts unselected by campuses. This
is so for two reasons. First, while a particular tract j may not be chosen by campus i in school year t
of my study period, it may be chosen by a campus other than i in school year t+1 (also within my
study period). Second, the baseline values of my explanatory factors vary from one school year to
another (e.g., the extent of gentrification since 2000 shifts between year t and year t+1). To address
these issues, Table 16 lists average values for the same tract measures as Table 15, but it does so for
selected tracts versus unselected tracts in school year 2012-13 only. As with Table 15, Table 16
cannot provide summary statistics for LEA-specific or campus-specific measures.
194
Table 15. Summary statistics for certain tract measures, across all census tracts and for all school
years between 2010-11 and 2020-21
Tract measure Average Minimum Maximum
Standard
deviation
Schoolchildren per square mile 2,852 0 14,267 2,202
Median gross rent (2018 USD) $1,369 $305 $3,650 $493
Indexed gentrification since 2000 0.28 -0.54 5.24 0.55
White alone, non-Hispanic share of
population
32% 0% 93% 30%
White alone, non-Hispanic share of
schoolchildren
24% 0% 100% 29%
Share of adults with at least a bachelor’s 49% 1% 96% 29%
Median household income (2018 USD) $78,428 $12,982 $250,000 $43,642
Rail station within 0.5 miles* 45% n/a n/a n/a
Vacant school building in tract* 2.2% n/a n/a n/a
Vacant non-school building in tract,
formerly occupied by other charter LEA*
1.8% n/a n/a n/a
Enrollment gap for all children (0-17 years) -4,043 -9,443 1,342 2,387
Blended math percentile of nearby charter
campuses
41
st
percentile
0
th
percentile
100
th

percentile
26%
Blended math percentile of nearby DCPS
campuses
50
th

percentile
0
th

percentile
100
th

percentile
23%
Note: These are indicator variables and so the average value represents share of census tracts with value of “1” over all
tracts and school years.
Building on Table 16, the remainder of section 4 graphically depicts relationships between
certain explanatory factors and new campus locations in school year 2012-13. I note that the
following descriptive figures are limited in their power and the insights they provide. They cannot
capture the multi-dimensional decision-making process that I hypothesize. Nonetheless, they help
present variations in campus locations, variations in explanatory factor values, and covariations
195
between the two in an intuitive way. For reference, I provide a map of District landmarks and
neighborhoods in Appendix E.
Table 16. Summary statistics for certain tract measures, selected versus unselected tracts, school
year 2012-13
Tract measure
2012-13 Average
for Selected
Tracts
2012-13 Average
for Unselected
Tracts
Number of observations 9 tracts 165 tracts
Schoolchildren per square mile 2,166 2,697
Median gross rent (2018 USD) $1,329 $1,256
Indexed gentrification since 2000 0.73 0.18
White alone, non-Hispanic share of population 27% 31%
White alone, non-Hispanic share of schoolchildren 19% 22%
Share of adults with at least a bachelor’s 45% 46%
Median household income (2018 USD) $75,011 $73,306
Rail station within 0.5 miles* 67% 44%
Vacant school building in tract* 67% 0%
Vacant non-school building in tract, formerly
occupied by other charter LEA*
44% 2%
Enrollment gap for all children (0-17 years) -4,604 -3,476
Blended math percentile of nearby charter campuses 49
th
percentile 40
th
percentile
Blended math percentile of nearby DCPS campuses 45
th
percentile 52
nd
percentile
Note: These are indicator variables and so the average value represents share of census tracts with value of “1” over all
tracts and school years.
4.1. Variations in New Charter Campus Locations
Figures 19a and 19b depict the locations of charter campuses opening between school years
2010-11 and 2020-21. Figure 19a plots these locations relative to where all charter campuses

196

operated within the study period, and Figure 19b plots these locations relative to where all DCPS
campuses operated within the study period.
Per Figure 19a, all charter campuses that opened during my study period did so in census
tracts located east of Rock Creek. Moreover, as Figure 19a demonstrates, this pattern in locational
choices actually mirrors the distribution of all charter campuses that operated during my study
period. All charter campuses that opened prior to school year 2010-11 and continued to operate
during my study period also selected census tracts east of Rock Creek.
In contrast, Figure 19b reveals that the locational pattern of new charter campuses does not
mirror the distribution of DCPS campuses operating during my study period. A number of DCPS
campuses operated west of Rock Creek between school years 2010-11 and 2020-21.
Subsequent Figures in this section imply potential mechanisms driving this distributional
discrepancy. Census tracts west of Rock Creek are not highly dense with children (Figure 20), have
not significantly gentrified since 2000 (Figure 21), are relatively expensive to occupy in terms of
gross rent levels (Figure 22), have not housed any vacant school buildings or non-school properties
recently vacated by other charter LEAs (Figure 23), and lie closest to almost all of the highest-quality
DCPS campuses (Figure 24b). In short, they comprise the wealthiest and whitest region of the
District, while tracts east of the Anacostia comprise the most impoverished and least white region of
the District (Asch & Musgrove, 2017; Burner v. Washington).

197

Figure 19a. New charter campus locations and all charter campus locations, school year 2010-11 to
2020-21

Sources: See sub-section 3.5.

Figure 19b. New charter campus locations and all DCPS campus locations, school year 2010-11 to
2020-21

Sources: See sub-section 3.5.
198
4.2. Variations in Tracts’ Non-Schooling Characteristics and New Campus Locations
In order to sensibly depict covariations between new campus locations and census tract
characteristics, Figures 20, 21, and 22 isolate observed openings and tract -level measurements from
school year 2012-13.
Figure 20. New charter campus locations and density of children per tract, school year 2012-13
Source: 2011 5-year ACS Table B01001.
As shown in Figure 20, new charter campus locations appear positively correlated with
tracts’ density of children levels. Figure 21 captures a similar prospective relationship between new
campus locations and tracts’ extent of gentrification since 2000. The highly-gentrified areas plotted
in Figure 21 are consistent with those identified in other studies of District demography (e.g.,
Richardson, Mitchell, & Franco, 2019). In contrast to Figures 20 and 21, Figure 22 does not show
any clear relationship between new campus locations and tracts’ median gross rent levels.
199
Figure 21. New charter campus locations and gentrification since 2000 per tract, school year 2012-
13
Source: See sub-section 3.3.2.
Figure 22. New charter campus locations and median gross rent per tract, school year 2012-13
Source: 2011 5-year ACS Table B25064.

200

4.3. Variations in Adequate Facility Availability and New Campus Locations
Of all the Figures presented in this section 4, Figure 23 presents the most compelling case
for an explanatory factor driving new charter campus locations. Specifically, Figure 23 plots the
availability of vacant school buildings, along with non-school properties recently vacated by charter
LEAs. The spatial locations of new charter campuses in school year 2012-13 correspond almost
perfectly to the distribution of available facilities.
In fact, of the 87 charter campus openings I observe in this study, 41 involve a charter
campus occupying a vacant or partially vacant school building (i.e., 47% of all openings), and another
19 involve a charter campus occupying a non-school building that was previously occupied by
another charter LEA (i.e., 22% of all openings). Overall, 69% of all new charter campuses in my
study period opened in a facility previously occupied by an educational institution.
Of the 41 instances where a new charter campus moved into a vacant or partially vacant
school building, 32 were cases where a campus moved into a vacant DCPS building, 2 were cases
where a campus moved into a school building constructed by a District charter LEA, and 7 were
cases where a campus moved into a vacant private school or postsecondary institution building. The
closure of DCPS campuses is actually the primary mechanism by which charters accessed a facility
during my study period, creating the requisite space for 37% of all charter campus openings.
Of the 19 instances where a new charter campus moved into a non-school building
previously occupied by another charter LEA, eight were cases where the non-school space became
available due to either: (a) the revocation of other LEA’s charter, or (b) the other LEA’s
relinquishment of the campus due to shrinking enrollment.

201
Figure 23. New charter campus locations and facilities available per tract, school year 2012-13
Sources: See sub-section 3.4.
4.4. Variations in Public School Quality Near Tracts and New Campus Locations
Figures 24a and 24b superimpose new campus locations over the blended math percentiles
of public school campuses close to each tract. To reiterate, I use these blended math percentiles as a
proxy for school quality (see sub-section 3.5.2). The two Figures shade tracts based on the math
percentiles of nearby charter campuses and nearby DCPS campuses, respectively.
Figure 24a does not evince any obvious association between new charter campuses locations
and the quality of nearby, existing charter campuses. Conversely, Figure 24b highlights a clear spatial
demarcation in the school quality of nearby DCPS campuses for census tracts west of Rock Creek
and census tracts east of Rock Creek. As no charter campuses opened in a census tract west of Rock
Creek, Figure 24b implies a strong and negative correlation between new charter campus locations
and the quality of nearby DCPS campuses.
202
Figure 24a. New charter campus locations and tracts’ nearby charter quality, school year 2012-13
Source: See sub-section 3.5.2.
Figure 24b. New charter campus locations and tracts’ nearby DCPS quality, school year 2012-13
Source: See sub-section 3.5.2.

203

5. Locational Regression Models

The descriptive statistics I present in section 4 are interesting and potentially illuminating.
Nevertheless, they cannot estimate the relationship between each factor and new campus locations
while controlling for all other factors. As a result, I develop a r egression framework for running a
number of locational models.
5.1. The Conditional Logistic Functional Form
To do so, I employ McFadden’s (1973) conditional logistic regression framework. Per
Hoffman and Duncan (1988), the conditional logistic form is we ll-suited for modeling choice when:
1. choices are across discrete alternatives;
2. choices are related to attributes of the alternatives rather than attributes of the choosers; and
3. irrelevant alternatives are independent of each other, i.e., removing an unselected alternative
from the choice set does not significantly alter the odds ratios between a selected alternative
and the remaining unselected alternatives for a given chooser.

My methodology feasibly meets each of these criteria: I am modeling charter LEAs’
locational choices across discrete census tracts; my explanatory factors concern the characteristics of
census tracts rather than characteristics of the charter LEAs; and unchosen tracts should be largely
independent of one another. Still, to confirm my models fulfill the third criterion, I conduct a
number of tests post-estimation (see sub-section 5.3).
To briefly present the conditional logistic functional form, assume a simple case, where I
individuals are choosing between two alternatives. Each individual picks an alternative based on the
value of a particular characteristic x, which varies across the two alternatives. Then the probability
that individual i chooses alternative j (of the two alternatives) can be expressed as:
𝑝𝑝 � 𝑌𝑌 𝑖𝑖 𝑗𝑗 = 1| 𝑥𝑥 , 𝑌𝑌 𝑖𝑖 1
+ 𝑌𝑌 𝑖𝑖 2
= 1 � =
𝑒𝑒 𝛽𝛽 1
𝑥𝑥 𝑖𝑖𝑖𝑖
𝑒𝑒 𝛽𝛽 1
𝑥𝑥 𝑖𝑖 1 + 𝑒𝑒 𝛽𝛽 1
𝑥𝑥 𝑖𝑖 2
Eq. 1
204
As with unconditional logistic probability functions, Equation 1 can be rearranged to express
the conditional logistic regression as a linear combination of explanatory factors. In this simple case,
the rearrangement simply yields:
𝑙𝑙 𝑙𝑙 𝑙𝑙 𝑖𝑖𝑡𝑡 [ 𝑝𝑝 ( 𝑌𝑌 𝑖𝑖 1
= 1, 𝑌𝑌 𝑖𝑖 2
= 0| 𝑥𝑥 , 𝑌𝑌 𝑖𝑖 1
+ 𝑌𝑌 𝑖𝑖 2
= 1)] = 𝛽𝛽 1
𝑥𝑥 𝑖𝑖 𝑗𝑗 Eq. 2
The conditional logistic probability function (Equation 1) appears quite similar to that of a
multinomial logistic regression. And in fact, the two forms even share the same log likelihood
function (Hoffman & Duncan, 1988). However, they differ in a crucial way as well.
Assume that there are I individuals, each with a single characteristic z, which varies in value
for each individual i. Also assume that there are J alternatives, from which each individual i can make
a single selection, and which have a single characteristic x that varies in value for each alternative j.
The conditional logistic form presented above assumes that values of characteristic x,
varying across J alternatives, drive the choice of alternative j made by individual i. Accordingly, a
conditional logistic regression would estimate a single coefficient for the characteristic x (e.g., 𝛽𝛽 1
),
which is applicable to all J alternatives. A multinomial logistic form, however, assumes that
variations in characteristic z, varying across I individuals, drive the choice of alternative j made by
individual i. As a result, a multinomial logistic regression would estimate J – 1 coefficients to capture
how the value of characteristic z influences the probability of each alternative being selected. These
coefficients apply to all I individuals.
Based on my research question and hypotheses, I construct the below conditional logistic
framework. To help test my hypotheses, I specify different sets of explanatory factors across
different locational models. These models also vary in their incorporation of aggregate versus grade-
specific measures of resident children and nearby schooling options. For more information on this
distinction, please see sub-sections 3.3.2, 3.5.1, and 3.5.2.

205

𝑙𝑙 𝑙𝑙 𝑙𝑙 𝑖𝑖𝑡𝑡 � 𝑝𝑝 � 𝑌𝑌 𝑖𝑖 𝑗𝑗 = 1, 𝑌𝑌 𝑖𝑖𝑖𝑖
= 0 ∀ 𝑘𝑘 ≠ 𝑗𝑗 | 𝒙𝒙 , 𝒘𝒘 , 𝒛𝒛 , 𝒗𝒗, ∑ 𝑌𝑌 𝑖𝑖 𝑗𝑗 𝐽𝐽 𝑗𝑗 = 1
= 1 � � = 𝒙𝒙 𝒊𝒊𝒊𝒊 𝜷𝜷 + 𝒘𝒘 𝒊𝒊 𝜸𝜸 + 𝒛𝒛 𝒊𝒊𝒊𝒊 𝜹𝜹 + 𝒗𝒗 𝒊𝒊𝒊𝒊 𝝆𝝆

where:
• Y ij is an indicator variable for whether new charter campus i opens in census tract j;
• x ij is a set of explanatory factors capturing characteristics of census tract j that are
relevant to charter campus i (e.g., when measuring the density of only children whose
ages approximately correspond to the grade levels of the new charter campus);
• w j is a set of explanatory factors capturing the availability of adequate facilities in census
tract j (subscript i is absent as availability does not change across new charter campuses);
• z ij is a set of explanatory factors capturing public schooling opportunities close to census
tract j that are relevant to charter campus i (e.g., when measuring the blended math
percentile of nearby schools based on only grade levels that approximately correspond to
those of the new charter campus);
• v ij is a set of explanatory factors capturing: (1) the distance from census tract j to other,
existing campuses operated by the same charter LEA opening campus i, and (2) if the
opening of campus i actually represents the relocation of a prior campus, the distance
from census tract j to the prior iteration of the campus.

Conducting conditional logistic regressions requires the pairing of chosen alternatives with
relevant, unchosen alternatives. In the context of my analysis, that means I pair: (a) the observation
of a particular campus i opening in a certain tract j, with (b) observations of that same campus i not
opening in tracts other than j. Furthermore, because campus i opened in school year t, the values for
unchosen alternatives should be associated with the same school year t.
From a computational perspective, it is impossible to pair each chosen tract j with all 173
tracts not chosen by campus i in school year t. Therefore, for each tract j chosen in school year t by
campus i, I use STATA’s pseudo-random number generator to randomly select 9 tracts not chosen
by campus i., also in school year t. Per Haab and McConnell (2002) and Parsons and Kealy (1992),
having a “stratum” (i.e., chosen alternative matched with unchosen alternatives) of size 10 is
sufficient for the consistent estimation of conditional logistic coefficients. I use STATA’s “seed”
206
function so that each regression model contains the same matched sets of observations. Hence,
coefficients, odds ratios, and levels of significance are comparable across models.
5.2. Locational Model Results
Within Table 17, I present the results of ten different locational Models. Each Model
employs the regression framework outlined in sub-section 5.1. The estimated coefficients from these
Models, which Table 17 lists as odds ratios, directly inform the findings I delineate below, and are
stable in magnitude and significance. Moreover, Models 3 – 10 indicate high levels of joint
significance for my specified factors as well as good levels of overall fit. Below, I group my findings
by type of explanatory factor, similar to how I organize Table 17.
5.2.1. Findings for Non -Schooling Characteristics of Tracts
My models yield minor evidence that locations of new charter campuses are significantly and
positively associated with the density of children across census tracts, but only when that density is
measured in grades specific to new charter campus (see Models 7 and 8). Yet they provide no
evidence of significant relationship between locations of new campuses and tracts’ degrees of
gentrification since 2000 or tracts’ median gross rent levels.
I also find limited evidence that locations of new charter campuses are significantly and
positive correlated with tracts’ proximity to rail transit stations (see Models 4 and 6). But the
explanatory power of proximity to rail becomes insignificant once a locational model accounts for
nearby charter campus quality and nearby DCPS campus quality separately (see Models 5 and 7).
5.2.2. Findings for the Availability of Adequate Facilities
The estimated coefficients for facility indicators are consistently significant and positive
when included in my models (see Models 3 – 10), and including these indicators significantly
improves model fit and joint significance. The odds ratio for vacant school buildings is over twice as
large as that for non-school buildings previously occupied by other charter LEAs. Regardless, both
207
results indicate that new campuses are significantly more likely to locate in tracts with either such
facility available.
5.2.3. Findings for Nearby Public Schooling Opportunities
Additionally, my models generate significant and positive coefficients for my enrollment gap
measure (see Models 2 – 10). New charter campuses are significantly more likely to locate in tracts
with a high number of resident children relative to students enrolled at nearby charter or DCPS
campuses. This is the case when calculating the gap using all children in a given tract and all students
enrolled in nearby schools, as well as when calculating the gap via age and grade ranges specific to
each opening charter campus (see Models 2 – 5 and Models 6 – 10).
Concerning school quality, my models produce strong evidence that a higher blended math
percentile for nearby charter campuses significantly raises the likelihood that a new charter campus
opens in a given tract (see Models 5, 7, and 8). They produce similar, albeit more moderate, evidence
that a higher blended math percentile for DCPS campuses close to a census tract significantly
diminishes that likelihood (see Models 7 and 8).
Based on the significant and opposing relationships between new charter campus locations
and the quality of nearby charter campuses versus DCPS campuses, it is unsurprising that my
Models generate minimal evidence of a significant coefficient for the quality of all nearby public
campuses combined (see Models 4 and 6). I also observe consistent insignificance for the estimated
coefficients of grade-specific blends of math percentiles, regardless of whether these blends are for
all public school campuses, charter campuses only, or DCPS campuses only (see Models 9 and 10).
5.2.4. Findings for Existing Campuses and Relocation Considerations
I find consistent evidence for the significance of a tract’s distance to existing campuses
under the same LEA and, if the new campus opening actually represents a relocation, the tract’s
distance to the campus’ former location (see Models 1 – 10). More specifically, the estimated odd
208
ratios indicate that new campuses are significantly more likely to open in tracts close to their charter
LEA’s existing campuses, as well as tracts close to their prior sites if relocating. Model 8 estimates
no significant difference in the coefficient for a tract’s distance to existing campuses educating
similar grades versus dissimilar grades.
5.2.5. Marginal Effect Sizes for Explanatory Factors
As a final note, the odds ratios I present communicate critical aspects of the relationships
between my explanatory factors and the likelihood of a charter campus opening in a particular tract.
However, it can also be quite useful to refer to these factors’ marginal effects, i.e., the extent to
which changes in tract characteristics change the likelihoods of charter campuses opening within
them. With this in mind, Table 18 lists marginal effect sizes for Model 7’s explanatory factors.
Table 17. Conditional logistic regression results (dependent variable: new charter campus opening in given census tract), school year 2010-
11 to 2020-21
Notes: + for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001 Observations: 870
All coefficients reported as odds ratios Zero outcomes: 783
Indicators for blended math percentile of all/DCPS campuses (non-grade-specific) are unnecessary Nonzero outcomes: 87
(1) (2) (3) (4) (5)
Non-schooling characteristics of census tracts
Density of all children (0-17 years) 0.9999 0.9999 1.0001 1.0001 1.0001
… Density of children in grades specific to new charter campus
Index of gentrification since 2000 0.8548 0.9478 0.9441 0.8785 0.9379
Median gross rent 0.9993* 0.9996 1.0002 1.0006 1.0007
Rail transit station within 0.5 miles of tract 1.3078 1.4671 1.7869 1.9286+ 1.6257
Availability of adequate facilities
Vacant school building in tract 51.2734*** 53.9760*** 54.2548***
Vacant non-school building in tract, formerly occupied by other charter LEA 23.8523*** 21.9122*** 19.9939***
(6) (7) (8) (9) (10)
Non-schooling characteristics of census tracts
Density of all children (0-17 years)
… Density of children in grades specific to new charter campus 1.0003 1.0004+ 1.0004* 1.0003 1.0003
Index of gentrification since 2000 0.7705 0.8334 0.7579 0.8223 0.8044
Median gross rent 1.0007 1.0007 1.0008 1.0004 1.0005
Rail transit station within 0.5 miles of tract 1.9483+ 1.6178 1.7632 1.8607 1.7493
Availability of adequate facilities
Vacant school building in tract 59.7039*** 59.9856*** 59.1267*** 57.9915*** 57.4906***
Vacant non-school building in tract, formerly occupied by other charter LEA 23.2616*** 22.0635*** 21.3666*** 27.0154*** 25.5845***
209
Table 17 cont’d. Conditional logistic regression results (dependent variable: new charter campus opening in given census tract), school
year 2010-11 to 2020-21
(1) (2) (3) (4) (5)
Nearby public schooling opportunities
Enrollment gap for all children (0-17 years) 1.0001+ 1.0003*** 1.0003*** 1.0004***
… Enrollment gap for children in grades specific to new campus
Blended math percentile of all nearby public campuses 0.1687+ 0.2023
… Availability of grade-specific blended math percentile of all campuses
… Grade-specific blended math percentile of all campuses
Availability of blended math percentile of nearby charter campuses 2.7831
Blended math percentile of nearby charter campuses 5.5059+
… Availability of grade-specific blended math percentile of charters
… Grade-specific blended math percentile of charters
Blended math percentile of nearby DCPS campuses 0.2034
… Availability of grade-specific blended math percentile of DCPS campuses
… Grade-specific blended math percentile of DCPS campuses
Existing campus and relocation considerations
Average distance from tract to existing campuses under same charter LEA 0.7817** 0.7583** 0.6632** 0.6672** 0.6492**
… Average distance to existing campuses teaching similar grades
… Average distance to existing campuses teaching dissimilar grades
If new charter campus is relocation, average distance from tract to previous
location
0.5882*** 0.5831*** 0.6614** 0.6704** 0.6667**
Converged Log Likelihood -166.854 -163.422 -86.498 -85.623 -79.663
Likelihood Ratio chi2 66.94 73.80 227.65 229.40 241.32
Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000
Pseudo R-squared 0.1671 0.1842 0.5682 0.5726 0.6023
210
Table 17 cont’d. Conditional logistic regression results (dependent variable: new charter campus opening in given census tract), school
year 2010-11 to 2020-21
(6) (7) (8) (9) (10)
Nearby public schooling opportunities
Enrollment gap for all children (0-17 years)
… Enrollment gap for children in grades specific to new campus 1.0008*** 1.0010*** 1.0010*** 1.0009*** 1.0011***
Blended math percentile of all nearby public campuses 0.1686
… Availability of grade-specific blended math percentile of all campuses 4.7583
… Grade-specific blended math percentile of all campuses 0.4751
Availability of blended math percentile of nearby charter campuses 1.8503 2.1326
Blended math percentile of nearby charter campuses 6.9544* 5.5325+
… Availability of grade-specific blended math percentile of charters 2.5987
… Grade-specific blended math percentile of charters 2.4177
Blended math percentile of nearby DCPS campuses 0.1320+ 0.1186+
… Availability of grade-specific blended math percentile of DCPS campuses 2.5896
… Grade-specific blended math percentile of DCPS campuses 0.3976
Existing campus and relocation considerations
Average distance from tract to existing campuses under same charter LEA 0.6621** 0.6613** 0.6601** 0.6570**
… Average distance to existing campuses teaching similar grades 0.7171*
… Average distance to existing campuses teaching dissimilar grades 0.5106+
If new charter campus is relocation, average distance from tract to previous
location
0.6164** 0.5938*** 0.6026** 0.6087*** 0.5820***
Converged Log Likelihood -84.356 -77.899 -77.019 -84.516 -80.434
Likelihood Ratio chi2 231.94 244.85 246.61 231.62 239.78
Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000
Pseudo R-squared 0.5789 0.6111 0.6155 0.5781 0.5985
211
Table 18. Marginal effects of Regression Model 7 explanatory factors
Explanatory Factor
Measurement
Scale
Mean Value
Across Sample
of Census
Tracts
Change in Probability
of Campus Opening
in Tract (i.e., Marginal
Effect Size)
Approximate Change in
Likelihood at Means
Density of children in grades specific to
new charter campus
Children / square
mile
823 children
per square mile
0.01%
0.6% for 100 more children
per square mile
Index of gentrification since 2000
Continuous
indexed measure
0.27 -2.86%
-0.3% for increase in
gentrification index of 0.10
Median gross rent U.S. dollars $1,318 0.01%
1.1% for $100 more
in median gross rent
Rail transit station within 0.5 miles of tract Indicator variable 45.5% of tracts 7.55% 7.55%
Vacant school building in tract Indicator variable 7.2% of tracts 64.25% 64.25%
Vacant non-school building in tract,
formerly occupied by other charter LEA
Indicator variable 4.5% of tracts 48.56% 48.56%
Enrollment gap for children in grades
specific to new campus
Children in tract
minus children
enrolled nearby
-1,533 children 0.02%
1.5% for 100 more children
in tract versus enrolled
nearby
Availability of blended math percentile of
nearby charter campuses
Indicator variable 86.2% of tracts 9.66% 9.66%
Blended math percentile of nearby charter
campuses
Percentile ranking 43
rd
percentile 30.44%
3.0% for 10 percentile point
increase
Blended math percentile of nearby DCPS
campuses
Percentile ranking 49
th
percentile -31.77%
-3.2% for 10 percentile point
increase
Average distance from tract to existing
campuses under same charter LEA
Decimal degrees
(in hundredths)
2.28 decimal
degrees
(in hundredths)
-6.49%
-6.5% for 1 one-hundredth
decimal degree increase
If new charter campus is relocation, average
distance from tract to previous location
Decimal degrees
(in hundredths)
2.32 decimal
degrees
(in hundredths)
-8.18%
-8.2% for 1 one-hundredth
decimal degree increase
Note: Marginal effects calculated at means. For indicator variables, marginal effect size at means is calculated as change in likel ihood when indicator changes from “0”
to “1”.
212

213

5.3. Independence of Irrelevant Alternatives
After running these models, I test the independence of irrelevant alternatives in two ways.
First, I conduct Hausman-McFadden tests that: (1) remove a single unselected (i.e., irrelevant)
alternative from my final dataset, (2) re-estimate a Model’s coefficients, and (3) compare that
Model’s estimated coefficients from the full dataset to the estimated coefficients from the limited
dataset (i.e., the dataset with a single alternative removed). These tests fail to reject the null
hypothesis that the irrelevant alternatives in my dataset are indeed independent.
I raise two caveats to the Hausman-McFadden test approach. First, computing power
restrictions force me to match each census tract j selected by new charter campus i in school year t
with 9 randomly selected census tracts also from school year t that campus i did not select. Because
these unselected (i.e., irrelevant) alternatives are randomly selected for each campus opening (i.e.,
stratum) i, every stratum contains a different set of irrelevant alternatives. Therefore, removing a
single irrelevant alternative from my dataset (i.e., a single census tract k in school year t) does not
actually remove that alternative from each stratum i, since not every stratum has been matched with
tract k through the random sampling process. It is not obvious how to resolve this complication.
Second, multiple methodological studies have concluded that the Hausman-McFadden test can be
biased and inconsistent in determinations of independence (e.g., Cheng & Long, 2007).
As an alternative to the Hausman-McFadden test, I also run my models using the nested
logistic regression form, which relaxes the assumption of independent irrelevant alternatives. If the
signs, estimated magnitudes, and levels of significance of the coefficients estimated via nested
logistic regression are similar to those estimated via conditional logistic regression, the irrelevant
alternatives in my conditional logistic models can be considered non -nested, i.e., independent.
I create a nested structure for my set of alternatives, grouping them by three geographic
regions: west of Rock Creek, between Rock Creek and the Anacostia River, and east of the
214
Anacostia River. The discrepancies across these three regions, captured in section 4’s Figures,
inform my nesting approach, as does the Henig and MacDonald (2002) paper.
The nested logistic framework assumes that campus-specific attributes drive campuses’
choice of geographic region, and that my alternative-specific factors drive campuses’ choice of
particular census tract afterward. I therefore generate two new variables that are specific not to each
alternative, but to each charter campus i: (1) whether it is considered a “comprehensive” charter
campus in terms of its type, and (2) whether it falls under a charter management organization
(CMO). I select these two campus-specific attributes based on Henig and MacDonald’s (2002) early
research on District charter campuses’ locational decisions. A discussion of charter typology is
located in Appendix D; a discussion of CMO status is located in Chapter 4.
As the nested logistic results are quite consistent with the conditional logistic results, I again
conclude that the irrelevant alternatives in my models are non-nested in nature and therefore
independent. For a comparison of Model 9’s results using the conditional logistic framework (see
Table 17) versus the nested logistic framework, please refer to Appendix R.
5.4. Other Explanatory Factors Considered
While not discussed in the body of this paper, I assess an array of alternative explanatory
factors for my locational models, each of which is deemed insignificant in a majority of them. For
one, I conduct a second PCA using only 2009 – 2018 ACS data for the four dimensions of
gentrification listed in sub-section 3.3.2. I then use that PCA’s first component to develop 3-year
and 5-year indexed measures of gentrification for each census tract. Testing these two measures
necessarily restricts my regression analysis to school years 2013-14 through 2020-21 and school years
2015-16 through 2020-21, respectively. Regardless, I find no significant difference in these 3-year
and 5-year indexed measures and my preferred indexed measure of gentrification since 2000.

