Fall 22
https://doi.org/10.1177/0895904820951124
Educational Policy 2022, Vol. 36(5) 1078 –1107
© The Author(s) 2020 Article reuse guidelines:
sagepub.com/journals-permissions DOI: 10.1177/0895904820951124
journals.sagepub.com/home/epx
Article
The effects of charter school enrollment losses and tuition reimbursements on school districts: Lifting boats or sinking them?
Bryan A. Mann1 and Paul Bruno2
Abstract We analyze a natural experiment in which policymakers in Pennsylvania first implemented, and later removed, reimbursements to districts for students exiting to brick and mortar and cyber charter schools. Generalized difference- in-difference models show that larger shares of students enrolling in charter schools predict decrements in spending, financial health, and achievement in sending districts; however, these relationships attenuate in years when districts receive reimbursements. After receiving reimbursements, districts with increased competition spent more on instruction and instructional support services, but not on facilities or non-instructional operations. Perhaps due to higher instructional expenditures, the relationship between competition and student achievement in reimbursement years is significantly less negative, and at times even positive, compared to non-reimbursement years. Cyber charter schools induce fewer instructional expenditures in districts than brick and mortar charter schools. The findings show clear policy choices can support traditional public systems experiencing competition.
1Department of Educational Leadership and Policy Studies, University of Kansas, Lawrence, KS, USA 2Department of Education Policy, Organization, and Leadership, University of Illinois at Urbana-Champaign, Champaign, IL, USA
Corresponding Author: Bryan A. Mann, Department of Educational Leadership and Policy Studies, University of Kansas, Joseph R. Pearson Hall, Rm. 404, 1122 West Campus Rd., Lawrence, KS 66045-3101, USA. Email: bryanmann@ku.edu
951124 EPXXXX10.1177/0895904820951124Educational PolicyMann and Bruno research-article2020
Mann and Bruno 1079
Keywords school choice, charter schools, competition, education policy, economics of education
Education reformers in the United States and across the world suggest that school choice policies incentivize competition between schools to create “a rising tide that lifts all boats,” meaning that choice improves outcomes for all students (Hoxby, 2001, 2003). This assumption drives policy adoption of various school choice models including charter schools, school vouchers, and inter- and intra-district choice schools, among others (Berends et al., 2011). School choice implementation may expand as policymakers at the highest levels of the U.S. government continue to argue that choice provides “more options” that “yield better results, for all students” (DeVos, 2017).
The ubiquity of charter schools makes them an ideal school type to con- sider when determining how the competitive effects logic unfolds in practice. Charter schools enroll a growing share of K-12 students in the United States, recently up to 3.2 million, and have received backing from policymakers across the political spectrum (National Alliance for Public Charter Schools, 2018; Wohlstetter et al., 2013).
Pressures associated with charter school competition may impose finan- cial and programmatic stress on nearby traditional public schools and dis- tricts (TPSDs). However, these patterns are inconsistent. Some studies find that TPSDs’ financial outlooks and academic outcomes suffer as charter sec- tors expand, while others show relationships vary in magnitude and direc- tion (Arsen & Ni, 2012; Bettinger, 2005; Bifulco & Reback, 2013; Bruno, 2019; Cordes, 2018; Epple et al., 2016; Imberman, 2011; Winters, 2012). A systemic analysis of competitive effects studies shows small positive effects of competition on achievement, but these effects depend on policy design and differ based on student demographics (Jabbar et al., 2019). Increased competition in some contexts depresses home values, which can undercut the ability of TPSDs to increase resources (Cook, 2018; Brehm et al., 2017). To extend the rising tide metaphor, competitive forces may create tides that lift all boats, but these forces may also sink the capacity of TPSDs by impos- ing unintended costs. Researchers need to continue to examine these issues, especially since TPSDs serve the majority of students in the United States (NCES, 2017).
One of the recommendations from past research is for state agencies to consider offsetting the diminished resources that competitive environments place on TPSDs (Cook, 2018, p. 61). The next step in this line of research is to identify and examine contexts where states have pursued this strategy.
1080 Educational Policy 36(5)
Additionally, the competitive effects of charter schools on TPSDs and dis- tricts may vary across policy and economic contexts. Yet one limitation of existing research is that studies examine specific contexts, such as cities or smaller regions, which vary from one another along many dimensions. Studies conducted in a single local context make it difficult to know whether observed TPSD responses to competition are discretionary (e.g., to appeal to parents) or are determined by constraints (e.g., due to fixed costs). These limitations hinder theoretical understandings of educational competition and prevent extant work from offering clear guidance to policymakers.
We examine a natural experiment in Pennsylvania in which state policy- makers introduced, and later removed, tuition reimbursements for TPSDs experiencing enrollment losses to charter schools. Using panel data on all TPSDs in the state over 16 years and detailed records of the number of stu- dents exiting each TPSD system for charter schools, we explore variation in the relationships between charter school enrollment losses and district opera- tions as districts enjoy greater or lesser degrees of financial insulation from competitive pressures. The length of the panel allows us to consider periods before, during, and after the reimbursement policy while controlling for dis- trict and labor market-by-year fixed effects.
Pennsylvania data also distinguish enrollments in brick and mortar charter schools from cyber charter schools. Brick and mortar charter schools educate students in physical buildings and resemble traditional public schools in many ways. Cyber charter schools are full time online educational programs with students attending synchronous or asynchronous virtual classes. Brick and mortar charter schools enroll students from mostly urban areas, while cyber charter schools capture enrollment from nearly every school district in the state (Mann & Baker, 2019). Our use of statewide data and statewide competitive effect patterns (due to the statewide spread of cyber charter schools) allow us to explore whether differences in reimbursement policy and the type of charter school change TPSDs’ competitive responses.
Research Questions
We consider four interrelated research questions:
1. How do enrollment losses to charter schools change the spending behavior of Pennsylvania school districts? Does spending change with the implementation of tuition reimbursements?
2. Do enrollment losses to charter schools reduce academic performance in Pennsylvania school districts? Do academic performance patterns change with the implementation of tuition reimbursements?
Mann and Bruno 1081
3. Do enrollment losses to charter schools hurt the financial health of Pennsylvania school districts? Does financial health improve with the implementation of tuition reimbursements?
4. Are the answers to RQs 1–3 different for brick and mortar and cyber charter school competition?
Competition, School District Finances, and Student Performance
Charter school implementation reflects the understanding that underpins school choice policy: competition for enrollment improves all schools in an educational ecosystem and forces schools to close if they do not improve (Chubb & Moe, 1990; Friedman, 1955). This logic assumes that schools of choice improve the educational experience for all students in a choice envi- ronment, including those who remain in TPSDs (Hoxby, 2001, 2003). However, three critiques have emerged that challenge the theory of action of how competition serves as a school improvement mechanism.
The first critique is that academic improvement in TPSDs only occurs if districts have the capacity to allocate resources in a manner that improves student outcomes. Such responses may not be possible if duplicating educa- tional services across sectors imposes excess costs on districts (e.g., growing per-pupil fixed costs; Bifulco & Reback, 2013). Indeed, competition causes many school districts to face financial strain (Arsen & Ni, 2012; Bruno, 2019; Cook, 2018; Brehm et al., 2017; Ladd & Singleton, 2018). However, financial strain findings are not universal. For example, school districts in Massachusetts saw increases in instructional spending over several years, especially related to reimbursements (Ridley & Terrier, 2018). Conversely, charter school entry into an educational ecosystem in Ohio decreased home values and limited districts’ ability to raise additional tax dollars (Cook, 2018).
A second critique is districts do not always perceive and react to enroll- ment losses in a manner aimed at enhancing instruction. Instead, they respond to competitive pressures in complicated ways that vary based on organiza- tional hierarchies and other factors (Jabbar, 2015). In one example, principals in Michigan perceived little competition from charter schools, and when they did perceive competition, it had little bearing on decision-making (NCSC, 2011). Traditional systems may also create barriers that “blunt the potentially damaging impact of competition” by encouraging TPSD administrators to pursue activities like lobbying for a charter school to close rather than com- peting instructionally (Arsen & Ni, 2008, pg. 5).
A third critique is that school organizations follow patterns of institu- tional isomorphism and avoid innovation (DiMaggio & Powell, 1983).
1082 Educational Policy 36(5)
Classroom practices in charter schools often are the same as those in TPSDs (Berends et al., 2010; Lubienski, 2003; Preston et al., 2012). Many charter schools innovate administratively rather than pedagogically (Lubienski, 2003). An analysis of charter school mission statements shows variety in specialist charter schools, but these schools are rare because most resemble TPSDs (Renzulli et al., 2015). As a result, often charter schools offer lim- ited incentives for TPSDs to innovate due to few differences between sectors.
These critiques help explain why competitive effects vary across contexts. Charter schools have positive effects on achievement and future career earn- ings in Florida (Sass et al., 2016), no effects in North Carolina (Bifulco & Ladd, 2006), and little effect on achievement gains and school practices in California and Michigan (Bettinger, 2005; Buddin & Zimmer, 2009). States with large private school sectors have better outcomes in public schools because states provide more resources to these schools to insulate them from competition (Arum, 1996). Perhaps a lack of a response, or a lack of capacity, undermine goals of competition leading to universal improvement.1
While differences in responses may explain mixed findings in previous research, there are differences in educational governance across states that may lead to varying effects of competition. Our study delves deeply into one state to see how differences in policy and incentives mediate differences in competitive effects. We also explore how different state-level policies on managing competitive effects change the results of implementation.
