As a Matter of Fact:
The National
Charter School
Study III 2023
LAST UPDATED 6/19/2023
As a Matter of Fact:
The National Charter
School Study III 2023
Executive Summary
Authors
Margaret E. Raymond, Ph.D.
James L. Woodworth, Ph.D., Lead Analyst- 31 State Study
Won Fy Lee, Ph.D., Lead Analyst- CMO Study
Sally Bachofer, Ed.M.
Contributors
Meghan E. Cotter Mazzola, M.S.
William D. Snow
Tzvetelina Sabkova, M.A.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023 1
© 2023 CREDO
Center for Research on Education Outcomes
Stanford University
Stanford, CA
https://credo.stanford.edu
CREDO, the Center for Research on Education Outcomes at Stanford University, aims to improve empirical
evidence about education reform and student performance at the primary and secondary levels. CREDO
at Stanford University supports education organizations and policy makers in using reliable research and
program evaluation to assess the performance of education initiatives. CREDO’s valuable insight helps
educators and policy makers strengthen their focus on the results of innovative programs, curricula, policies,
and accountability practices.
Acknowledgments
CREDO gratefully acknowledges the support of the state education agencies that contributed their data to
this partnership. Our data access partnerships form the foundation of CREDO’s work, without which studies
like this would be impossible. We strive daily to justify the condence placed in us.
The research presented here uses condential data from state departments of education. The views
expressed herein do not necessarily represent the positions or policies of the organizations noted above.
No ocial endorsement of any product, commodity, service or enterprise mentioned in this publication is
intended or should be inferred. In addition:
> The research presented here utilizes SLDS Data from the Idaho State Board of Education (SBOE) and the
Idaho State Department of Education. Any research errors are the sole responsibility of the author(s).
> This research result used data structured and maintained by the MERI-Michigan Education Data Center
(MEDC). MEDC data is modied for analysis purposes using rules governed by MEDC and is not identical
to data collected and maintained by the Michigan Department of Education (MDE) and/or Michigan’s
Center for Educational Performance and Information (CEPI). Results, information and opinions solely
represent the analysis, information and opinions of the author(s) and are not endorsed by, or reect the
views or positions of, grantors, MDE and CEPI or any employee thereof.
> Data for this report was provided by the Missouri Department of Elementary and Secondary Education.
> The conclusions of this research do not necessarily reect the opinions or ocial position of the Texas
Education Agency, the Texas Higher Education Coordinating Board, or the State of Texas.
The analysis and conclusions contained herein are exclusively those of the authors and are not endorsed by
any of CREDO’s supporting organizations, their governing boards, or the state governments, state education
departments or school districts that participated in this study. All errors are attributable to the authors.
CREDO also acknowledges the Walton Family Foundation and The City Fund for supporting this research.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 32
Project Description
As a Matter of Fact: The National Charter School Study III 2023 (NCSSIII) is the third national study by CREDO
evaluating the academic progress of students enrolled in charter schools in the United States. The current
report presents ndings from 2014 to 2019, which yields four periods of year-to-year student growth as
measured by state achievement tests. It includes data from 29 states plus Washington, D.C., and New
York City, which for convenience we report as 31 states. In addition, because we have used a common
methodology across the three studies, we can combine results into trends to support insights of the
performance of students enrolled in charter schools over the past 15 years.
To organize the extensive body of this current research eort, CREDO separated the analysis into two parts
and produced two reports: (1) Charter School Performance in 31 States (CSP31) and (2) Charter Management
Organization 2023 (CMO23). CSP31 examines the performance of the full set of charter school students
and schools, while CMO23 analyzes the dierence in academic growth between students attending charter
schools associated with charter management organizations (CMOs) and those attending stand-alone
charterschools (SCS).
1
We present this combined Executive Summary for both reports as well as common
Summary of Findings, Conclusions and Implications to ensure we present the fullest picture of performance
in charterschools.