215

In addition, rather than an indexed measure of gentrification, I consider the four dimensions
of gentrification individually. As with my preferred indexed measure, these four dimensions are not
statistically significant when included in my models.
Separately, rather than calculating changes in these four dimensions over time, I simply
observe their values in each school year. Given Hankins (2007) work, I also test including the shares
of schoolchildren considered white – a measure not captured within my gentrification index. Results
from specifications including the above dimensions suggest that each is an insignificant predictor of
new charter campus locations. The insignificance of tracts’ racial compositions, in particular, is
surprising and contradicts Henig and MacDonald’s (2002) and Hankins (2007) findings.
Furthermore, I test the inclusion of distance to the District’s central business district
(DDOT, n.d), as well the numbers of nearby charter or DCPS campuses and their total enrollment
levels. The insignificance of the distance to the central business district factor contradicts Henig and
MacDonald’s (2002) findings also.
Finally, I proxy nearby schools’ quality with reading proficiency percentiles instead of math
proficiency percentiles. I find no significant difference in using one measure over the other.
6. Discussion

This paper generates a number of noteworthy results, some of which respond to existing
research, and others of which represent novel contributions to the literature.
6.1. Support for and Extensions of Prior Findings
More than anything else, this paper documents the primacy of adequate facilities in charter
LEAs’ locational decisions. This is a finding consistent with both Henig and MacDonald (2002) and
the REACH Project (2019). Further, I find significant and positive explanatory power for two
distinct types of adequate facilities: (1) vacant school buildings, and (2) non -school buildings
formerly occupied by another charter LEA. Unquestionably, District charter LEAs have an
216
overriding preference to occupy space where another school once operated, confirming my
Hypothesis 3.
I also conclude that nearby public school quality drives new charter campus locations, but
not in the way I anticipate (see Hypothesis 5) or Henig and MacDonald (2002) report. Namely, the
quality of proximate public schooling is insignificant when considered in aggregate, because high-
quality charter campuses attract new charter campuses while high-quality DCPS campuses repel
them. This is my most intriguing result, and I posit two reasons for the positive correlation between
the location of high-quality charter campuses and new charter campuses.
For one, a legitimacy mechanism may be at play (Paino, Boylan, & Renzulli, 2017).
Communities with high-performing charter campuses may view the institution of charter schools as
more legitimate than communities with lower-performing charter campuses. In turn, charter LEAs
may site new campuses in areas where they anticipate feeling welcomed rather than scrutinized.
Two, charter LEAs may believe that the population of children living close to high-performing
charter campuses is responsible for their superior results. Hence, charter LEAs may locate their new
campuses nearby, hoping to draw from the same population and produce superior results as well.
Conversely, new charter campuses may systematically avoid areas with high-performing
DCPS campuses because they sense these neighborhoods are satisfied with their traditional public
schools. Notably, DCPS uses geographic attendance zones to determine which students can attend
which of its campuses by right. As a result, its schools’ sociodemographic compositions typically
mirror the those of their surrounding neighborhoods (Orfield & Ee, 2017).
High rent levels, when combined with this attendance zoning mechanism and residential
segregation, act as an especially effective barrier. They help ensure that wealthy neighborhoods have
unchallenged access to campuses with wealthy students, which also tend to be the highest-
performing ones (see section 4; see also Black, 1999). Put less politely: high rents prevent less wealthy

217

families from living in such neighborhoods and therefore prevent them from attending such
campuses by right.
It is highly probable that District charter operator s are aware of this dynamic and therefore
avoid the District’s wealthiest and most expensive neighborhoods, whose residents already have
guaranteed access to the vast majority of high-performing DCPS campuses.
6.2. Contradictions to Prior Findings
Unlike the availability of facilities and nearby school quality, this paper fails to generate
support for the non-schooling characteristics of neighborhoods identified in other studies (see sub-
section 2.1). I find no evidence that new charter campuses tend to locate in gentrifying
neighborhoods, and I find very limited evidence for a rail transit effect, disconfirming my
Hypotheses 1 and 2, respectively.
Regarding the insignificance of rail transit, the primacy of vacant school buildings may be to
blame: the locations of vacant school buildings are correlated with those of rail transit stations,
probably an intended outcome of city planning.
Meanwhile, the insignificance of gentrifying census tracts helps qualify earlier work of mine
(Eisenlohr, 2020), wherein I documented a significant association between District census tracts’
exposure to charter campuses and their levels of gentrification between 2000 and 2017. If this
association truly does exist, then this paper suggests the locations of charter campuses are causing
the gentrification of District neighborhoods, rather than responding to it.
6.3. Novel Contributions to the Literature
This paper also uncovers three previously-unarticulated mechanisms driving new charter
campus locations. First, charter LEAs prefer for new campuses to be located near their existing
campuses, even when such the new and existing campuses teach similar grades. The portion of this
finding related to similar campuses is surprising and contradicts part of my Hypothesis 6. Perhaps
218
clustering campuses is so beneficial from an administrative perspective that it overcomes the
disadvantage of multiple campuses competing for the same students. Whatever the case may be,
additional work is necessary to understand why clustering campuses is a dominant strategy.
Second, charter LEAs strongly prefer to relocate campuses over as short a distance as
possible, confirming my Hypothesis 7. Presumably, LEAs do not want to risk lowered enrollment by
drastically changing their students’ travel times post-relocation. Yet they may also minimize
relocation distances to avoid the costs of “re-marketing” themselves to a new community. As with
existing campuses, additional work is necessary to understand the precise logics undergirding this
preference.
Third, and as Hypothesis 4 predicts, charter LEAs strongly prefer to open new campuses in
neighborhoods with high enrollment gaps, where it should be easiest to maximize enrollment levels.
6.4. Study Limitations
Before concluding, I recognize that my findings are subject to the following limitations. First
and foremost, this paper examines new charter campus locations in a single jurisdiction with a
certain set of rules and regulations. It is probable that analyses in other cities and states unveil other
dynamics due to contextual differences.
Second, it is impossible for me to properly identify counterfactual cases when defining the
availability of both vacant school buildings and properties formerly occupied by a charter LEA. The
availability of and leasing arrangements for these properties is highly idiosyncratic, and I cannot
know with any certainty that: (a) a vacant school building or property was available for charter use,
and (b) a charter LEA opted not to use that space.
Third, my dataset of opening charter campuses does not contain enough observations to
construct nested logit models that account for the many variations in charter LEAs’ attributes, such
as “type” of charter school (A+ Colorado, 2018). Accounting for such factors would be quite

219

interesting given the qualitative evidence that a feedback loop exists between a charter LEA’s type
and the neighborhood or community it serves (Langhorne, 2019).
And finally, this paper does not control for the presence of District private schools. Future
research can add to the literature by incorporating these data in locational models, as well as by
exploring how households substitute between private schools, charter schools, and traditional public
schools within a given district. As aforementioned, this is an exhaustive endeavor, requiring
extensive data on factors like tuition levels over time, voucher programs, comparable and
standardized outcomes for students across school types, institutional identities, etc.
7. Conclusion

From a macroscopic view, what does this paper say about the legacy of charter reform in the
District? Has Congress’ laissez faire approach to the planning and oversight of District charter LEAs
been beneficial or deleterious? Have the locational decisions of charter LEAs augmented schooling
options for the city’s residents in an equitable way?
Perhaps unsurprisingly, this paper tells a familiar story in its response to these questions: one
of differentiated policy outcomes arising from entrenched and spatially-defined social inequalities.
As section 4 broadly sketches, the District is segregated across three regions, in terms of
both its physical and sociodemographic geography. Its western region, lying west of Rock Creek, is
easily its most affluent, and has by far the most residents considered white alone, non-Hispanic. The
District’s eastern portion, lying east of the Anacostia River, contains its most impoverished
households, and its residents are almost exclusively black American. The District’s middle band,
lying between Rock Creek and the Anacostia, is an area of relative diversity. It has neighborhoods of
wealthy and white households, wealthy and non-white households, and working class and non-white
households; some of these areas are demographically stable, while others are undergoing substantial
gentrification (see Figure 21).
220
Of the 87 new charter campuses that opened during my study period, 27 (i.e., 30%) opened
east of the Anacostia, 60 (i.e., 70%) opened in the District’s middle band, and no new charter
campuses opened west of Rock Creek. This distribution of charter campuses sharply contrasts with
the distribution of the District’s children. In my study period’s typical school year, 43% of a ll
students living east of Rock Creek were living east of the Anacostia.
To truly understand whether charter campuses’ locational decisions have supplemented
households’ options rather than replacing them, we must account for whether a new charter campus
remained open for the remainder of my study period, whether it occupied a vacant DCPS building
or a private elementary or secondary school building, and also whether it represented a brand new
charter campus or the relocation of an existing one.
Not including campuses that opened in the final year of my study period (i.e., school year
2020-21), of the 24 campuses that opened east of the Anacostia, only 5 of them were: (1) operating
by the end of my study period, (2) did not represent the relocation of an existing campus that was
operating prior to my study period, and (3) did not occupy a vacant DCPS building or private
elementary or secondary school building. That is, only 21% of the charter campuses that opened east
of the Anacostia represented a true net increase to these communities’ educational options. Not
including Early Learning campuses, since they naturally enroll fewer children, these figures decrease
to three out of 20 campuses ( i.e., 15%).
In contrast, and again not including campuses that opened in the final year of my study
period, of the 50 campuses that opened between Rock Creek and the Anacostia, 21 of them –
equivalent to 42% – fulfilled the same three criteria and therefore represent a true net increase to
these communities’ educational options. Without Early Learning campuses, the calculation yields 17
out of 43 campuses (i.e., 40%). Regardless, on a per-new-campus basis, the District’s middle band of

221

communities have received over twice as many new educational options as communities east of the
Anacostia. In gross terms, they have received over four times as many.
Taken in totality, then, this paper’s suggests that charter campuses’ locational behaviors,
intersected with the spatial inequalities of the District, have produced three distinc t meanings for
charter reform, one for each of the District’s regions. In the communities west of Rock Creek,
charter schools are an irrelevant intervention. Households residing in these neighborhoods have
guaranteed and exclusionary access to the highest-quality DCPS campuses, and because of their
performance levels, these campuses’ buildings will not become available for charter use in the
foreseeable future.
In the communities east of the Anacostia, new charter campuses are largely a backfilling
mechanism. They serve to reactivate educational spaces in neighborhoods whose public schooling
options have dwindled after successive waves of DCPS closures and a number of charter closures or
relocations too.
Finally, in the communities between Rock Creek and the Anacostia, the opening of new
charter campuses is more complicated. Much like those east of the Anacostia, many charter
campuses in the District’s middle band have moved into unused school buildings and properties
vacated by other charters. Yet critically, a sizeable portion of charter campuses in this region actually
do represent brand new locations of and options for public schooling.
This paper’s locational models hint that spatial variations in school performance levels are
driving these regionally-defined discrepancies. And if the locations of new charter campuses
continue to go unplanned, the gulf in new educational opportunities west and east of the Anacostia
is likely to deepen for the following reasons. First, the majority of high -performing charter campuses
are located between Rock Creek and the Anacostia. Second, the District’s charter authorizer actively
encourages high-performing charter LEAs to open new campuses (PCSB, 2019, pp. 7-8). Third, my
222
models demonstrate that charter LEAs prefer to locate new campuses close to their existing ones.
Fourth and finally, my models also reveal that charter LEAs prefer to locate new campuses close to
other, high-performing charter campuses.
Yet this path dependency and the inequitable access it engenders are not inexorable. To
elude them, though, requires a fundamental reframing of how jurisdictions should manage their
charter operators. Although the District publicly funds charter campus facilities, it does not publicly
control their locations. A shift to public control of locations is possible, however, if other policies
change in tandem.
First, the District could approve new charter LEAs and approve the expansions of existing
LEAs conditional upon new campuses opening in the most underserved communities. Second, and
to support the first policy change, the District’s Public Charter School Board or another District
agency could assume facilities underwriting responsibilities. That is, the District could proactively
search for private facilities in underserved communities, underwrite their leases, and then provide
those facilities to new charter LEAs or expanding LEAs. This underwriting would eliminate the
“chicken-and-egg” problem that frequently plagues charter operators in the District and elsewhere:
to secure a facilities lease requires a charter being granted, but charter applications typically ask
where an operator will locate beforehand (although again, authorizers do not punish operators for
ultimately opening a campus elsewhere; see REACH Project, 2019). It would also eliminate the need
for per-pupil facilities allotments.
It is easier to envision these policy changes in the District, which has a single authorizer and
already provides public facilities funding, than in other states and school districts. Nevertheless, it is
possible for these other jurisdictions to adopt similar rules, at least if sufficient political will exists.
In the end, it seems reasonable to encourage a high-performing charter operator to expand,
just as it seems reasonable to provide vacant school buildings for charter use. But policies like these

223

cannot be blank slates left for individual operators to act upon. If state and local governments truly
want charter reform to distribute educational opportunities in a more equitable way, it is clear they
must actively plan for that vision. They must proactively serve their constituents, rather than relying
on third parties to do so for them.

224
CHAPTER 6
Concluding Remarks
As long as a program is perceived as benefiting solely those with little power in society, it is
unlikely to prosper for long.
–James Forman Jr., “The Rise and Fall of School Vouchers”, 2006
Across this dissertation’s chapters, I have attempted to tell the story of charter reform in the
District of Columbia. The outcomes of this story are probably not new or surprising to readers.
Many articles and books point to the continuation of America’s social inequalities, ones often
defined by race, place, and income. What is new to this story, what I hope inspires and surprises
readers, is the way that these outcomes have materialized.
As I outline in Chapter 1, arguments for expanding school choice typically fall into one of
two camps. New choice mechanisms like charter schools will either boost academic outcomes – in
school of choice and/or traditional neighborhood schools – or they will make access to high-
performing schools more equitable. However, these two types of claims are fundamentally linked.
The notions of opportunity and quality undergird them both.
So far, American researchers and policymakers have expended most of their efforts
corroborating the claims concerning academic outcomes. Yet the empirical evidence on choice
mechanisms and academic outcomes is quite mixed, underscoring the need for greater scrutiny of
claims in the second camp of “equity”. In this regard, a range of studies identify heightened student
segregation across schools of choice, and they contemplate some prospective factors driving this
division. These factors almost exclusively relate to the behavior of individual households, such as
outgroup avoidance, schooling preferences, and social and informational networks. What such

225

studies fail to tell us is whether the designs of choice mechanisms themselves innately undermine
their ability to produce more equitable access to high-performing schools.
Here, then, is where I believe my dissertation provides true value. I identify three distinct
channels, specific to charter schools, through which charter reform has reinforced the District’s
stark social divide: (1) an association with gentrification ( see Chapter 3); (2) a regressive and
discriminatory school closure policy (see Chapter 4); and (3) an absence of planning for the locations
of new charter campuses (see Chapter 5).
The District’s policymakers and planners can directly influence at least two of these
channels. Public agencies can help drive the locations of new charter campuses by guiding the
facilities acquisition process. They can also redefine the criteria for charter closure so that it becomes
a last resort rather than an annual occurrence. Charter reform’s association with gentrification is
more complex. If gentrifying neighborhoods attract new charter campuses, then the earlier solution
to locational decisions applies nicely. If charters cause gentrification though, as my results
collectively imply (see Chapters 3 & 5), then coordinated housing policies become vital as well.
Nevertheless, even if the above strategies prove effective, they will not guarantee all
households equal access to high-performing schools. As long as race, place, and income are
intertwined, discrepancies will persist in who learns where. As this dissertation posits in its
introduction, the above strategies are necessary but not sufficient endeavors in the pursuit of equity.
This dissertation focuses on describing implications for equitable access, but the design of
District charter reform exhibits many other troubling components. As one example, over the entire
history of the District’s charter reform, only one charter operators’ teachers have successfully
unionized, at Mundo Verde Public Charter School in 2019 (Austermuhle, 2019). The District has
continually allowed individual charter operators to aggressively combat and punish employee efforts
at unionization (Associated Press, 2019; Austermuhle, 2017). As a result, m any charter teachers lack
226
stability in areas that teachers in DCPS and other traditional school districts take for granted. Rules
on sick leave and substitute teachers are delegated to individual charter operators (DC Scholars
Public Charter School, 2019; The Next Step Public Charter School, 2018), and almost all charter
teachers do not have access to union-affiliated legal representation.
Unsurprisingly, teachers at District charter schools also earn less than their DCPS
counterparts on average (Cohen, 2019). And although District officials centrally determine
compensation for DCPS leadership, they do not control executive salaries at District charter
operators. Some of these leaders receive more than $300,000 per year, earning more than the DCPS
chancellor despite overseeing the educations of far fewer students. Such high levels of compensation
are an especially visible diversion of public funding away from the teachers and students who need it
most (Cohen, 2019; DCHR, 2020). Taken collectively, the above points raise important questions
about the quality of teachers attracted to District charter schools versus DCPS.
As a separate and broader concern, it is plausible that as cities gentrify, the compositions of
their government agencies and elected officials do too. Who will emphasize the District’s poorest
and most marginalized communities, when only a few city leaders resemble those communities’
residents? This is an understudied question that merits serious consideration going forward.
In the end, what does this dissertation tell readers about the legacy of charter reform as an
instrument of equity? Unquestionably, the District’s version of charter reform and the District’s
historical and structural inequalities do not perfectly match the settings of other jurisdictions. Still,
my analyses raise a complicated and concerning set of implications that may be quite relevant to
policymakers and planners in other cities and states.
More than anything else, this dissertation shows that the design of District charter reform
has produced differing results for disparate communities. It has been an irrelevant development for
the vast majority of households west of Rock Creek, who inhabit the District’s whitest and
227
wealthiest neighborhoods, and who have by-right access to many of the District’s highest-
performing traditional public schools and financial access to t he most prestigious private schools.
In contrast, it has been mostly a backfilling mechanism for the majority of households east
of the Anacostia, who inhabit the District’s poorest and most heavily black American
neighborhoods. This region’s charter campuses have largely served as replacements for closed or
otherwise departed traditional public, charter, and private schools. It is fair to wonder how many
charter campuses would open east of the Anacostia if the District did not aggressively and
continuously close existing ones.
Compared to the above two groups, it is clear that households in the District’s middle band,
between Rock Creek and the Anacostia, have benefitted the most from charter reform. It is in these
neighborhoods, which have gentrified the most over the past 20 years, that charter campuses have
opened in a way that supplements other schooling options rather than replacing them. This region also
houses many of the District’s top-performing charter campuses, an expected outcome based on the
literature, and these campuses are by definition most accessible to nearby households. Perhaps
charter schools are increasingly popular in the District precisely because they have mostly benefited
gentrifying households between Rock Creek and the Anacostia.
This, however, brings up the higher order question: what presence will charter operators
have in the District in 20 years? What about 50 years? If the District continues to gentrify, if many
Washingtonians continue to emigrate to more affordable counties in Maryland and Virginia, if
historically marginalized communities continue to have lesser access to high -quality schools, will
households in the District still even want charter schools? Is the true story of charter reform merely
one of transition, in which reform helped economically revitalize the District, alter its
sociodemographic characteristics, and effectively repopulate its traditional public school system?

228

If that is the ending for the District’s charter reform, it will be a very tragic one; it will
represent an utter failure to meet the goals of policymakers and the dire needs of many District
residents. But we must remember that these are only potential outcomes. It is not too late. We can
still make the story of charter reform in the District – and in other places – one of more equitable
opportunities. To do so, though, we must truly want more equitable opportunities. We must
prioritize the needs of the forgotten and the invisible, and we must revise our rules and plans in their
interests above all others.

229

A Final Story of the District

I am invisible, understand, simply because people refuse to see me.
–Ralph Ellison, Invisible Man

In the summer of 2019, work began on a new mixed-use complex right by the Rhode Island
Avenue Metro stop, easily visible from the Red Line’s elevated tracks. In and of itself, the project
was unremarkable, one of the many new and large developments that have risen in the District over
the past two decades, now looming over the older townhomes nearby. Before the construction team
could build, though, they needed to demolish a large strip mall occupying a hill within the project
site.
The mall was home to a number of businesses, many of which tend to be thought of as
“working class”: a Forman Mills, a Rainbow, a Popeyes. The businesses had their “everything must
go” sales and the demolition began. As it progressed, a single word of graffiti bloomed on an
exposed cinderblock wall facing the Metro stop. “GENTRIFY” it said, in all caps at least 20 feet tall,
easily legible to the thousands of Red Line passengers riding past it every day.
The message was an encapsulation of neighborhood change to-date, and it was also a
harbinger of greater change to come. Within two months of its appearance, it was already gone,
obscured by a layer of new gray paint (presumably applied by the development company).
In the near future, residents will move in to the new mixed-use complex by Rhode Island
Avenue, where the strip mall once stood. Many of these residents will be white. Many of them will
earn six-figure salaries. And many of them will have graduate degrees. In short, many of the new
residents will look nothing like those who continue to live in the townhomes nearby. It is also likely
230
the broader neighborhood will shift in response, with new businesses catering to the inmoving
demographic and further redevelopment in a feedback loop of “revitalization ”.
It is debatable whether these changes will be a net positive or negative for the District and its
people. Yet that ambiguity does not excuse the physical and symbolic whitewashing of those who
may have lost something from this change by those who likely gained. There is a “cultural genocide”
that lies behind that word of graffiti and its quick erasure (Crockett, 2020), one legitimized by the
infallibility of market-based change. It is a pervasive experience for longtime Washingtonians, who
have watched their city morph into something unrecognizable over the past two decades, in ways
that they cannot control.
With almost every law or policy enacted, there are winners and losers, sufferers and gainers.
When we make changes in the name of equity, we cannot avert our gaze from those in pain, especially
when our solutions seem to make things worse. The biggest issue with District charter reform is not
that charter schools exist. Rather, it is that the public institutions who legalized and oversee charter
schools have turned a blind eye to those who may not benefit from them. By doing so, they have
enabled the inequalities that existed prior to the School Reform Act – severe residential segregation,
gaps in employment and household wealth, acute divisions in public and private school access, all
squarely defined by race and income – to reproduce themselves in the arena of charter reform.
In their behavior, these public institutions are no different from the developers of that new
mixed-use complex off Rhode Island Avenue. But instead of affecting some city blocks, they are
shaping the educations and lives of innumerable children.

231

REFERENCES

A+ Colorado. (2018). Unequal choices: School model diversity in Denver Public Schools. Denver, CO.
http://apluscolorado.org/reports/unequal-choices-school-model-diversity-in-denver-public-
schools/
Abdulkadiroğlu, A., Angrist, J. D., Dynarski, S. M., Kane, T. J., & Pathak, P. A. (2011).
Accountability and flexibility in public schools: Evidence from Boston's charters and pilots.
The Quarterly Journal of Economics, 126(2), 699-748.
Abdulkadiroğlu, A., Angrist, J., & Pathak, P. (2014). The elite illusion: Achievement effects at
Boston and New York exam schools. Econometrica, 82(1), 137-196.
Abdulkadiroglu, A., Pathak, P. A., Schellenberg, J., & Walters, C. R. (2017). Do parents value school
effectiveness? (NBER Working Paper No. 23912). National Bureau of Economic Research.
Abramowitz, M. (1990a, April 4). Mayor signs 1991 D.C. budget with complaints. The Washington
Post, D3.
Abramowitz, M. (1990b, March 6). D.C. schools knew error last year. The Washington Post, B1.
Ahmed, S. (2007). The language of diversity. Ethnic and Racial Studies, 30(2), 235-256.
Ainsworth, J. W. (2002). Why does it take a village? The mediation of neighborhood effects on
educational achievement. Social Forces, 81(1), 117-152.
Alexander, M. (2010). The new Jim Crow: Mass incarceration in the age of colorblindness. The New Press.
American Association of University Women (AAUW). (2018). The simple truth about the gender pay gap.
https://files.eric.ed.gov/fulltext/ED596219.pdf
Asch, C. M., & Musgrove, G. D. (2017). Chocolate city: A history of race and democracy in the nation's capital.
UNC Press Books.
232
Asch, C., & Musgrove, D. (2018, April 3). Opinion: The origins of ‘east of the river’. National Public Radio
– WAMU (Washington, DC affiliate). https://wamu.org/story/18/04/03/opinion-origins-
east-river/
Associated Press. (2019, March 8). Teachers say charter school is closing over union efforts.
https://apnews.com/d04ea9bcfe12494c850fdae6fe2f08c5
Austermuhle, M. (2013, November 15). Four decades after getting home rule, the fight in D.C. goes on.
National Public Radio – WAMU (Washington, DC affiliate).
https://wamu.org/story/13/11/15/four_decades_on_dc_continues_fighting_for_home_ru
le/
Austermuhle, M. (2016, September 23). When blacks fled the South, D.C. became home for many from North
Carolina. National Public Radio – WAMU (Washington, DC affiliate).
https://wamu.org/story/16/09/23/when_blacks_fled_the_south_dc_became_home_for_
many_from_north_carolina/
Austermuhle, M. (2017, March 30). Efforts to unionize teachers at D.C. charter school called off after divisive
campaign. National Public Radio – WAMU (Washington, DC affiliate).
https://wamu.org/story/17/03/30/effort-unionize-teachers-d-c-charter-school-called-off-
divisive-campaign/
Austermuhle, M. (2019, November 14). Mundo Verde will be the city’s first charter school with a teachers
union. DCist. https://dcist.com/story/19/11/14/mundo-verde-will-be-the-citys-first-
charter-school-with-a-teachers-union/
Baker, D. M., & Lee, B. (2019). How does light rail transit (LRT) impact gentrification? Evidence
from fourteen US urbanized areas. Journal of Planning Education and Research, 39(1), 35-49.
Baker, J., & Cangemi, J. (2016). Why are there so few women CEOs and senior leaders in corporate
America?. Organization Development Journal, 34(2), 31-43.

233

Balingit, M., & McLaren, M. (2017, August 17). Despite test gains, fewer than a third of D.C.
students rated “college and career ready”. The Washington Post.
https://www.washingtonpost.com/news/education/wp/2017/08/17/despite-test-gains-
only-a-third-of-d-c-students-rated-college-and-career-ready/?utm_term=.53e32ce5
Ball, S. J. (1993). Education markets, choice and social class: The market as a class strategy in the UK
and the USA. British Journal of Sociology of Education, 14(1), 3-19.
Ball, S. J., & Vincent, C. (1998). ‘I heard it on the grapevine’: ‘Hot’ knowledge and school choice. British
Journal of Sociology of Education, 19(3), 377-400.
Ballou, D., Goldring, E., & Liu, K. (2006). Magnet schools and student achievement (Working Paper).
National Center for the Study of Privatization in Education, Teachers College, Columbia
University.
Barnum, M. (2019, February 5). Five things we’ve learned from a decade of research on school closures. Chalkbeat.
https://www.chalkbeat.org/posts/us/2019/02/05/school-closure-research-review
Baum-Snow, N., & Hartley, D. A. (2017). Accounting for central neighborhood change, 1980-2010 (Working
Paper No. WP2016-09). Federal Reserve Bank of Chicago.
https://fraser.stlouisfed.org/files/docs/historical/frbchi/workingpapers/frbchi_workingpa
per_2016-09.pdf
Bell, D. A., Jr. (1979). Brown v. Board of Education and the interest -convergence dilemma. Harvard
Law Review, 93(3), 518-533.
Beracha, E., & Hardin, W. G., III (2019). The housing price premium associated with charter schools (Working
Paper). Real Estate Economics.
Bickford, A., & Massey, D. S. (1991). Segregation in the second ghetto: Racial and ethnic segregation
in American public housing, 1977. Social Forces, 69(4), 1011-1036.
234
Bifulco, R., & Ladd, H. F. (2007). School choice, racial segregation, and test ‐score gaps: Evidence
from North Carolina's charter school program. Journal of Policy Analysis and Management, 26(1),
31-56.
Black, S. E. (1999). Do better schools matter? Parental valuation of elementary education. The
Quarterly Journal of Economics, 114(2), 577-599.
Blagg, K., & Chingos, M. (2016, May 23). Does gentrification explain rising student test scores in Washington,
DC? Urban Institute. https://www.urban.org/urban-wire/does-gentrification-explain-rising-
student-scores-washington-dc
Booker, K., Gilpatric, S. M., Gronberg, T., & Jansen, D. (2008). The effect of charter schools on
traditional public school students in Texas: Are children who stay behind left behind? Journal
of Urban Economics, 64(1), 123-145.
Bouker, J. (2008, December). Building the best capital city in the world: Appendix one. The Brookings
Institution. https://www.brookings.edu/wp-content/uploads/2016/07/appendix-1.pdf
Brehm, M., Imberman, S. A., & Naretta, M. (2017). Capitalization of charter schools into residential
property values. Education Finance and Policy, 12(1), 1–27.
Broadway, B. (2000, December 17). Ground broken for Metro station in NE. The Washington Post,
C3.
Brooks, L., & Valadez, C. (2017). African Americans in the greater DC area: 1930 to the present. Center for
Washington Area Studies, GW Institute of Public Policy, The George Washington
University. https://cpb-us-
e1.wpmucdn.com/blogs.gwu.edu/dist/7/677/files/2017/09/Final-Version-Mapping-
Segregation-Policy-Brief-1tjali5.pdf
Bross, W., Harris, D. N., & Liu, L. (2016, October). The effects of performance-based school closure and
charter takeover on student performance. Education Research Alliance for New Orleans.