Pennsylvania Charter School Policy and Implementation
It is appropriate to study competitive effects in Pennsylvania for two rea- sons. The first is that the history of Pennsylvania charter school policy offers two different financial realities for school districts over time: One where districts were provided reimbursements for charter school tuition expendi- tures (from 2002–03 to 2010–11) and one where the tuition reimbursements were not offered (prior to 2002–03 and 2011–12 to present). The second reason is that the policy environment in Pennsylvania has offered two types of charter schools: brick and mortar and cyber. While many states have vir- tual or cyber charter schools, Pennsylvania’s cyber charter school sector has tens of thousands of students (Mann & Baker, 2019). The scale of cyber charter schooling puts a heightened level of pressure on TPSDs not seen in other charter school states, particularly because all districts in Pennsylvania face competitive pressure from schools that can enroll even students that live far away from the schools’ offices.
Mann and Bruno 1083
Pennsylvania lawmakers enacted charter schools through Act 22 of 1997 and, as shown in Figure 1, by 2015–16 charter schools enrolled approxi- mately eight percent of the state’s public school students, with cyber charter schools accounting for roughly 30% of that enrollment. The Pennsylvania Department of Education (PDE) lists goals for the programs outlined by advocates of choice: improve schooling through choice, innovation, and accountability (2004).
A charter school in Pennsylvania begins when a school receives a charter from an authorizer, which usually are local school boards in Pennsylvania, though some districts like Philadelphia have had a state-created school reform panel that controls district operations and possesses charter-authorizing authority (Pew, 2015).2 Cyber charter schools must receive authorization from the PDE. Once authorized, all charter schools operate with greater autonomy than TPSDs, are not subject to conventional governance by elected school boards, and have flexibility in hiring and staffing (Pew, 2015).
The majority of funding for TPSDs in Pennsylvania comes from local tax dollars, as does funding for charter schools. When a student leaves for a char- ter school, the TPSD to which they are zoned (the “sending district”) must pay the charter school an associated tuition. The tuition amounts to approxi- mately 70% of the district’s per-pupil expenditure after making deductions for funds and services not related to charter school operation, including some federal aid, adult education, transportation, debt service, and building con- struction (Hardy, 2015). The tuition payments occur on a monthly basis and
Figure 1. Percentage of students enrolled in charter schools in Pennsylvania. Authors’ calculations from Pennsylvania Department of Education data.
1084 Educational Policy 36(5)
represent the number of students enrolled in each charter school from the district during the previous month. This allows school district leaders con- tinually to be aware of the number of the students in their district enrolled in a charter school. This funding allocation also leads to varied tuition rates. Using 2014’s tuition rates for illustrative purposes, sending district tuition rates range from $6,600 to $17,000 for a non-special education student and $13,000–$43,000 for a special education student (Hardy, 2015).
As shown in Figure 2, brick and mortar or cyber charter schools enroll- ment differs substantially based on sending district composition. Brick and mortar charter schools tend to operate and enroll students in urban areas, while cyber charter schools enroll students from across the state (Mann & Baker, 2019). Cyber charter schools have more urban students than rural stu- dents, but an individual rural student is more likely to enroll in a cyber charter school (Mann et al., 2016). Consistent with differences between charter school types, we find the share of district students enrolled in cyber charter schools correlates with the share enrolled in brick and mortar charter schools only weakly (r = .19).
Beginning with the budget for FY 2003 after an intense round of negotia- tions, the legislature enacted a measure that would provide a 30 percent reim- bursement (additional to the aforementioned deductions) to Pennsylvania school districts for tuition payments to charter and cyber charter schools
Figure 2. Change in the percentage of students enrolled in charter schools in Pennsylvania districts between 2000–01 and 2015–16. Marker size proportional to average district student membership. Urbanicity based on 2000–01. Authors’ calculations from Pennsylvania Department of Education data.
Mann and Bruno 1085
(Fitzgerald et al., 2002). The reimbursements reflected prior year enrollment, so districts could plan and make changes to their budgets (Griffith, 2014). There could have been a slight lag effect due to districts waiting for the expected amount of reimbursements to get deposited into their accounts (these deposits happened six times annually), but districts generally could develop reasonable estimations of the reimbursement based on their charter school enrollment statistics from one year to the next. The legislature elimi- nated this reimbursement for the 2011–12 school year after enacting austerity measures following the 2008 financial crisis. These reimbursements remain a point of debate between lawmakers (Langley, 2016).
Understanding the effect of charter school reimbursement cuts on TPSDs in Pennsylvania is important because more than 90% of Pennsylvania stu- dents attend TPSDs. All schools face some level of competitive pressure because Pennsylvania features cyber charter schools that have drawn enroll- ment from 499 out of 500 school districts in the state (Mann & Baker, 2019). Lawmakers need to have reliable information on reimbursement funding as they continue to debate this policy in current and future legislative sessions, and in other states.
Data and Methods
Our data come primarily from public files released annually by PDE. These files cover the universe of Pennsylvania school districts, 500 districts, between the 2000–01 and 2015–16 school years.3 These data include detailed revenue, expenditure, and budget reserve information, local property tax millage rates, English learner enrollment, and shares of students meeting aca- demic proficiency standards on statewide end-of-year exams in math and reading. The number of students in a district enrolling in charter schools is our measure of competitive pressure faced by TPSDs. Because charter school enrollments are provided at the district-charter school level, we can distin- guish enrollments in cyber charter schools from enrollments in brick-and- mortar charter schools.
We make three adjustments to these data. First, because tuition payments by sending districts to charter schools are categorized as instructional expen- ditures, we subtract these payments from both total and instructional expen- diture figures. This prevents TPSD per-student spending amounts from being inflated by tuition costs for students exiting for the charter sector and thus no longer enrolled in the district itself. Prior to 2003–04 data, it is necessary to infer these tuition payment amounts from districts’ per-pupil tuition rates and charter school enrollments, as actual payments were not released, and for some districts this is not possible in 2000–01 because tuition rates were not
1086 Educational Policy 36(5)
published. In other years, when it is possible to adjust total and instructional expenditures using actual and inferred tuition payments, those figures corre- late at r = .99 and differ from one another by less than one-half of one percent on average. This suggests that inferred and actual payments are similar, and in results available upon request, we find expenditure results are nearly iden- tical if we exclude school years prior to 2003–04.
Second, charter school enrollment data show an anomalous drop in charter school enrollment in 2008–09. Twenty-nine charter schools reporting enroll- ments in previous and subsequent years are not present in the data file in that year, so we impute enrollments from each district as the mean of those enroll- ments in 2007–08 and 2009–10. This produces charter school enrollment fig- ures consistent with the overall time trend. Finally, in 2009–10 the Appalachian Area School District4, newly formed by the merger of two other school dis- tricts, reports charter school enrollments far below what is observed in the next year or what is observed in its component districts in the previous year. On the assumption this reflects a temporary disruption in reporting, we impute this district’s charter school enrollment in this year as the mean of its charter enrollment share in 2010–11 and the enrollment-weighted mean of the charter enrollments of the merged districts in 2008–09. In results avail- able upon request, excluding Appalachian Area School District in 2009–10 and all districts in 2008–09 has virtually no effect on our estimates.
We supplement the data from the PDE with data from three additional sources. First, we link school districts to labor markets using comparable wage index files from the National Center for Education Statistics (NCES; Taylor & Fowler, 2006). While we adjust all financial figures for inflation (to 2016 dollars), linking school districts to specific labor markets allows us to adjust for geographic variation in the costs of labor and other economic fac- tors. Controlling for unobserved differences between labor markets helps to alleviate concerns that our results are biased by local factors, such as contem- poraneous policy changes or changes in economic conditions that do not vary consistently statewide. Second, we use district total and special education enrollment figures and other student demographic data, as well as data on districts’ debt levels, from the NCES Common Core of Data. Finally, while the PDE’s student proficiency rate data have the advantage of spanning the length of our dataset, it may be that the shares of students meeting academic standards do not adequately capture changes in student achievement within different proficiency bands, particularly as testing regimes have changed. Therefore, we supplement Pennsylvania proficiency rate data with data from the Stanford Education Data Archive (SEDA; Reardon et al., 2017). Though only available for the 2008–09 through 2014–15 school years and including some charter school achievement data with that of their local or affiliated
Mann and Bruno 1087
TPSDs, SEDA produces estimates of districts’ mean student achievement in math and English/language arts (ELA) on a scale that is comparable across tests and years by combining proficiency rate data from state tests with data from the National Assessment of Educational Progress.
Summary statistics for our variables of interest are presented in Table 1. There is considerable variation in the shares of students that districts send to charter schools. Some districts send none in some years and others send more than half of their potential students. Further, there is a considerable positive skew in our revenue, expenditure, and debt figures even when mea- sured on a per-pupil basis. We therefore subject these figures to an inverse hyperbolic sine transformation prior to conducting the analyses. This gives our results a similar interpretation as would a natural log transformation, but allows us to retain observations in which districts report values of zero (Burbidge et al., 1988).
Analytical Strategy
We estimate a series of models for each achievement or fiscal outcome Y in district d in labor market l and year t:
Y charter reimbursed
charter reimbursed
dlt dlt t
dlt t
= +
+ ×
β β
β 1 2
3
%
% + + + +δ ψ εd lt dltDdltΩΩ (1)
%charter represents the percentage share of the district’s students enrolling in charter schools, and reimbursed is a dummy variable equal to one in years in which the state offered aforementioned tuition reimbursements (2002–03 through 2010–11), and zero otherwise. Because we interact these two vari- ables, β1 estimates the relationship between charter enrollment shares and our outcomes in non-reimbursement years (i.e., when reimbursed = 0) and β3 estimates the extent to which those relationships differ in reimbursement years (i.e., when reimbursed = 1).