Our work deliberately focuses on a specic outcome: the annual progress that students make over an
academic year. In this report, we look at students in charter schools compared to the experience they
would have had in the traditional public schools (TPS) they would otherwise have attended. One notable
limitation of this approach is that we have limited line of sight “under the hood” and into the role that
localized environmental, regulatory and organizational factors play on individual school performance. Our
contribution to the K-12 education research and practice landscape is to test fundamental questions of the
eectiveness of charter schools and highlight outcomes and trends rooted in academic progress.
A study of the academic impacts of charter schools on their students is timely. Insights about the educational
eectiveness of schools, school operators, K-12 academic programs and education policy are valuable
today more than ever. The 2022 results from the National Assessment of Educational Progress removed any
ambiguity about student learning after the COVID-19 pandemic. As a country, student academic performance
has regressed by two decades in math and fallen steeply in reading, with the most severe performance
declines found among minority, poverty and special needs populations that were already struggling before
the pandemic. The need for evidence-backed approaches to sustained academic success for students
transcends demographic, economic and political divides. As school and district leaders, policy makers,
teachers, families and philanthropists build and implement plans to address pandemic-accelerated declines
in student learning, they need analysis of school and system achievement presented here to guide and
support their eorts.
1 The CMO study does not include Idaho, Maryland, and Ohio.
Contents
Table of Figures ................................................................................ 2
Project Description ............................................................................. 3
Methodology ................................................................................... 4
Summary of Findings ............................................................................ 5
Do All Students Benet? ......................................................................... 6
Where Is Positive Academic Growth Happening? .................................................... 6
What Can We Learn from CMOs? ................................................................. 8
Variations in Charter School Performance ......................................................... 8
Charter School Growth and Achievement ......................................................... 10
Exceptional Performance in Charter Schools ...................................................... 11
Evidence of Improvement over Time ............................................................. 12
Conclusions ................................................................................... 12
Implications ................................................................................... 16
Table of Figures
Figure 1: Annual Academic Growth of Charter School Students, Reading and Math ...................... 5
Figure 2: Annual Academic Growth of Charter School Students
by Charter School Type, Reading and Math ......................................................... 7
Figure 3: Academic Growth of Charter Schools Compared to Their Local TPS, Reading ................... 9
Figure 4: Academic Growth of Charter Schools Compared to Their Local TPS, Math ..................... 9
Figure 5: Academic Growth and Achievement 2015 to 2018, Reading .................................10
Figure 6: Academic Growth and Achievement 2015 to 2018, Math .................................... 11
Figure 7: Annual Academic Growth of Charter School Students across Three National Studies ........... 12
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 54
Summary of Findings
Looking at year-to-year academic progress from 2015 to 2019, the typical charter school student in our
national sample had reading and math gains that outpaced their peers in the traditional public
schools (TPS) they otherwise would have attended. We report these dierences as marginal days of
additional (or fewer) days of learning on a learning benchmark of 180 days of learning each school year for
matched TPS students. In math, charter school students, on average, advanced their learning by an additional
six days in a year’s time, and in reading added 16 days of learning.
Figure 1: Annual Academic Growth of Charter School Students, Reading and Math
** Signicant at p ≤ 0.01
Figure above originally appears as Figure 1.7 in CSP31.
These average eects are across all students, all schools, for all time periods. There is considerable variation
around these averages and this variation forms the foundation for additional analyses and ndings in our
two papers.
This growth represents accelerated learning gains for tens of thousands of students across the country. Each
student and each school is a proof point that shows that it is possible to change the trajectory of learning
for students at scale, and it is possible to dramatically accelerate growth for students who have traditionally
been underserved by traditional school systems.
Methodology
This research depends on data-sharing partnership agreements with state education agencies. One common
requirement across all agreements is that the processing, analysis and security of the student-level data must
meet the Federal Education Rights and Privacy Act (FERPA) requirements. This study complies with FERPA
regulations as interpreted by each state providing data.
Using both student and school level data, our resulting data set included 81 percent of tested public school
students in the United States, making it one of the largest data sets of student-level observations created
to date. We used this information to create a matched student data set with over 6,500,000 student-level
observations from over 1,853,000 charter students and a matched comparison group.