235

https://educationresearchalliancenola.org/files/publications/Bross-Harris-Liu-The-Effects-
of-Performance-Based-School-Closure-and-Charter-Takeover-on-Student-Performance.pdf
Brown, D. (1995a, December 1). Many D.C. students lack a basic necessity: textbooks. The
Washington Post, A1.
Brown, D. (1995b, August 3). GOP offers ‘principals’ for reform. The Washington Post, A10.
Brown, D. (1995c, September 21). D.C. Board approves plan to privatize some schools. The
Washington Post, A1.
Brown, E. (2012, September 11). D.C. auditor: Rhee -era school closures cost millions more than
originally reported. The Washington Post. https://www.washingtonpost.com/blogs/dc-
schools-insider/post/audit-says-rhee-era-school-closures-cost-millions-more-than-originally-
reported/2012/09/11/c90c5c80-fc3b-11e1-8adc-
499661afe377_blog.html?utm_term=.89bdb1187294
Brown, E. (2013, January 17). D.C. to close 15 underenrolled schools. The Washington Post.
https://www.washingtonpost.com/local/education/chancellor-kaya-henderson-names-15-
dc-schools-on-closure-list/2013/01/17/e04202fa-6023-11e2-9940-6fc488f3fecd_story.html
Brown, E. (2016, August 8). Prosecutors decline to bring criminal charges after investig ation into
D.C. charter school. The Washington Post. https://www.washingtonpost.com/news/education
Brown University. (n.d). Longitudinal tract database (LTDB) downloads.
https://s4.ad.brown.edu/projects/diversity/researcher/LTBDDload/DataList.aspx
Browne, I., & Misra, J. (2003). The intersection of gender and race in the labor market. Annual Review
of Sociology, 29, 487-513.
Brummet, Q., & Reed, D. (2019). The effects of gentrification on the well-being and opportunity of original
resident adults and children (Working Paper WP 19-30). Federal Reserve Bank of Philadelphia.

236

Brunner, E. J., Imazeki, J., & Ross, S. L. (2010). Universal vouchers and racial and ethnic
segregation. The Review of Economics and Statistics, 92(4), 912-927.
Buckley, J., & Schneider, M. (2006). Are charter school parents more satisfied with schools? Evidence
from Washington, DC. Peabody Journal of Education, 81(1), 57-78.
Bulkley, K. E. (2013). Conceptions of equity: How influential actors view a contested concept.
Peabody Journal of Education, 88(1), 10-21.
Burdick ‐Will, J., Keels, M., & Schuble, T. (2013). Closing and opening schools: The association between
neighborhood characteristics and the location of new educational opportunities in a large urban
district. Journal of Urban Affairs, 35(1), 59-80.
Business Insider (2020). Dow Jones Index, DJIA.
https://markets.businessinsider.com/index/dow_jones
Cai, Z., & Heathcote, J. (2018). College tuition and income inequality (Centre for Economic Policy
Research Discussion Paper DP13101).
https://repec.cepr.org/repec/cpr/ceprdp/DP13101.pdf
Capps, K. (2015, September 2). How the federal government built white suburbia. CityLab.
https://www.citylab.com/equity/2015/09/how-the-federal-government-built-white-
suburbia/403321/
Center for Research on Education Outcomes (CREDO) (2013). National charter school study: Executive
summary.
http://credo.stanford.edu/documents/NCSS%202013%20Executive%20Summary.pdf
Center for Research on Education Outcomes (CREDO) (2015). Urban charter school study: Report on 41
regions.
https://urbancharters.stanford.edu/download/Urban%20Charter%20School%20Study%20
Report%20on%2041%20Regions.pdf
237
Chakrabarti, R. (2008). Can increasing private school participation and monetary loss in a voucher
program affect public school performance? Evidence from Milwaukee. Journal of Public
Economics, 92(5-6), 1371-1393.
Chakrabarti, R., & Roy, J. (2010, September 1). Do charter schools crowd out private school enrollment?
Evidence from Michigan (Staff Report No. 472). Federal Reserve Bank of New York.
https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr472.html
Chandler, M. A. (2015a, May 12). Many D.C. Public School students travel less than a mile to
school. The Washington Post. https://www.washingtonpost.com/local
Chandler, M.A. (2015b, December 15). Friendship charter schools founder Donald Hense to retire.
The Washington Post. https://www.washingtonpost.com/local/education/friendship-charter-
schools-founder-donald-hense-to-retire/2015/12/15/d710d280-a281-11e5-9c4e-
be37f66848bb_story.html
Chapman, T. K., & Donnor, J. K. (2015). Critical race theory and the proliferation of US charter
schools. Equity & Excellence in Education, 48(1), 137-157.
Cheng, A., Trivitt, J. R., & Wolf, P. J. (2016). School choice and the branding of Milwaukee private
schools. Social Science Quarterly, 97(2), 362-375.
Cheng, S., & Long, J. S. (2007). Testing for IIA in the multinomial logit model. Sociological Methods &
Research, 35(4), 583-600.
Chetty, R., Friedman, J. N., & Rockoff, J. E. (2014). Measuring the impacts of teachers II: Teacher
value-added and student outcomes in adulthood. American Economic Review, 104(9), 2633-79.
Chubb, J. E., & Moe, T. M. (1990). Politics, markets and America's schools. Brookings Institution Press.
Clark, D. L., & Amiot, M. A. (1981). The impact of the Reagan administration on federal education
policy. The Phi Delta Kappan, 63(4), 258-262.

238

Clark, M. A., Gleason, P. M., Tuttle, C. C., & Silverberg, M. K. (2015). Do charter schools improve
student achievement? Educational Evaluation and Policy Analysis, 37(4), 419-436.
Clotfelter, C. T., Ladd, H. F., & Vigdor, J. L. (2009). The academic achievement gap in grades 3 to 8.
The Review of Economics and Statistics, 91(2), 398-419.
Cohn, D. (1991, March 10). D.C. area grows up, spreads out. The Washington Post, A1.
Cohn, D., & Sanchez, R. (1994, April 28). Migration from D.C. is booming. The Washington Post, A1.
Cohen, R. M. (2017, November 30). Does D.C. charter schools’ autonomy come at the cost of
public accountability? Washington City Paper.
https://www.washingtoncitypaper.com/news/article/20984282/does-dc-charter-schools-
autonomy-come-at-the-cost-of-public-accountability
Cohen, R. M. (2019, January 30). D.C. charter administrators have some of the highest school
salaries in town; their teachers, some of the lowest. Washington City Paper.
https://www.washingtoncitypaper.com/news/city-desk/article/21045319/dc-charter-
administrators-have-some-of-the-highest-school-salaries-in-town-their-teachers-some-of-the-
lowest?fbclid=IwAR2BJ0nGDYoeyCs3Xb7c6MLZNGHKUkM0348bvwu_ISu3pHDpWV
YBMAIpr30
Collinson, R. (2011). Rental housing affordability dynamics, 1990- 2009. Cityscape, 13(2), 71-103.
Covington, K. L. (2009). Spatial mismatch of the poor: An explanation of recent declines in job
isolation. Journal of Urban Affairs, 31(5), 559-587.
Crisman, J., & Kim, A. (2019). Property outlaws in the Southland: The potential and limits of
guerrilla urbanism in the cases of arts gentrification in Boyle Heights and street vending
decriminalization in Los Angeles. URBAN DESIGN International, 24(3), 159-170.

239

Crockett, S. A., Jr. (2020, February 12). Gentrification remains undefeated: DC cultural landmark Horace &
Dickies closing after 30 years. The Root. https://www.theroot.com/gentrification-remains-
undefeated-dc-cultural-landmark-1841636609
Cullen, J. B., Jacob, B. A., & Levitt, S. D. (2005). The impact of school choice on student outcomes: An
analysis of the Chicago Public Schools. Journal of Public Economics, 89(5-6), 729-760.
Cullen, J. B., Jacob, B. A., & Levitt, S. (2006). The effect of school choice on participants: Evidence
from randomized lotteries. Econometrica, 74(5), 1191-1230.
Cunningham, L., & Remy, D. (1993, March 7). Hollowing out D.C. The Washington Post, C8.
Davis, T., & Oakley, D. (2013). Linking Charter School Emergence to Urban Revitalization and
Gentrification: A socio ‐spatial Analysis of Three Cities. Journal of Urban Affairs, 35(1), 81-102.
DC Council. (n.d). D.C. home rule. https://dccouncil.us/dc-home-rule/
DC Public Charter School Board (PCSB). (2010, October). Public Charter School Board delays
implementation of charter school evaluation. https://dcpcsb.org/public-charter-school-board-
delays-implementation-charter-school-evaluation
DC Public Charter School Board (PCSB). (2013, September). Performance management framework:
Guidelines and technical guide. http://dcpcsb.org/sites/default/files/Release--
2012_2013_PMF%20Guidelines%20revised%209%2012.pdf
DC Public Charter School Board (PCSB). (2014, January). 2013-14 performance management framework:
Guidelines and technical guide.
http://dcpcsb.org/sites/default/files/2014TechnicalGuide%2001%2010%2014%20FINAL.
pdf
DC Public Charter School Board (PCSB). (2015, August). 2014-15 performance management framework:
Policy and technical guide. http://dcpcsb.org/sites/default/files/Revised%202014-
15%20PMF_Policy__Tech_Guide%20Approved%20%208%2017%2015%20FINAL1.pdf
240
DC Public Charter School Board (PCSB). (2016, November). 2015-16 DC public charter school quality
reports. http://dcpcsb.org/sites/default/files/2016-12-12%20Final%202015-
16%20PMF%20Book.pdf
DC Public Charter School Board (PCSB). (2017, November). 2016-17 DC public charter school quality
reports. https://dcpcsb.egnyte.com/dl/30mlhJPiWv/
DC Public Charter School Board (PCSB). (2018a). 2018 annual report.
https://www.dcpcsb.org/sites/default/files/report/42329%20DCPCSB%202018%20Annu
al%20Report-8.5.18.pdf
DC Public Charter School Board (PCSB). (2018b). DC public charter schools serve higher percentages of at-
risk students and high needs special education students than DC’s traditional public schools.
https://www.dcpcsb.org/blog/at-risk-special-ed-students
DC Public Charter School Board (PCSB). (2018c). School quality report [2018 reports].
https://www.dcpcsb.org/schoolquality
DC Public Charter School Board (PCSB). (2018d, Ja nuary). Financial analysis report.
https://dcpcsb.org/financial-analysis-report
DC Public Charter School Board (PCSB). (2019a, February). All closed DC public charter schools.
https://dcpcsb.egnyte.com/dl/sK1pkdcndM/
DC Public Charter School Board (PCSB). (2019b, August 19). Test Results – PARCC [2019].
https://dcpcsb.org/test-results-parcc
DC Public Charter School Board (PCSB). (2019c, November). 2019-20 Performance management
framework: Policy & technical guide.
https://www.dcpcsb.org/sites/default/files/media/file/2019.11.13%20Vote%202019-
20%20PMF%20Policy%20%26%20Technical%20Guide%20November%20Update%20ZA
TT%20A%20Clean%20BBF.pdf

241

DC Public Charter School Board (PCSB). (2019d, February). All approved local education agency, 1996-
2019. https://dcpcsb.egnyte.com/dl/Rfj8UyEEUh/
DC Public Charter School Board (PCSB). (2020a). Performance management framework (PMF).
https://dcpcsb.org/performance-management-framework-pmf
DC Public Charter School Board (PCSB). (2020b). DC PCSB policies. https://www.dcpcsb.org/dc-
pcsb-policies
DC Public Charter School Board (PCSB). (2020c). Qualitative site reviews.
https://dcpcsb.org/qualitative-site-reviews
DC Scholars Public Charter School. (2019, August). Employee handbook.
https://dcpcsb.org/sites/default/files/media/file/2019-
2020%20Employee%20Policies%20DC%20Scholars%20PCS.pdf
Deming, D. J., Hastings, J. S., Kane, T. J., & Staiger, D. O. (2014). School choice, school quality, and
postsecondary attainment. American Economic Review, 104(3), 991-1013.
District of Columbia Department of Human Resources (DCHR). (2020, March 31). DC government
employee listing: Quarter 2.
https://dchr.dc.gov/sites/default/files/dc/sites/dchr/publication/attachments/public_bod
y_employee_information_200331.pdf
District Department of Transportation (DDOT). (2018). School transit subsidy program.
https://ddot.dc.gov/page/school-transit-subsidy-program
District Department of Transportation (DDOT). (n.d). DC central business district.
https://ddot.dc.gov/sites/default/files/dc/sites/ddot/publication/attachments/dc_central
_business_district_bikes.pdf
District of Columbia Office of Planning (DC Planning). (n.d). Comprehensive plan.
https://planning.dc.gov/node/637932
242
District of Columbia Office of Revenue Analysis (ORA). (2015, November 5). The role of demographics in
NAEP and PARCC scores. https://districtmeasured.com/2015/11/05/the-role-of-
demographics-in-naep-and-parcc-scores/
District of Columbia Office of the Chief Financial Officer (OCFO). (2018). District of Columbia public
charter schools (GC0) FY 2019 proposed budget.
https://cfo.dc.gov/sites/default/files/dc/sites/ocfo/publication/attachments/gc_dcpcs_chapt
er_2019m.pdf
District of Columbia Office of the Deputy Mayor for Education (DME). (2019, August 30). Citywide
landscape of former DCPS facilities remaining in educational use or government owned.
https://dme.dc.gov/publication/citywide-landscape-former-dcps-facilities-remaining-
educational-use-or-government-owned
District of Columbia Office of the State Superintendent of Education (OSSE). (2008, January 30).
The uniform per student funding formula.
https://osse.dc.gov/sites/default/files/dc/sites/osse/release_content/attachments/12538/
PowerpointDraftJanuary30_2008_without_Draft.pdf
District of Columbia Office of the State Superintendent of Education (OSSE). (2011, July). 2011
DC CAS results.
https://osse.dc.gov/sites/default/files/dc/sites/osse/release_content/attachments/22137/
2011_DC_CAS_Results_updated.pdf
District of Columbia Office of the State Superintendent of Education (OSSE). (2018). 2017-18 school
year enrollment audit report and data. https://osse.dc.gov/node/ 1306796
District of Columbia Office of the State Superintendent of Education (OSSE). (2019). 2018-19 PARCC
results and resources. https://osse.dc.gov/page/2018-19-parcc-results-and-resources

243

District of Columbia Office of the State Superintendent of Education (OSSE). (2020). 2019-20 school
year enrollment audit report and data. https://osse.dc.gov/page/2019-20-school-year-enrollment-
audit-report-and-data
District of Columbia Office of the State Superintendent of Education (OSSE). (n.d.a). For parents:
Transportation for students with disabilities. https://osse.dc.gov/service/ parents-transportation-
students-disabilities
District of Columbia Office of the State Superintendent of Education (OSSE). (n.d.b). Enrollment audit
data. https://osse.dc.gov/enrollment
District of Columbia Office of the State Superintendent of Education (OSSE). (n.d.c). State
assessments. https://osse.dc.gov/assessments
District of Columbia Public Schools (DCPS). (n.d.a). Home website. https://dcps.dc.gov/
District of Columbia Public Schools (DCPS). (n.d.b). Thaddeus Stevens Early Learning Center: School year
2020-21 (fiscal year 2021) DCPS initial allocation.
https://www.dcpsdatacenter.com/assets/docs/pdfs/fy21initialallocation_Stevens%20Early
%20Learning%20Center.pdf
Dolan, G. (2018, August 15). Why can’t the middle class afford Catholic school anymore? EducationNext.
https://www.educationnext.org/why-cant-middle-class-afford-catholic-school-anymore/
EdChoice. (2020). Fast facts [on private school choice]. https://www.edchoice.org/engage/fast-
facts/#voucher-fast-facts
Education Commission of the States (ECS). (2017, October). Does the state have open enrollment
programs. http://ecs.force.com/mbdata/MBQuestNB2n?rep=OE1701
Eisenlohr, A. (2020). Assessing the relationship between charter schools and neighborhood gentrification in the
District of Columbia (Working Paper).

244

Ellen, I. G., Horn, K. M., & Reed, D. (2019). Has falling crime invited gentrification? Journal of
Housing Economics, 46, [101636].
Engberg, J., Gill, B., Zamarro, G., & Zimmer, R. (2012). Closing schools in a shrinking district: Do
student outcomes depend on which schools are closed? Journal of Urban Economics, 71(2), 189-
203.
Ewing, R., Tian, G., Goates, J. P., Zhang, M., Greenwald, M. J., Joyce, A., Kircher, J., & Greene, W.
(2015). Varying influences of the built environment on household travel in 15 diverse
regions of the United States. Urban Studies, 52(13), 2330–2348.
Fahle, E. M., Reardon, S. F., Kalogrides, D., Weathers, E. S., & Jang, H. (2020). Racial segregation
and school poverty in the United States, 1999–2016. Race and Social Problems, 12(1), 42-56.
Fauntroy, M. K. (2003). Home rule or house rule: Congress and the erosion of local governance in the District of
Columbia. University Press of America.
Figlio, D. N., & Hart, C. (2010). Competitive effects of means -tested school vouchers (No. w16056).
National Bureau of Economic Research.
Finn, C. E., Jr., Manno, B. V., & Vanourek, G. (1998, July 7). Charte rs: after Marcus Garvey. The
Washington Post. https://www.washingtonpost.com/archive/opinions/1998/07/07/charters-
after-marcus-garvey/df001055-e76b-4a84-bd35-4329d908879c/?utm_term=.5ab3bb7891e8
Fisher, R., Ury, W. L., & Patton, B. (2011). Getting to yes: Negotiating agreement without giving in. Penguin.
Fiss, O. M. (1965). Racial imbalance in the public schools: The Constitutional concepts. Harvard Law
Review, 78(3), 564-617.
Flynn, S. L., & Arguelles, L. (2001). The story of a charter school closure. Paths of Learning: Options for
Families & Communities, 9, 44-51.
Forman, J., Jr. (2004). The secret history of school choice: How progressives got there first. The
Georgetown Law Journal, 93, 1287-1319.
245
Forman, J., Jr. (2006). The rise and fall of school vouchers: A story of religion, race, and politics.
UCLA Law Review, 54, 547-604.
Forman, J., Jr. (2007). Do charter schools threaten public education: Emerging evidence from fifteen
years of a quasi-market for schooling. University of Illinois Law Review, 3, 839-880.
Forstall, R. L. (1996, March). Population of states and counties of the United States: 1790 to 1990.
Population Division, U.S. Census Bureau, U.S. Department of Commerce.
https://www.census.gov/population/www/censusdata/PopulationofStatesandCountiesofth
eUnitedStates1790-1990.pdf
Frankenberg, E., Siegel-Hawley, G., & Wang, J. (2010). Choice without equity: Charter school segregation
and the need for civil rights standards. Civil Rights Project/Proyecto Derechos Civiles.
https://civilrightsproject.ucla.edu/research/k-12-education/integration-and-
diversity/choice-without-equity-2009-report/frankenberg-choices-without-equity-2010.pdf
Freeman, L. (2005). Displacement or succession? Residential mobility in gentrifying
neighborhoods. Urban Affairs Review, 40(4), 463–491.
Freeman, L., & Braconi, F. (2004). Gentrification and displacement: New York City in the 1990s. Journal
of the American Planning Association, 70(1), 39–52.
Frey, J. (2002, July 6). Language of the heart. The Washington Post.
https://www.washingtonpost.com/archive/lifestyle/2002/07/06/language-of-the-
heart/18783190-8524-42e5-b266-f15d636b1d60/?utm_term=.f00ba8f7d0b6
Friedman, B. M. (2015). Work and consumption in an era of unbalanced technological advance. In
A. Chai & C. M. Baum (Eds.), Demand, complexity, and long-run economic evolution (pp. 17-35).
Springer.
Friedman, M. (1962). The role of government in education. In Capitalism and freedom. University of
Chicago Press.

246

Friends of Choice in Urban Schools (FOCUS). (2018). About us. https://focusdc.org/about-us
Fryer Jr, R. G., & Levitt, S. D. (2004). Understanding the black- white test score gap in the first two
years of school. Review of Economics and Statistics, 86(2), 447-464.
Fryer Jr, R. G., & Levitt, S. D. (2006). The black- white test score gap through third grade. American
Law and Economics Review, 8(2), 249-281.
Gaines, P., & Ly, P. (1996, August 16). New charter schools race opening bell. The Washington Post.
https://www.washingtonpost.com/archive
Gallagher, M. (2019, October 11). New DC education data shows how school choice plays out across Wards.
Washington, DC: Urban Institute. https://greaterdc.urban.org/blog/new-dc-education-data-
show-how-school-choice-plays-out-across-wards
Gathright, J. (2019, June 12). D.C. charter school leaders campaign for more space. National Public Radio –
WAMU (Washington, DC affiliate). https://wamu.org/story/19/06/12/d-c-charter-school-
leaders-campaign-for-more-space/
Gibson, C., & Jung, K. (2002, September). Historical census statistics on population totals by race, 1790 to
1990, and by Hispanic origin, 1970 to 1990, for the United States, regions, divisions, and states
(Working Paper No. 56). Population Division, U.S. Ce nsus Bureau.
https://www.census.gov/content/dam/Census/library/working-papers/2002/demo/POP-
twps0056.pdf
Gilens, M., & Page, B. I. (2014). Testing t heories of American politics: Elites, interest groups, and
average citizens. Perspectives on Politics, 12(3), 564-581.
Glomm, G., Harris, D., & Lo, T. F. (2005). Charter school location. Economics of Education Review,
24(4), 451-457.
Glomm, G., & Ravikumar, B. (2003). Public education and income inequality. European Journal of
Political Economy, 19(2), 289-300.

247

Goldstein, S. (2018, June 5). Nine out of 10 new jobs are going to those with a college degree. MarketWatch.
https://www.marketwatch.com/story/nine-out-of-10-new-jobs-are-going-to-those-with-a-
college-degree-2018-06-04
Gordon, M. F., de la Torre, M., Cowhy, J. R., Moore, P. T., Sartain, L., & Knight, D. (2018, May).
School closings in Chicago: Staff and student experiences and academic outcomes. University of Chicago
Consortium on School Research. https://consortium.uchicago.edu/sites/default/files/2018-
10/School%20Closings%20in%20Chicago-May2018-Consortium.pdf
Green, T. L., & Gooden, M. A. (2016). The shaping of policy: Exploring the context, contradictions,
and contours of privilege in “Milliken v. Bradley”, over 40 years later. Teachers College Record,
118(3), 1-30.
Greenberg, E. (2018, November 13). New measures of student poverty: Replacing free and reduced-price lunch
status based on household forms with direct certification. Washington, DC: Urban Institute.
https://www.urban.org/research/publication/new-measures-student-poverty
Grove, L. (1995, January 27). The sermon on the mount. The Washington Post, C1.
Guerra, E., Cervero, R., & Tischler, D. (2012). Half -mile circle: Does it best represent transit station
catchments? Transportation Research Record, 2276(1), 101–109.
Guerrieri, V., Hartley, D., & Hurst, E. (2013). Endogenous gentrification and housing price
dynamics. Journal of Public Economics, 100, 45–60.
Gunderson, S. (1995a, August 10). Lawmaker offers schools plan. The Washington Post, DC3.
Gunderson, S. (1995b, October 9). World class schools for Washington. The Washington Post, A27.
Haab, T. C., & McConnell, K. E. (2002). Valuing environmental and natural resources: The econometrics of
non-market valuation. Edward Elgar Publishing.
Hackworth, J. (2002). Postrecession gentrification in New York City. Urban Affairs Review, 37(6), 815–
843.
248
Hankins, K. B. (2007). The final frontier: Charter schools as new community institutions of
gentrification. Urban Geography, 28(2), 113-128.
Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2009). New evidence about Brown v. Board of
Education: The complex effects of school racial composition on achievement. Journal of
Labor Economics, 27(3), 349-383.
Hanushek, E. A., Kain, J. F., Rivkin, S. G., & Branch, G. F. (2007). Charter school quality and
parental decision making with school choice. Journal of Public Economics, 91(5-6), 823-848.
Harris, H. (1994, February 3). School privatization is tough sell. The Washington Post, DC1.
Hastings, J. S., & Weinstein, J. M. (2008). Information, school choice, and academic achievement:
Evidence from two experiments. The Quarterly Journal of Economics, 123(4), 1373-1414.
Haynes, V. D. (2005, November 6). A costly charter school failure. The Washington Post.
http://www.washingtonpost.com/wp-
dyn/content/article/2005/11/05/AR2005110501395.html
Haynes, V. D. (2008, July 3). Rhee seeks tenure -pay swap for teachers. The Washington Post.
http://www.washingtonpost.com/wp-
dyn/content/article/2008/07/02/AR2008070203498.html
Heinrich, C. J., & Marschke, G. (2010). Incentives and their dynamics in public sector performance
management systems. Journal of Policy Analysis and Management, 29(1), 183-208.
Hendey, L. (2017, December). Racial inequalities in the Washington, DC, region. Washington, DC: Urban
Institute.
https://www.urban.org/sites/default/files/publication/95401/2017.12.28_regional_equity_
finalized_0.pdf
Henig, J. R. (1995). Rethinking school choice: Limits of the market metaphor. Princeton University Press.

249

Henig, J. R., Holyoke, T. T., Lacireno-Paquet, N., & Moser, M. (2001 , February). Growing pains: An
evaluation of charter schools in the District of Columbia, 1999-2000. The Center for Washington
Area Studies, The George Washington University.
Henig, J. R., & MacDonald, J. A. (2002). Locational decisions of charter schools: Probing the market
metaphor. Social Science Quarterly, 83(4), 962-980.
Hicks, B. (2017, March 9). New gentrifiers? Studying the effects of charter schools on neighborhoods in Los Angeles.
Center for Cities + Schools, University of California, Berkeley .
http://citiesandschools.berkeley.edu/blog/new-gentrifiers-study-the-effects-of-charter-schools-
on-neighborhoods-in-LA
Hirschman, A. O. (1970). Exit, voice, and loyalty: Responses to decline in firms, organizations, and states.
Harvard University Press.
Hoffman, S. D., & Duncan, G. J. (1988). Multinomial and conditional logit discrete -choice models
in demography. Demography, 25(3), 415-427.
Holder, S. (2020, February 28). For those living in public housing, it’s a long way to work. CityLab.
https://www.citylab.com/equity/2020/02/jobs-near-me-commute-public-housing-
locations-applicant-data/607084/
Holme, J. J. (2002). Buying homes, buying schools: School choice and the social construction of
school quality. Harvard Educational Review, 72(2), 177-205.
Horowitz, J., Keil, S., & Spector, L. (2009). Do charter schools affect property values? The Review of
Regional Studies, 39(3), 297–316.
Horwitz, S. (1993, June 12). D.C. Education Board appears ready to shut 9 of 10 schools on list. The
Washington Post, B1.
Horwitz, S. (1994, March 5). A public defeat on privatization. The Washington Post, B1.
Horwitz, S. (1995a, December 8). In D.C. schools, it’s tough going. The Washington Post, C1.