We use a generalized difference-in-difference approach, making within- district comparisons over time. δd is thus a set of district fixed effects to control for unobserved, time-invariant differences between districts. We fur- ther include labor market-by-year fixed effects (ψlt ) , where labor markets are Census-defined place-of-work areas (Taylor & Fowler, 2006). This allows us not only to control for statewide changes over time, but also to compare only districts operating in a similar economic context (e.g., with similar local labor costs) in a given year. D is a vector of time-varying district characteris- tics that are likely to be related to finances and achievement, including the natural log of district enrollment to control for economies of scale, as well as
1088 Educational Policy 36(5)
Table 1. Summary Statistics.
N Mean SD Min Max
Enrollment 7977 3389.4 8003.5 124 201190 % Charter 7977 2.0 3.2 0.0 60.3 % Cyber Charter 7977 1.3 1.1 0.0 12.9 % Brick & Mortar Charter 7977 0.8 2.8 0.0 52.2 % FRL 7962 32.5 18.3 0.0 100 % SPED 7975 16.1 3.8 0.0 45.2 % White 7977 87.6 17.5 1.1 100 % EL 7977 1.1 2.1 0.0 23.1 Achievement % Proficient or Advanced in Math 7975 66.3 16.7 6.0 97.3 % Proficient or Advanced in Reading 7975 69.1 11.9 9.1 94.9 Mean Cohort-Standardized
Achievement in Math 3480 0.2 0.3 −1.2 1.2
Mean Cohort-Standardized Achievement in ELA
3480 0.2 0.3 −1.4 1.2
Revenue State Revenue 7977 6068 2733 1627 45399 Local Revenue 7977 7999 3690 2001 26491 Real Estate Taxes 7977 6088 3443 913 24944 Real Estate Tax Rate (Mills) 6973 48 47 7 616 Expenditures Total Expenditures 7893 14570 3248 7800 46736 Instruction 7893 8125 1584 4816 25846 Instructional Support Services 7977 4307 979 2213 13942 Pupil Support 7977 412 155 0 2665 Instructional Staff Support 7977 426 215 0 2503 Administration Support 7977 898 236 334 2695 Pupil Health Services 7977 149 52 48 731 Business Services 7977 203 96 0 1591 Plant Services 7977 1236 293 483 3782 Student Transportation 7977 802 314 64 3202 Non-Instructional Operations 7977 283 127 0 1301 Facilities 7977 45 216 0 6920 Financial Health Undesignated Fund Balance 7977 1461 1357 −7320 17342 Long-Term Debt 6979 7 10 0 605
Note. Data in this table combine annual observations of 500 unique districts from 2000–1 through 2015–16. Revenues, expenditures, fund balances, and debt are per-pupil and expressed in 2016 dollars.
Mann and Bruno 1089
the shares of district students who are white, English language learners, or eligible for free or reduced-price lunch or special education services. ε is an error term, and we cluster standard errors on districts. These approaches align with methodological strategies used in similar previous education research on charter schools and their influence on school finance (e.g., Arsen & Ni, 2012; Bruno, 2019).
Results
RQ 1: TPSD Expenditures
Higher charter enrollments are associated with lower per-pupil spending in TPSDs, particularly when tuition reimbursements are not provided to dis- tricts. Table 2 presents results from models predicting per-pupil spending as a function of charter school enrollment shares. Because these expenditure outcomes have been subjected to an inverse hyperbolic sine transformation, coefficients can be interpreted (as in the case of a natural log transformation) roughly as semi-elasticities, or as a percent change in spending associated with a percentage-point increase in charter school enrollment.
In non-reimbursement years, a percentage point increase in charter school enrollment is associated with lower per-pupil spending in the sending TPSD of about 0.5% (column 1). A district going from one to eight percent of its students enrolled in charter schools – approximately the statewide change during this time period – would thus be predicted to spend roughly 3.3%, or $485, less per student in the absence of tuition reimbursements.
The coefficient on the interaction term indicates that that relationship is substantially less negative in years in which districts receive state reimburse- ment for their tuition payments, suggesting that districts spend, rather than save, some of these additional revenues. In reimbursement years, the rela- tionship between charter competition and total spending per student is indis- tinguishable from zero (p = .73).
It is perhaps surprising that larger charter school enrollment shares are associated with lower per-pupil spending in sending districts because dis- tricts losing enrollment would tend also to lose economies of scale. Moreover, as discussed above, districts’ tuition payments to charter schools for each pupil are substantially less than the district’s per-pupil expenditure; this should increase a district’s per-pupil spending on remaining students when a student exits for the charter sector. However, because we control for the district’s TPS enrollment, we account for changes in economies of scale over time. Our results are thus relationships between charter school enroll- ment and district expenditures net of the fact that districts will tend to lose
1090
T ab
le 2
. Fi
xe d
Ef fe
ct R
eg re
ss io
ns P
re di
ct in
g Pe
r- Pu
pi l S
pe nd
in g.
In st
ru ct
io na
l s up
po rt
s er
vi ce
s
T
ot al
In st
ru ct
io n
A ll
Pu pi
l Su
pp or
t
In st
ru ct
io na
l St
af f
Su pp
or t
A dm
in .
Su pp
or t
Pu pi
l H
ea lth
Se
rv ic
es Bu
si ne
ss
Se rv
ic es
Pl an
t Se
rv ic
es St
ud en
t T
ra ns
po rt
N on
-In st
ru ct
io na
l O
pe ra
tio ns
Fa ci
lit ie
s
(1
) (2
) (3
) (4
) (5
) (6
) (7
) (8
) (9
) (1
0) (1
1) (1
2)
% ag
e C
ha rt
er −
0. 00
5* **
(0 .0
01 )
− 0.
00 7*
** (0
.0 01
) −
0. 00
5* **
(0 .0
01 )
− 0.
02 2*
(0 .0
09 )
− 0.
02 4*
(0 .0
11 )
− 0.
00 6*
* (0
.0 02
) −
0. 00
8* (0
.0 03
) −
0. 00
2 (0
.0 07
) −
0. 00
6* **
(0 .0
02 )
0. 01
0* **
(0 .0
03 )
− 0.
02 3*
(0 .0
10 )
− 0.
06 2*
(0 .0
26 )
R ei
m bu
rs ed
x
% ag
e C
ha rt
er 0.
00 4*
** (0
.0 01
) 0.
00 3*
(0 .0
01 )
0. 00
5* *
(0 .0
02 )
0. 00
9 (0
.0 08
) 0.
00 3
(0 .0
06 )
0. 00
4* *
(0 .0
01 )
0. 00
6* **
(0 .0
02 )
0. 00
9+
(0 .0
05 )
0. 00
6* **
(0 .0
02 )
0. 00
3* *
(0 .0
01 )
0. 00
0 (0
.0 12
) 0.
03 3
(0 .0
26 )
A dj
. R -s
q 0.
79 0.
93 0.
93 0.
82 0.
73 0.
86 0.
81 0.
78 0.
88 0.
94 0.
71 0.
44 D
is tr
ic ts
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
O bs
er va
tio ns
78 76
78 76
79 60
79 60
79 60
79 60
79 60
79 60
79 60
79 60
79 60
79 60
N ot
e. S
ta nd
ar d
er ro
rs c
lu st
er ed
o n
di st
ri ct
s. A
ll m
od el
s in
cl ud
e di
st ri
ct a
nd la
bo r
m ar
ke t-
by -y
ea r
fix ed
e ffe
ct s
an d
co nt
ro l f
or t
he n
at ur
al lo
g of
d is
tr ic
t en
ro llm
en t
an d
th e
sh ar
es o
f d is
tr ic
t st
ud en
ts w
ho a
re w
hi te
, e lig
ib le
fo r
fr ee
- or
r ed
uc ed
-p ri
ce lu
nc h,
a re
E ng
lis h
le ar
ne rs
, o r
re ce
iv e
sp ec
ia l e
du ca
tio n
se rv
ic es
. D ep
en de
nt v
ar ia
bl es
ar
e in
ve rs
e hy
pe rb
ol ic
s in
e- tr
an sf
or m
ed . +
p <
.1 , *
p <
.0 5,
* *p
< .0
1 **
*p <
.0 01
.
Mann and Bruno 1091
economies of scale when enrollments fall ceteris paribus, and will thus tend to spend more per pupil. In results available upon request, when we no lon- ger control for the log of district enrollment, the main effect of charter school enrollment on total per-pupil spending switches signs (β = 0.003, p = .07), though the coefficient on the interaction term is essentially unchanged. Additionally, the negative relationship between charter school enrollment and per-pupil spending previews our revenue results below indicating that, as in other contexts, charter school enrollment is associated with decreased local property tax revenue, which tends to put downward pressure on spending.
Financial reports provided by the state segment district spending into sev- eral more specific categories of activity, including instruction, instructional support services, non-instructional operations, and facilities management. Considering these categories separately offers insight not only into how district spending changes are associated with charter competition, but also on how dis- tricts use the additional resources that tuition reimbursements provide.
As shown in Table 2, in non-reimbursement years an additional percentage point of students enrolled in charter schools is associated with significantly lower per-pupil spending on all major categories of expenditure, including instruction (−0.7 percent, column 2), instructional support (−0.5 percent, col- umn 3), non-instructional operations (−2.3 percent, column 11), and facilities (−6.2 percent). One might expect that these relationships would all become more positive (or less negative) in the presence of tuition reimbursements, but this is not what we observe.