To create rigorous tests of our research questions, we need to compare charter school students’ experience
with an alternative, in this case the learning that occurs in nearby TPS. We match each charter student whose
records appear in the data with records of traditional public school students with identical traits and aligned
prior test scores who enrolled in schools that the charter student would have attended if not at their charter
school. This approach, the Virtual Control Record protocol, creates a “virtual twin” to a charter school student.
For research purposes, the virtual twin diers from the charter student only in the school attended.
This study approach mirrors the one used in the 2009 and 2013 studies. The only change to the method was
to rematch the charter school students to a new set of TPS students each year.
2
The data collected for this
study consisted of student-level demographics, school enrollment and achievement test scores in reading/
English language arts (ELA) and math To assure accurate estimates of charter school impacts, we use
statistical methods to control for dierences in student demographics and eligibility for categorical program
support such as free or reduced-price lunch eligibility and special education. In this way, we have created the
analysis so that dierences in the academic growth between the two groups are a function of which schools
they attended.
In these 2023 studies, we present our ndings about learning outcomes measured in days of learning. The
measure uses a benchmark of learning: the average student in TPS will obtain a year’s learning in a year’s
time. Computationally, the benchmark student attends school for 180 days in a year and advances their
learning by 180 days. If another student makes more (or less) progress in the same period of time, we present
that as additional (or fewer) days of learning.
2 This change meets the new standards of the What Works Clearinghouse at the National Center for Education Evaluation.
0
2
4
6
8
10
12
14
16
18
MathReading
16**
6**
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 76
but 10 days less growth in math than their TPS peers. These results are strongly hampered by the
performance of virtual charter schools; despite having only six percent of charter school students
enrolled, their impact on student progress of 58 fewer days of learning in reading and 124 fewer days in
math has damaging consequences for students and exerts a outsized drag on overall national results.
> Grade conguration – charter schools serving elementary, middle, and high school students had
statistically positive growth in both reading and math. Results for multilevel charter schools were
negative in math and similar to the TPS comparison groups in reading. Seeing growth in all grade
spans helps us understand that trends in the national aggregate performance are not concentrated in
particular grades.
> Continuous Enrollment – charter students overcome an initial learning dip associated with a school
change, and by their fourth year in their charter school, they show 45 days stronger growth in reading
than their TPS peers and 39 additional days of learning per year in math. The longer a student stays
enrolled in a charter school, the better the student’s academic outcomes are.
> School Management – students who attend a charter school that is part of a charter management
organization (CMO) experience signicantly accelerated growth compared to students enrolled in stand-
alone charter schools (SCS). Even so, CMO schools and SCS provide stronger learning than TPS in reading,
and CMOs do so in math. CMO-aliated students advanced by 27 additional days in reading and 23
more days in math over TPS, both of which are statistically signicant. Stand-alone charter schools still
grew signicantly more than TPS in reading by 10 additional days of learning, but were no dierent in
math. Given that SCS serve two-thirds of all students enrolled in charter schools, soft math performance
in these schools taints the otherwise decisive results in other parts of the study.
Figure 2: Annual Academic Growth of Charter School Students by Charter School Type, Reading and Math
Figure above originally appears as Figure 2.4 in CMO23.
Do All Students Benet?
When we probe these results to determine if all students benet, we nd positive results are not only present
in the aggregate, but also across student race/ethnicity groups:
> Black and Hispanic students in charter schools advance more than their TPS peers by large margins in
both math and reading.
> Multiracial, Native American, and White students in charter schools show equivalent progress to
their TPS peers in reading, but had weaker growth than their TPS peers in math.
> Asian students in charter schools showed similar growth to their TPS peers.
When we examined academic growth for special populations of students, we found that, compared with their
TPS peers:
> Charter school students in poverty had stronger growth
> English-language learner students attending charter schools had stronger growth
> Students receiving special education services had signicantly weaker growth in both math and
reading on average, though CMO-aliated students with Special Education needs have learning on par
with their TPS Special Education peers.