250

Horwitz, S. (1995b, June 12). Wide discontent with schools puts D.C. board under siege. The
Washington Post, B1.
Horwitz, S., & Jordan, M. (1992, September 17). D.C. dropout rate among worst in U.S. The
Washington Post, A1.
Hoxby, C. M. (1996). Are efficiency and equity in school finance substitutes or complements? Journal
of Economic Perspectives, 10(4), 51-72.
Hoxby, C. M. (2000a). Does competition among public schools benefit students and taxpayers?
American Economic Review, 90(5), 1209-1238.
Hoxby, C. (2000b). Peer effects in the classroom: Learning from gender and race variation (Working Paper No.
7867). National Bureau of Economic Research.
Hoxby, C. M. (2003). School choice and school productivity. Could school choice be a tide that lifts
all boats? In C. M. Hoxby (Ed.), The economics of school choice (pp. 287-342). University of
Chicago Press.
Hoxby, C. M., Murarka, S., & Kang, J. (2009). How New York City’s charter schools affect achievement. New
York City Charter Schools Evaluation Project.
Hunt, D. B. (2001). What went wrong with public housing in Chicago? A history of the Robert
Taylor homes. Journal of the Illinois State Historical Society (1998-), 94(1), 96-123.
Hwang, J., & Lin, J. (2016). What have we learned about the causes of recent gentrification? Cityscape,
18(3), 9–26.
Hyra, D., & Rugh, J. S. (2016). The US great recession: Exploring its association with black
neighborhood rise, decline and recovery. Urban Geography, 37(5), 700–726.
Imberman, S. A. (2011a). Achievement and behavior in charter schools: Drawing a more complete
picture. The Review of Economics and Statistics, 93(2), 416-435.
251
Imberman, S. A. (2011b). The effect of charter schools on achievement and behavior of public
school students. Journal of Public Economics, 95(7-8), 850-863.
Jacobs, N. (2013). Understanding school choice: Location as a determinant of charter school racial,
economic, and linguistic segregation. Education and Urban Society, 45(4), 459-482.
Johnson, C. (2015). Gentrifying New Orleans: thoughts on race and the movement of capital. Souls,
17(3-4), 175-200.
Jolliffe, I. T., & Cadima, J. (2016). Principal component analysis: A review and recent developments.
Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences,
374(2065), 20150202.
Jordan, R., & Gallagher, M. (2015, August). Does school choice affect gentrification? Washington, DC: Urban
Institute. https://www.urban.org/sites/default/files/publication/65841/2000374-Does-School-
Choice-Affect-Gentrification.pdf
Kahlenberg, R., & Potter, H. (2014, August 30). The original charter school vision. The New York
Times. https://www.nytimes.com/2014/08/31/opinion/sunday/albert-shanker-the-original-
charter-school-visionary.html
Keating, D., & Haynes, V. D. (2007, June 10). Can D.C. schools be fixed? The Washington Post.
http://www.washingtonpost.com/wp-
dyn/content/article/2007/06/09/AR2007060901415.html
Keels, M., Burdick ‐Will, J., & Keene, S. (2013). The effects of gentrification on neighborhood public
schools. City & Community, 12(3), 238-259.
Kemple, J. J. (2015). High school closures in New York City: Impacts on students’ academic outcomes, attendance,
and mobility. Research Alliance for New York City Schools.
https://research.steinhardt.nyu.edu/scmsAdmin/media/users/sg158/PDFs/hs_closures/H
ighSchoolClosuresinNewYorkCity_ResearchAllianceforNYCSChools_pdf.pdf

252

Kerr, S. (2012). Public education in an era of privatization: A spatial examination of the relationship between charter
school clusters and gentrification in Washington, D.C. and Brooklyn, NY (Unpublished Doctoral
Dissertation). University of Florida, Miami.
Kijakazi, K., Brooks Atkins, R. M., Paul, M., Price, A. E., Hamilton, D. , Darity, W. A., Jr. (2016,
November). The color of wealth in the nation’s capital. Washington, DC: Urban Institute, Duke
University, The New School, & The Insight Center for Community Economic Development.
https://www.urban.org/sites/default/files/publication/85341/2000986-the-color-of-wealth-in-
the-nations-capital.pdf
King, C. (1995, July 1). Finally, honest talk about D.C. schools. The Washington Post, A27.
Kingdon, J. W. (2011). Agendas, alternatives, and public policies (2nd ed.). Glenview, IL: Pearson
Education.
Kitali, A. E., Kidando, E., Sando, T., Moses, R., & Ozguven, E. E. (2017). Evaluating aging
pedestrian crash severity with Bayesian complementary log–log model for improved
prediction accuracy. Transportation Research Record, 2659(1), 155-163.
Kleitz, B., Weiher, G. R., Tedin, K., & Matland, R. (2000). Choice, charter schools, and household
preferences. Social Science Quarterly, 81(3), 846-854.
Knoll, E. (1959). The truth about desegregation in the Washington, D.C. public schools. The Journal
of Negro Education, 28(2), 92-113.
Kozol, J. (1991). Savage inequalities: Children in America's schools. Broadway Books.
Krueger, A. B. (1999). Experimental estimates of education production functions. The Quarterly
Journal of Economics, 114(2), 497-532.
Labaree, D. F. (1997). Public goods, private goods: The American struggle over educational goals.
American Educational Research Journal, 34(1), 39-81.

253

Lacireno-Paquet, N., & Holyoke, T. T. (2007). Moving forward or sliding backward: The evolution
of charter school policies in Michigan and the District of Columbia. Educational Policy , 21(1),
185-214.
Ladd, H. F., Clotfelter, C. T., & Holbein, J. B. (2017). The growing segmentation of the charter school
sector in North Carolina. Education Finance and Policy, 12(4), 536-563.
Lamarche, C. (2008). Private school vouchers and student achievement: A fixed effects quantile
regression evaluation. Labour Economics, 15(4), 575-590.
Langhorne, E. (2019, February 28). The real faces behind the ‘corporate reform’ of America’s public
schools. Forbes. https://www.forbes.com
Lankford, H., & Wyckoff, J. (1992). Primary and secondary school choice among public and
religious alternatives. Economics of Education Review, 11(4), 317-337.
Lankford, H., & Wyckoff, J. (2006). The effect of school choice and residential location on the racial
segregation of students. In T. J. Gronberg & D. W. Jansen (Eds.), Improving School
Accountability (pp. 185–239). Emerald Group Publishing Limited.
Library of Congress. (n.d). Today in history – September 20. https://www.loc.gov/item/today-in-
history/september-20/
Lin, J. (2002). Gentrification and transit in northwest Chicago. Transportation Quarterly, 56(4), 175-
191.
Loeb, S., & McEwan, P. (2006). An economic approach to education policy implementation. In M.
Honig (Ed.), New directions in education policy implementation: Confronting complexity (pp. 169-186).
Albany, NY: SUNY Press.
Loeb, V. (1995, October 29). Students, teachers tell Barry of a crisis in D.C. schools. The Washington
Post, B1.

254

Loh, T. H. (2018, December 18). Where the Washington region is zoned for single-family homes: An update.
Greater Greater Washington. https://ggwash.org/view/70232/ washington-region-single-
family-zoning-an-update
Loh, T. H. (2019, January 30). How well is the region doing at planning for growth near Metro? Greater
Greater Washington. https://ggwash.org/view/70464/transit-oriented-delopment-are-we-
upzoning-near-metro-rail-washington-region
Lubienski, C., Gulosino, C., & Weitzel, P. (2009). School choice and competitive incentives:
Mapping the distribution of educational opportunities across local education
markets. American Journal of Education, 115(4), 601–647.
Lytton, M. (2011, August). Have all the costs of closing a school been considered? OECD Centre for
Effective Learning Environments. https://www.oecd-
ilibrary.org/docserver/5kg5c8bwqs31-
en.pdf?expires=1592837116&id=id&accname=guest&checksum=70C57D0ACD20F3D064
9B867F3A0D64E3
Mann, L. (n.d). The complexities of slavery in the nation’s capital. White House Historical Association.
https://www.whitehousehistory.org/the-complexities-of-slavery-in-the-nations-capital
Marcuse, P. (1985). Gentrification, abandonment, and displacement: Connections, causes, and policy
responses in New York City. Washington University Journal of Urban and Contemporary Law, 28, 195-
240.
Marsh, J. (2012). The micropolitics of implementing a school- based bonus policy: The case of New
York City’s compensation committees. Educational Evaluation and Policy Analysis, 34(2), 164-
184.
Martuzzi, M., & Elliott, P. (1998). Estimating the incidence rate ratio in cross -sectional studies using a
simple alternative to logistic regression. Annals of Epidemiology, 8(1), 52-55.
255
Masur, K. (2011, March 28). Capital injustice. The New York Times.
https://www.nytimes.com/2011/03/29/opinion/29masur.html
McDonnell, L. M., & Elmore, R. F. (1987). Getting the job done: Alternative policy instruments.
Educational Evaluation and Policy Analysis, 9(2), 133-152.
McEachin, A., & Atteberry, A. (2017). The impact of summer learning loss on measures of school
performance. Education Finance and Policy, 12(4), 468-491.
McFadden, D. (1973). Conditional logit analysis of qualitative choice behavior. In P. Zarembka (ed.),
Frontiers in economics (pp. 105-135). New York: Wiley.
McKelvey, R. D., & Zavoina, W. (1975). A statistical model for the analysis of ordinal level
dependent variables. Journal of Mathematical Sociology, 4(1), 103–120.
McLaughlin, M. W. (1987). Learning from experience: Lessons from policy
implementation. Educational Evaluation and Policy Analysis, 9(2), 171-178.
Mead, S. (2005, October 12). Capital campaign: Early returns on District of Columbia charter schools.
Progressive Policy Institute. https://files.eric.ed.gov/fulltext/ED491209.pdf
Meghjani, T. (2020, March 5). Student achievement is on the rise, but critical gaps persist. D.C. Policy Center.
https://www.dcpolicycenter.org/publications/achievement-gaps/
Mikulecky, M. T. (2013, June). Open enrollment is on the menu – But can you order it? Denver, CO:
Education Commission of the States.
https://www.ecs.org/clearinghouse/01/07/96/10796.pdf
Mock, B. (2019, November 18). How go-go music became kryptonite for gentrification in D.C. CityLab.
https://www.citylab.com/equity/2019/11/go-go-official-music-washington-dc-culture-
gentrification/601966/
My School DC. (2019). School finder SY 2019-2020. http://find.myschooldc.org/
My School DC. (2020). School finder SY 2020-2021. http://find.myschooldc.org/

256

Myers, D., & Park, J. (2019). A constant quartile mismatch indicator of changing rental affordability
in U.S. metropolitan areas, 2000 to 2016. Cityscape, 21(1), 163-200.
National Bureau of Economic Research (NBER). (2010, September 20). Untitled [announcement
regarding Great Recession]. https://www.nber.org/cycles/sept2010.pdf
National Center for Education Statistics (NCES). (2020, May). Public charter school enrollment.
https://nces.ed.gov/programs/coe/indicator_cgb.asp
National Center for Education Statistics (NCES). (n.d). Elementary/secondary information system [for
Common Core of Data]. https://nces.ed.gov/ccd/elsi/
National Center for Research on Education Access and Choice Project (REACH Project). (2019).
Improving policy, implementation, and outcomes for disadvantaged students (Ongoing research project,
University of Southern California, Michigan State University, Tulane University, University
of Texas at Austin, Florida State University, Montclair State University).
National Charter School Resource Center. (2018, March). An analysis of the charter school facility landscape
in Colorado. U.S. Department of Education.
https://charterschoolcenter.ed.gov/sites/default/files/files/field_publication_attachment/
Analysis%20of%20the%20Facilities%20Landscape%20in%20Colorado.pdf
Neubeck, K. J., & Cazenave, N. A. (2002). Welfare racism: Playing the race card against America's poor.
Routledge.
Newman, K., & Wyly, E. K. (2006). The right to stay put, revisited: Gentr ification and resistance to
displacement in New York City. Urban Studies, 43(1), 23–57.
New York Times News Service. (1995, February 16). D.C. credit rating gets junk bond status. The
Baltimore Sun. http://articles.baltimoresun.com/1995-02-16/news/1995047014_1_junk-
bond-credit-rating-city-rating

257

Nichols, A., & Özek, U. (2010). Public school choice and student achievement in the District of Columbia
(CALDER Working Paper No. 53). National Center for Analysis of Longitudinal Data in
Education Research.
Office of the District of Columbia Auditor (DC Auditor). (2004, May 17). Review of the financial
operations of the Village Learning Center Public Charter School. https://dcauditor.org/wp-
content/uploads/2018/07/DCA1304.pdf
Office of the District of Columbia Auditor (ODCA). (2014, September 5). DC Public Education
Reform Amendment Act (PERAA) (Report No. 4). Washington, DC: Georgetown University
EdCore.
Oprysko, C., & Kimbel-Sannit, A. (2019, August 30). Trump again flirts with easing capital gains
taxes. Politico. https://www.politico.com/story/2019/08/30/trump-capital-gains-taxes-
1478882
Orfield, G. (1978). Research, politics and the antibusing debate. Law and Contemporary Problems, 42(4),
141-173.
Orfield, G. (2014). Tenth annual Brown lecture in education research: A new civil rights agenda for
American education. Educational Researcher, 43(6), 273-292.
Orfield, G., & Ee, J. (2017). Our segregated capital: An increasingly diverse city with racially polarized schools.
Civil Rights Project/Proyecto Derechos Civiles.
https://escholarship.org/content/qt3m9315v6/qt3m9315v6.pdf
Osborne, D. (1990, December 9). Ten ways to turn D.C. around. The Washington Post, SM21.
Osborne, D. (2012). Improving charter school accountability: The challenge of closing failing schools. Washington,
DC: Progressive Policy Institute. https://files.eric.ed.gov/fulltext/ED533181.pdf
Osborne, D. (2015, September). A tale of two systems: Education reform in Washington D.C. Progressive
Policy Institute. https://files.eric.ed.gov/fulltext/ED570145.pdf
258
Paglin, C. (2001). Why charter schools stumble—and sometimes fall. Northwest Education, 6(3), 20-25.
Paino, M., Boylan, R. L., & Renzulli, L. A. (2017). The closing door: The effect of race on charter
school closures. Sociological Perspectives, 60(4), 747-767.
Parsons, G. R., & Kealy, M. J. (1992). Randomly drawn opportunity sets in a random utility model
of lake recreation. Land Economics, 68(1), 93-106.
Pearman, F. A., & Swain, W. A. (2017). School choice, gentrification, and the variable significance of
racial stratification in urban neighborhoods. Sociology of Education, 90(3), 213-235.
Perry, A. M. (2018, April 24). Where charter schools are built shows our commitment to integration. The Avenue,
Brookings Institution. https://www.brookings.edu/blog/the-avenue/2018/04/23/where-
charter-schools-are-built-shows-our-commitment-to-integration/
Persell, C. H., & Cookson Jr, P. W. (1985). Chartering and bartering: Elite education and social
reproduction. Social Problems, 33(2), 114-129.
Pfeffer, F. T., Danziger, S., & Schoeni, R. F. (2013). Wealth disparities before and after the Great
Recession. The ANNALS of the American Academy of Political and Social Science, 650(1), 98-123.
Potter, H., & Quick, K. (2018, May 15). Diverse-by-design charter schools. The Century Foundation.
https://tcf.org/content/report/diverse-design-charter-schools/?session=1&agreed=1
Rawls, J. (2001). Justice as fairness: A restatement. Belknap Press.
Reardon, S. F. (2011). The widening academic achievement gap between the rich and the poor: New
evidence and possible explanations. In G. J. Duncan & R. J. Murnane (Eds.), Whither
opportunity? Rising inequality, schools, and children’s life chances (pp. 91-116). Russell Sage
Foundation.
Reardon, S. F. (2016). School segregation and racial academic achievement gaps. RSF: The Russell
Sage Foundation Journal of the Social Sciences, 2(5), 34-57.

259

Reay, D., & Ball, S. J. (1998). ‘Making their minds up’: Family dynamics of school choice. British
Educational Research Journal, 24(4), 431-448.
Renzulli, L. A., & Evans, L. (2005). School choice, charter schools, and white flight. Social Problems,
52(3), 398-418.
Richardson, J., Mitchell, B., & Franco, J. (2019, March 19). Shifting neighborhoods: Gentrification and cultural
displacement in American cities. Washington, DC: National Community Reinvestment Coalition.
https://ncrc.org/gentrification/
Richardson, L., & Ragland, J. (1991, June 20). 67 D.C. school administrators get pink slips. The
Washington Post, A1.
Rinas, E. J. (1996). A Constitutional analysis of race -based limitations on open enrollment in public
schools. Iowa Law Review, 82, 1501-1534.
Rogers, E. M. (2010). The diffusion of innovations. New York, NY: Free Press.
Ross, C. E., & Wu, C. L. (1995). The links between education and health. American Sociological Review,
60, 719-745.
Rothstein, R. (2017). The color of law: A forgotten history of how our government segregated America. Liveright
Publishing Corporation.
Rouse, C. E. (1998). Private school vouchers and student achievement: An evaluation of the Milwaukee
Parental Choice Program. The Quarterly Journal of Economics, 113(2), 553-602.
Rumberger, R. W., & Willms, J. D. (1992). The impact of racial and ethnic seg regation on the
achievement gap in California high schools. Educational Evaluation and Policy Analysis, 14(4),
377-396.
Ryan, J. E., & Heise, M. (2002). The political economy of school choice. The Yale Law Journal, 111(8),
2043-2136.
Sanchez, R. (1993). D.C. budget seen as a tourniquet. The Washington Post, B1.

260

Saporito, S. (2003). Private choices, public consequences: Magnet school choice and segregation by race
and poverty. Social Problems, 50(2), 181-203.
Sass, T. R. (2006). Charter schools and student achievement in Florida. Education Finance and Policy, 1(1),
91-122.
Schneider, A., & Ingram, H. (1990). Behavioral assumptions of policy tools. The Journal of Politics, 52(2),
510-529.
Schneider, A., & Ingram, H. (1993). Social construction of target populations: Implications for
politics and policy. American Political Science Review, 87(2), 334-347.
Schneider, H. (1995, November 3). House approves plan to slash D.C. budget. The Washington Post,
A1.
Schneider, H., & Sanchez, R. (1995, December 20). For vouchers, a key test. The Washington Post, D1.
Schneider, H., & Vise, D. (1995a, February 2). Barry says D.C. deficit now $722 million: Mayor to
ask Congress to cut city wages. The Washington Post, A1, A13.
Schneider, H., & Vise, D. (1995b, August 2). Speaker comes out of the House to air his ideas for the
District. The Washington Post, D1, D6.
Schneider, H. & Vise, D. (1995c, June 24). D.C. seeks more money. The Washington Post, A1.
Schneider, H., & Yang, J. E. (1996, March 28). D.C. budget deadlock nears end. The Washington Post,
B1-B2.
Schneider, M., & Buckley, J. (2003). Making the grade: Comparing DC charter schools to other DC
public schools. Educational Evaluation and Policy Analysis, 25(2), 203-215.
Scott, J.C. (1998). Seeing like a state: How certain schemes to improve the human condition have failed. Yale
University Press.
Semega, J., Kollar, M., Creamer, J., & Mohanty, A. (2019, September). Income and poverty in the United
States: 2018 (Report No. P60-266). Current Population Reports, U.S. Census Bureau, U.S.

261

Department of Commerce.
https://www.census.gov/content/dam/Census/library/publications/2019/demo/p60-
266.pdf
Sen, A. K. (1999). Development as freedom. Alfred A. Knopf Publishing.
Shaw, K. S., & Hagemans, I. W. (2015). ‘Gentrification without displacement' and the consequent
loss of place: The effects of class transition on low ‐income residents of secure housing in
gentrifying areas. International Journal of Urban and Regional Research, 39(2), 323-341.
Simons, S. (2018, April 4). Black, white, and Asian – Three reflections on the 1968 D.C. riots. National
Public Radio – WAMU (Washington, DC affiliate).
https://wamu.org/story/18/04/04/black-white-asian-three-reflections-1968-d-c-riots/
Smith, A. G. (1995). Public school choice and open enrollment: Implication s for education,
desegregation, and equity. Nebraska Law Review, 74, 255-303.
Smith, F. L. (1992, July 12). Give D.C. kids a choice. The Washington Post, C1.
Smith, N. (1979). Toward a theory of gentrification: A back to the city movement by capital, not
people. Journal of the American Planning Association, 45(4), 538–548.
Snyder, T., & Musu-Gillette, L. (2015, April 16). Free or reduced price lunch: A proxy for poverty? Washington,
DC: National Center for Education Statistics. https://nces.ed.gov/blogs/nces/post/free-or-
reduced-price-lunch-a-proxy-for-poverty
Spillane, J.P., Reiser, B.J, & Gomez, L.M. (2006). Policy implementation and cognition: The role of
human, social, and distributed cognition in framing policy. In M. Honig (Ed.), New directions
in education policy implementation: Confronting complexity (pp. 47-64). Albany, NY: SUNY Press.
Stein, P. (2017, October 12). The healthy D.C. economy is leaving behind longtime black residents, new
study finds. The Washington Post. https://www.washingtonpost.com/local
262
Stein, P. (2018a). One of the city’s oldest charter schools is so broke that it can’t make payroll. The
Washington Post. https://www.washingtonpost.com/local/education/one-of-the-citys-oldest-
charter-schools-is-so-broke-that-it-cant-make-payroll/2018/03/13/81b775d2-264b-11e8-
874b-d517e912f125_story.html?utm_term=.ea41537306ea
Stein, P. (2018b). City charter board votes to shut down D.C.’s only all- girls public school. The
Washington Post. https://www.washingtonpost.com/local/education/city-charter-board-
votes-to-shut-down-dcs-only-all-girls-public-school/2018/01/12/a3a6176e-f7b8-11e7-beb6-
c8d48830c54d_story.html?utm_term=.a4c33d12f6a8
Stewart, F. (2009). Horizontal inequality: Two types of trap. Journal of Human Development and
Capabilities, 10(3), 315-340.
Stone, D. (2012). Policy paradox (3rd ed.). New York, NY: WW Norton & Company.
Strauss, V. (1995a, November 3). House moves to reshape D.C. schools. The Washington Post, B1.
Strauss, V. (1995b, November 4). House loads school plan with options. The Washington Post, B1.
Strauss, V., & Slevin, P. (1998). After Garvey decision, doubts a bout charter schools. The Washington
Post, D1, D4.
Sturdivant-Sani, C. (2018, September 13). The real estate struggle is real for DC charter schools. Commercial
Observer. https://commercialobserver.com/2018/09/the-real-estate-struggle-is-real-for-dc-
charter-schools/
Su, J. G., Jerrett, M., McConnell, R., Berhane, K., Dunton, G., Shankardass, K., Reynolds, K.,
Chang, R., & Wolch, J. (2013). Factors influencing whether children walk to school. Health
Place, 22, 153-161.
Sunderman, G. L., Coghlan, E., & Mintrop, R. (2017, May). School closure as a strategy to remedy low
performance. National Education Policy Center.
https://nepc.colorado.edu/publication/closures

263

Taylor, A. (2018, April 3). The riots that followed the assassination of Martin Luther King Jr. The
Atlantic. https://www.theatlantic.com/photo/2018/04/the-riots-that-followed-the-
assassination-of-martin-luther-king-jr/557159/
Teske, P. E., Fitzpatrick, J. L., & O'Brien, T. (2009). Drivers of choice: Parents, transportation, and school
choice. Center on Reinventing Public Education, University of Washington Bothell.
The Next Step Public Charter School. (2018, June). Employee handbook.
https://static1.squarespace.com/static/5910f3ec86e6c0368e043abc/t/5e14ad15141f9b6e39
1bd688/1578413334848/Employee+Handbook+amended+final+revision+8-21-18-
TAM.pdf
Thornton, P. H., & Ocasio, W. (1999). Institutional logics and the historical contingency of power in
organizations: Executive succession in the higher education publishing industry, 1958-1990.
American Journal of Sociology, 105(3), 801-843.
Tiebout, C. M. (1956). A pure theory of local expenditures. Journal of Political Economy, 64(5), 416-424.
Toth, K. (2015, February 13). Longtime Bed-Stuy residents have had it with hipster bars invading
their neighborhood. Village Voice. https://www.villagevoice.com
Tucker, C. (2015, February 10). D.C. charter school protests possible closure over misuse of funds allegation.
WJLA. https://www.wjla.com/news/local
Turque, B. (2009). Residential charter school for teen mothers to close. The Washington Post.
http://www.washingtonpost.com/wp-
dyn/content/article/2009/05/14/AR2009051403993.html
University of Illinois at Chicago (UIC). (2014). The socioeconomic change of Chicago’s community areas
(1970-2010): gentrification index. Chicago, IL: Nathalie P. Vorhees Center for Neighborhood
and Community Improvement. https://voorheescenter.red.uic.edu/wp-
content/uploads/sites/122/2017/10/Voorhees-Center-Gentrification-Index-Oct-14.pdf