While the coefficients on the interaction term are never negative–indicat- ing that charter enrollment is not associated with additional decrements to spending in reimbursement years–they are only significantly positive in the case of instruction and instructional support services. These relative increases in instruction-related spending are consistent with competition inducing instructional improvements in TPSDs. However, the decrements to all cate- gories of spending in non-reimbursement years suggests that the fiscal pres- sures associated with competition from charter schools may constrain districts looking to make such improvements.5
We are unable to link district tuition payments to charter schools to spe- cific types of instructional expenditure (e.g., regular education vs. vocational education), and are thus unable to adjust those expenditure categories as we do for instruction as a whole. However, tuition payments are not attributed to instructional support services. Specific categories of this service spending therefore do not require such adjustment and can be disaggregated to under- stand why districts facing additional charter competition spend relatively more on these activities in reimbursement years. This is seen in columns 4–10 of Table 2.
1092 Educational Policy 36(5)
Reductions in instructional support service spending associated with char- ter school enrollment in non-reimbursement years are driven by lower spend- ing on pupil support (e.g., guidance services; column 4), staff support (e.g., professional development; column 5), administrative support (e.g., legal costs; column 6), pupil health services (e.g., vision screenings; column 7), and plant services (e.g., grounds keeping; column 9). Charter school enroll- ment is not associated with lower per-pupil spending on business services (e.g., payroll services; column 8) or student transportation between home and school or between schools (column 10), consistent with such costs being rela- tively fixed for districts.
Again, in no case are these relationships more negative in reimbursement years, and they are often significantly more positive. In particular, when tuition reimbursements are available, districts facing additional charter school enrollment appear to spend more on administration, pupil health services, plant services, and student transportation; the coefficient on the relevant inter- action term is also marginally significant for business service spending.
RQ 2: TPSD Student Achievement
We find that higher levels of charter school competition associate with lower achievement in district schools. These relationships attenuate–and even reverse–in years when districts receive reimbursements for enrollment losses.
Results from models predicting student achievement outcomes are pre- sented in Table 3. As shown in column 1, in years when districts do not receive tuition reimbursements an additional percentage point of students enrolled in charter schools is associated with 0.19 percentage points fewer TPSD students meeting state proficiency standards in reading. The coeffi- cient on the interaction term indicates that the relationship is significantly less negative in years in which districts receive reimbursements. Results are similar in math (column 3), including a significantly more-positive relation- ship between charter competition and student achievement in reimbursement years relative to non-reimbursement years, though the estimated relationship during non-reimbursement years is not significantly different from zero.
Results using cohort-standardized student scale scores provided by SEDA (columns 2 and 4) are similar in both math and ELA, suggesting results are not simply an artifact of the use of proficiency rate measures of achievement. In non-reimbursement years, a percentage point of students exiting TPSDs for charter schools is associated with student achievement in districts that is lower by 0.006 standard deviations in ELA (column 5) and 0.009 standard deviations in math (column 6). Again, these relationships significantly atten- uate, in each case by approximately two-thirds, when districts receive tuition
Mann and Bruno 1093
reimbursements. These results are consistent with charter competition impos- ing an operational strain on nearby TPSDs and suggest that competitive pres- sures, if they improve outcomes at all, do so only when associated fiscal pressures are ameliorated.6
RQ 3: TPSD Revenue and Financial Health
Greater charter school enrollment losses are associated with financial strain in TPSDs, even in the presence of tuition reimbursements, due at least in part to decreases in revenue. As shown in Table 4, lower overall spending levels in districts with larger charter enrollment shares appear to be partially driven by lower revenues from state and local sources, which collectively represent more than 90% of district revenues in Pennsylvania. Larger charter school enrollment shares are not significantly related to higher per-pupil state reve- nues in non-reimbursement years (column 1), though, as expected, that rela- tionship increases in magnitude in reimbursement years. Total local revenues (column 2) are, if anything, negatively related to charter enrollment shares, and these relationships are likely to be positively biased in some sense by the fact that districts sending larger numbers of students to charter schools are obligated to redirect larger amounts of revenue to those schools in the form of tuition payments.7
Table 3. Fixed Effect Regressions Predicting Student Achievement.
Percentage proficient or advanced Mean cohort- standardized achievementReading Math
All SEDA Sample All
SEDA Sample ELA Math
(1) (2) (3) (4) (5) (6)
%age Charter −0.190** (0.063)
−0.280** (0.099)
−0.142 (0.110)
−0.316* (0.123)
−0.006** (0.002)
−0.009** (0.003)
Reimbursed × 0.274*** 0.233*** 0.320* 0.227* 0.004* 0.006*** %age Charter (0.072) (0.060) (0.127) (0.103) (0.002) (0.002) Adj. R-sq 0.91 0.94 0.93 0.95 0.95 0.90 Districts 500 497 500 497 497 497 N 7958 3471 7958 3471 3471 3471
Note. Standard errors clustered on districts. All models include district and labor market-by-year fixed effects and control for the natural log of district enrollment and the shares of district students who are white, eligible for free- or reduced-price lunch, are English learners, or receive special education services. +p < .1, *p < .05, **p < .01 ***p < .001.
1094 Educational Policy 36(5)
Much as was the case with total spending, it is perhaps surprising that we find a negative relationship between charter school enrollment and local rev- enue. Again, however, this is due in part to the fact that we control for (the natural log of) district enrollment, accounting for the fact that falling enroll- ment will tend to increase per-pupil revenue, all else being equal, because revenues are distributed across a smaller number of students. Moreover, our local revenue results are consistent with previous literature. This result is driven primarily by a negative relationship between charter school enroll- ment and local property tax revenue (column 3). We are unable to establish why this happens, but it does not appear to be a result of local property tax effort; as shown in column 4. Larger charter enrollment shares are associated with (insignificantly) higher property tax rates. This suggests that lower property tax revenues may be driven by declines in local property tax value. While perhaps surprising, this is consistent with studies finding that plausibly exogenous expansions of charter schooling can decrease local property val- ues (Brehm et al., 2017; Cook, 2018). This could be the case, for example, if charter schools are interpreted by homebuyers as a signal of poor local TPSD quality.
Given that we observe both lower spending and lower revenues in districts as local charter sectors expand, the net impact on districts’ overall financial health is ambiguous. The results in Table 4 suggest that these impacts are
Table 4. Fixed Effect Regressions Predicting Per-Pupil Revenue and Financial Health Measures.
State revenuea
Local revenue
Real estate tax rate (Mills)
Undesignated fund balance
Long-term debtaAlla
Real estate taxesa
(1) (2) (3) (4) (5) (6)
%age Charter 0.003 (0.002)
−0.004 (0.003)
−0.006* (0.003)
0.726 (0.538)
−81.107*** (20.631)
0.019* (0.009)
Reimbursed × 0.003+ −0.001 −0.001 0.203 9.420 −0.008 %age Charter (0.002) (0.001) (0.001) (0.329) (15.140) (0.006) Adj. R-sq 0.99 0.98 0.98 0.74 0.51 0.74 Districts 500 500 500 500 500 500 Observations 7960 7960 7960 6958 7960 6962
Note. Standard errors clustered on districts. All models include district and labor market-by-year fixed effects and control for the natural log of district enrollment and the shares of district students who are white, eligible for free- or reduced-price lunch, English learners, or receive special education services. aInverse hyperbolic sine-transformed. +p < .1, *p < .05, **p < .01 ***p < .001.
Mann and Bruno 1095
negative on balance. A percentage point increase in the share of district stu- dents enrolled in charter schools is associated with a decrease in per-pupil undesignated fund balances of roughly $80 per pupil (column 5), or more than five percent of the average balance we observe in districts during this period. Since these balances represent amounts available to districts to spend on any purpose, this suggests districts’ finances are strained by competitive pressure from charter schools, and declines in spending are insufficient to fully offset declines in available revenue.
The relationship between charter enrollment and districts’ undesignated fund balances is not significantly different between reimbursement and non- reimbursement years. This again suggests that reimbursements, when avail- able, are used primarily to bolster spending rather than to stabilize budgets. Consistent with this, the same percentage point increase in charter enrollment is associated with an increase in per-pupil long-term debt levels of roughly two percent (column 6).
RQ 4: Brick and Mortar vs. Cyber Charter Schools
Table 5 presents results from models similar to those presented earlier, but distinguishes the enrollment shares of brick and mortar charter schools from those of cyber charter schools. Results are qualitatively similar for both types of charter school but differences–while not always statistically significant– are generally consistent with cyber charter schools inducing relatively fewer competitive improvements in TPSDs. For example, while relationships between charter school enrollments and overall TPSD spending are similar for both types of charter school (column 1), in non-reimbursement years a percentage point increase in the share of students enrolling in charter schools is associated with a decrease in per-pupil instructional spending of roughly 0.6% if those students enroll in brick and mortar charter schools, but 1.4% if those students enroll in cyber charter schools (column 2).
Perhaps one reason TPSDs do not change spending is because of reports of poor instructional quality in the cyber charter school sector (Ahn & McEachin, 2017; CREDO, 2015). Further, coefficients on their respective interaction terms suggest that patterns substantially attenuate in the presence of tuition reimbursements. Differences in other major categories of expendi- ture between cyber and brick and mortar charter school enrollments are gen- erally not significant (columns 3–5), though they suggest that competition from brick and mortar charter schools is associated with decreases in spend- ing on non-instructional operations (e.g., student activities and community services). Competition from cyber charter schools is also associated with lower facilities spending.