In the past, a common claim asserted that positive academic results in charter schools arise from advantages
that their students bring to their schooling. In some cases the claim focused on students having more
motivated parents. Another version suggests targeting behavior on the part of the school results in a student
body that is better prepared academically, a practice commonly referred to as “cherry picking” or “cream
skimming”. If true, the students in charter schools would show higher academic achievement at the point of
enrollment. In multiple analyses, we do not see signicant evidence of an undue advantage to charter schools.
In fact, we nd the opposite is true: charter schools enroll students who are disproportionately lower achieving
than the students in their former TPS.
Where Is Positive Academic Growth Happening?
Deeper into our analysis, we examine where student learning gains are occurring, and nd that positive and
strong eects exist in charter schools that vary widely by location and conguration.
> States – 18 states in the NCSS3 study produced signicantly stronger growth for students enrolled in
their charter schools when compared with their TPS peers; in 12 states, growth was similar to TPS peers.
Students attending charter schools had weaker reading growth than their TPS peers in only one state,
Oregon. In 12 states, charter school students had signicantly stronger growth in math than their peers
in TPS. In 16 states, math growth was similar between charter students and their TPS peers. Only three
states showed weaker growth for charter students compared to their peers.
> Locale – compared to their TPS peers, urban charter school students had 29 additional days of growth
per year in reading and 28 additional days of growth in math, both of which were signicant. Suburban
charter school students also had stronger growth in reading (+14 days) and in math (+3 days). Rural
students enrolled in charter schools had the equivalent of ve additional days of learning in reading,
10**
-3
27**
23**
Days of Learning
SCS CMO
-5
0
5
10
15
20
25
30
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 98
Figure 3: Academic Growth of Charter Schools Compared to Their Local TPS, Reading
Figure above draws from Figure 1.22 in CSP31, and Figure 36 in CMO23.
In math, more charter schools have weaker results than they do in reading, as presented in the gure below.
As the share of charter schools with growth greater than their TPS peers is comparable with the same
growth in reading across all categories, the driver of the overall weaker performance in math is the greater
percentage of charter schools (all, CMO-aliated and stand-alone charter schools) that perform worse than
their TPS peers. Stand-alone charter schools have the largest share of schools with lower growth in math in
comparison to their local TPS.
Figure 4: Academic Growth of Charter Schools Compared to Their Local TPS, Math
Figure above draws from Figure 1.22 in CSP31, and Figure 37 in CMO23.
These encouraging results require a note of caution. Since the reference point in these comparisons is the
growth that equivalent students in the local TPS realize, this comparison does not reveal if the dierence
is modest or large, nor does it indicate where in the range of absolute achievement the dierence occurs.
Positive dierences at the lowest levels of achievement may not be sucient to move students ahead fast
enough to reach long-term outcomes such as academic prociency or post-secondary readiness. Similarly,
a charter school may post growth results that are considered outsized for any school but still lag behind
What Can We Learn from CMOs?
Comprising one-quarter of the schools, but serving 37 percent of students in our national data set, Charter
Management Organizations (CMOs) are producing much of the learning gains we observed for charter school
students.
As with our national top-line results, we nd robust results for CMOs when we grouped their students by
race/ethnicity, special populations, where the CMOs are located, grade spans of the schools in the network
and how long a student enrolls in the school. As with all schools, there is a range of performance for CMOs,
and we share their student impacts in Appendix A.
Our analysis uncovered additional ways that CMOs are returning more positive, and often gap-busting,
results:
> New CMOs and new schools in existing CMOs open with strong results, in both cases delivering
stronger average gains for their students than their local TPS. The student gains in new CMOs are not as
strong initially as their older CMO peers. New schools started by mature CMOs deliver positive gains in
their early years that were none the less smaller than the older CMO schools.
> Size or age of a CMO does not relate to their quality, which means some CMOs are growing poorly
performing networks of schools.
> Clustering of CMOs’ schools within a single state returns signicantly more days of learning for their
students than in CMOs that operate schools in more than one state.