264

Urquiola, M. (2005). Does school choice lead to sorting? Evidence from Tiebout variation. American
Economic Review, 95(4), 1310-1326.
U.S. Census Bureau. (2019). 2019 TIGER/Line shapefiles: census [for the District of Columbia].
https://www.census.gov/
U.S. Government Accountability Office (GAO). (1995). District of Columbia: financial crisis: Statement of
John W. Hill, Jr. (Report No. GAO/T-AIMD-95-88).
https://www.gao.gov/assets/110/105812.pdf
U.S. Government Accountability Office (GAO). (2003a). District of Columbia: structural imbalance and
management issues (Report No. GAO-03-666). https://www.gao.gov/new.items/d03666.pdf
U.S. Government Accountability Office (GAO). (2003b). New charter schools across the country and in the
District of Columbia face similar start-up challenges (Report No. GAO-03-899).
https://www.gao.gov/products/GAO-03-899
U.S. Government Accountability Office (GAO). (2017, February). District of Columbia charter schools:
Multi-agency plan needed to continue progress addressing high and disproportionate discipline rates [accessible
version] (Report No. GAO-17-165). https://www.gao.gov/assets/690/682673.pdf
U.S. Government Printing Office. (1973). Ninety-third Congress: Congressional pictorial directory.
Washington, DC.
U.S. House of Representatives. (n.d). Black-American representatives and senators by Congress, 1870 – present.
https://history.house.gov/Exhibitions-and-Publications/BAIC/Historical-Data/Black-
American-Representatives-and-Senators-by-Congress/
U.S. Navy. (n.d.a). History [of Joint Base Anacostia-Bolling].
https://www.cnic.navy.mil/regions/ndw/installations/jbab/about/history.html
265
U.S. Navy. (n.d.b). Washington Navy Yard: Manufacturing: WWII to 1962.
https://www.history.navy.mil/content/history/museums/nmusn/explore/photography/wa
shington-navy-yard/manufacturing/wwii-1962.html
U. S. News & World Report. (n.d). Best high schools in District of Columbia.
https://www.usnews.com/education/best-high-schools/district-of-columbia
Valant, J.., & Lincove, J. A. (2018, September). New Orleans students’ commute times by car, public transit, and
school bus. Urban Institute:
https://www.urban.org/sites/default/files/publication/99051/new_orleans_students_commute
_times_by_car_public_transit_and_school_bus_0.pdf
Vigdor, J. L. (2002). Does gentrification harm the poor? [with Comments]. Brookings-Wharton Papers on
Urban Affairs, 133–182.
Vise, D. (1995, June 9). GOP looking to sell D.C. on school plans. The Washington Post, D3.
Vise, D., & Horwitz, S. (1995, May 13). District school chief backs voucher plan. The Washington Post,
C1.
Vise, D., & Schneider, H. (1995a, March 31). More reform ahead for D.C., GOP says. The
Washington Post, A1.
Vise, D., & Schneider, H. (1995b, December 17). Who’s in charge? The Washington Post, N13.
Vise, D., & Schneider, H. (1995c, August 4). Bipartisan partnership works for Speaker and D.C.
delegate. The Washington Post, C3.
Vise, D., & Schneider, H. (1995d, November 8). Control Board summons Barry, says city isn’t
cooperating. The Washington Post, C1.
Walters, P. B. (2001). Educational access and the state: Historical continuities and discontinuities in
racial inequality in American education. Sociology of Education, (74), 35-49.
266
Washington, DC Investor Relations. (n.d). District of Columbia investor relations.
https://www.dcbonds.com/the-district-of-columbia-dc/i192
Washington Lawyers’ Committee for Civil Rights. (2005, March). Separate and unequal: The state of the
District of Columbia Public Schools fifty years after Brown and Bolling. A Parents United for the D.C.
Public Schools Civic Leader Advisory Committee Report.
http://www.washlaw.org/pdf/Separate_and_Unequal_Report.pdf
Washington Metropolitan Area Transit Authority (WMATA). (2014). Metro facts 2014.
https://web.archive.org/web/20160222003449/
http://www.wmata.com/about_metro/docs/Metro%20Facts%202014.pdf
Washington Monthly. (2016, June 30). The monthly interview: The man who reinvented public
housing. Washington Monthly. https://washingtonmonthly.com/2016/06/30/the-monthly-
interview-the-man-who-reinvented-public-housing/
Washington Post. (1993, June 26). A report card on D.C. schools. The Washington Post, A24.
Washington Post. (1994, August 3). D.C. schools: ‘Do not enter’? The Washington Post, A16.
Washington Post (1995, June 28). Mayor Barry, vouchers, and more. The Washington Post, A20.
Weatherley, R., & Lipsky, M. (1977). Street-level bureaucrats and institutional innovation:
Implementing special-education reform. Harvard Educational Review, 47(2), 171-197.
Weeden, J., & Jacobson, J. (2017, October 11). How can we close our persistent education gaps in DC?
Greater Greater Washington. https://ggwash.org/view/65113/closing-our-persistent-
education-achievement-gaps-in-dc
Weiher, G. R., & Tedin, K. L. (2002). Does choice lead to racially distinctive schools? Charter schools
and household preferences. Journal of Policy Analysis and Management, 21(1), 79-92.
Wharton, A. S. (2009). The sociology of emotional labor. Annual Review of Sociology, 35, 147-165.
267
Wilgoren, D. (1996, December 16). D.C. charter school born of confusion. The Washington Post, A1,
A14.
Wilgoren, D. (1997, June 25). A fresh start for charter schools in D.C. The Washington Post.
https://www.washingtonpost.com/archive/local/1997/06/25/a-fresh-start-for-charter-
schools-in-dc/2d5f2cbc-df37-4cff-88cf-2ca6762b06da/?utm_term=.bf07dff92f13
Wilgoren, D., & Strauss, V. (1999, November 7). On the learning curve. The Washington Post.
https://www.washingtonpost.com/wp-srv/WPcap/1999-11/07/081r-110799-idx.html
Wilson, T. S. (2016). Contesting the public school: Reconsidering charter schools as counterpublics.
American Educational Research Journal, 53(4), 919-952.
Winters, M. A. (2012). Measuring the effect of charter schools on public school student achievement in
an urban environment: Evidence from New York City. Economics of Education Review, 31(2), 293-
301.
Wirt, F.M., & Kirst, M. (2005). The political dynamics of American education (3rd ed.). Richmond, CA:
McCutchan Publishing Corporation.
Wolfe, A. (Ed.). (2009). School choice: The moral debate. Princeton University Press.
Yearby, R. (2018). The impact of structural racism in employment and wages on minority women's
health. Human Rights Magazine.
https://www.americanbar.org/groups/crsj/publications/human_rights_magazine_home/th
e-state-of-healthcare-in-the-united-states/minority-womens-health/
Ziebarth, T., & Palmer, L. (2018, January). Measuring up to the model: A ranking of state public charter school
laws (Ninth Annual Edition). National Alliance for Public Charter Schools.
https://www.publiccharters.org/sites/default/files/documents/2018-
01/2018%20MEASURING%20UP%20TO%20THE%20MODEL%20Final_0.pdf
268
Zuk, M., Bierbaum, A. H., Chapple, K., Gorska, K., & Loukaitou-Sideris, A. (2018). Gentrification,
displacement, and the role of public investment. Journal of Planning Literature, 33(1), 31-44.
Zukin, S., Trujillo, V., Frase, P., Jackson, D., Recuber, T., & Walker, A. (2009). New retail capital
and neighborhood change: Boutiques and gentrification in New York City. City &
Community, 8(1), 47-64.
Appendix A: Select Literature on School Choice Mechanisms and Academic Outcomes
Choice
Mechanism
Studied
Paper Geography
Period of
Study
Grades
Studied
Analytical Framework
Testing for
Direct Effect or
Spillover Effect,
or Both
Empirical Findings
Private school
vouchers
Rouse (1998) Milwaukee,
WS
1990-1993 Grades K-8 Estimates the difference in
standardized assessment
performance for successful
voucher applicants (i.e.,
lottery winners) relative to
unsuccessful voucher
applicants (i.e., lottery losers),
plus students in Milwaukee
Public Schools who did not
apply for vouchers
Treatment Effect Mixed. Statistically
significant gains for
successful voucher
applicants in
standardized assessment
math scores relative to
two control groups.
Cannot reject null
hypothesis for reading
scores.
Private school
vouchers
Lamarche
(1998)
Milwaukee,
WS
1990-1993 Unclear but
likely Grades
K-8
Estimates the difference in
standardized assessment
performance for successful
voucher applicants to
unsuccessful voucher
applicants, plus students in
Milwaukee Public Schools
who did not apply for
vouchers. Uses quantile
regressions based on prior
standardized assessment
scores to see if there are
differential effects across the
distribution of students’ prior
standardized assessment
scores.
Treatment Effect Mixed. Statistically
significant gains for
successful voucher
applicants in math scores
relative to two control
groups, larger effect for
lower-performing
students (lower
quantiles). Statistically
significant gains for
successful applicants in
readings scores in low
quantiles, yet significant
losses in reading gains
for successful applicants
in high quantiles.
Private school
vouchers
Chakrabarti
(2008)
Milwaukee,
WS
1987-2002 Grades 3-5 Estimates the association
between a traditional public
school’s share of enrolled
Spillover Effect on
Traditional Public
School System
Positive. Post-
treatment, Milwaukee
Public Schools exhibited
269
students who are considered
in poverty (a proxy for
competition with private
schools over students eligible
for vouchers) and that
school’s change in
standardized assessment
performance levels before
and after 1998 court order
expanding voucher program
to include religious schools.
significant increases in
standardized assessment
scores. This is especially
notable for reading and
language arts scores,
where the effect size is
larger as the public
school’s share of
students in poverty
increases.
Private school
vouchers
Figlio & Hart
(2010)
State of
Florida
2001-2002
school year
Grades K-12 Estimates the association
between various measures of
“competition” a public
school feels from private
schools (including nearest
private school, number of
private schools within five
miles, diversity of private
schools) and the
performance of traditional
public school students on
standardized assessments in
the year leading up to the
implementation of the state’s
voucher program (the law
was announced in 2001 and
went into effect in the 2002-
2003 school year).
Spillover Effect on
Traditional Public
School System
Positive. In the school
year between
announcement of
voucher policy and its
implementation, observe
statistically significant
increases in standardized
assessment scores in
both reading and math
based on the level of
“competition” a
traditional public school
experiences. Observe
greater effects in schools
with the “most to lose”
from voucher policy–
elementary and middle
schools (where vouchers
for private school cover
greater percentage of
cost) and schools at the
threshold of eligibility
for federal Title I dollars.
Magnet
schools
Abdulkadiroğlu,
Angrist, &
Pathak (2014)
Boston, MA
and New
York City,
NY
Boston:
1997-2009
Boston:
Grades 7, 9
Using fuzzy regression
discontinuity (both reduced
form and two-stage least
squares), compares the
Treatment Effect Mixed. Fail to find
broad evidence that
enrollment in a particular
magnet school causes a
270
New York
City: 2004-
2007
New York
City: Grade 9
PSAT/SAT/AP and
standardized test scores of
students who are just above
magnet schools’ cut-off
criteria (in terms of
applicants’ prior academic
performance) and receive an
offer of admission with those
students who are just below
the criteria.
borderline student’s
performance on tests
improves. The sole
exception to this is the
10
th
grade state
standardized assessment
scores at Boston’s two
less prestigious magnet
schools, and AP exam
scores at Boston’s least
prestigious magnet
school.
Magnet
schools
Ballou,
Goldring, &
Liu (2006)
Mid-sized
district in
Tennessee
1999-2003 Grades
6, 7, 8
Estimate the difference in
standardized assessment
scores on math and reading
tests for magnet school
lottery winners relative to
those of magnet school
lottery losers.
Treatment Effect Mixed. After controlling
for student demographic
characteristics and prior
test performance, for all
but one magnet school
(and only in math) the
authors are unable to
reject the null hypothesis
that admission to a
magnet school has no
effect on a middle
school student’s
standardized test scores.
Intra-district
open
enrollment
Cullen, Jacob,
& Levitt (2005)
Chicago, IL 1993-1995 Grade 9 Estimate the effect of a
student “opting out” of
his/her neighborhood school
on his/her chances of
completing 10
th
grade,
completing 11
th
grade, and
graduating on time. The
authors control for
sociodemographic
characteristics, qualitative
data taken from student
surveys (meant to inform
measures of motivation,
Treatment Effect Negative. Once the
authors instrument for a
student’s residential
distance to nearby public
high schools (an
instrument I find not
exogenous to grade
completion/graduation
given its likely
correlation with
households’
sociodemographic
characteristics), they find
271
“grit”, etc.), and the type of
public high school “opted
into” (high performing, a
vocational/career tech-
oriented Career Academy, or
other).
that: (a) Career Academy
schools appear to boost
outcomes for students
regardless of whether or
not they opt in, and (b)
those students opting
into high performing
high schools or “other”
high schools
systematically differ on
qualitative survey
responses relative to
students who attend
those schools who did
not “opt in”
Intra-district
open
enrollment
Cullen, Jacob,
& Levitt (2006)
Chicago, IL 2000-2001 Grade 9 Estimate the traditional
academic outcomes
(standardized assessment
scores, high school
graduation probabilities,
attendance rates, etc.) and
other outcomes (disciplinary
incidences and probability of
being arrested) for students
who won lotteries for over-
subscribed high schools
relative to those students
who lost lotteries, in
Chicago’s open enrollment
system.
Treatment Effect Negative. Fail to reject
the null hypothesis that
lottery winners to over-
subscribed high schools
experience significant
increases in any
traditional academic
outcome measure;
however, the authors do
find significant
reductions in students’
disciplinary incidences
and chances of being
arrested outside of
school. However, it is
possible that there is
endogeneity of discipline
data with school
characteristics.
Intra-district
open
enrollment
Deming,
Hastings, Kane,
& Staiger
(2014)
Charlotte-
Mecklenburg
School
District
2002 Grades 9-12 Estimate the academic
outcomes (attendance at 2 or
4-year college, attendance at
“very competitive college”,
secondary school GPA, “on
track” status in school, SAT
scores) of students who won
Treatment Effect Mixed. Of particular
note is the authors’
bifurcation of findings
by student gender.
Among female students,
they concluded that
those who won open
272
lottery to their first choice
high schools in open
enrollment system relative to
those students who did not
win lottery to their first
schools.
enrollment lottery to
their first choice school
were significantly more
likely to attend a 4-year
college, significantly
more likely to attend a
“very competitive
college”, possessed
significantly higher
GPAs, and were
significantly more likely
to stay on track in
secondary math course
enrollment. The authors
do not observe
statistically significant
effects on boys. They
find some statistically
significant effects when
not considering gender.
Intra-district
open
enrollment
Hastings &
Weinstein
(2008)
Charlotte-
Mecklenburg
School
District
2006-2007 Grades K-12 Use two different methods,
and in each look at the effect
of the standardized
assessment scores of the
school a student “opted into”
on that student’s own scores:
(1) Via two-stage least
squares, estimate the
standardized assessment
scores (reading and math) of
students who randomly
received an information sheet
on each school’s academic
performance relative to those
who did not receive an
information sheet.
(2) Estimate the standardized
assessment scores of students
who won lotteries relative to
Treatment Effect Positive. In each of the
methods, find that there
is a statistically
significant positive
association between a
student’s own
standardized math and
reading scores and the
standardized assessment
score level of the school
that student “opted
into”.
273
those who lost lotteries for
open enrollment options.
Charter
schools
Bifulco & Ladd
(2007)
State of
North
Carolina
1996-2002 Grades 4-8 Use a student-level fixed-
effects panel dataset to
estimate changes in
standardized assessment
scores for students who
switch from traditional public
schools to charter schools.
Treatment Effect Negative. Charter
schools significantly
depress black students’
math test scores but do
not significantly alter
white students’ math
scores. The authors
attribute this to
segregated charter
schools’ effects on black
students. It would be
interesting to see the
analysis of segregation in
the charter sector’s
effects in terms of
income as well as race.
Charter
schools
Ladd,
Clotfelter, &
Holbein (2017)
State of
North
Carolina
2003-2011 Grades 4-8 Use a student-level fixed-
effects panel dataset to
estimate changes in
standardized assessment
scores for students who
switch from traditional public
schools to charter schools.
The authors additionally
bifurcate analysis from 2003-
2007 and 2008-2011 to
identify whether the
effectiveness of charter
schools has changed as the
charter sector in North
Carolina has matured.
Treatment Effect Negative. On average,
charter schools were
significantly less
effective than traditional
public schools in
promoting test score
gains in math and
reading for the 2003-
2007 period. Between
2008-2011, the
researchers could not
reject the null hypothesis
that charter schools were
as effective as traditional
public schools in
promoting gains in
reading and math. The
authors additionally
report that the test
scores of charters that
274
close are, on average,
lower than those that
stay open (not a
surprising finding).
Charter
schools
Hanushek,
Kain, Rivkin, &
Branch (2007)
State of
Texas
1996-2002 Grades 4-8 The authors use a student-
level fixed effects panel
dataset (effectively taking
first differences) to estimate
the effect of charter school
attendance on standardized
assessment scores in math
and reading for students who
switch from traditional public
schools to charter schools.
The authors break down
their analysis of charters’
estimated effects by years
that the charter schools have
been operating. The authors
also use an OLS framework
to estimate the differences in
traditional public school
students to charter school
students but, due to issues of
endogeneity in student
populations (differences in
average student traits across
the two sectors), they do not
lean heavily on their OLS
analyses.
Treatment Effect Negative. The authors
find that charters open
for 1-3 tend to have
statistically significant
negative effects on
students’ composite
math and reading scores.
Charters open for 4 or
more years do not
appear to have any
significant effect on
students’ composite
scores. This is true for
both OLS and first-
difference models. When
adding school-level
fixed-effects, the authors
no longer find significant
negative effects for
charter schools
operating for 3 years.
The authors also find
that charter parents are
statistically significantly
more likely to remove
their child from a charter
school based on negative
performance
information than are
parents whose children
are in traditional public
schools.
Charter
schools
CREDO
Report (2013)
27 States 2006-2011 Grades K-12 Utilize a propensity-score
matching system, what they
term a Virtual Control
Record (VCR), to match a
charter school student with a
Treatment Effect Mixed. The authors
report mixed results in
charter schools’ effects
on math and reading
scores. They find that
275
similar traditional public
school student in the same
school district. The authors
then contrast the
standardized assessment
score growth of charter
school students with their
VCR “twins” to determine
charter effectiveness.
the bulk of charter
schools exhibited no
significant effect relative
to traditional public
schools, a quarter to a
third (25% to 29% for
reading and math,
respectively)
outperformed traditional
public schools, and
about a quarter to a third
(19% to 31% for reading
and math, respectively)
under-performed relative
to traditional public
schools. The authors do
not control for
variations in school-level
characteristics in this
analysis.
Charter
schools
CREDO
Report (2015)
41 School
Districts
across the
U.S. (22
States)
2006-2012 Grades K-12 Utilize a propensity-score
matching system, what they
term a Virtual Control
Record (VCR), to match a
charter school student with a
similar traditional public
school student in the same
school district. The authors
then contrast the
standardized assessment
score growth of charter
students with their VCR
“twins” to determine charter
effectiveness.
Treatment Effect Mixed. The authors find
that in 25 out of 41
school districts, charter
schools significantly raise
math scores relative to
traditional public
schools; yet in 9 out of
41 school districts,
charter schools
significantly decrease
math scores relative to
traditional public
schools. The distribution
for reading scores is very
similar. Again, the
authors do not control
for variations in school-
level characteristics in
this analysis.
Charter
schools
Sass (2006) State of
Florida
1999-2002 Grades 3-10 For testing the effect of
charter school enrollment on
Both Mixed. Sass concludes
that although first-year
276
students, the author uses a
student-level fixed-effects
panel dataset to estimate
changes in standardized
assessment scores for
students moving from
traditional public schools to
charter schools and vice
versa.
For testing the competitive
effect of charter schools’
presence on traditional public
schools’ performance levels,
Sass estimates the association
between: (a) measures of
competition between charters
and traditional public schools
(number of charter schools
and enrollment share of
charter schools in 2.5, 5, and
10 mile radii of each
traditional public school) and
gains in traditional public
school students’ standardized
assessment scores.
charter schools produce
significantly lower test
scores relative to
traditional public
schools, charter schools
achieve parity with
traditional public schools
in subsequent years per
both math and reading
scores. Sass finds that
charter schools serving
special education
students significantly
decrease those students’
score growth.
Furthermore, Sass
reports a significant and
positive association with
the number of charter
schools that lie within
2.5 miles of a traditional
public school and that
traditional public
school’s performance (p.
24). The results are
relatively robust to
expanding the measure
of charter proximity and
considering private
schools.
Charter
schools
Clark, Gleason,
Tuttle, &
Silverberg
(2015)
33 charter
middle
schools (13
States)
2005-2009 Grades 4-8 The authors estimate the
effect of charter schools on
standardized assessment
scores 2 years later (reading
and math scores) for students
who won lotteries to attend
charter middle schools
relative to students who lost
lotteries to attend charter
middle schools. The authors
then average the effects they
Treatment Effect Mixed. The authors fail
to find broad evidence
that attendance at a
charter school
significantly effects
students’ test scores.
They find a significant
and positive effect on
FRPL students’ math
scores 2 years after
enrollment, but also find
277
find across all charter schools
to develop an estimate of the
grand mean effect, while also
comparing the fixed-effect
estimate for each charter
school.
significant and negative
effects on non-FRPL
students’ math and
reading scores 2 years
after enrollment as well
as math scores 1 year
after enrollment. For
other significant results
by sub-group, see
paper’s Table 4.
Charter
schools
Imberman
(2011a)
School
District in
American
Southwest
1994-2007 Grades 1-12 The author uses a student-
level fixed-effects panel
dataset to estimates the
effects of students’ switching
from traditional public
schools to charter schools.
As dependent variables, the
authors uses both traditional
academic outcomes
(standardized assessment
scores in math, reading, and
language) as well as “non-
cognitive” outcomes
(attendance rates, discipline
incidences). The author
divides analysis across types
of charter school (start-up
versus conversion from
traditional public school) as
well as years the charter
school has been operating,
suspecting that conversion
charter schools and new
charter schools have different
effects than start-up and
experienced charter schools,
respectively. The author also
analyzes differences in
estimated effects by
elementary, middle, and high
schools.
Treatment Effect Mixed. The author finds
both significant negative
and positive effects of
charter schools,
depending on the model
specification. The author
finds mixed results of
charter schools on
traditional academic
outcomes. Notably, the
author more consistently
finds significant positive
effects of charter schools
on “non-cognitive”
outcomes. However, the
author finds that these
non-cognitive benefits
(improved attendance
and lower rates of
discipline incidents) tend
to disappear quickly for
students returning to
traditional public
schools, complicating
the notion that charters
convey long-term
benefits.
278
Charter
schools
Imberman
(2011b)
School
District in
American
Southwest
1993-2005 Grades 1-12 The author uses an
instrumental variable
approach to account for the
potentially endogenous
presence of charter schools
near traditional public
schools (instrument is
effectively the amount of
commercial space available
near a traditional public
school). Using a student-level
fixed-effects panel dataset,
the author then estimates the
relationship between a
measure of the competition a
traditional public school
experiences from charter
schools (a measure of the
enrollment share captured by
charters within 1 or 1.5
miles) and changes in the
standardized assessment
scores as well as attendance
and discipline outcomes for
that traditional public
school’s students. The author
breaks down his analysis to
understand how the effects
may vary based on how long
a traditional public school
has been exposed to
competition from a charter
school/charter schools, as
well as by student grade
levels and sociodemographic
characteristics.
Spillover Effect on
Traditional Public
School System
Mixed. The author finds
both significant negative
and positive effects of
charter schools,
depending on the model
specification. For
instance, the author
finds significant negative
effects of charter school
competition on
traditional public school
students’ math and
language scores; when
analyzing by grade levels,
the author finds these
negative effects to be
concentrated in
elementary school
grades. Nevertheless, the
author finds that
competition from
charter schools
significantly reduces
traditional public
schools’ rates of student
discipline for middle and
high school students.
The author also finds
that the negative effects
of charter schools on
traditional elementary
school students’
standardized assessment
scores dissipate once the
traditional public school
has experienced charter
competition for multiple
years. Additionally, the
author finds significantly
negative effects for
279
female students and
black American students.
Charter
schools
Winters (2012) New York
City, NY
2005-2009 Grades 3-8 The authors constructs a
fixed-effects panel dataset to
estimate the relationship
between the share of a
student’s peers leaving for
charter schools in a given
year and the change in
his/her standardized
assessment scores in math
and reading (ELA) from that
year to the next. The a uthor
tests how different scales of
fixed-effects (at the school,
joint school-student, and
student levels) result in
different findings as well as
how the relationship changes
based on regressions
including data for only
certain sociodemographic
sub-groups of traditional
public school students.
Spillover Effect on
Traditional Public
School System
Positive. Regressions
using student-level fixed
effects indicate a
significant and positive
relationship between the
share of a student’s peers
who leave for public
charter schools and the
change in that student’s
standardized assessment
scores – both math and
reading – between that
and the next year (no
significance when using
school or joint student-
school fixed-effects).
When the authors run
student-level fixed
effects for
sociodemographic sub-
groups, they find a
significant effect for
Latino/a students’ math
scores only as well as for
black American and
Latino/a students’
reading scores.
Charter
schools
Booker,
Gilpatric,
Gronberg, &
Jansen (2008)
State of
Texas
1993-2004 Grades 3-8 The authors estimate the
effect of charter schools’
presence, experienced by
individual traditional public
schools and by traditional
public school districts, on
standardized assessment
scores in math and reading.
Their measures for
competition experienced by
individual traditional public
schools are the number of
Spillover Effect on
Traditional Public
School System
Positive. The authors
report that when
separate regressions are
run, one set with
individual school
competitive measures
and the other set with
district-level competitive
measures, each set
indicates significant and
positive effects of
charter schools on
280
charter schools within 5
miles of a traditional public
school campus and number
of charter schools within 6-
10 miles of a traditional
public school campus. Their
measure for competition at
the district level is the share
of district students attending
charter schools. The authors
additionally explore effects
for sociodemographic sub-
groups and traditional public
schools’ historical
standardized assessment
performance levels.
traditional public school
students’ math and
reading scores. When
running regressions that
consider
sociodemographic sub-
groups and traditional
public schools’ historical
performance levels, the
authors find significant
and positive effects for
black American and
Latino/a in traditional
public schools whose
historical performances
are in the bottom half of
the historical
performance
distribution.
Charter
schools
Abdulkadiroğlu,
Angrist,
Dynarski,
Kane, & Pathak
(2011)
Boston, MA 2002-2008 Grades 6-8
and Grade
10
The authors estimate the
effect of number of years
enrolled in a charter school
on students’ standardized
assessment scores in ELA
and math (and writing for
high school students) for
students who won lotteries to
charter schools relative to
students who lost lotteries to
charter schools. As with
other lottery regressions, the
authors use a two-stage least
squares where the instrument
is lottery victory.
Treatment Effect Positive. The authors
report a statistically
significant, positive
effect for years of
charter attendance for
middle school students
who won charter
lotteries, for both ELA
and math standardized
assessment scores. The
authors also tend to find
statistically significant
and positive effects from
charter school
attendance on high
school students’
assessment scores (see
paper’s Table IV).
Charter
schools
Hoxby,
Murarka, &
Kang (2009)
New York
City, NY
2000-2008 Grades 3-8 The authors estimate the
effect of number of years
enrolled in a charter school
on students’ standardized
Treatment Effect Positive. The authors
report a statistically
significant and positive
effect for years of
281
assessment scores in math
and English, as well as on
New York Regents Exam
scores, for students who won
lotteries to charter schools
relative to students who lost
lotteries to charter schools.
charter attendance for
students who won
charter lotteries, for
math and English
standardized assessment
scores and Regents
Exam scores. The
authors do not find
differences across socio-
demographic groups.
Furthermore, the
researchers find that
gains made in charter
schools persist for
students who leave the
charter schools for
traditional public
schools.
282
283
Appendix B: Results of Principal Component Analysis, Changes in Census Tracts 2000-2017
Component eigenvalue difference proportion cumulative
C1 3.4473 2.5338 0.6895 0.6895
C2 0.9135 0.5189 0.1827 0.8722
C3 0.3945 0.2315 0.0789 0.9511
C4 0.1630 0.0814 0.0326 0.9837
C5 0.0816 n/a 0.0163 1.0000
Dimension of change for census tract, 2000-17 C1 C2 C3 C4 C5
white alone, non-Hispanic population share 0.4930 0.2580 -0.3786 0.0386 -0.7386
white alone, non-Hispanic school-age population share 0.3371 0.7243 0.5617 0.0905 0.1949
share of population with at least a bachelor's degree 0.4905 0.0626 -0.5768 -0.1227 0.6385
growth in median household income 0.4530 -0.4117 0.3963 -0.6795 -0.0802
tract’s percentile rank for median household income 0.4444 -0.4851 0.2265 0.7166 0.0485
Sources: 2000 total population comes from 2000 Census Table SF1/P001; 2000 white alone, non- Hispanic total and
school-age population come from 2000 Census Table SF1/P012I; 2000 school-age population comes from 2000 Census
Table SF1/DP1; 2000 share of population with at least a bachelor’s degree comes from 2000 Census Table SF3/DP2;
and 2000 median household income comes from 2000 Census Table SF3/DP3. 2017 total population comes from 2017
5-year ACS Table B01003; 2017 white alone, non- Hispanic total and school-age population come from 2017 5-year ACS
Table B01001H; 2017 school-age population comes from 2017 5-year ACS Table B01001; 2017 share of population with
at least a bachelor’s degree comes from 2017 5-year ACS Table S1501; and 2017 median household income comes from
2017 5-year ACS Table S1901.
Note: Adjustments for inflation made using the U.S. Bureau of Labor Statistics’ Consumer Price Index (CPI) for All
Urban Consumers in the Baltimore-DC Metropolitan Statistical Area.
284
Appendix C: Constructing a Longitudinal Dataset of Charter Campus Locations
Types of resources consulted include:
• documents published by the District of Columbia Public Charter School Board (e.g., annual
reports that list charter school campuses operating in a given school year, charter review and
renewal reports, etc.);
• campus-level, annual enrollment audit reports released by District of Columbia Office of the
State Superintendent of Education;
• charter school directories provided by the District of Columbia Office of the State
Superintendent of Education and a charter school financial management firm;
• charter school facility reports from the District of Columbia Office of the Deputy Mayor for
Education;
• charter school websites;
• procurement solicitations released by the District of Columbia Office of Contracting and
Procurement or reported in the District of Columbia Register;
• federal government reports on charter schools in the District of Columbia, including
directories from the U.S. Department of Education and publications from the federal
Government Accountability Office;
• charter school facilities and purchase order rosters from the District of Columbia Office of
the Chief Financial Officer;
• hearing minutes from a session of the District of Columbia Neighborhood Council 6A
Committee on Economic Development and Zoning Committee;
• an academic journal article that identifies charter school campuses in the District o f
Columbia (Özek, 2011);
• Washington Post news articles; and
• District of Columbia neighborhood-based or community-based newspaper articles,
newsletters, and blog posts.
Charter schools operating in SY1996-97
Gaines, P., & Ly, P. (1996, August 16). New char ter schools race opening bell. The
Washington Post. https://www.washingtonpost.com/archive
Charter schools operating in SY1997-98
Asim, J. (1997, April 6). Mixed media: Artists in the classroom. The Washington Post.
https://www.washingtonpost.com/archive
DC Public Charter School Board. (2014). Annual report: school year 2013-14 [for Community
Academy Public Charter Schools].
http://www.dcpcsb.org/sites/default/files/CCPCSfinal.pdf
Gottfredson, G. D. (2013). Changing course: preventing gang membership (Report No. NCJ
243471). Washington, DC: U.S. Department of Justice and U.S. Department of
Health and Human Services.
285
Manno, B. V. (1997, August 31). Charting a new course. The Washington Post.
https://www.washingtonpost.com/archive
Charter schools operating in SY1998-99
Charry, R. (1998, December 21). Charter school accountability lacking. The Common
Denominator. http://www.thecommondenominator.com/122198_news2.html
Charry, R. (1999, May 3). Charter schools put on probation. The Common Denominator.
http://wwww.thecommondenominator.com/050399_news1.html
District of Columbia Office of the Chief Financial Officer. (2000). Public charter schools (GC0)
[fiscal year 2001 budget proposal].
http://app.cfo.dc.gov/budget/budget/pdf/gc.pdf
Mathews, J. (1998, September 9). Charter schools open to high demand. The Washington Post,
B3.
Washington Post Staff. (1998, July 17). Metro in brief.
https://www.washingtonpost.com/archive
Wilgoren, D., & Strauss, V. (1999, November 7). Curriculum of confidence: On the learning
curve. The Washington Post, A1, C1, C8, C9.
Charter schools operating in SY1999-00
Charry, R. (1998, December 21). Charter school accountability lacking. The Common
Denominator. http://www.thecommondenominator.com/122198_news2.html
Charry, R. (1999, August 23). 11 new charter schools set to open. The Common Denominator.
http://www.thecommondenominator.com/082399_news7.html
District of Columbia Office of the Chief Financial Officer. (2000). Public charter schools (GC0)
[fiscal year 2001 budget proposal].
http://app.cfo.dc.gov/budget/budget/pdf/gc.pdf
McDowell, L. M., & Sietsema, J. P. (2002, March). Directory of public elementary and secondary
education agencies, 1999-2000 (National Center for Education Statistics No. NCES-
2002-314). U.S. Department of Education.
Wilcoren, D., & Strauss, V. (1999, November 7). Curriculum of confidence: on the learning
curve. The Washington Post, A1, C1, C8, C9.
Charter schools operating in SY2000-01
District of Columbia Office of the Chief Financial Officer. (2000). Public charter schools (GC0)
[fiscal year 2001 budget proposal].
http://app.cfo.dc.gov/budget/budget/pdf/gc.pdf
Lang, L. (2001, February 2). Learning the hard way. Washington City Paper.
https://www.washingtoncitypaper.com
McDowell, L. M., & Sietsema, J. P. (2002, November). Directory of public elementary and secondary
education agencies, 2000-2001 (National Center for Education Statistics No. NCES-
2003-310). U.S. Department of Education.
Charter schools operating in SY2001-02
McDowell, L. M., & Sietsema, J. P. (2003, September). Directory of public elementary and secondary
education agencies, 2001-02 (National Center for Education Statistics No. NCES-2003-
351). U.S. Department of Education.