1096
T ab
le 5
. R
es ul
ts D
is tin
gu is
hi ng
B ri
ck a
nd M
or ta
r C
ha rt
er S
ch oo
ls fr
om C
yb er
C ha
rt er
S ch
oo ls
.
Pe r-
pu pi
l s pe
nd in
ga Pe
rc en
ta ge
p ro
fic ie
nt o
r ad
va nc
ed Pe
r- pu
pi l l
oc al
re
ve nu
ea
R ea
l es
ta te
ta
x ra
te
(M ill
s) U
nd es
ig na
te d
fu nd
b al
an ce
Lo ng
-t er
m
de bt
a T
ot al
In st
ru ct
io n
In st
ru ct
io na
l su
pp or
t se
rv ic
es
N on
- In
st ru
ct io
na l
op er
at io
ns Fa
ci lit
ie s
R ea
di ng
M at
h Lo
ca l
re ve
nu e
R ea
l es
ta te
ta
xe s
(1 )
(2 )
(3 )
(4 )
(5 )
(6 )
(7 )
(8 )
(9 )
(1 0)
(1 1)
(1 2)
% ag
e Br
ic k
&
M or
ta r
− 0.
00 5*
* (0
.0 02
) −
0. 00
6* **
(0 .0
02 )
− 0.
00 6*
** (0
.0 01
) −
0. 02
6* (0
.0 11
) −
0. 04
2 (0
.0 28
) −
0. 15
7* (0
.0 65
) −
0. 13
3 (0
.1 12
) −
0. 00
5 (0
.0 03
) −
0. 00
7* (0
.0 03
) 0.
39 6
(0 .5
56 )
− 77
.0 20
** *
(2 2.
97 1)
0. 00
8 (0
.0 09
) R
ei m
bu rs
ed ×
%
ag e
B& M
0. 00
4* **
(0 .0
01 )
0. 00
1 (0
.0 01
) 0.
00 5*
(0 .0
02 )
0. 00
1 (0
.0 12
) 0.
02 9
(0 .0
27 )
0. 23
2* *
(0 .0
72 )
0. 23
3+
(0 .1
22 )
− 0.
00 1
(0 .0
01 )
− 0.
00 1
(0 .0
01 )
0. 17
3 (0
.3 21
) 18
.1 86
(1 4.
68 0)
− 0.
00 3
(0 .0
07 )
% ag
e C
yb er
− 0.
00 6
(0 .0
04 )
− 0.
01 4*
** (0
.0 03
) −
0. 00
5 (0
.0 03
) −
0. 00
7 (0
.0 19
) −
0. 18
3* (0
.0 88
) −
0. 44
4* (0
.1 77
) −
0. 34
2 (0
.2 33
) −
0. 00
1 (0
.0 05
) −
0. 00
0 (0
.0 05
) 2.
30 1*
(0 .9
35 )
− 89
.9 20
(5 8.
63 2)
0. 08
6* *
(0 .0
29 )
R ei
m bu
rs ed
×
% ag
e C
yb er
0. 00
7+
(0 .0
04 )
0. 01
5* **
(0 .0
03 )
0. 00
6+
(0 .0
03 )
− 0.
00 6
(0 .0
22 )
0. 07
7 (0
.0 91
) 0.
64 5*
** (0
.1 91
) 1.
07 0*
** (0
.2 65
) 0.
00 1
(0 .0
04 )
− 0.
00 2
(0 .0
04 )
0. 11
0 (0
.8 83
) −
65 .8
41 (5
7. 65
4) −
0. 04
4+
(0 .0
23 )
p( M
ai n
Ef fe
ct s)
0. 70
0. 04
0. 78
0. 40
0. 13
0. 13
0. 41
0. 43
0. 25
0. 06
0. 84
0. 01
p( In
te ra
ct io
ns )
0. 50
0. 00
0. 78
0. 75
0. 61
0. 04
0. 00
0. 72
0. 73
0. 94
0. 16
0. 10
A dj
. R -s
q 0.
79 0.
93 0.
93 0.
71 0.
44 0.
91 0.
93 0.
98 0.
98 0.
74 0.
51 0.
74 D
is tr
ic ts
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
50 0
O bs
er va
tio ns
78 76
78 76
79 60
79 60
79 60
79 58
79 58
79 60
79 60
69 58
79 60
69 62
N ot
e. S
ta nd
ar d
er ro
rs c
lu st
er ed
o n
di st
ri ct
s. A
ll m
od el
s in
cl ud
e di
st ri
ct a
nd la
bo r
m ar
ke t-
by -y
ea r
fix ed
e ffe
ct s
an d
co nt
ro l f
or t
he n
at ur
al lo
g of
d is
tr ic
t en
ro llm
en t
an d
th e
sh ar
es o
f d is
tr ic
t st
ud en
ts w
ho a
re w
hi te
, e lig
ib le
fo r
fr ee
- or
r ed
uc ed
-p ri
ce lu
nc h,
E ng
lis h
le ar
ne rs
, o r
re ce
iv e
sp ec
ia l e
du ca
tio n
se rv
ic es
. p (M
ai n
Ef fe
ct s)
a nd
p(
In te
ra ct
io ns
) ar
e tw
o- si
de d
p- va
lu es
fr om
W al
d te
st s
of t
he e
qu al
ity o
f t he
c oe
ffi ci
en ts
o n
th e
m ai
n ef
fe ct
s an
d in
te ra
ct io
ns , r
es pe
ct iv
el y,
fo r
br ic
k an
d m
or ta
r an
d cy
be r
ch ar
te r
en ro
llm en
t sh
ar es
. a In ve
rs e
hy pe
rb ol
ic s
in e-
tr an
sf or
m ed
. + p <
.1 , *
p <
.0 5,
* *p
< .0
1 **
*p <
.0 01
.
Mann and Bruno 1097
Results considering student achievement (columns 6 and 7) are consistent with instructional expenditures being an important mechanism by which charter schools affect TPSD student outcomes. In non-reimbursement years, a percentage point increase in the share of students enrolled in brick and mor- tar charter schools is associated with 0.16 percentage points fewer TPSD students who are proficient in reading, but a similar increase in cyber charter school enrollment is associated with 0.44 percentage points fewer TPSD stu- dents who are proficient in reading. Similarly, the coefficients on respective interaction terms suggest these relationships attenuate at higher levels in cyber charter schools relative to brick and mortar charter schools during reimbursement years. The pattern is similar in math.
As discussed, previous research shows that charter schools are associated with reduced property tax revenue. As shown in Table 5, we observe this pattern in Pennsylvania only for brick and mortar charter school enrollment (column 9). This is consistent with brick and mortar charter schools signal- ing low school quality to homeowners or potential homeowners. These rela- tionships are not significantly different from one another (p = .25) and cyber charter school enrollment shares are positively related to local property tax rates (column 10). We thus cannot rule out that cyber charter schools also relate to declines in local property values that are offset by increases in mill- age rates.
Both cyber and brick and mortar charter school enrollments are associated with lower per-pupil undesignated fund balances in TPSDs (column 11), sug- gesting similar financial stress associated with both types of charter school. Only cyber charter school enrollments are associated with TPSD debt levels (column 12). This may seem odd given the fund balance results, but districts may take out long-term debt for a variety of reasons that may not always indicate fiscal duress (e.g., necessary capital improvements), and as shown in Appendix Table A1, our long-term debt results are sensitive to the inclusion of district time trends. These results should thus be interpreted with caution.
Robustness Checks
Our use of district and labor market-by-year fixed effects affords us consider- able protection from bias due to unobserved factors. Our estimates may nev- ertheless be biased by unobserved factors varying within districts over time that are associated with charter school enrollment shares and outcomes of interest if those factors do not vary uniformly within labor markets. In Appendix Table A1 we present results from models predicting our outcomes of primary interest that also include quadratic or cubic district-specific time trends. If these time trends are driven by districts’ charter enrollment, their
1098 Educational Policy 36(5)
inclusion in the model will tend to bias estimates toward zero, making these results conservative. Moreover, if these trends explain virtually all of the residual variation in our outcomes, the resulting estimates may be unstable and identified primarily from measurement error.
Nevertheless, including these time trends serves as a useful test of whether our results are driven to a substantial degree by differing pre-existing district trends. The results controlling for the cubic trends are of particular interest because our panel spans not only a period of economic growth leading up to the Great Recession, but also the subsequent downturn and recovery. These changes likely had major implications for district budgets and outcomes. If these implications differed for districts within a labor market in ways corre- lated with charter school enrollment, our results may be biased even net of our fixed effects. For example, the elimination of tuition reimbursements was just one part of a large set of education budget cuts enacted for the 2011–12 school year, and those other cuts may disproportionately have affected dis- tricts receiving more reimbursements. If so, allowing districts’ time trends to differ flexibly (e.g., not simply linearly) should reveal whether unobserved factors substantially affect our estimates.
Appendix Table A1 suggests that these trends are not a major concern in most cases. Our student achievement, total spending, and financial health results change very little in the presence of linear time trends, particularly when we do not attempt to distinguish enrollments in different types of char- ter school. Our results for specific categories of expenditure are somewhat sensitive to these checks, particularly for spending on non-instructional oper- ations and facilities, but these represent small portions of districts’ budgets. Estimates of the relationship between charter enrollment shares and districts’ property tax revenues also change in these specifications, but because these models explain virtually all of the variation in property tax revenue, these coefficients may not be well identified.