> CMOs that took on “turn-around” schools, absorbing those schools into their portfolios, positively
impacted results for students who remained enrolled in the turn-around school. In addition, the balance
of the CMO portfolio did not experience a downturn in student learning.
> The Charter School Growth Fund serves as a case study of charter school growth accelerators. CMOs
that the Growth Fund chooses to support have dramatically larger pre-funding learning gains than other
CMOs. The schools that existed at the time of selection remain strong. New CMO schools also open with
dramatically larger learning gains in both subjects judged against their TPS comparisons.
> Excellence at Scale puts dozens of CMOs at the forefront of eorts to provide education that is both
equitable and eective in moving student achievement to give their students full preparation for their
next steps.
Variations in Charter School Performance
In our reports, we analyze school-level performance, in addition to student-level performance, continuing to
report on growth as the outcome variable. Not every charter school provides quality academic programming
or an eective learning environment for students. Across all charter schools in our study, 36 percent have
greater growth, 47 percent have equivalent growth and 17 percent have lower growth relative to their local
TPS. CMO-aliated charter schools display stronger performance, with 43 percent having greater growth,
42 percent having equivalent growth, and 15 percent having lower growth in comparison to their local TPS.
Stand-alone charter schools have slightly more moderate results.
READING
BetterSameWorse
STAND-ALONE CHARTERS
CMO CHARTER SCHOOLS
ALL CHARTER SCHOOLS 17% 47% 36%
15% 42% 43%
18% 50% 32%
MATH
BetterSameWorse
STAND-ALONE CHARTERS
CMO CHARTER SCHOOLS
ALL CHARTER SCHOOLS 25% 39% 36%
22% 34% 44%
27% 42% 31%
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 1110
Schools that have average student
achievement above the state average
(above the 50th percentile) are presented
in the top half of the gure. In reading,
43 percent of all schools have average
performance in the upper half in their
respective states, with a majority
of those high achievement schools
also having stronger growth than
their local TPS. Zeroing in on the low-
growth/low-achievement quadrant,
207 schools (4.1 percent) in our study
have lower academic growth than their
local alternatives and have student
achievement that is below the 30th percentile of state achievement at the end of the school year.
Figure 6: Academic Growth and Achievement 2015 to 2018, Math
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.2% 2.0% 4.9% 3.8%
70th Percentile
50th Percentile
30th Percentile
1.0% 8.6% 12.0% 7.5%
4.9% 14.3% 13.8% 6.2%
7.1% 7.5% 5.3% 1.3%
Low Growth,
Low Achievement
High Growth,
Low Achievement
Figure above originally appears as Figure 1.26 in CSP31.
In math, above average achievement exists in 40 percent of charter schools, while 60 percent of schools
have achievement that is lower than their state averages. Twenty-eight percent of schools in the data set are
high-growth/high-achievement schools, returning great gains for their students. Zeroing in again on the low-
growth/low-achievement quadrant, 348 schools (7.1 percent) have lower academic growth than their local
alternatives and have student achievement that is below the 30th percentile of state achievement at the end
of the school year.
The number of schools in the low-growth/low-achievement quadrant, though smaller in reading than in math,
remains a key concern.
the community schools in achievement. Simultaneous consideration of student academic growth and
achievement is the only way to get the complete picture of charter school performance.
Charter School Growth and Achievement
Student academic growth measures how much students advance their learning in a year’s time, and student
achievement measures the stock of their knowledge at the end of the year. We believe it is critical to examine
both growth and achievement in order to understand how well schools prepare students for next steps in
school and life. We map each school’s average growth and average achievement against the the growth of
matched TPS students and average state performance. Examining both measurements for all schools in our
national data set during the most recent growth period, we present ndings in four basic categories of school
performance:
> High Growth—High Achievement: schools that exceed the growth of their local options and whose
students are above the state average in overall achievement
> High Growth—Low Achievement: schools that exceed the growth of their local options but with overall
student achievement below the
state average
> Low Growth—High Achievement:
schools whose students exceed the
state average on achievement but
do not advance as much yearly as
their comparisons
> Low Growth—Low Achievement:
schools with lower academic
growth than their local alternatives
and whose students’ achievement
is lower than the state average at
the end of a school year.