286

Charter schools operating in SY2002-03
McDowell, L. M., & Sietsema, J. P. (2005, March). Directory of public elementary and secondary
education agencies, 2002-03 (National Center for Education Statistics No. NCES-2005-
315). U.S. Department of Education.
U.S. Government Accountability Office. (2003, September). New charter schools across the county
and in the District of Columbia face similar start-up challenges (GAO No. GAO-03-899).
https://www.gao.gov

Charter schools operating in SY2003-04
DC Public Charter School Board. (2005, August). Annual report: academic year 2004-2005.
https://www.dcpcsb.org/sites/default/files/report/2005%20Annual%20Report%2
0%281%29.pdf
District of Columbia Register. (2003, September 12). Invitation to bid [for Community
Academy Public Charter Schools].
http://dcregisterarchives.dc.gov/sites/default/files/dc/sites/OS/release_content/at
tachments/14747/07%20BOARDS%20COMMISSIONS%20AND%20AGENCIE
S.pdf
Jair Lynch Real Estate Partners. (n.d). Kingsman Academy Public Charter School.
https://www.jairlynch.com/projects/kingsman-academy-public-charter-school-3/

Charter schools operating in SY2004-05
DC Public Charter School Board. (2005, August). Annual report: academic year 2004-2005.
https://www.dcpcsb.org/sites/default/files/report/2005%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2006, July 30). Annual report: 2006.
https://www.dcpcsb.org/sites/default/files/report/2006%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2012, January 24). 2012-13 charter renewal report: Carlos
Rosario International Public Charter School.
https://www.dcpcsb.org/sites/default/files/data/images/rosario%20renewal%20fi
nal%20report.pdf
District of Columbia Office of the State Superintendent of Education. (2005). 2004
attachment 9 – summary of residency verification by school.
https://osse.dc.gov/publication/2004-attachment-9-summary-residency-
verification-school
District of Columbia Register. (2004, July 23). Invitation to bid [for Cesar Chavez Public
Charter Middle School for Public Policy].
http://dcregisterarchives.dc.gov/sites/default/files/dc/sites/OS/release_content/at
tachments/14463/09%20BOARDS%20COMMISSIONS%20AND%20AGENCIE
S.pdf
Jair Lynch Real Estate Partners. (n.d). Kingsman Academy Public Charter School.
https://www.jairlynch.com/projects/kingsman-academy-public-charter-school-3/
U.S. Senate. (2005, May 25). Special hearing before a Subcommittee of the Committee on Appropriations
Unites States Senate, one hundred ninth Congress, first session: conditions and improvements to the
District of Columbia public school system (Senate Hearing 109-285). Washington, DC: U.S.
Government Printing Office. https://www.govinfo.gov/content/pkg/CHRG-
109shrg24940/html/CHRG-109shrg24940.htm

287
Charter schools operating in SY2005-06
ANC 6A Economic Development & Zoning Committee. (2006, February 28). Agenda.
http://anc6a.org/wp-content/uploads/EDZAFeb2806.pdf
DC Public Charter School Board. (2005, August). Annual report: academic year 2004-2005.
https://www.dcpcsb.org/sites/default/files/report/2005%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2006, July 30). Annual report: 2006.
https://www.dcpcsb.org/sites/default/files/report/2006%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2007, July 30). Annual report: 2007.
https://www.dcpcsb.org/sites/default/files/report/2007%20Annual%20Report%2
0%281%29.pdf
District of Columbia Office of the State Superintendent of Education. (2006). 2005 AUDIT
attachment 6 – summary of students for whom enrollment and residency were verified by school and
grade. https://osse.dc.gov/publication/2005-audit-attachment-6-summary-students-
whom-enrollment-and-residency-were-verified
Marks, J. (2005, December 11). DCPS, charter schools in Washington DC.
https://productforums.google.com/d/topic/gec-students/v59SXxLLUd8
Charter schools operating in SY2006-07
DC Public Charter School Board. (2006, July 30). Annual report: 2006.
https://www.dcpcsb.org/sites/default/files/report/2006%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2007, July 30). Annual report: 2007.
https://www.dcpcsb.org/sites/default/files/report/2007%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2008, July 30). Annual report: 2008.
https://www.dcpcsb.org/sites/default/files/report/2008%20Annual%20Report%2
0%281%29.pdf
District of Columbia Office of the State Superintendent of Education. (2007). 2006 AUDIT
attachments 1-4. https://osse.dc.gov/publication/2006-audit-attachments-1-4
District of Columbia Office of the State Superintendent of Education. (n.d). NSLP public
release for SY 2006-2007.
https://osse.dc.gov/sites/default/files/dc/sites/osse/release_content/attachments/
10161/NSLP%20Schools%202006-2007.pdf
Labbé, T. (2006, August 24). Six charter schools opening with unique outlooks studies
include Latin and ESL. The Washington Post.
https://www.washingtonpost.com/archive
Charter schools operating in SY2007-08
DC Public Charter School Board. (2007, July 30). Annual report: 2007.
https://www.dcpcsb.org/sites/default/files/report/2007%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2008, July 30). Annual report: 2008.
https://www.dcpcsb.org/sites/default/files/report/2008%20Annual%20Report%2
0%281%29.pdf

288

DC Public Charter School Board. (2009, July 30). Annual report: 2009.
https://www.dcpcsb.org/sites/default/files/report/2009%20Annual%20Report%2
0%281%29.pdf
District of Columbia Office of the State Superintendent of Education. (2008). 2007 audit
attachments 1-4. https://osse.dc.gov/publication/2007-audit-attachments-1-4

Charter schools operating in SY2008-09
DC Public Charter School Board. (2008, July 30). Annual report: 2008.
https://www.dcpcsb.org/sites/default/files/report/2008%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2009, July 30). Annual report: 2009.
https://www.dcpcsb.org/sites/default/files/report/2009%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2010, July 30). Annual report: 2010.
https://www.dcpcsb.org/sites/default/files/report/2010%20Annual%20Report%2
0%281%29.pdf
District of Columbia Office of the State Superintendent of Education. (2009). 2008
attachment 9 summary of residency verification by school.
https://osse.dc.gov/sites/default/files/dc/sites/osse/publication/attachments/200
8%20Attachment%209%20Summary%20of%20Residency%20Verification%20by%
20School.pdf
Ozek, U. (2011). Public school choice in the District of Columbia: A descriptive analysis (Brief 13).
National Center for Analysis of Longitudinal Data in Education Research.

Charter schools operating in SY2009-10
DC Public Charter School Board. (2009, July 30). Annual report: 2009.
https://www.dcpcsb.org/sites/default/files/report/2009%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2010, July 30). Annual report: 2010.
https://www.dcpcsb.org/sites/default/files/report/2010%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2011, July 30). Annual report: 2011.
https://www.dcpcsb.org/sites/default/files/report/2011%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2009, December 23). Press release: City Collegiate Public
Charter School relinquishes charter, Public Charter School Board approves other requests &
proposals. https://www.dcpcsb.org/city-collegiate-public-charter-school-relinquishes-
charter
District of Columbia Office of the State Superintendent of Education. (2010). 2009
attachments 1-13.
https://osse.dc.gov/sites/default/files/dc/sites/osse/publication/attachments/200
9_attachments_1_13.pdf
Hyde Schools. (2010, December 13). Hyde Leadership Public Charter School in DC to hold ribbon-
cutting, grand opening ceremony for new building.
https://www.hyde.edu/blog/2010/12/13/hyde-leadership-public-charter-school-in-
dc-to-hold-ribbon-cutting-grand-opening-ceremony-for-new-building/