While we should be cautious about interpreting specific differences in the composition of districts’ revenues and expenditures, our main results are robust to relaxing the assumption that districts would have parallel trends in their outcomes in the absence of changes in charter school enrollment. Some estimates even increase in magnitude in these specifications. Whether we allow districts their own time trends or not, larger shares of students enrolled in charter schools are associated with greater financial strain and lower stu- dent achievement in TPSDs. The relationship with student achievement attenuates when tuition reimbursements are available, but the relationship with financial health does not, perhaps because districts use the additional revenue to bolster their instructional programs rather than maintaining their reserves or paying down debts.
Mann and Bruno 1099
Another potential issue we encountered was that district enrollments in brick and mortar charter schools are weakly correlated with enrollments in cyber charter schools. It may therefore be that the different relationships we observe for the different kinds of charter schools reflect differences between districts that experience competition only from one or the other type of char- ter school. However, we observed similar results, available upon request, when considering only the 346 districts for which both cyber and brick and mortar charter school enrollment shares were larger in 2015–16 than in 2000–01 (i.e., the districts for which there was net enrollment growth in both types of charter school during this time).
The final robustness check we employed was to run the analyses without Philadelphia’s school district in the dataset. The reason for this check is that Philadelphia has long been a central focus of education reformers, has had unique charter authorization arrangements, and it is the largest district in the state. These outlier possibilities could skew findings. Our findings remained the same when removing Philadelphia.
Discussion and Policy Recommendations
Our study shows that larger shares of students enrolling in charter schools across Pennsylvania associate with decrements to spending, financial health, and student achievement in sending districts. However, we show some of these relationships substantially or completely attenuate during years in which dis- tricts receive partial reimbursement for revenue losses associated with charter school enrollment. Districts facing charter competition spend relatively more on instruction and instructional support services when they receive reimburse- ments, but they spend no more on facilities or non-instructional operations.
This research adds to a growing body of literature indicating that, what- ever the other benefits, school choice schemes can impose financial hardship on TPSDs, especially if there are no available supplementary sources of rev- enue such as tuition reimbursements. These factors should be considered in jurisdictions where school choice programs are expanding. We find some of the first direct evidence that competitive hardship can be alleviated by state- level policymakers.
Another noteworthy aspect of our findings is that academic achievement relates to changes in reimbursements and spending. The relationship between charter school enrollment and student achievement in TPSDs is less nega- tive–or even positive–when reimbursements are offered, though relationships with district financial health are unchanged. These results are unchanged–and in some cases increase in magnitude–when we relax the parallel trends assumption common to difference-in-difference estimation strategies. We
1100 Educational Policy 36(5)
also find suggestive evidence that cyber charter schools impose somewhat different, and perhaps weaker, competitive pressures on TPSDs than brick and mortar counterparts. Previous studies show inconsistent results in regards to competitive effects in school choice environments (Arum, 1996; Bettinger, 2005; Buddin & Zimmer, 2009). Our findings show the influence of school type and policy construction (particularly related to funding) on whether competitive effects occur in choice environments.
These findings have implications for both policy and the theory of action of school choice. A key metaphor in explaining the theory of action is that “a rising tide lifts all boats” (Hoxby, 2001, 2003). We remind leaders and poli- cymakers that they need to carefully navigate this tide to ensure lifting, or else the imposition of competitive markets has the potential to sink the capa- bilities and outcomes of TPSDs. One key element in Pennsylvania is extra funding ensures that TPSDs can effectively compete with charter schools.
Another interesting finding is that results change when considering cyber charter schools. This may happen for a few reasons. One possibility is that because they are neither generally authorized by a district nor physically located in a district’s boundaries, cyber charters are less salient than brick and mortar charters to TPSD administrators. Similarly, to the extent that cyber charter schools have relatively poor student achievement outcomes (e.g., Ahn & McEachin, 2017; CREDO, 2015), TPSD administrators may be inclined to compete with them in ways unrelated to instructional quality (e.g., aforemen- tioned non-instructional operations).
These findings suggest that the nature of policymaking for cyber and brick and mortar charter schools should reflect the differences between these types of schools. We are unable to fully explain these differences here because we are only able to observe the response in terms of resource allocations. School districts may use other tactics not captured through observations of finances. One example, as shown in previous research, is districts have used a variety of restructuring strategies to create their own online schools, but still could not recapture students (Mann, 2019). Due to this lack of success, perhaps TPSDs do not allocate additional finances because they believe students leave anyway. Understanding strategies like these requires more research to determine the full range of responses, but based on our study the responses are different to brick and mortar charters.
Taken together, our findings suggest TPSDs make competitive improve- ments to their educational programs in the presence of charter schools only if they have additional financial resources to do so. This means funding effective competitive environments is costly. These findings help to reconcile inconsis- tent results from prior studies because we provide some of the first evidence that straightforward policy choices help sustain and improve TPSD systems
Mann and Bruno 1101
experiencing competition. This evidence has implications for policymakers and for researchers seeking to understand how competition affects statewide systems. States considering or experiencing expansions of school choice pro- grams may wish to consider adopting policies, such as tuition reimbursements, that help to insulate districts from associated competitive pressure. However, the costs of such policies need to be weighed carefully against their benefits.
The recommendation of re-adopting tuition reimbursements in Pennsylvania begs the question: Is it possible for a state to adopt such an expensive reform? One of the promises of school choice is that the market is more cost efficient because it offers a cost-efficient mechanism for effectively determining qual- ity schools and seeing others close or fail (Hoxby, 2001). Our study shows effective choice environments are more expensive than other arrangements because maintaining multiple school sectors requires duplicating programs and services. In short, excess cost could be a natural byproduct of choice, not the other way around. Honest conversations about the cost of effectively implementing choice plans are necessary and timely. Noting that traditional public schools may require more resources to compete should be part of this conversation.
In our conversations with school leaders across school types (charter schools and traditional public schools), there seems to be general agreement about reinstating tuition reimbursements. The reason for this is not ideologi- cal, but rather practical. There is less tension between organizations when one is not perceived as taking resources from the other, and charter advocates may feel less political pressure from traditional public school advocacy groups if choice happens in such a way where traditional public school dis- tricts do not financially suffer. This means there is potential for a broad coali- tion of support for a policy that offers choice with financial guardrails for the organizations participating in the competitive environment.
However, reimbursement policy likely may not please fiscal conservatives seeking to enhance market logics while reducing costs of school services. Winning supporters of reimbursement policy from this group is challenging, but Pennsylvania did it in the past with some Republicans supporting reim- bursement early in charter school adoption. Additionally, more Republicans favor increasing per-pupil spending than reducing it in recent education spending polls (EducationNext, 2020). It was the austerity-minded politics after the 2008 recession that led to the reimbursements being cut. As the state’s financial conditions improve and lawmakers push for more school options, it may be politically possible to support financially struggling school districts through reinstating reimbursements. These reimbursements could alleviate resource drains and elevate possibilities for healthier, and perhaps even effectively competitive, choice environments.
1102
A pp
en di
x A
– R
ob us
tn es
s C
he ck
s
T ab
le A
1. F
ix ed
E ffe
ct R
eg re
ss io
ns w
ith D
is tr
ic t-
Sp ec
ifi c
T im
e T
re nd
s.
Pe r-
Pu pi
l E xp
en di
tu re
sa Pe
rc en
ta ge
P ro
fic ie
nt o
r A
dv an
ce d
T
ot al
In st
ru ct
io n
In st
ru ct
io na
l Su
pp or
t Se
rv ic
es
N on
- In
st ru
ct io
na l
O pe
ra tio
ns Fa
ci lit
ie s
R ea
di ng
M at
h
R ea
l E st
at e
T ax
es
Pe r
Pu pi
la
U nd
es ig
na te
d Fu
nd
Ba la
nc e
Pe r
Pu pi
l
Lo ng
-T er
m
D eb
t Pe
r Pu
pi la
(1
) (2
) (3
) (4
) (5
) (6
) (7
) (8
) (9
) (1
0) (1
1) (1
2) (1
3) (1
4) (1
5) (1
6) (1
7) (1
8) (1
9) (2
0)
% ag
e C
ha rt
er −
0. 00
4* (0
.0 02
) −
0. 00
4* (0
.0 02
) −
0. 00
6* **
(0 .0
01 )
− 0.
00 6*
** (0
.0 01
) −
0. 00
3+
(0 .0
02 )
− 0.
00 3+
(0 .0
02 )
− 0.
00 2
(0 .0
10 )
− 0.
00 2
(0 .0
10 )
− 0.
01 9
(0 .0
31 )
− 0.
01 9
(0 .0
32 )
− 0.
19 5*
(0 .0
92 )
− 0.
19 4*
(0 .0
96 )
− 0.
25 2*
(0 .1
23 )
− 0.
25 2+
(0 .1
28 )
0. 00
3 (0
.0 02
) 0.
00 3
(0 .0
03 )
− 10
0. 06
** (3
2. 80
) −
10 0.
04 **
(3 4.
14 )
0. 01
4 (0
.0 10
) 0.
01 4
(0 .0
10 )
R ei
m bu
rs ed
x
% ag
e C
ha rt
er
0. 00
4* **
(0 .0
01 )
0. 00
4* **
(0 .0
01 )
0. 00
2 (0
.0 01
) 0.
00 2
(0 .0
01 )
0. 00
4* (0
.0 02
) 0.
00 4*
(0 .0
02 )
− 0.
00 1
(0 .0
10 )
− 0.
00 1
(0 .0
11 )
0. 03
2 (0
.0 25
) 0.
03 2
(0 .0
26 )
0. 25
8* **
(0 .0
72 )
0. 25
8* **
(0 .0
75 )
0. 31
1* (0
.1 33
) 0.
31 1*
(0 .1
38 )
− 0.