Figure 5: Academic Growth and Achievement 2015 to 2018, Reading
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.1% 1.5% 5.8% 2.8%
70th Percentile
50th Percentile
30th Percentile
0.7% 9.1% 17.0% 6.1%
3.1% 12.3% 17.6% 6.4%
4.1% 6.8% 5.8% 1.1%
Low Growth,
Low Achievement
High Growth,
Low Achievement
Figure above originally appears as Figure 1.25 in CSP31.
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in
Figure5. These values appear on the study website as a
layer of the chart—the user can see the quadrant totals
and then drill down to see the inner-quadrant values.
11.4 31.7
26.3 30.9
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in Figure
6. These values appear on the study website as a layer
of the chart—the user can see the quadrant totals and
then drill down to see the inner-quadrant values.
11.8 28.2
33.8 26.4
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 1312
1. In both reading and math, charter schools provide students with stronger learning compared with the
learning in the traditional public schools that are otherwise available to them.
Across the broad range of charter schools, the evidence suggests that they are a robust education option
under many conditions. Whether stand-alone or networked, charter schools operate by law mainly on
their own, making decisions they expect will serve their students well. According to our latest ndings, the
autonomy given to them usually yields positive results. The majority of charter schools provide better year-
to-year outcomes for students compared to their traditional public-school options. Most of these schools
perform better to such a degree that the dierence is statistically signicant.
The results stand up to deeper investigation. Charter schools produce superior student gains despite
enrolling a more challenging student population than their adjacent TPS. They move Black and Hispanic
students and students in poverty ahead in their learning faster than if they enrolled in their local TPS. They
are more successful than the local public school alternatives across most grade spans and community
settings. These results show that charter schools use their exibility to be responsive to the local needs of
their communities.
These ndings generalize into lessons for policy leaders, educators, and funders. Knowing that the average student
in the average charter school can outperform their TPS peers raises important questions about the priority placed
on student outcomes in education decisions in many communities.
2. Some charter schools provide less student learning than their local district schools, although a larger
proportion delivers better learning outcomes. The latter group includes over 1,000 charter schools
managing stang and resources to deliver superior academic results that eliminate the learning gap
across student groups.
Vital lessons also come from the distribution of school performance around the average. Over the past 30
years, small, large, urban, rural, networked or stand-alone charter schools, autonomous and independent of
each other, have arrived at their own solutions for giving their students stronger learning experiences. The
discretion that charter schools enjoy does not guarantee that each school or every charter network realizes
strong student outcomes. Our study illuminated the range of learning across schools.
Despite declining shares, there remain a concerning number of charter schools with weaker student
outcomes. While lower-performing schools make up a larger share of stand-alone charter schools, CMOs and
networks also have a substantial share that produces low gains for their students. This study has profound
implications for charter schools and charter networks that do not support student learning. Charter boards
and authorizers are the accountability side of the charter school equation. They evaluate school performance
and, if necessary, dictate remedies. As our analysis shows, disturbing numbers of charter schools and
networks have low learning levels. There are brick-and-mortar, online, networked, and stand-alone charter
schools with sub-par results.
The number of school closures we observed in the years of this study was small compared to the counts of
schools with the lowest student growth and academic achievement. Since primary and secondary education
is essential to the social contract, providing a foundation for future opportunities, the claim of “choice” cannot
justify derailing students’ preparation. Especially in the post-COVID era, the need for charter boards and
authorizers to address under-performance in their schools has never been more critical.
Exceptional Performance in Charter Schools
Perhaps the most revealing nding of our study is that more than 1,000 schools have eliminated learning
disparities for their students and moved their achievement ahead of their respective state’s average
performance. We refer to these schools as “gap-busting” charter schools. They provide strong empirical
proof that high-quality, high-equality education is possible anywhere. More critically, we found that dozens
of CMOs have created these results across their portfolios, demonstrating the ability to scale equitable
education that can change lives.