289
Charter schools operating in SY2010-11
DC Public Charter School Board. (2010, July 30). Annual report: 2010.
https://www.dcpcsb.org/sites/default/files/report/2010%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2011, July 30). Annual report: 2011.
https://www.dcpcsb.org/sites/default/files/report/2011%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2012, July 30). Annual report: 2012.
https://www.dcpcsb.org/sites/default/files/report/2012%20Annual%20Report%2
0%281%29.pdf
Charter schools operating in SY2011-12
DC Public Charter School Board. (2011, July 30). Annual report: 2011.
https://www.dcpcsb.org/sites/default/files/report/2011%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2012, July 30). Annual report: 2012.
https://www.dcpcsb.org/sites/default/files/report/2012%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2013). Annual report: 2013.
https://www.dcpcsb.org/sites/default/files/report/2013%20Annual%20Report%2
0%281%29.pdf
Charter schools operating in SY2012-13
DC Public Charter School Board. (2012, July 30). Annual report: 2012.
https://www.dcpcsb.org/sites/default/files/report/2012%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2013). Annual report: 2013.
https://www.dcpcsb.org/sites/default/files/report/2013%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2014). Annual report: 2014.
https://www.dcpcsb.org/sites/default/files/report/2014%20Annual%20Report%2
0%281%29.pdf
Staff Reports. (2012, August 20). D.C. charter school start dates – 2012. The Washington Post.
https://www.washingtonpost.com/local
Charter schools operating in SY2013-14
DC Public Charter School Board. (2013). Annual report: 2013.
https://www.dcpcsb.org/sites/default/files/report/2013%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2014). Annual report: 2014.
https://www.dcpcsb.org/sites/default/files/report/2014%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2015). On the road to success: 2015.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202015%20Annual
%20Report%20%281%29.pdf
290
Charter schools operating in SY2014-15
DC Public Charter School Board. (2014). Annual report: 2014.
https://www.dcpcsb.org/sites/default/files/report/2014%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2015). On the road to success: 2015.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202015%20Annual
%20Report%20%281%29.pdf
DC Public Charter School Board. (2016). 2016 annual report.
https://www.dcpcsb.org/sites/default/files/report/2016.07.27-dcpcsb-annual-
report-single-page%20%281%29.pdf
District of Columbia Deputy Mayor for Education. (2014). Appendix B: public charter facility
data sheet, SY2014-15 [2014 Master Facilities Plan Annual Supplement].
https://dme.dc.gov/sites/default/files/dc/sites/dme/publication/attachments/Ap
pendix%20B-%20Public%20Charter%20Facility%20Data%2C%20SY14-15.pdf
Charter schools operating in SY2015-16
DC Public Charter School Board. (2015). On the road to success: 2015.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202015%20Annual
%20Report%20%281%29.pdf
DC Public Charter School Board. (2016). 2016 annual report.
https://www.dcpcsb.org/sites/default/files/report/2016.07.27-dcpcsb-annual-
report-single-page%20%281%29.pdf
DC Public Charter School Board. (2017). 2017 annual report.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202017%20Annual
%20Report%20.pdf
District of Columbia Deputy Mayor for Education. (2015). Appendix B: SY15-16 public charter
school facility data [2015 Master Facilities Plan Annual Supplement].
https://dme.dc.gov/sites/default/files/dc/sites/dme/publication/attachments/Cop
y%20of%20SY15-16%20MFP%20Public%20Charter%20School%20DataPCSB.pdf
District of Columbia Office of the State Superintendent of Education. (n.d). School directory for
school year 2015 – 2016.
https://sledtraining.osse.dc.gov/vPage/SchoolDirectory/298/91189?grda3d73920-
e561-4c76-8832-25f8731fd1d8-page=3&grda3d73920-e561-4c76-8832-
25f8731fd1d8-sort=Telephone_Number-asc
Charter schools operating in SY2016-17
DC Public Charter School Board. (2016). 2016 annual report.
https://www.dcpcsb.org/sites/default/files/report/2016.07.27-dcpcsb-annual-
report-single-page%20%281%29.pdf
DC Public Charter School Board. (2017). 2017 annual report.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202017%20Annual
%20Report%20.pdf
DC Public Charter School Board. (2018). 2018 annual report.
https://www.dcpcsb.org/sites/default/files/report/42329%20DCPCSB%202018%
20Annual%20Report-8.5.18.pdf
District of Columbia Deputy Mayor for Education. (2016). Appendix B: PCS SY 2016-17
enrollment data [2016 Master Facilities Plan Annual Supplement].
291
https://dme.dc.gov/sites/default/files/dc/sites/dme/publication/attachments/Ap
pendix%20B%20-%20PCS%20SY2016-17%20Enrollment%20Data_0.xlsx
District of Columbia Office of the State Superintendent of Education. (2017). 2016-17 school
year school-by-school enrollment audit UPSFF data.
https://osse.dc.gov/sites/default/files/dc/sites/osse/publication/attachments/201
6-17%20School%20Year%20School-by-
School%20Enrollment%20Audit%20UPSFF%20Data_0.xlsx
Charter schools operating in SY2017-18
DC Public Charter School Board. (2017). 2017 annual report.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202017%20Annual
%20Report%20.pdf
DC Public Charter School Board. (2018). 2018 annual report.
https://www.dcpcsb.org/sites/default/files/report/42329%20DCPCSB%202018%
20Annual%20Report-8.5.18.pdf
Charter schools operating in SY2018-19
DC Public Charter School Board. (2018). 2018 annual report.
https://www.dcpcsb.org/sites/default/files/report/42329%20DCPCSB%202018%
20Annual%20Report-8.5.18.pdf
DC Public Charter School Board. (2018). 2018-19 school directory.
https://www.dcpcsb.org/sites/default/files/report/2018-
19%20School%20Directory.pdf
Charter schools operating in SY2019-20
My School DC. (2019). School finder SY 2019-20. http://find.myschooldc.org/
Charter schools operating in SY2020-21
My School DC. (2020). School finder SY 2020-21. http://find.myschooldc.org/
Sources documenting multiple years of charter school operation
21
st
Century School Fund. (n.d). Untitled [list of traditional public and public charter school
facilities]. http://www.21csf.org/csf-home/DocUploads/DataShop/DS_342.xlsx
21
st
Century School Fund. (n.d). Untitled [list of traditional public and public charter school
facilities]. http://www.21csf.org/csf-home/DocUploads/DataShop/DS_420.xlsx
Center for Education Reform. (2011, December). Appendix D. Closed charter schools by state:
national data 2011. https://www.edreform.com/wp-
content/uploads/2011/12/CER_FINALClosedSchools2011-1.pdf
DC Public Charter School Board. (2016, February). Beating the odds.
https://www.dcpcsb.org/sites/default/files/report/AA%20Report%20V4.pdf
DC Public Charter School Board. (2018, September). All expanded public charter campuses.
https://www.dcpcsb.org/sites/default/files/All%20Expanded%20Public%20Charte
r%20Campuses_0918.xls
District of Columbia Deputy Mayor for Education. (2017, July). Citywide landscape of formerly
closed DCPS facilities.
https://dme.dc.gov/sites/default/files/dc/sites/dme/publication/attachments/City
wide%20Landscape%20of%20Formerly%20Closed%20DCPS%20Facilities_Append
ix_Updated%20July2017_0.xlsx
292
District of Columbia Office of the State Superintendent of Education. (n.d). Public charter
school historic enrollment audit information 2001 – SLED [through Fall 2009].
https://sled.osse.dc.gov/DOCS/PCS_School-by-Grade-Enrollment-
Audited_and_Oct_Cert_2001-2009.xls
Henig, J., Holyoke, T. T., Lacireno-Paquet, N., & Moser, M. (2001, February). Growing pains:
an evaluation of charter schools in the District of Columbia, 1999-2000 (ERIC Report No.
451-303). https://files.eric.ed.gov/fulltext/ED451303.pdf
Mead, S. (2005, October 12). Capital campaign: Early returns on District of Columbia charter schools.
Progressive Policy Institute. https://files.eric.ed.gov/fulltext/ED491209.pdf
U.S. Government Accountability Office. (2005, November 17). D.C. charter schools:
strengthening monitoring and process when schools close could improve accountability and ease
student transitions (GAO No. GAO-06-73). https://www.gao.gov
Sources documenting timeline for single charter local educational agency
DC Public Charter School Board. (2018, March 19). 2017-18 twenty-year charter review report:
Carlos Rosario International Public Charter School.
https://www.dcpcsb.org/sites/default/files/report/2018-04-
02%20Carlos%20Rosario%2020%20Year%20Review%20Report%20SY17-18.pdf
Friendship Public Charter Schools. (n.d). 20 years of achievement: A timeline.
https://www.friendshipschools.org/about/history/
Office of the District of Columbia Auditor. (2004, May 17). Review of the financial operations of
the Village Learning Center Public Charter School. https://dcauditor.org/wp-
content/uploads/2018/07/DCA1304.pdf
KIPP DC. (n.d). Our story. https://www.kippdc.org/about/our-story/
Latin American Montessori Bilingual Public Charter School. (n.d). Our school: History.
https://www.lambpcs.org/history
Lerner, M. (2017, October 17). Exclusive interview with Daniela Anello, head of school DC Bilingual.
Parents Have School Choice Kids Win.
https://parentshaveschoolchoicekidswin.com/2017/10/17/exclusive-interview-
with-daniela-anello-head-of-school-dc-bilingual/
Maya Angelou Schools. (n.d). Our beginnings. https://www.seeforever.org/about-foundation/
Rich, W. (2014, May 4). South by west. Capital Community News.
http://www.capitalcommunitynews.com/content/south-west-2
Shin, A. (2002, September 20). Cracking the books. Washington City Paper.
https://www.washingtoncitypaper.com/news/article/13024991/cracking-the-books
Untitled [Eagle Academy Public Charter School application for U.S. Department of
Education PR/Award # U282M170038]. (2017).
https://innovation.ed.gov/files/2017/09/eagleacademypubliccharterNAR.pdf
Washington Latin Public Charter School. (2016, August). Family handbook.
http://www.dcpcsb.org/sites/default/files/report/2016-
2017%20Student%20Handbook%280WFC%29%28WashinLatinPCS%28LEA%29
%29.pdf
293
Appendix D: Constructing a Dataset of Charter School Types
I consulted each of the following resources in assigning a type – and when necessary, a
secondary type – to each charter LEA that operated in the District prior to school year 2018-19. I
have organized these resources by time of publication, rather than alphabetization, to demonstrate
coverage of charter LEAs across the school years in my dataset. I note that I also considered the
legal name of each charter LEA, as this is an important component of a charter LEA’s “brand” and
frequently communicates the core philosophy, curricular focus, or main educational programming
offered by the charter LEA (e.g., Shining Stars Montessori Academy PCS or The School for Arts in
Learning PCS), or the specific population the charter LEA intends to serve (e.g., Academy of Hope
Adult PCS).
Determining the types of District charter schools and classifying them accordingly is an
iterative and inexact exercise. To begin, I reviewed the A+ Colorado report to understand the
typologies considered by other researchers in the field. Next, I defined a preliminary typology of
District charter schools by cross-referencing an array of historical and present-day resources. For
school years 2006-07 and after, I relied heavily on annual charter sector reports and annual school
performance/quality reports from the District’s Public Charter School Board, which frequently
categorize charter LEAs or include statements from the charter LEAs themselves regarding their
instructional approach and models. For school years prior to 2006-07, I drew upon a greater variety
of resources, notably federal government documents and articles in local newspapers. The
preliminary typology I developed per these resources often aligned with the one developed by A+
Colorado (2018). This Appendix provides an exhaustive list of resources consulted.
I then used my preliminary typology along with historical documentation to classify each of the
charter LEAs in my longitudinal dataset; and as necessary, I modified my typology to ensure a better
fit with the data I was reviewing. At the end of this process, I identified eleven “types” of charter
schools that have operated or are currently operating within the District: Adult, Blended Learning,
Comprehensive, Creative Arts, Dual Language, Early Childhood, Global Culture, Learner Centered,
Postsecondary/Vocational, Specific Population, STEM.
I note that KIPP DC PCS and Friendship PCS, which are the largest charter LEAs in terms
of students enrolled and number of campuses, do offer more varied educational experiences at their
campuses than other District charter LEAs. Nonetheless, I assign each of these LEAs and their
campuses a uniform type for two reasons. First, each LEA’s campus-level diversification falls within
a broader brand of offering a “comprehensive” or rigorous education to all students. Second, these
LEAs are large enough that they operate elementary, middle, and high school campuses. And
because continuously enrolled students will naturally matriculate across these diversified campuses, it
is unlikely that households opt to attend KIPP DC or Friendship PCS because of multiple, specific
types of education available. In a way, these large charter LEAs have become increasingly similar to
District of Columbia Public Schools as they have grown, offering a somewhat diversified educational
experience but trying to remain able to serve all District families (i.e., a “something for everyone”
approach).
294
List of referenced resources, organized from earliest to most recent date of
publication
Manno, B. V. (1997, August 31). Charting a new course. The Washington Post.
https://www.washingtonpost.com/archive
Mathews, J. (1998, September 9). Charter schools open to high demand. The Washington Post,
B3.
Charry, R. (1998, December 21). Charter school accountability lacking. The Common
Denominator. http://www.thecommondenominator.com/122198_news2.html
Charry, R. (1999, August 23). 11 new charter schools set to open. The Common Denominator.
http://www.thecommondenominator.com/082399_news7.html
Wilgoren, D., & Strauss, V. (1999, November 7). Curriculum of confidence: On the learning
curve. The Washington Post, A1, C1, C8, C9.
U.S. House of Representatives. (2003, May 9). Hearing before the Committee on Government Reform
House of Representatives, one hundred eighth Congress, first session: in search of educational
excellence in the nation’s capital: a review of academic options for students and parents in the
District of Columbia (Hearing 108-30). Washington, DC: U.S. Government Printing
Office. https://archive.org/details/gov.gpo.fdsys.CHRG-108hhrg88196
U.S. Government Accountability Office. (2005, November 17). D.C. charter schools:
strengthening monitoring and process when schools close could improve accountability and ease
student transitions (GAO No. GAO-06-73). https://www.gao.gov
DC Public Charter School Board. (2005, August). Annual report: academic year 2004-2005.
https://www.dcpcsb.org/sites/default/files/report/2005%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2006, July 30). Annual report: 2006.
https://www.dcpcsb.org/sites/default/files/report/2006%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2007, July 30). Annual report: 2007.
https://www.dcpcsb.org/sites/default/files/report/2007%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2008, July 30). Annual report: 2008.
https://www.dcpcsb.org/sites/default/files/report/2008%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2009, July 30). Annual report: 2009.
https://www.dcpcsb.org/sites/default/files/report/2009%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2010, March). 2009 School performance reports.
http://www.dcpcsb.org/sites/default/files/data/images/pcsb_spr_2009_webfinal.p
df
DC Public Charter School Board. (2010, July 30). Annual report: 2010.
https://www.dcpcsb.org/sites/default/files/report/2010%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2011, July 30). Annual report: 2011.
https://www.dcpcsb.org/sites/default/files/report/2011%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2011, December). 2011 charter school performance reports.
http://www.dcpcsb.org/sites/default/files/data/images/pcsb%20book_dec1.pdf
295
DC Public Charter School Board. (2012, July 30). Annual report: 2012.
https://www.dcpcsb.org/sites/default/files/report/2012%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2013). 2012 DC public charter school performance reports.
http://www.dcpcsb.org/sites/default/files/report/2012-11-
15%20PMF%20Book%20SY11-12.pdf
DC Public Charter School Board. (2013). Annual report: 2013.
https://www.dcpcsb.org/sites/default/files/report/2013%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2013). 2013 DC public charter school performance reports.
https://dcpcsb.egnyte.com/dl/1GR2UjXUPf/
DC Public Charter School Board. (2014). Annual report: 2014.
https://www.dcpcsb.org/sites/default/files/report/2014%20Annual%20Report%2
0%281%29.pdf
DC Public Charter School Board. (2014, November). 2013-14 DC public charter school
performance reports. https://dcpcsb.egnyte.com/dl/hdsUo9pDbh/
DC Public Charter School Board. (2015). On the road to success: 2015.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202015%20Annual
%20Report%20%281%29.pdf
DC Public Charter School Board. (2016). 2016 annual report.
https://www.dcpcsb.org/sites/default/files/report/2016.07.27-dcpcsb-annual-
report-single-page%20%281%29.pdf
DC Public Charter School Board. (2016, November 22). 2015-16 DC public charter school
quality reports. http://www.dcpcsb.org/sites/default/files/report/2016-12-
12%20Final%202015-16%20PMF%20Book.pdf
DC Public Charter School Board. (2017). 2017 annual report.
https://www.dcpcsb.org/sites/default/files/report/DCPCSB%202017%20Annual
%20Report%20.pdf
DC Public Charter School Board. (2017, November 30). 2016-17 DC public charter school
quality reports. http://www.dcpcsb.org/sites/default/files/2017-12-
20%20Final%202016-17%20PMF%20Book%20%282%29.pdf
DC Public Charter School Board. (2018). 2018 annual report.
https://www.dcpcsb.org/sites/default/files/report/42329%20DCPCSB%202018%
20Annual%20Report-8.5.18.pdf
DC Public Charter School Board. (2018). School quality report [2018 reports].
https://www.dcpcsb.org/schoolquality
DC Public Charter School Board. (2019). Find a charter school [live charter school directory
function on the Public Charter School Board’s official website].
https://www.dcpcsb.org/find-a-school
My School DC. (2019). School finder SY 2019-2020. http://find.myschooldc.org/
My School DC. (2020). School finder SY 2020-2021. http://find.myschooldc.org/
Based on these resources, I assigned the following type(s) for each charter LEA (see next page).
Charter LEA name Primary type Secondary type Specific population
Global
culture
information
Academia Bilingue de la Communidad PCS (ABC) Dual language n/a n/a n/a
Academy for Learning Through the Arts PCS (ALTA) Creative arts n/a n/a n/a
Academy for Renewal in Education PCS (ARE) Specific population n/a
Neglected, abused, or
delinquent
n/a
Academy of Hope Adult PCS Adult n/a n/a n/a
Achievement Prep Academy PCS Postsec./voc. n/a n/a n/a
AppleTree PCS Early childhood n/a n/a n/a
Arts and Technology PCS Creative arts Comprehensive n/a n/a
Barbara Jordan PCS Comprehensive n/a n/a n/a
BASIS DC PCS Comprehensive n/a n/a n/a
Booker T. Washington PCS Postsec./voc. n/a n/a n/a
Breakthrough Montessori PCS Learner centered n/a n/a n/a
Bridges PCS Early childhood n/a n/a n/a
Briya PCS (form. Education Strengthens Families) Adult Specific population Adult and EL n/a
Capital City PCS Comprehensive Creative arts n/a n/a
Carlos Rosario International PCS Adult Specific population Adult and EL n/a
Cedar Tree Academy PCS (form. Howard Road Academy)***
Comprehensive /
Early childhood
Creative arts n/a n/a
Center City PCS Comprehensive n/a n/a n/a
Cesar Chavez PCS for Public Policy Postsec./voc. n/a n/a n/a
Children’s Studio PCS Creative arts n/a n/a n/a
City Arts & Prep PCS (form. William E. Doar Jr.) Creative arts n/a n/a n/a
City Collegiate PCS Comprehensive n/a n/a n/a
City Lights PCS Specific population n/a SPED and Alternative n/a
Community Academy PCS Comprehensive n/a n/a n/a
Community College Prep Academy PCS Adult n/a n/a n/a
Creative Minds International PCS Creative arts Global culture n/a non-specific
DC Bilingual PCS Dual language n/a n/a n/a
DC Prep PCS Postsec./voc. n/a n/a n/a
DC Scholars PCS Postsec./voc. n/a n/a n/a
Democracy Prep Congress Heights PCS Postsec./voc. n/a n/a n/a
296
District of Columbia International School PCS Dual language Global culture n/a non-specific
E.L. Haynes PCS Postsec./voc. n/a n/a n/a
Eagle Academy PCS Early childhood STEM n/a n/a
Early Childhood Academy PCS Early childhood STEM n/a n/a
Elsie Whitlow Stokes Community Freedom PCS Dual language n/a n/a n/a
Excel Academy PCS Specific population n/a Girls n/a
Friendship PCS Comprehensive n/a n/a n/a
Goodwill Excel Center PCS Adult n/a n/a n/a
Harmony DC PCS STEM n/a n/a n/a
Hope Academy PCS Postsec./voc. n/a n/a n/a
Hope Community PCS Comprehensive n/a n/a n/a
Hospitality High PCS (form. Marriot Hospitality) Postsec./voc. n/a n/a n/a
Howard University Middle School PCS STEM n/a n/a n/a
IDEA PCS Postsec./voc. n/a n/a n/a
Ideal Academy PCS Comprehensive STEM n/a n/a
Imagine Southeast PCS Comprehensive n/a n/a n/a
Ingenuity Prep PCS Comprehensive Early childhood n/a n/a
Inspired Teaching Demonstration PCS Comprehensive n/a n/a n/a
Jos-Arz PCS Specific population n/a SPED n/a
Kamit Institute for Magnificent Achievers PCS (KIMA) Global culture n/a n/a Afrocentric
Kingsman Academy PCS Specific population n/a
SPED, alternative, and
at-risk
n/a
KIPP DC PCS Comprehensive n/a n/a n/a
Latin American Montessori Bilingual PCS (LAMB) Dual language Learner centered n/a n/a
LAYC Career Academy PCS Adult Specific population Adult and EL n/a
Lee Montessori PCS Learner centered n/a n/a n/a
Marcus Garvey PCS Global culture Specific population At-risk Afrocentric
Mary McLeod Bethune PCS Dual language n/a n/a n/a
Maya Angelou PCS*** Adult/Specific Population n/a
At-risk and alternative
needs
n/a
Meld Even/Start PCS (MEI Futures) Specific population n/a Teenage mothers n/a
Meridian PCS Comprehensive n/a n/a n/a
297
Monument Academy PCS Specific population n/a
At-risk, foster care, and
homeless
n/a
Mundo Verde Bilingual PCS Dual language n/a n/a n/a
National Collegiate Prep PCS Postsec./voc. n/a n/a n/a
New School for Enterprise and Development PCS Postsec./voc. Specific population Low-income teenagers n/a
New Vistas PCS Specific population n/a At-risk n/a
Nia Community PCS Global culture n/a n/a Afrocentric
Options PCS Specific population n/a SPED and At-risk n/a
Paul PCS Comprehensive n/a n/a n/a
Perry Street Prep PCS (form. Hyde Leadership) Postsec./voc. n/a n/a n/a
Potomac Prepratory PCS (form. Potomac Lighthouse) Creative arts n/a n/a n/a
Richard Milburn PCS Specific population n/a At-risk n/a
Richard Wright PCS for Journalism and Media Arts Creative arts Postsec./voc. n/a n/a
Rocketship DC PCS Blended Learning n/a n/a n/a
Roots PCS Global culture n/a n/a Afrocentric
Sasha Bruce PCS Specific population n/a At-risk and SPED n/a
SEED PCS of Washington, D.C. Specific population n/a At-risk n/a
Sela PCS Dual language n/a n/a n/a
Septima Clark PCS Specific population n/a Black-American boys n/a
Shining Stars Montessori Academy PCS Learner centered n/a n/a n/a
Somerset Prep DC PCS Postsec./voc. n/a n/a n/a
SouthEast Academy PCS Postsec./voc. n/a n/a n/a
St. Coletta PCS Specific population n/a Special Education n/a
Sustainable Futures PCS Adult Specific population
At-risk and alternative
needs
n/a
Techworld PCS STEM n/a n/a n/a
The Children's Guild DC PCS Comprehensive n/a n/a n/a
The Next Step / El Proximo Paso PCS Adult Specific population At-risk, Adult, EL n/a
The School for Arts in Learning PCS (SAIL) Creative arts n/a n/a n/a
Thea Bowman Preparatory Academy PCS Postsec./voc. n/a n/a n/a
Thurgood Marshall Academy PCS Postsec./voc. n/a n/a n/a
Tree of Life PCS Comprehensive Global culture n/a Afrocentric
Tri-Community PCS Comprehensive n/a n/a n/a
298
Two Rivers PCS Comprehensive n/a n/a n/a
Village Learning Center PCS Comprehensive n/a n/a n/a
Washington Academy PCS Comprehensive n/a n/a n/a
Washington Global PCS Postsec./voc. Global culture n/a non-specific
Washington Latin PCS Dual language Comprehensive n/a n/a
Washington Leadership Academy PCS Comprehensive Postsec./voc. n/a n/a
Washington Mathematics Science & Technology PCS STEM n/a n/a n/a
Washington Yu Ying PCS Dual language n/a n/a n/a
World PCS Comprehensive n/a n/a n/a
Young America Works PCS Postsec./voc. n/a n/a n/a
Young Technocrats PCS STEM n/a n/a n/a
YouthBuild PCS Adult Specific population Adult and EL n/a
Notes:
1. In school years prior to 2013-14, Cedar Tree Academy PCS (formerly Howard Road Academy PCS) and its campuses are counted as Comprehensive; for school
year 2013-14 and after, they are counted as Early Childhood.
2. For school years prior to 2012-13, Maya Angelou PCS and its campuses are counted as Specific Population; for school year 2012 -13 and after, it is counted as
both an Adult and Specific Population LEA due to its Young Adult Learning Center being added to its existing Evans Campus.
299
300
Appendix E: Map of Landmarks and Neighborhoods in the District of Columbia
Note: Census tract boundaries traced in light gray; neighborhood names shown in black text; major geographic features
listed in bolded black text.
Appendix F: Patterns in Charter Sector Growth by School Type, School Year 2000-01 to 2017-18
School
year
Charter
LEAs
Adult
Blended
learning
Compre-
hensive
Creative
arts
Dual
language
Early
childhood
Global
culture
Learner
centered
Postsec.
/ voc.
Specific
population
STEM
1996-97 2 0 0 0 0 0 0 1 0 0 1 0
1997-98 4 0 0 1 1 0 0 1 0 0 1 0
1998-99 18 2 0 4 2 1 0 0 0 2 4 3
1999-00 27 2 0 6 3 1 0 1 0 6 6 2
2000-01 34 2 0 9 3 1 0 2 0 7 8 2
2001-02 36 2 0 11 3 1 0 2 0 8 7 2
2002-03 35 2 0 12 3 1 0 2 0 8 6 1
2003-04 37 2 0 12 3 2 1 2 0 9 5 1
2004-05 42 2 0 12 4 4 1 2 0 11 5 1
2005-06 52 3 0 14 6 5 4 2 0 10 6 2
2006-07 55 4 0 15 6 6 4 3 0 9 6 2
2007-08 56 4 0 15 6 6 4 3 0 9 7 2
2008-09 60 4 0 15 6 7 4 3 0 11 8 2
2009-10 57 4 0 14 6 6 4 3 0 12 6 2
2010-11 52 4 0 13 4 6 4 2 0 11 6 2
2011-12 53 4 0 14 4 7 4 1 1 10 6 2
2012-13 57 6 0 15 5 7 4 1 1 11 6 2
2013-14 61 7 0 15 5 9 5 1 1 12 5 2
2014-15 61 8 0 14 4 9 5 1 2 11 5 3
2015-16 62 8 0 13 4 9 5 1 2 12 6 3
2016-17 65 9 1 14 3 9 5 1 3 12 6 3
2017-18 66 10 1 14 3 9 5 1 3 12 6 3
Sources: See sub-section 3.5 of Chapter 3 and Appendix D.
Notes:
1. In school years prior to 2013-14, Cedar Tree Academy PCS (formerly Howard Road Academy PCS) and its campuses are counted as Comprehensive; for school
year 2013-14 and after, they are counted as Early Childhood.
2. For school years prior to 2012-13, Maya Angelou PCS and its campuses are counted as S pecific Population; for school year 2012-13 and after, it is counted as
both an Adult and Specific Population LEA due to its Young Adult Learning Center being added to its existing Evans Campus.
301
School
year
Charter
campuses
Adult
Blended
learning
Compre
-hensive
Creative
arts
Dual
language
Early
childhood
Global
culture
Learner
centered
Postsec.
/ voc.
Specific
population
STEM
1996-97 2 0 0 0 0 0 0 1 0 0 1 0
1997-98 4 0 0 1 1 0 0 1 0 0 1 0
1998-99 19 2 0 5 2 1 0 0 0 2 4 3
1999-00 31 2 0 8 3 1 0 2 0 6 7 2
2000-01 39 2 0 12 3 1 0 3 0 7 9 2
2001-02 41 2 0 14 3 1 0 3 0 8 8 2
2002-03 40 2 0 16 3 1 0 3 0 8 6 1
2003-04 44 2 0 18 3 2 1 3 0 9 5 1
2004-05 52 2 0 18 4 5 1 3 0 12 6 1
2005-06 66 3 0 22 7 6 5 3 0 11 7 2
2006-07 74 5 0 26 8 7 5 4 0 10 7 2
2007-08 82 5 0 28 7 8 7 4 0 12 9 2
2008-09 96 5 0 37 7 10 7 4 0 15 9 2
2009-10 95 5 0 37 7 9 8 4 0 16 7 2
2010-11 92 5 0 35 5 10 9 3 0 16 7 2
2011-12 96 6 0 35 4 12 12 2 1 15 7 2
2012-13 101 8 0 35 5 12 13 2 1 16 7 2
2013-14 103 10 0 33 5 12 14 2 1 18 6 2
2014-15 105 12 0 34 4 12 14 2 2 17 5 3
2015-16 108 13 0 34 4 11 14 2 2 19 6 3
2016-17 110 16 1 35 3 11 12 2 3 18 6 3
2017-18 112 16 2 35 3 12 12 2 3 18 6 3
Sources: See sub-section 3.5 of Chapter 3 and Appendix D.
Notes:
1. In school years prior to 2013-14, Cedar Tree Academy PCS (formerly Howard Road Academy PCS) and its campuses are counted as Comprehensive; for school
year 2013-14 and after, they are counted as Early Childhood.
2. For school years prior to 2012-13, Maya Angelou PCS and its campuses are counted as Specific Population; for school year 2012 -13 and after, it is counted as
both an Adult and Specific Population LEA due to its Young Adult Learning Center being added to its existing Evans Campus.
302
303
Appendix G: Spatial Proliferation of Charters by Type, School Years 2000-01 through 2017-18
Note: Because the first Blended Learning charter school opened in the District in school year 2016-
17, it is not possible to plot their charter -years attributable to tracts in a meaningful way.
Adult charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
304
Comprehensive charter-years attributable to each census tract, school years 2000-01 through 2017-
18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
Creative Arts charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
305
Dual Language charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
Early Childhood charter-years attributable to each census tract, school years 2000-01 through 2017-
18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
306
Global Culture charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
Learner Centered charter-years attributable to each census tract, school years 2000-01 through 2017-
18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
307
Postsecondary/Vocational charter-years attributable to each census tract, school years 2000-01
through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
Specific Population charter-years attributable to each census tract, school years 2000-01 through
2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
308
STEM charter-years attributable to each census tract, school years 2000-01 through 2017-18
Sources: See sub-sections 3.4 and 3.5 of Chapter 3 for information on charter campus locations and their types.
Appendix H: Descriptive Statistics for Measures Used in Main Regression Analyses, Chapter 3
Note: Land area of tracts included for informational purposes only; it is not a variable included in the main regression analysis.
Measure Average Median
Standard
deviation
Land area (square miles) 0.32 0.22 0.29
Relative Scale score 1.61 1.00 1.61
Absolute Index score 0.36 0.16 0.61
TOTAL charter-years 97.81 94.00 71.20
...Adult charter-years 11.01 2.00 22.94
...Blended Learning charter-years 0.16 0.00 0.52
...Comprehensive charter-years 33.11 35.00 24.62
...Creative Arts charter-years 5.98 0.00 7.87
...Dual Language charter-years 9.06 1.00 13.79
...Early Childhood charter-years 8.27 0.00 10.78
...Global Culture charter-years 2.43 0.00 8.79
...Learner Centered charter-years 0.96 0.00 1.57
...Postsecondary/Vocational charter-years 15.36 12.00 15.00
...Specific Population charter-years 9.06 5.00 10.97
...STEM charter-years 2.39 0.00 4.89
Distance to central business district (decimal degrees) 0.05 0.05 0.02
Between Rock Creek and Anacostia 61.1% n/a n/a
East of Anacostia 26.3% n/a n/a
Within 0.5 miles of Metrorail station 45.1% n/a n/a
Proximity of population to 75% non-WANH in 2000 9.15 4.19 31.16
Housing units per square mile in 2000 7,776 5,815 6,225
Share of occupied units that were rented in 2000 59.1% n/a n/a
Sources: See section 3 of Chapter 3 for a full discussion of data sources.
Note: Land area of census tracts included in Table for informational purposes only; it is not a variable included in the main regre ssion analysis.
309
Appendix I: Results of Sensitivity Analyses, Chapter 3
Table I.1. Sensitivity analysis 1: charter-year catchment area reduced from 1.0-mile radius to 0.5-mile radius
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(1) (2) (3) (4) (5) (6)
Gentrification measure
Relative
Scale
Relative
Scale
Absolute
Index
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used Ordinal probit Ordinal probit OLS OLS OLS OLS
TOTAL charter-years 0.0074** 0.0026** -0.0003
...Adult charter-years 0.0058 0.0070* -0.0019
...Blended Learning charter-years -3.2104 -0.0111 0.0140
...Comprehensive charter-years 0.0099 0.0064+ -0.0007
...Creative Arts charter-years 0.0091 0.0101 0.0048
...Dual Language charter-years 0.0050 -0.0055 0.0006
...Early Childhood charter-years 0.0060 0.0090 0.0005
...Global Culture charter-years -0.0434+ -0.0102+ -0.0013
...Learner Centered charter-years 0.0386 0.0090 -0.0170
...Postsecondary/Vocational charter-years 0.0162 -0.0088 0.0007
...Specific Population charter-years 0.0792** 0.0300** 0.0092**
...STEM charter-years 0.0302 0.0432* -0.0014
Distance to central business district -24.6930** -22.1514** -8.513** -7.5274** -2.0845** -2.0999*
Between Rock Creek and Anacostia 1.1345** 1.0948** 0.2850** 0.2205* 0.2509** 0.2283**
East of Anacostia -0.4435 -0.7138 -0.0440 -0.0951 0.1830** 0.1480**
Within 0.5 miles of Metrorail station 0.2352 0.2334 0.2187** 0.2444** 0.0163 0.0173
Proximity of population to 75% non-WANH in 2000 0.0113 0.0097 0.0009 -0.0014 0.0006* -0.0002
310
Housing units per square mile in 2000 -0.0000 -0.0000 -0.0000* -0.0000+ 0.0000 0.0000
Share of occupied units that were rented in 2000 -0.9412+ -1.0410+ 0.2697 0.1692 -0.1351* -0.1414*
cut1 -1.3559 -1.2344
cut2 -0.5726 -0.4078
cut3 -0.1413 0.0678
cut4 0.6100 0.9124
constant 0.4448 0.4151 -0.0282 -0.0246
n (# of census tracts) 175 175 175 175 175 175
Wald statistic or F-statistic 148.39 167.09 23.44 13.29 16.22 10.84
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.2866 0.3228 0.3803 0.4439 0.3354 0.3674
Notes: Standard errors in Regressions 1 and 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 3, 4, 5, and 6 are
adjusted for heteroskedasticity using the Huber-White sandwich estimator.
311
Table I.2. Sensitivity analysis 2: rail station catchment area expanded from 0.5- mile radius to 1.0-mile radius
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(1) (2) (3) (4) (5) (6)
Gentrification measure
Relative
Scale
Relative
Scale
Absolute
Index
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used Ordinal probit Ordinal probit OLS OLS OLS OLS
TOTAL charter-years 0.0047** 0.0020** -0.0004+
...Adult charter-years 0.0098 0.0040+ 0.0012
...Blended Learning charter-years 0.3691 0.0832 -0.0165
...Comprehensive charter-years 0.0159* 0.0064** 0.0010
...Creative Arts charter-years -0.0129 0.0086 -0.0011
...Dual Language charter-years -0.0138 -0.0094* -0.0033*
...Early Childhood charter-years 0.0060 0.0041 -0.0007
...Global Culture charter-years -0.0442** -0.0100* -0.0032*
...Learner Centered charter-years 0.2259+ 0.0001 0.0149
...Postsecondary/Vocational charter-years -0.0011 0.0026 -0.0007
...Specific Population charter-years 0.0214+ 0.0065 0.0016
...STEM charter-years -0.0134 0.0104 -0.0091**
Distance to central business district -22.9879** -13.0037 -8.3079** -2.2296 -2.0953** -1.8633*
Between Rock Creek and Anacostia 0.8486* 0.8593* 0.1605 0.1410 0.2743** 0.2722**
East of Anacostia -0.7253+ -1.4044* -0.1699 -0.4632+ 0.1937** 0.1731**
Within 1.0 miles of Metrorail station 0.3018 0.4204 0.0866 0.0548 0.0272 0.0064
Proximity of population to 75% non-WANH in 2000 0.0124 0.0113 0.0013+ 0.0002 0.0007* 0.0004
Housing units per square mile in 2000 -0.0000 -0.0000 -0.0000* -0.0000* 0.0000 0.0000
Share of occupied units that were rented in 2000 -0.6514 -0.9296+ 0.4124 0.3711 -0.1360* -0.1455*
cut1 -1.0083 -0.5030
cut2 -0.2251 0.3381
cut3 0.2215 0.8467
cut4 0.9875 1.7329
312
constant 0.3887 0.0656 -0.0396 -0.0236
n (# of census tracts) 175 175 175 175 175 175
Wald statistic or F-statistic 150.03 174.40 23.54 15.92 15.69 8.61
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.2898 0.3369 0.3645 0.4201 0.3437 0.3598
Notes: Standard errors in Regressions 1 and 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 3, 4, 5, and 6 are
adjusted for heteroskedasticity using the Huber-White sandwich estimator.
313
Table I.3. Sensitivity analysis 3: charter-year catchment area reduced from 1.0-mile radius to 0.5-mile radius, and rail station catchment
area expanded from 0.5-mile radius to 1.0-mile radius
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(1) (2) (3) (4) (5) (6)
Gentrification measure
Relative
Scale
Relative
Scale
Absolute
Index
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used Ordinal probit Ordinal probit OLS OLS OLS OLS
TOTAL charter-years 0.0075* 0.0031** -0.0003
...Adult charter-years 0.0069 0.0098* -0.0017
...Blended Learning charter-years -3.0593 -0.0523 0.0111
...Comprehensive charter-years 0.0100 0.0067+ -0.0007
...Creative Arts charter-years 0.0127 0.0166 0.0053
...Dual Language charter-years 0.0043 -0.0092 0.0004
...Early Childhood charter-years 0.0063 0.0093 0.0005
...Global Culture charter-years -0.0464* -0.0114+ -0.0014
...Learner Centered charter-years 0.0153 -0.0180 -0.0189
...Postsecondary/Vocational charter-years 0.0178 -0.0060 0.0009
...Specific Population charter-years 0.0752** 0.0276** 0.0091**
...STEM charter-years 0.0292 0.0380* -0.0017
Distance to central business district -24.1308** -21.9005** -8.9616** -8.2169** -2.0341** -2.1518*
Between Rock Creek and Anacostia 1.0939** 1.0682** 0.2661** 0.2304** 0.2474** 0.2291**
East of Anacostia -0.5156 -0.7681+ -0.0864 -0.1239 0.1761** 0.1460**
Within 1.0 miles of Metrorail station 0.3005 0.2589 0.0919+ 0.0392 0.0233 0.0022
Proximity of population to 75% non-WANH in 2000 0.0117 0.0101 0.0012 -0.0009 0.0007* -0.0002
Housing units per square mile in 2000 -0.0000 -0.0000 -0.0000* -0.0000+ 0.0000 0.0000
Share of occupied units that were rented in 2000 -0.8212 -0.9611+ 0.3475 0.2257 -0.1281+ -0.1374*
cut1 -1.1509 -1.1032
cut2 -0.3764 -0.2852
cut3 0.0556 0.1909
314
cut4 0.8129 1.0405
constant 0.4545 0.5002 -0.0432 -0.0180
n (# of census tracts) 175 175 175 175 175 175
Wald statistic or F-statistic 148.05 166.61 22.74 13.19 16.32 10.86
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.2860 0.3218 0.3551 0.4120 0.3357 0.3658
Notes: Standard errors in Regressions 1 and 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 3, 4, 5, and 6 are
adjusted for heteroskedasticity using the Huber-White sandwich estimator.
315
Table I.4. Sensitivity analysis 4: secondary types for charter schools substituted in for primary types, when applicable
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(2) (4) (6)
Gentrification measure
Relative
Scale
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used Ordinal probit OLS OLS
TOTAL charter-years n/a n/a n/a
...Adult charter-years 0.1519+ -0.0087 0.0181**
...Blended Learning charter-years 0.5709+ 0.1016 -0.0128
...Comprehensive charter-years 0.0150* 0.0076** 0.0002
...Creative Arts charter-years -0.0145 0.0084 -0.0026
...Dual Language charter-years -0.0011 -0.0134+ -0.0010
...Early Childhood charter-years 0.0166 0.0010 -0.0003
...Global Culture charter-years -0.0196 -0.0139** -0.0004
...Learner Centered charter-years -0.0184 -0.0086 -0.0018
...Postsecondary/Vocational charter-years 0.0121 0.0023 -0.0005
...Specific Population charter-years 0.0137+ 0.0059+ 0.0007
...STEM charter-years -0.0235 0.0056 -0.0031*
Distance to central business district -18.7425* -1.8888 -2.4779**
Between Rock Creek and Anacostia 0.9410* 0.2380* 0.2723**
East of Anacostia -1.9037** -0.3792 0.1763**
Within 0.5 miles of Metrorail station 0.3564+ 0.2324** 0.0169
Proximity of population to 75% non-WANH in 2000 0.0127 -0.0003 0.0005+
Housing units per square mile in 2000 -0.0000 -0.0000** 0.0000
Share of occupied units that were rented in 2000 -1.2964* 0.2525 -0.1649*
cut1 -1.1691
cut2 -0.3103
cut3 0.1851
cut4 1.0581
316
constant 0.0536 0.0091
n (# of census tracts) 175 175 175
Wald statistic or F-statistic 177.48 17.93 7.73
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.3428 0.4469 0.3605
Notes: Standard errors in Regression 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 4 and 6 are adjusted for
heteroskedasticity using the Huber-White sandwich estimator.
317
Table I.5. Sensitivity analysis 5: charter-years weighted by school year-specific enrollment for each charter school campus (using charter-
years and enrollment data for school years between 2004-05 and 2017-18)
+ for p < 0.10; * for p < 0.05; ** for p < 0.01
(1) (2) (3) (4) (5) (6)
Gentrification measure
Relative
Scale
Relative
Scale
Absolute
Index
Absolute
Index
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Change in Share
of White Alone,
non-Hispanic
Schoolchildren
Regression model used Ordinal probit Ordinal probit OLS OLS OLS OLS
TOTAL charter-years 0.5959** 0.2193** -0.0193
...Adult charter-years 1.0772* 0.3867* 0.1281
...Blended Learning charter-years 37.2698 0.9085 -1.1299
...Comprehensive charter-years 1.6649** 0.6682** 0.0906
...Creative Arts charter-years -0.0840 0.2190 0.0479
...Dual Language charter-years -0.9468 -1.0597 -0.5697*
...Early Childhood charter-years -1.3653 0.7318 -0.0451
...Global Culture charter-years -17.2228** -4.7799** -0.9042+
...Learner Centered charter-years 44.9523 5.7552 5.4575
...Postsecondary/Vocational charter-years -0.1311 -0.2651 -0.0828
...Specific Population charter-years 2.3870 0.1821 0.1619
...STEM charter-years -1.9275 0.9620 -1.0816**
Distance to central business district -24.4797** -18.6100* -8.1403** -4.4221 -2.1431** -2.0546*
Between Rock Creek and Anacostia 0.8195* 0.8187* 0.1592+ 0.1862+ 0.2584** 0.2547**
East of Anacostia -0.8400* -1.5410* -0.1828 -0.2956 0.1938** 0.1521*
Within 0.5 miles of Metrorail station 0.2050 0.3090 0.1914** 0.2190** 0.0176 0.0130
Proximity of population to 75% non-WANH in 2000 0.0126 0.0088 0.0008 -0.0003 0.0006* 0.0003
Housing units per square mile in 2000 -0.0000 -0.0000 -0.0000* -0.0000* 0.0000 0.0000
Share of occupied units that were rented in 2000 -0.8467+ -1.1972* 0.3251 0.2292 -0.1415* -0.1554*
cut1 -1.3207 -1.1676
cut2 -0.5152 -0.3031
cut3 -0.0553 0.2008
318
cut4 0.7265 1.0641
constant 0.4077 0.2279 -0.0217 0.0008
n (# of census tracts) 175 175 175 175 175 175
Wald statistic or F-statistic 155.27 175.22 25.75 16.81 15.64 7.72
Prob > Chi-squared or Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Adjusted or Pseudo R-squared 0.2999 0.3385 0.4046 0.4379 0.3360 0.3667
Sources: Total enrollment figures for each charter campus, used to weight charter -years in this regression, come from OSSE’s official enrollment audit reports, available
for school years 2004-05 through 2017-18.
Notes: Standard errors in Regressions 1 and 2 are adjusted for heteroskedasticity using the observed information matrix; standard errors in Regressions 3, 4, 5, and 6 are
adjusted for heteroskedasticity using the Huber-White sandwich estimator.
319
320
Appendix J: Enrollment Statistics by Charter School Type, School Years 2004-05 through 2017-18
Note: Because the first Blended Learning charter campus opened in the District in school year 2016-
17, it is not possible to calculate this type’s enrollment statistics due to insufficient observations.
Enrollment statistics for charter campuses by type, school year 2004-05 through 2017-18.
Charter school type
Average
enrollment
Median
enrollment
Standard
deviation of
enrollment
Number of
observations
comprehensive 427 340 277 444
postsecondary/vocational 385 353 196 213
adult 343 161 463 111
creative arts 284 236 186 73
STEM 278 298 89 31
specific population 252 238 160 95
dual language 236 189 175 137
early childhood 152 108 155 133
learner centered 128 118 64 13
global culture 88 59 56 39
Total 1,289
Sources: Total enrollment figures for each charter campus come from OSSE’s official enrollment audit reports, available
for school year 2004-05 through 2017-18.
Note: Table rows sorted in descending order per average enrollment; Blended Learning type omitted from table due to
insufficient observations.
321
Appendix K: The Community Eligibility Provision
Until recently, all traditional public and public charter schools determined student eligibility
for the FRPL program on an individual basis, via a student application indicating household income
level and enrollment in the following federal programs: Temporary Assistance for Needy Families
(TANF), Supplemental Nutrition Assistance Program (SNAP), and Food Distribution Program on
Indian Reservations (FDPIR) (USDA, 2013). And so historically, if a school reported that 48
percent of its 100 students were enrolled in the FRPL program, that meant the school had reviewed
and approved FRPL paperwork for 48 of its students (i.e., 48 percent of 100 students).
In school year 2011-12, though, the U.S. Department of Agriculture (USDA) began
implementing a significant adjustment to its FRPL program called the Community Eligibility
Provision (CEP) (USDA, n.d). The Provision, which state and then local education agencies
individually opt into, dramatically changes how schools determine eligibility for the FRPL program
by encouraging them to pull data directly from federal and state government agencies. Under the
CEP, students are automatically eligible if they or their households are listed in the following
databases: SNAP, TANF, or FDPIR; McKinney-Vento Act homeless children; Runaway and
Homeless Act children and youth; Title I, Part C migrant children; foster-care; federally-funded
Head Start program or state-funded Head Start or pre-kindergarten program (USDA, 2016).
If at least 40 percent of a school’s students are “directly certifiable” in this way, then its share
of students that are FRPL-eligible is calculated as: (1) the number 1.6, multiplied by (2) the
percentage of students “directly certified”, and (3) no more than 100 percent of students (USDA,
2016). And so, if 48 percent of a school’s 100 students are “directly certified”, then 76.8 percent of
its students (i.e., 1.6 multiplied by 48 percent) are deemed FRPL -eligible. If fewer than 40 percent of
a school’s students are “directly certified”, then the school must continue to determine eligibility on
a student-by-student basis via the application form (USDA, 2016).
As the example makes clear, if any school within a district qualifies for and participates in the
CEP, their participation makes it infeasible for researchers to rely on the FRPL percentages of that
district’s schools as a consistent proxy for those schools’ levels of poverty. Accordingly, and as other
researchers have concluded (GAO, 2017), District charter campuses’ FRPL percentages have been
an unreliable proxy for student poverty since school year 2012-13, when District charter LEAs first
began using the CEP (USDA, 2016).
Additional references
U.S. Department of Agriculture (USDA). (2013, August 13). Applying for free and reduced price school
meals. https://www.fns.usda.gov/school-meals/applying-free-and-reduced-price-school-
meals
U.S. Department of Agriculture (USDA). (2016, September). Community eligibility provision (CEP):
Planning & implementation guidance. https://fns-
prod.azureedge.net/sites/default/files/cn/SP61-2016-CEP_Guidance.pdf
U.S. Department of Agriculture (USDA). (n.d). Community eligibility provision.
https://www.fns.usda.gov/school-meals/community-eligibility-provision
Appendix L: Descriptive Tables for Non-Adult Charter Enrollment, by Geography, School Year (SY) 2010 -11 to 2018-19
Total enrollment, all campuses
Geographic location SY10-11 SY11-12 SY12-13 SY13-14 SY14-15 SY15-16 SY16-17 SY17-18 SY18-19
West of Anacostia River 16,322 17,503 18,761 19,989 21,081 21,226 22,075 22,978 23,402
East of Anacostia River 10,679 11,642 12,803 12,902 12,968 13,642 14,679 15,375 14,975
… Ward 7 6,323 6,298 6,415 6,125 5,867 6,201 6,383 6,674 6,836
… Ward 8 4,356 5,344 6,388 6,777 7,101 7,441 8,296 8,701 8,139
27,001 29,145 31,564 32,891 34,049 34,868 36,754 38,353 38,377
Distribution of students (in percentage points), all campuses
Geographic location SY10-11 SY11-12 SY12-13 SY13-14 SY14-15 SY15-16 SY16-17 SY17-18 SY18-19
West of Anacostia River 60 60 59 61 62 61 60 60 61
East of Anacostia River 40 40 41 39 38 39 40 40 39
… Ward 7 23 22 20 19 17 18 17 17 18
… Ward 8 16 18 20 21 21 21 23 23 21
100 100 100 100 100 100 100 100 100
Share of students considered black American (in percentage points), all campuses
Geographic location SY10-11 SY11-12 SY12-13 SY13-14 SY14-15 SY15-16 SY16-17 SY17-18 SY18-19
West of Anacostia River 78.9 76.9 73.7 71.8 70.0 69.3 67.6 66.6 64.9
East of Anacostia River 98.7 98.7 97.4 98.0 98.0 98.1 97.4 97.3 97.0
… Ward 7 98.0 98.0 96.7 97.1 97.1 97.2 96.6 96.5 96.1
… Ward 8 99.6 99.5 98.2 98.9 98.8 98.8 98.1 98.0 97.8
86.7 85.6 83.3 82.1 80.6 80.5 79.5 78.9 77.5
322
Share of students considered Hispanic/Latino (in percentage points), all campuses
Geographic location SY10-11 SY11-12 SY12-13 SY13-14 SY14-15 SY15-16 SY16-17 SY17-18 SY18-19
West of Anacostia River 14.5 15.9 16.8 17.6 17.4 16.6 18.0 17.7 17.9
East of Anacostia River 0.8 1.0 1.3 1.3 1.2 1.3 1.7 1.7 1.9
… Ward 7 1.3 1.6 1.9 2.0 2.1 1.8 2.6 2.6 2.6
… Ward 8 0.2 0.2 0.6 0.6 0.5 0.8 1.1 1.1 1.3
9.1 9.9 10.5 11.2 11.3 10.6 11.5 11.3 11.6
Share of students considered white (in percentage points), all campuses
Geographic location SY10-11 SY11-12 SY12-13 SY13-14 SY14-15 SY15-16 SY16-17 SY17-18 SY18-19
West of Anacostia River 4.7 5.4 6.4 7.2 8.5 9.7 9.8 10.8 11.5
East of Anacostia River 0.3 0.2 0.3 0.3 0.3 0.2 0.3 0.3 0.2
… Ward 7 0.4 0.3 0.4 0.4 0.4 0.3 0.2 0.3 0.4
… Ward 8 0.1 0.1 0.2 0.2 0.2 0.1 0.3 0.3 0.1
3.0 3.3 3.9 4.5 5.4 6.0 6.0 6.6 7.1
323
324
Appendix M: Summary of Discrete Hazard Modeling
In the text that follows, I review both the logistic and complementary log-log functional forms, both
of which can be used in binomial regression analysis. I first provide an overview of the logistic
functional form and its use in regression analysis. I then introduce the complementary log-log
functional form, and I explore its potential benefits in the context of hazard modeling. While the
complementary log-log functional form may be theoretically advantageous to the logistic functional
form when incorporating baseline hazard rates, the two forms’ regression results are approximately
equivalent when the probability of the hazard event approaches 0.
An overview of logistic functional form and regression analysis
Take a discrete outcome 𝑦𝑦 𝑖𝑖 that:
• equals 1 with probability 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ), the latter of which by definition is bounded by 0 and 1,
• equals 0 with probability 1 − 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ), also bounded by 0 and 1, and
• depends upon the value of a single explanatory factor, 𝑥𝑥 1
.
The odds ratio (OR) for outcome 𝑦𝑦 𝑖𝑖 can be expressed as:
𝑂𝑂𝑂𝑂
𝑖𝑖 =
𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
1 − 𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
Eq. 1
The odds ratio is bounded by 0 and ∞, approaching 0 as 𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 ) → 0 and approaching ∞ as
𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 ) → 1. Taking the logarithm of OR yields:
ln[ 𝑂𝑂𝑂𝑂
𝑖𝑖 ] = ln �
𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
1 − 𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
� Eq. 2
The ln[OR] is bounded by − ∞ and ∞.
This type of function, which transforms a probability bounded by 0 and 1 into a continuous measure
bounded by − ∞ and ∞, is known as a “log-link function”.
The logistic regression assumes that the natural logarithm of the odds ratio, i.e., the log-link function,
is equal to the linear combination of some parameters and explanatory factors. Again, for ease of
interpretation, let us assume a constant term 𝛼𝛼 and a single explanatory factor 𝑥𝑥 1
. Then the logistic
regression’s functional form can be expressed as:
325
ln[ 𝑂𝑂𝑂𝑂
𝑖𝑖 ] = ln �
𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
1 − 𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )
� = 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 Eq. 3
Rearranging the terms gives the following equation for probability 𝑝𝑝 :
𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ) =
𝑒𝑒 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 1 + 𝑒𝑒 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 =
1
1 + 𝑒𝑒 −( 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 )