00 3*
** (0
.0 01
) −
0. 00
3* *
(0 .0
01 )
19 .8
9 (1
4. 81
) 19
.8 8
(1 5.
41 )
− 0.
00 8
(0 .0
06 )
− 0.
00 8
(0 .0
06 )
A dj
. R -s
q 0.
79 0.
77 0.
96 0.
95 0.
95 0.
95 0.
82 0.
80 0.
50 0.
45 0.
93 0.
92 0.
94 0.
94 0.
99 0.
99 0.
63 0.
60 0.
86 0.
84
M od
el s
D ist
in gu
ish in
g En
ro llm
en ts
in B
ric k
an d
M or
ta r
an d
Cy be
r Ch
ar te
r Sc
ho ol
s
% ag
e Br
ic k
& M
or ta
r −
0. 00
3+
(0 .0
02 )
− 0.
00 3+
(0 .0
02 )
− 0.
00 5*
** (0
.0 01
) −
0. 00
5* **
(0 .0
02 )
− 0.
00 3
(0 .0
02 )
− 0.
00 3
(0 .0
02 )
− 0.
00 3
(0 .0
10 )
− 0.
00 3
(0 .0
11 )
0. 01
6 (0
.0 28
) 0.
01 6
(0 .0
29 )
− 0.
10 9
(0 .0
91 )
− 0.
10 8
(0 .0
94 )
− 0.
12 7
(0 .1
31 )
− 0.
12 7
(0 .1
37 )
0. 00
3 (0
.0 03
) 0.
00 3
(0 .0
03 )
− 83
.6 8*
(3 5.
99 )
− 83
.6 5*
(3 7.
46 )
0. 01
3 (0
.0 09
) 0.
01 3
(0 .0
09 )
R ei
m bu
rs ed
x
% ag
e B&
M 0.
00 4*
** (0
.0 01
) 0.
00 4*
** (0
.0 01
) 0.
00 1
(0 .0
01 )
0. 00
1 (0
.0 01
) 0.
00 4*
(0 .0
02 )
0. 00
4* (0
.0 02
) 0.
00 1
(0 .0
11 )
0. 00
1 (0
.0 12
) 0.
03 1
(0 .0
26 )
0. 03
1 (0
.0 28
) 0.
21 4*
* (0
.0 75
) 0.
21 4*
* (0
.0 78
) 0.
20 1
(0 .1
30 )
0. 20
1 (0
.1 35
) −
0. 00
2* (0
.0 01
) −
0. 00
2* (0
.0 01
) 27
.2 7+
(1 5.
87 )
27 .2
5+
(1 6.
52 )
− 0.
00 7
(0 .0
06 )
− 0.
00 7
(0 .0
07 )
% ag
e C
yb er
− 0.
00 5
(0 .0
05 )
− 0.
00 5
(0 .0
05 )
− 0.
00 9*
* (0
.0 03
) −
0. 00
9* *
(0 .0
03 )
− 0.
00 3
(0 .0
03 )
− 0.
00 3
(0 .0
03 )
0. 00
8 (0
.0 17
) 0.
00 8
(0 .0
18 )
− 0.
11 4
(0 .0
97 )
− 0.
11 4
(0 .1
01 )
− 0.
57 8*
* (0
.1 80
) −
0. 57
7* *
(0 .1
87 )
− 0.
96 5*
** (0
.2 32
) −
0. 96
4* **
(0 .2
42 )
0. 00
4 (0
.0 04
) 0.
00 4
(0 .0
04 )
− 11
9. 95
* (5
3. 81
) −
11 9.
93 *
(5 6.
04 )
0. 02
1 (0
.0 28
) 0.
02 1
(0 .0
29 )
R ei
m bu
rs ed
x
% ag
e C
yb er
0. 00
4 (0
.0 05
) 0.
00 4
(0 .0
05 )
0. 01
1* **
(0 .0
03 )
0. 01
1* **
(0 .0
03 )
0. 00
4 (0
.0 03
) 0.
00 4
(0 .0
03 )
− 0.
01 6
(0 .0
18 )
− 0.
01 6
(0 .0
19 )
0. 01
3 (0
.1 05
) 0.
01 3
(0 .1
10 )
0. 59
7* *
(0 .1
94 )
0. 59
7* *
(0 .2
02 )
1. 22
7* **
(0 .3
03 )
1. 22
6* **
(0 .3
15 )
− 0.
00 8+
(0 .0
04 )
− 0.
00 8+
(0 .0
04 )
− 58
.8 8
(5 6.
71 )
− 58
.9 4
(5 9.
07 )
− 0.
01 3
(0 .0
23 )
− 0.
01 3
(0 .0
24 )
Q ua
d. T
re nd
X X
X X
X X
X X
X X
C ub
ic T
re nd
X X
X X
X X
X X
X X
p( M
ai n
Ef f.)
0. 78
0. 79
0. 21
0. 23
0. 81
0. 82
0. 53
0. 55
0. 21
0. 22
0. 02
0. 02
0. 00
0. 00
0. 82
0. 82
0. 54
0. 55
0. 76
0. 77
p( In
te ra
ct io
ns )
0. 96
0. 96
0. 00
0. 00
0. 87
0. 88
0. 38
0. 40
0. 87
0. 87
0. 07
0. 08
0. 00
0. 00
0. 25
0. 26
0. 15
0. 17
0. 82
0. 83
A dj
. R -s
q 0.
79 0.
77 0.
96 0.
95 0.
95 0.
95 0.
82 0.
80 0.
50 0.
46 0.
93 0.
92 0.
94 0.
94 0.
99 0.
99 0.
63 0.
60 0.
86 0.
84
D is
tr ic
ts 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0 50
0
O bs
er va
tio ns
78 76
78 76
78 76
78 76
79 60
79 60
79 60
79 60
79 60
79 60
79 58
79 58
79 58
79 58
79 60
79 60
79 60
79 60
69 62
69 62
N ot
e. S
ta nd
ar d
er ro
rs c
lu st
er ed
o n
di st
ri ct
s. A
ll m
od el
s in
cl ud
e di
st ri
ct a
nd la
bo r
m ar
ke t-
by -y
ea r
fix ed
e ffe
ct s
an d
co nt
ro l f
or t
he n
at ur
al lo
g of
d is
tr ic
t en
ro llm
en t
an d
th e
sh ar
es o
f d is
tr ic
t st
ud en
ts w
ho a
re w
hi te
, e lig
ib le
fo r
fr ee
- or
r ed
uc ed
-p ri
ce lu
nc h,
E ng
lis h
le ar
ne rs
, o r
re ce
iv e
sp ec
ia l e
du ca
tio n
se rv
ic es
a nd
, i f i
nd ic
at ed
, a d
is tr
ic t-
sp ec
ifi c
lin ea
r tim
e tr
en d.
M od
el s
ar e
ot he
rw is
e sp
ec ifi
ed a
s de
sc ri
be d
in T
ab le
s 2–
5. a In
ve rs
e hy
pe rb
ol ic
s in
e- tr
an sf
or m
ed . +
p <
.1 , *
p <
.0 5,
* *p
< .0
1 **
*p <
.0 01
.
Mann and Bruno 1103
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publi- cation of this article.
ORCID iD
Bryan A. Mann https://orcid.org/0000-0003-1907-8804
Notes
1. While differences in response may lead to differences in results, variance in find- ings may also relate to research design (Betts, 2009). For example, Jinnai (2014) shows positive impacts of competition in overlapping grades, something noticed as a potential flaw in previous research. In New York City, proximity, including colocation, also played a role in uncovering positive spillover effects from char- ter schools (Cordes, 2018).
2. The state-created reform panel arrangement is rare. However, it is noteworthy because Philadelphia, which is by far the largest district in Pennsylvania, had this arrangement and did not revert to local control until 2017 (after the end of our study). For this reason, as shown later, we run the analysis without Philadelphia in the models. We observe consistent results.
3. Our dataset includes 500 school districts prior to 2009–2010. In that year two districts merged to form a new district, leaving 499 districts in our dataset. This would result in 501 unique school districts, though in most of our analyses we use labor market-by-year fixed effects, resulting in the exclusion of one district that is the only Pennsylvania district in a labor market that extends into New Jersey.
4. This is a pseudonym. 5. Our finding that charter school enrollment is associated with proportionally large
decreases in spending on facilities is somewhat at odds with previous work that shows that such capital-related costs are largely fixed for districts losing enroll- ment to charter schools (e.g., Cook, 2018). However, as shown in Table 1, dis- tricts in Pennsylvania report spending only an average of $45 per pupil per year on facilities during this time. Indeed, most districts report no such spending in most years and many districts never report facilities spending at all. Moreover, as shown in Appendix Table A1, our estimates of facilities spending are sensitive to the inclusion of district-specific time trends. We thus believe our facilities results should be interpreted with caution.
6. A disadvantage of the SEDA data is that they include some charter schools in district achievement figures. This may contaminate results insofar as we are
1104 Educational Policy 36(5)
interested in student performance in TPSDs, and not in charter schools that oper- ate within the district. However, SEDA’s district estimates do exclude charter schools from district figures when they offer primarily online instruction. This alleviates some concerns about contamination. Additionally, our results using SEDA data are consistent with our results using uncontaminated proficiency rate data, which suggests any remaining contamination is small in magnitude.
7. While districts’ tuition payments to charter schools are accounted as instructional expenditures, they are not attributed to specific revenue sources. This means there is no obvious way to adjust district revenues for these payments.
References
Ahn, J., & McEachin, A. (2017). Student enrollment patterns and achievement in Ohio’s online charter schools. Educational Researcher, 46(1), 44–57.