Evidence of Improvement over Time
Findings from this study take on even more weight when considered in the historical context of the 15 years
of CREDO studies on student academic progress in charter schools. Between the 2009 and 2023 studies,
against a backdrop of at performance for the nation as a whole, the trend of learning gains for students
enrolled in charter schools is both large and positive.
Figure 7: Annual Academic Growth of Charter School Students across Three National Studies
** Signicant at p ≤ 0.01
Figure above originally appears as Figure 1.8 in CSP31.
Conclusions
The outcomes of these studies are largely positive and support several conclusions about the current
landscape of charter schools across America. Perhaps more importantly, the opportunity to position these
ndings in the larger body of research leads to a number of implications about the fundamental policies and
practices of charter schooling at a more global level.
-6**
-17**
Days of Learning
-25
-20
-15
-10
-5
0
5
10
15
20
25
3rd Study2nd Study1st Study
-3
6**
6**
16**
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 1514
or portfolio size, we see high- and low-impact CMOs and networks. This further supports earlier CREDO
research that showed that CMOs only replicate the quality they already have. The implications of replicating
schools with weak results is clear. The big upside is the ability of dozens of CMOs to scale their gap-busting
performance. Additionally, CMOs that concentrate their operations within a single state have stronger gains
than multistate CMOs, though both groups do well by their students.
Programs of external funding and support to CMOs to grow their networks, represented here by the Charter
School Growth Fund, focus on some of the stronger CMOs and networks in our study. After high-performing
CMOs receive endorsement, the learning of students in those CMO schools rises in reading but holds steady
in math.
The majority of new CMO schools are no better or worse than the parent organization has already produced, so
decisions to approve applications by CMOs to open new schools must consider the contributions to student learning
of schools in the existing portfolio.
CMO growth accelerators help augment board and authorizer reviews through their extensive selection process; the
growth of their grant-receiving CMOs maintains the strong student learning that led to their selection. The expansion
of these high-quality schools and networks benets more students and communities.
4. Charter schools and networks improve over time, as do the systems that oversee them.
Insights about improvement in schools and networks stem from this study and CREDO’s prior multistate
studies.
In the years of this study, student growth in charter schools was the strongest observed in any of CREDO’s
multistate studies. Added to the results from the previous two studies, a strong trend of improvement
becomes clear. We see substantial increases in student learning in CMOs in both tested subjects and in
reading for stand-alone charter schools. Even the nding of no dierence in math learning in stand-alone
charter schools vis a vis TPS, a decline from the 2017 study results, still marks an improvement from the
statistically signicant negative results in the rst CMO vs. stand-alone comparisons in 2013.
A better understanding of the improvement in the sector comes from two dierent ndings. The rst is
that the largest share of improvement comes from existing charter schools. Compared to the National
Assessment of Education Progress (NAEP) trend, evidence of schools getting better over time is welcome
news.
Second, new schools opened with stronger results than at any time in the past. Growth in the number of
CMOs since the last study plays a role. Many stand-alone charter schools also pushed their results upward.
Strengthening authorizer standards and practices, a drive that took root in the 2010s, also sets a higher bar
that resulted in better schools opening.
Finding ways to improve student academic outcomes is an ambition shared by policy and community leaders,
educators, funders and parents. Charter school results show that change for the better is possible in the larger
education system. The key to improvement lies outside any particular school or network model, though many are
worthy of emulation. It is simply not possible to drive single solutions through the diverse landscape that is U.S.
public education. Lessons from the charter school experience and results may be helpful in charting a future course
in public education.
Closure is not the sole remedy. As we learned from our special investigation, the “takeover“ of
underperforming schools by strong CMOs led to improved student learning for the students who remained
enrolled before and after the transfer. The gains did not adversely aect student academic progress in the
rest of the CMOs’ schools. This policy tool may have broader utility than previously realized.
At the high end of the performance range, good news exists in the growing share of schools outpacing
learning in their local TPS. In both subjects and for both CMO and stand-alone schools, larger shares are
“better than” and a smaller share is “weaker than” compared to earlier work.