Eq. 4
Again, note that 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ) is bounded by 0 and 1. It is also symmetric around predicted probability 0.5,
approaching a probability of 1 and a probability of 0 at equal rates.
Equation 4 can be generalized into a Bernoulli probability mass function as follows:
𝑃𝑃 ( 𝑌𝑌 = 𝑦𝑦 | 𝑋𝑋 = 𝑥𝑥 ) = �
1
1 + 𝑒𝑒 −( 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1
)

�
𝑦𝑦 ∗ � 1 −
1
1 + 𝑒𝑒 −( 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1
)

�
( 1 − 𝑦𝑦 )
Eq. 5
The likelihood function for all 𝑛𝑛 observations can then be written as:
𝐿𝐿 ( 𝜃𝜃 ) = ∏ �
1
1 + 𝑒𝑒 − � 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 �

�
𝑦𝑦 𝑖𝑖 ∗ � 1 −
1
1 + 𝑒𝑒 − � 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 �

�
( 1 − 𝑦𝑦 𝑖𝑖 )
𝑛𝑛 𝑖𝑖 = 1
Eq. 6
Finally, taking the logarithm of this likelihood functi on produces the log-likelihood function for the
logistic regression, with which maximum likelihood estimation (MLE) can be used to develop
estimated parameters 𝛼𝛼 � and 𝛽𝛽 1
�
:
𝐿𝐿𝐿𝐿 ( 𝜃𝜃 ) = ∑ 𝑦𝑦 𝑖𝑖 ∗ log �
1
1 + 𝑒𝑒 − � 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 �

� + ∑ (1 − 𝑦𝑦 𝑖𝑖 ) ∗ log � 1 −
1
1 + 𝑒𝑒 − � 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 �

�
𝑛𝑛 𝑖𝑖 = 1

𝑛𝑛 𝑖𝑖 = 1
Eq. 7
Extending to the complementary log-log function form and regression analysis
Let us take the same assumptions used for the logistic regression. Namely, take a discrete outcome
𝑦𝑦 𝑖𝑖 that:
• equals 1 with probability 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ), the latter of which by definition is bounded by 0 and 1,
• equals 0 with probability 1 − 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ), also bounded by 0 and 1, and
• depends upon the value of a single explanatory factor, 𝑥𝑥 1
.
The complementary log-log functional form is also a “log-link” function. However, instead of
assuming that the logarithm of the odds ratio equals a linear combination of some parameters and

326

explanatory factors – as a logistic regression does – a complementary log-log regression assumes
that:

ln( −ln[1 − 𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 )]) = 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 Eq. 8

As with the logistic regression’s functional form, the above can be rearranged to give the following
equation for probability p:

𝑝𝑝 ( 𝑥𝑥 1 𝑖𝑖 ) = 1 − 𝑒𝑒 − 𝑒𝑒 𝛼𝛼 + 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 Eq. 9

Again, and as with the logistic function, note that 𝑝𝑝 ( 𝑥𝑥 𝑖𝑖 ) is bounded by 0 and 1. However, unlike the
logistic regression, the complementary log-log function is asymmetric: it approaches 0 much more
slowly than it approaches 1, and therefore is not “centered” at probability of 0.5 as the logistic
regression is. As a result, it appears well-suited for application in instances where positive outcomes
(i.e., 𝑦𝑦 𝑖𝑖 = 1) occur rarely (Kitali, Kidando, Sando, Moses, & Ozguven, 2017).

Figure M.1. Logistic function versus complementary log-log function.

Finally, the Bernoulli probability mass function can be utilized to construct a log-likelihood function
for the complementary log-log regression in the same way as for the logistic regression.

0.0
0.5
1.0
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
9
10
Probability of Event
Value of X
Logistic function Complementary log-log function
327
Incorporating baseline hazard rates in a complementary log-log model
As Martuzzi and Elliott (1998) discuss, the complementary log -log functional form is well-suited for
application as a hazard model (i.e., values of 𝑦𝑦 𝑖𝑖 = 0 and 𝑦𝑦 𝑖𝑖 = 1, respectively), due to the following
assumption made by hazard models:
𝑝𝑝 ( 𝑥𝑥 ) = 1 − 𝑒𝑒 −𝜆𝜆 𝑡𝑡 Eq. 10
where λ represents a baseline hazard rate, common for all individuals, assumed to be constant over
period 𝑡𝑡 . Note the similarities between Equation 9 and Equation 10.
And in fact, as Jenkins (2005) demonstrates, the complementary log-log functional form can be
modified to explicitly incorporate baseline hazard rates for modeling survivorship over discrete time
periods. Accordingly, adjusting Equation 9 to include a baseline hazard rate yields:
𝑝𝑝 𝑗𝑗 ( 𝑥𝑥 1 𝑖𝑖 ) = 1 − 𝑒𝑒 − 𝑒𝑒 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 + 𝛾𝛾 𝑖𝑖 Eq. 11
where:
• 𝛾𝛾 𝑗𝑗 is the log of the difference in baseline hazard between time point 𝑗𝑗 and time point 𝑗𝑗 − 1
(i.e., in general, a measure of the baseline hazard rate for the discrete period beginning at 𝑗𝑗 −
1 and ending at 𝑗𝑗 ), and
• 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 is the risk associated with each individual due to that individual’s characteristics.
Note that the constant 𝛼𝛼 has been excluded to allow for the inclusion of baseline hazards for each
discrete time period. Note also that Equation 11 is equivalent to:
ln � −ln � 1 − 𝑝𝑝 𝑗𝑗 ( 𝑥𝑥 1 𝑖𝑖 ) � � = 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 + 𝛾𝛾 𝑗𝑗 Eq. 12
From a practical regression standpoint, because the characteristics of individual 𝑖𝑖 may change over
the 𝑗𝑗 , the data used for complementary log-log regression analysis must be organized in a “pooled”
fashion, as how ordinary least squares can incorporate the pooling of panel data.
Also from a practical regression standpoint, in a discrete time period context, 𝛾𝛾 𝑗𝑗 in Equation 12 can
be expressed as a series of dummy variables indicating each time period in the analysis:
ln( −ln[1 − 𝑝𝑝 𝑡𝑡 ( 𝑥𝑥 1 𝑖𝑖 )]) = 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 + 𝛾𝛾 𝑗𝑗 𝐷𝐷 𝑗𝑗 + 𝛾𝛾 𝑗𝑗 − 1
𝐷𝐷 𝑗𝑗 − 1
+ ⋯ + 𝛾𝛾 1
𝐷𝐷 1
Eq. 13
where 𝑡𝑡 represents a time period between 1 and 𝑗𝑗 .
328
Similarities between the complementary log-log and logistic regressions that incorporate baseline
hazard rates
As Jenkins (2005) also demonstrates, the logistic regression framework (Equation 3 above) can
accommodate a baseline hazard similarly:
ln �
𝑝𝑝 𝑖𝑖 ( 𝑥𝑥 1 𝑖𝑖 )
1 − 𝑝𝑝 𝑖𝑖 ( 𝑥𝑥 1 𝑖𝑖 )
� = 𝛽𝛽 1
𝑥𝑥 1 𝑖𝑖 + 𝛼𝛼 𝑗𝑗 Eq. 14
where 𝛼𝛼 𝑗𝑗 is a logistic function of the baseline hazard rate between time point 𝑗𝑗 and time point 𝑗𝑗 − 1,
and can be expressed as dummy variables as in Equation 13 above.
And in fact, it can be shown that the results of the complementary log-log and logistic regressions
with hazard rates approximate each other as the probability of the hazard event occurring
approaches 0 (Jenkins, 2005).
Additional references
Jenkins, S. P. (2005). Survival analysis. Institute for Social and Economic Research (Unpublished
Manuscript). University of Essex, Colchester, UK, 42, 54-56.
Appendix N: Logistic Regression Results, Chapter 4
Notes: + for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001
All coefficients reported as odds ratios, and charter LEAs closed for non-academic reasons excluded Observations: 389
Charter LEAs with Adult, Early Childhood, or Alternative Accountability campuses excluded Zero outcomes: 375
Charter LEAs that operated in the District for only one school year excluded Nonzero outcomes: 14
Student-based
regression
Ward-based regressions
Explanatory factor (1) (2) (3) (4)
Student population >= 95% Black American 3.178+ - - 0.986
Majority of students east of Anacostia River (Wards 7 or 8) - 3.930* - -
… Majority of students in Ward 7 - - 2.664 2.684
… Majority of students in Ward 8 - - 7.698** 7.778*
Proximate tracts >= 80
th
percentile (black American schoolchildren concentration) - - - -
Proximate tracts >= 80
th
percentile (impoverished schoolchildren concentration) - - - -
… Proximate census >= 80
th
percentile (BOTH concentrations) - - - -
Enrollment share < 1% 2.753 4.066* 4.314* 4.326*
Member of charter management organization (CMO) 1.947 2.114 3.175 3.187
Authorized by District’s Board of Education 0.698 0.966 1.009 1.010
Operated 6 to 10 years 2.396 3.008 3.341 3.348
Operated 11 to 15 years 2.926 3.371 6.059* 6.076*
Operated 16 to 20 years 0.482 0.393 0.685 0.687
Period 1 (SY10-11 to SY12-13) 0.002*** 0.002*** 0.001*** 0.001***
Period 2 (SY13-14 to SY15-16) 0.008*** 0.007*** 0.004*** 0.004***
Period 3 (SY16-17 to SY18-19) 0.011*** 0.008*** 0.005*** 0.005***
Converged Log Likelihood -54.169 -53.382 -51.504 -51.504
Wald chi2 118.15 115.33 107.68 107.69
Prob > chi2 0.0000 0.0000 0.0000 0.0000
329
Proximate census tract-based regressions
Explanatory Factor (5) (6) (7) (8) (9) (10)
Student population >= 95% Black American - - - - 1.505 1.251
Majority of students east of Anacostia River (Wards 7 or 8) - - - - - -
… Majority of students in Ward 7 - - - - - -
… Majority of students in Ward 8 - - - - - -
Proximate tracts >= 80
th
percentile (black American schoolchildren conc.) 1.592 4.775** - - 3.725+ -
Proximate tracts >= 80
th
percentile (impoverished schoolchildren conc.) 3.541 - 5.256** - - -
… Proximate census >= 80
th
percentile (BOTH concs.) - - - 6.150** - 5.388*
Enrollment share < 1% 3.773* 3.568+ 3.860* 3.728* 3.339+ 3.604+
Member of charter management organization (CMO) 2.696 2.263 2.804 2.841 2.087 2.691
Authorized by District’s Board of Education 0.746 0.770 0.738 0.715 0.760 0.714
Operated 6 to 10 years 3.190 3.090 3.181 3.294 2.914 3.181
Operated 11 to 15 years 5.052+ 4.709+ 4.959+ 5.316* 4.420+ 5.137+
Operated 16 to 20 years 0.634 0.490 0.666 0.726 0.478 0.697
Period 1 (SY10-11 to SY12-13) 0.002*** 0.002*** 0.001*** 0.002*** 0.002*** 0.002***
Period 2 (SY13-14 to SY15-16) 0.005*** 0.007*** 0.005*** 0.005*** 0.006*** 0.005***
Period 3 (SY16-17 to SY18-19) 0.007*** 0.007*** 0.007*** 0.006*** 0.007*** 0.006***
Converged Log Likelihood -52.079 -52.682 -52.158 -51.506 -52.556 -51.468
Wald chi2 110.75 112.59 110.83 109.01 111.92 108.70
Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
330
Appendix O: Supplementary Model Results, School Year 2010-11 to 2018-19, Chapter 4
Notes: + for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001
Functional form of Models 1 – 4 is complementary log-log (CLL), with z-scores reported
Functional form of Models 5 – 8 is Ordinary Least Squares (OLS), with t-scores reported
Charter campuses that operated in the District for only one school year excluded, and charter campuses east of the Anacostia excluded
Dep. Var:
Campus Tier
increase
Dep. Var:
Campus Tier
decrease
Dep. Var:
Black American
share change
Dep. Var:
White share
change
(1) (2) (3) (4) (5) (6) (7) (8)
Functional Form CLL CLL CLL CLL OLS OLS OLS OLS
Change in black American share of schoolchildren -2.55* - -0.01 - - - - -
Change in white share of schoolchildren - 2.32* - -0.39 - - - -
Campus Tier ranking increase (2
nd
prior to prior year) - - - - 0.42 - 0.43 -
Campus Tier ranking decrease (2
nd
prior to prior year) - - - - - -0.13 - 1.69+
Growth in total enrollment -1.18 -1.20 -1.06 -1.05 - - - -
Change in Ward 8 share of enrollment - - - - 1.58 1.63 -1.39 -1.67+
Member of charter management organization (CMO) 0.47 0.37 1.09 1.10 - - - -
Authorized by District's Board of Education 1.11 0.59 1.21 1.26 - - - -
Operated 6 to 10 years -1.73+ -2.02* -1.56 -1.57 -1.14 -1.15 1.42 1.48
Operated 11 to 15 years -1.36 -1.79+ -1.08 -1.10 -0.47 -0.49 1.25 1.34
Operated 16 to 20 years 0.55 0.48 0.23 0.25 -0.45 -0.40 0.54 0.51
Observations 331 331 331 331 208 208 208 208
Nonzero outcomes 39 39 38 38 - - - -
Wald chi2 or F-statistic 152.03 153.15 156.56 156.33 0.55 1.16 2.55 2.43
Prob > chi2 or Prob > F 0.0000 0.0000 0.0000 0.0000 0.7355 0.3285 0.0288 0.0362
331
332
Appendix P: Results of Principal Component Analysis, Changes in Census Tracts 2000-2018
Component eigenvalue difference proportion cumulative
C1 3.2085 2.6801 0.8021 0.8021
C2 0.5284 0.3572 0.1321 0.9342
C3 0.1712 0.0793 0.0428 0.9770
C4 0.0919 n/a 0.0230 1.0000
Dimension of change for census tract, 2000-18 C1 C2 C3 C4
white alone, non-Hispanic population share 0.4963 -0.5660 0.0846 0.6528
share of population with at least a bachelor's degree 0.5168 -0.4159 -0.0548 -0.7463
growth in median household income 0.4901 0.5273 0.6941 -0.0054
tract’s percentile rank for median household income 0.4964 0.4782 -0.7128 0.1296
Sources: 2000 Total Population comes from 2000 U.S. Census Table SF1 / P0001; 2000 White Alone, Non-Hispanic
Total comes from 2000 U.S. Census Table SF1 / P012I; 2000 Share of Population with At Least Bachelor’s Degree
comes from 2000 U.S. Census Table SF3 / DP2; 2000 Median Household Income comes from 2000 U.S. Census Table
SF3 / DP3. 2018 Total Population comes from 5-year 2018 ACS Table B01003; 2018 White Alone, Non-Hispanic Total
comes from 5-year 2018 ACS Table B01001H; 2018 Share of Population with At Least Bachelor’s Degree comes from
5-year 2018 ACS Table S1501; 2018 Median Household Income comes from 5-year 2018 ACS Table S1901.
Note: Adjustments for inflation made using the U.S. Bureau of Labor Statistics' Consumer Price Index (CPI) for All
Urban Consumers in the Washington-Arlington-Alexandria Metropolitan Statistical Area.
333
Appendix Q: Sources Used in Verifying non-DCPS School Buildings
For DC Prep PCS occupying building of former Our Lady of Perpetual Health Catholic
School
Parents Have School Choice Kids Win: Education Reform in the Nation’s Capital. (2019, December
2). Exclusive interview with Lauren Maestas, CEO DC Prep PCS.
https://parentshaveschoolchoicekidswin.com/2019/12/02/exclusive-interview-with-lauren-
maestas-ceo-dc-prep-pcs/
For Washington Yu Ying PCS occupying building of former Marist School Seminary
Parents Have School Choice Kids Win: Education Reform in the Nation’s Capital. (2016, March 1).
Exclusive interview with Maquita Alexander, head of Washington Yu Ying PCS.
https://parentshaveschoolchoicekidswin.com/2016/03/01/exclusive-interview-with-maquita-
alexander-head-of-washington-yu-ying-pcs/
For Lee Montessori PCS and Washington Leadership Academy PCS occupying building of
former St. Paul’s College
District of Columbia Office of Planning. (2017, November 16). Historic preservation review board: St.
Paul’s College.
https://planning.dc.gov/sites/default/files/dc/sites/op/publication/attachments/Historic%20Lan
dmark%20Nomination%20Staff%20Report%20%203105%203025%204th%20Street%20NE%20%
20St.%20Pauls%20College%20Cases%2017%2014%20and%2017%2021%20.pdf
For Lee Montessori PCS occupying school building constructed by Eagle Academy PCS
Goncalves, D. (2020, January 10). ‘After all the damage, now they want to sell?’ DC charter school leaving
before it opens. CBS: WUSA9. https://www.wusa9.com/article/news/education/eagle-academy-
charter-school-built-without-permits-leaving-before-opens/65-c89178db-8bd6-4245-a42a-
22d9d6f25179
For Statesmen Prep Academy for Boys PCS opening in Rocketship DC PCS-constructed
space
DC Public Charter School Board. (2018, June 25). Notice of vote on full charter approval – Statesmen College
Preparatory Academy for Boys PCS (formerly NorthStar PCS). https://dcpcsb.org/notice-vote-full-charter-
approval-statesmen-college-preparatory-academy-boys-pcs-formerly-northstar
Shinberg.Levinas. (2016). Rocketship Rise Academy. https://www.shinberglevinas.com/portfolio-
view/rocketship-rise-academy/
For Inspired Teaching Demonstration PCS and Mundo Verde Bilingual PCS occupying
building of former Catholic Sisters College
District of Columbia Office of Planning. (n.d). DC Government historic properties – public charter & related
schools.
https://planning.dc.gov/sites/default/files/dc/sites/op/publication/attachments/List%20of%20D
334
C%20Charter%20and%20Related%20Schools%20with%20Preservation%20Considerations%2002
%20%202018.pdf
For Latin American Montessori Bilingual PCS (LAMB) moving into building of former
private Kingsbury School
Goncalves, D. (2019, April 10). Area’s oldest school for children with learning differences slated to shut down but
extending school year by 5 weeks. CBS: WUSA9. https://www.wusa9.com/article/news/areas-oldest-
school-for-children-with-learning-differences-slated-to-shut-down-but-extending-school-year-by-5-
weeks/65-865eff98-d9b0-4cb4-9d01-ce8dd9bc5d71
Appendix R: Summary of Nested Logistic Modeling Comparisons
Notes: + for p<0.10; * for p<0.05; ** for p<0.01; *** for p<0.001 Observations: 870
Estimated coefficients displayed Zero outcomes: 783
Nonzero outcomes: 87
Conditional Nested
Non-schooling characteristics of census tracts
Density of all children (0-17 years)
… Density of children in grades specific to new charter campus 0.0003 0.0003
Index of gentrification since 2000 -0.1957 -0.2389
Median gross rent 0.0004 0.0005
Rail transit station within 0.5 miles of tract 0.6209 0.4305
Availability of adequate facilities
Vacant school building in tract 4.0603*** 3.4895**
Vacant non-school building in tract, formerly occupied by other charter LEA 3.2964*** 2.7751**
Nearby public schooling opportunities
Enrollment gap for all children (0-17 years)
… Enrollment gap for children in grades specific to new campus 0.0009*** 0.0009***
Blended math percentile of all nearby public campuses
… Availability of grade-specific blended math percentile of all campuses 1.5599 1.2634
… Grade-specific blended math percentile of all campuses -0.7443 -0.4134
Availability of blended math percentile of nearby charter campuses
Blended math percentile of nearby charter campuses
… Availability of grade-specific blended math percentile of charters
335
… Grade-specific blended math percentile of charters
Blended math percentile of nearby DCPS campuses
… Availability of grade-specific blended math percentile of DCPS campuses
… Grade-specific blended math percentile of DCPS campuses
Existing campus and relocation considerations
Average distance from tract to existing campuses under same charter LEA -0.4154** -0.3856**
… Average distance to existing campuses teaching similar grades
… Average distance to existing campuses teaching dissimilar grades
If new campus is relocation, average distance from tract to previous location -0.4964*** -0.4570**
Converged Log Likelihood -84.516 -81.593
Chi2 231.62 22.98
Prob > chi2 0.0000 0.0846
336

Abstract (if available)

Abstract In its analysis, this dissertation assesses a specific policy, the District of Columbia School Reform Act of 1995 (Pub. L. 104-134), which most notably legalized charter schools in the District of Columbia. The research questions it pursues, however, are broader and relevant to the planning, public policy, and education fields. First, how do components of a policy intersect with historical and structural inequalities associated with race and place? And second, how do these interactions influence a policy’s ability to effect greater equity of access? In pursuing answers to the above questions, this dissertation identifies: (1) a significant relationship between the presence of charter schools and the gentrification of District neighborhoods, albeit one that is dependent on type of charter school (e.g., STEM)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Property and labor formalization in the age of the sharing economy: Airbnb, housing affordability, and entrepreneurship in Havana

PDF

The impact of demographic shifts on automobile travel in the United States: three empirical essays

PDF

Productive frictions and urbanism in transition: planning lessons from traffic flows and urban street life in Ho Chi Minh City, Vietnam

PDF

Location of warehouses and environmental justice: Three essays

PDF

Environmental justice in real estate, public services, and policy

PDF

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

PDF

Tough conversations and missed opportunities: implementing district policies for racial equity

PDF

State policy as an opportunity to address Latinx transfer inequity in community college

PDF

Essays on the economics of cities

PDF

Household mobility and neighborhood impacts

PDF

Does organizational culture play a role in aviation safety? A qualitative case study analysis

PDF

Role of a principal supervisor in fostering principal development to support instructional improvement and a positive school climate

PDF

The development implications of China’s Belt and Road Initiative for Russia, Kazakhstan and Belarus

PDF

Urban consumer amenities and their accessibility

PDF

Unraveling decentralization of warehousing and distribution centers: three essays

PDF

Three essays on housing demographics: depressed housing access amid crisis of housing shortage

PDF

Navigating the struggle: neoliberal meritocracy, Latinx charter school graduates, and the college transition

PDF

The demand for reliable travel: evidence from Los Angeles, and implications for public transit policy

PDF

Constructing the achievement index: an assessment to help people achieve high-performance leadership in the workplace

PDF

Faculty learning and agency for racial equity

Asset Metadata

Creator Eisenlohr, Andrew (author)

Core Title Who learns where: understanding the equity implications of charter school reform in the District of Columbia

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Urban Planning and Development

Publication Date 11/20/2020

Defense Date 08/24/2020

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

Tag charter schools,equity of access,OAI-PMH Harvest,policy analysis,racial segregation,school choice,urban demography

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Boarnet, Marlon (committee chair), Marsh, Julie (committee member), Myers, Dowell (committee member)

Creator Email andy.eisenlohr@gmail.com,eisenloh@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-398566

Unique identifier UC11666174

Identifier etd-EisenlohrA-9135.pdf (filename),usctheses-c89-398566 (legacy record id)

Legacy Identifier etd-EisenlohrA-9135.pdf

Dmrecord 398566

Document Type Dissertation

Rights Eisenlohr, Andrew

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