Arsen, D., & Ni, Y. (2008). The competitive effect of school choice policies on perfor- mance in traditional public schools. National Education Policy Center.
Arsen, D., & Ni, Y. (2012). The effects of charter school competition on school dis- trict resource allocation. Educational Administration Quarterly, 48(1), 3–38.
Arum, R. (1996). Do private schools force public schools to compete? American Sociological Review, 61, 29–46.
Berends, M., Cannata, M., & Goldring, E. B. (2011). School choice and school improvement. Harvard Education Press.
Berends, M., Goldring, E. B., Stein, M., & Cravens, X. (2010). Instructional condi- tions in charter schools and students’ mathematics achievement gains. American Journal of Education 116(3), 303–335.
Bettinger, E. P. (2005). The effect of charter schools on charter students and public schools. Economics of Education Review, 24(2), 133–147.
Betts, J. R. (2009). The competitive effects of charter schools on traditional pub- lic schools. In M. Berends, M. G. Springer, Dl Ballou, & H. J. Walberg (Eds.), Handbook of research on school choice (pp. 195–208). Routledge.
Bifulco, R., & Ladd, H. (2006). The impacts of charter schools on student achieve- ment: Evidence from North Carolina. Education Finance and Policy, 1(1), 50–90.
Bifulco, R., & Reback, R. (2013). Fiscal impacts of charter schools: Lessons from New York. Education Finance and Policy, 9(1), 86–107.
Brehm, M., Imberman, S. A., & Naretta, M. (2017). Capitalization of charter schools into residential property values. Education Finance and Policy, 12(1) 1–27. https://doi.org/10.1162/EDFP_a_00192
Bruno, P. (2019). Charter competition and district finances: Evidence from California. Journal of Education Finance, 44(4), 361–384.
Buddin, R., & Zimmer, R. (2009). Is charter school competition in California improv- ing the performance of traditional public schools? Public Administration Review, 69(5), 831–845.
Burbidge, J. B., Magee, L., & Robb, A. L. (1988). Alternative transformations to han- dle extreme values of the dependent variable. Journal of the American Statistical Association, 83(401), 123–127. https://doi.org/10.2307/2288929.
Mann and Bruno 1105
Center for Research on Educational Outcomes (CREDO). (2015). Online charter school study. CREDO.
Chubb, J. E., & Moe, T. M. (1990). America's public schools: Choice is a panacea. The Brookings Review, 8(3), 4–12.
Cook, J. B. (2018). The effect of charter competition on unionized district revenues and resource allocation. Journal of Public Economics, 158, 48–62.
Cordes, S. A. (2018). In pursuit of the common good: The spillover effects of charter schools on public school students in New York City. Education Finance and Policy, 13(4), 484–512.
DeVos, B. (2017). Prepared Remarks by Secretary DeVos to JFK Jr. Forum at Harvard Kennedy School. United States Department of Education. Retrieved from https://www.ed.gov/news/speeches/prepared-remarks-secretary-devos- jfk-jr-forum-harvard-kennedy-school.
DiMaggio, P., & Powell, W. W. (1983). The iron cage revisited: Collective rational- ity and institutional isomorphism in organizational fields. American Sociological Review, 48(2), 147–160.
EducationNext. (2020). Results from the 2019 Education Next Poll. EducationNext. https://www.educationnext.org/2019-ednext-poll-interactive/
Epple, D., Romano, R., & Zimmer, R. (2016). Charter schools: A survey of research on their characteristics and effectiveness. Handbook of the Economics of Education, 5, 139–208.
Fitzgerald, T., Worden, A., & Wiggins, O. (2002, June 29). Crunch time for Pennsylvania budget; The deadline is tomorrow. A draft calls for an additional $75 million to Phila. schools and tripling the cigarette tax. The Philadelphia Inquirer. Retrieved March 6, 2018.
Friedman, M. (1955). The role of government in education. In R. A. Solo (Ed.), Economics and the public interest (pp. 127–134). Rutgers University Press.
Griffith, B. K. (2014). The financial impact of charter school’s enrollment on tra- ditional public school expenditures, resource allocation and programming (Publication No. 3647450) [Doctoral dissertation, Penn State University]. ProQuest Dissertations Publishing.
Hardy, D. (2015, April 24). How are charter schools funded in Pennsylvania? Keystone Crossroads. Retrieved from http://crossroads.newsworks.org/index.php/local/ keystone-crossroads/81092-multiple-choices-how-are-charter-schools-funded
Hoxby, C. M. (2001). Rising tide. Education Next, 1(4), 68–74. Hoxby, C. M. (2003). School choice and school productivity (or, could school choice
be a tide that lifts all boats?). In C.M. Hoxby (Ed.), The economics of school choice (pp. 287–341). University of Chicago Press.
Imberman, S. (2011). The effect of charter schools on achievement and behavior of public school students. Journal of Public Economics, 95(7–8), 850–863.
Jabbar, H. (2015). “Every kid is money” Market-like competition and school leader strategies in New Orleans. Educational Evaluation and Policy Analysis, 34(4), 638–659.
1106 Educational Policy 36(5)
Jabbar, H., Fong, C. J., Germain, E., Li, D., Sanchez, J., Sun, W., & Devall, M. (2019). “The competitive effects of school choice on student achievement: A systematic review. Educational Policy, https://doi.org/10.1177/0895904819874756.
Jinnai, Y. (2014). Direct and indirect impact of charter schools’ entry on traditional public schools: New evidence from North Carolina. Economics Letters, 124(3), 452–456. https://doi.org/10.1016/j.econlet.2014.07.016
Ladd, H. F., & Singleton, J. D. (2018). The Fiscal Externalities of Charter Schools: Evidence from North Carolina (SSRN Scholarly Paper No. ID 3082968). Social Science Research Network. Retrieved from https://papers.ssrn.com/ abstract=3082968
Langley, K. (2016, January 7). Charter school advocates criticize Gov. Wolf’s dis- tribution of grant money. Pittsburgh Post-Gazette. Retrieved March 06, 2018.
Lubienski, C. (2003). Innovation in education markets: Theory and evidence on the impact of competition and choice in charter schools. American Educational Research Journal, 40(2), 395–443.
Mann, B. (2019). Compete, conform, or both? School district responses to statewide cyber charter schools. Journal of School Choice, 1–26.
Mann, B., & Baker, D. (2019). Cyber charter schools and growing resource inequal- ity among public districts: Geospatial patterns and consequences of a statewide choice policy in Pennsylvania, 2002-2014. American Journal of Education, 125(2).
Mann, B., Kotok, S., Frankenberg, E., Fuller, E., & Schafft, K. (2016). Choice, cyber charter schools, and the educational marketplace for rural school districts. The Rural Educator, 37(3).
National Alliance for Public Charter Schools (2018). Our work. Retrieved September 25, 2018, from https://www.publiccharters.org/our-work
National Center for Education Statistics (2017). Public Charter School Enrollment. (March). Retrieved March 6, 2018 from https://nces.ed.gov/programs/coe/ indicator_cgb.asp.
National Center on School Choice (NCSC). (2011). How do principals respond to charter school competition? Research brief. Vanderbilt University, M. Cannata.
Pennsylvania Department of Education. (2004). “Charter Schools.” Retrieved August 8, 2016, from http://www.education.pa.gov/Documents/Codes%20 and%20Regulations/Basic%20Education%20Circulars/Purdons%20Statutes/ Charter%20Schools.pdf.
Pew Charitable Trusts. (2015). How Charter School Governance in Pennsylvania and Philadelphia Measures Up. The Pew Charitable Trusts. Retrieved from https:// www.pewtrusts.org/-/media/assets/2015/12/overseeingcharterschools_v5.pdf
Preston, C., Goldring, E., Berends, M., & Cannata, M. (2012). School innovation in district context: Comparing traditional public schools and charter schools. Economics of Education Review, 31(2), 318–330.
Reardon, S. F., Ho, A. D., Shear, B. R., Fahle, E. M., Kalogrides, D., & DiSalvo, R. (2017). Stanford Education Data Archive (Version 2.0). http://purl.stanford.edu/ db586ns4974.
Mann and Bruno 1107
Renzulli, L. A., Barr, A. B., & Paino, M. (2015). Innovative education? A test of spe- cialist mimicry or generalist assimilation in trends in charter school specialization over time. Sociology of Education, 88(1), 83–102.
Ridley, M., & Terrier, C. (2018). Fiscal and Education Spullovers from Charter School Expansion (Discussion Paper #2018.02). School Effectiveness & Inequality Initiative.
Sass, T., Zimmer, R., Gill, B., & Booker, K. (2016). Charter high school’s effect on educational attainment and earnings. Journal of Policy Analysis and Management, 35(3), 683–706.
Taylor, L. L., & Fowler, W. J. J. (2006). A Comparable Wage Approach to Geographic Cost Adjustment (No. NCES 2006-321). U.S. Department of Education, National Center for Education Statistics.
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.
Wohlstetter, P., Smith, J., & Farrell, C. C. (2013). Choices and challenges: Charter school performance in perspective. Harvard Education Press.
Author Biographies
Bryan Mann is Assistant Professor in the Department of Educational Leadership and Policy Studies at the University of Kansas. His research focuses on education policy and geography, specifically on school enrollment patterns as they relate to segregation and diversity, school choice, and alternative models of education.
Paul Bruno is Assistant Professor in the Department of Education Policy, Organi- zation, and Leadership at the University of Illinois at Urbana-Champaign. He studies a range of K-12 education policy issues related to teacher quality, teacher hiring, school finance, and school choice.