The real surprise of the study is the number of charter schools that have achieved educational equity for
their students: we call them “gap-busting” schools. Ensuring equivalent yearly growth across student groups
has two critical consequences. First, ensuring minority and poverty students learn on par with or better
than their White peers interrupts or reduces the achievement gap. It happens regularly in a large swath of
charter schools. More critically, there is strong evidence that these gap-busting schools can be scaled. Added
to the traditional district schools that achieve similar results, this is the life-transforming education that so
many students need. Second, these schools deliver hundreds of independent proof points that learning gaps
between student groups are not structural or inevitable; better results are possible.
Charter schools function as a portfolio, and their varied impacts on student learning are expected. Charter
school boards and authorizers are responsible for ensuring students perform well. Evidence shows that the
charter school enterprise benets students, and its positive outliers (e.g., gap busters) can pressure the rest
of the system.
The near-term implication for charter school boards and authorizers is two-pronged. Addressing chronic and/or
severe underperformance is necessary and imperative in the current education climate. Identifying high-impact
exemplars for probationary charter schools to study and emulate is possible. Transfer of sub-par schools to higher-
performing operators could be part of a larger incentive for growth and replication. At the same time, authorizers
might consider longer charter terms for charter schools that consistently demonstrate outstanding student learning
success.
Education leaders and policy makers need to understand that in eorts to improve, some failure is inevitable. Any
subsequent failure to address the poor performance compounds the damage. It also blocks constructive learning for
the future. Strong examples of authorizing exist and should be emulated.
Leadership and responsibility demand embracing practices and policies that lead to better results for students, not
maintaining the status quo.
3. The larger scale of Charter Management Organizations does not guarantee high performance—but on
balance, it helps.
When taken as a whole, schools managed by Charter Management Organizations and charter networks bring
a greater learning benet to students compared to stand-alone charter schools. Despite the dierences, both
groups of charter schools have had larger student success than traditional public schools with respect to
reading. We note, however, that math gains in stand-alone charter schools were equivalent to TPS learning.
Our analysis highlights attributes of higher-performing CMOs and networks that could be useful in future
discussions. Size or age of the CMO does not relate to student learning: at every increment of CMO age
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Executive Summary | Last Updated 6/19/2023 1716
Poorly performing charter schools are often ignored. A number of these schools were observed during this
study window. There is data to assess policy leaders and authorizers to hold them accountable for protecting
children’s futures. As tough as closing schools is, the disservice of not closing poorly performing schools has
large and lingering ripple eects.
Our results show that the framework of charter schools helps current students and strengthens public
education overall. We contend these incentives have broader applicability in public schools and see signs
of their spread. Collaborations between charter schools and local district schools have grown over time.
Some states, including Kentucky and Maine, have adopted policies to give educators freedom in adjusting
instruction and boosting performance. However, uptake is slow.
In the year 2023, the importance of strong academic achievement among America’s students has never
been greater. The students hit hardest by school closures during the coronavirus pandemic are precisely
those whom this research illuminates as being able to benet the most from charter schools. In this study
thousands of charter schools have proved that we can do better for our students. The current number of
students beneting from these schools is 3.7 million, but the number could drastically increase if more
schools agreed to the same arrangement. Whether it be termed “charter school” or something else, the
deduction from this data is that when both sides of the equation—exibility and accountability—are working
together for more schools, more students’ academic results will improve.
References
Boast, L., Cliord, B., & Doyle, D. (2020). Learning in real time: How charter schools served students during
COVID-19 closures. National Alliance for Public Charter Schools. Retrieved from https://www.publiccharters.
org/our-work/publications/how-charter-schools-served-students-during-covid-19-closures
Childs, J., Grooms, A., & Mozley, M. P. (2022). Hidden in (virtual) plain sight: A charter district’s focus on
attendance during COVID-19. Education and Urban Society. https://doi.org/10.1177/00131245211065414
CREDO. (2022). Charter Schools’ Response to the Pandemic in California, New York and Washington State.
The Center for Research on Education Outcomes. https://credo.stanford.edu/wp-content/uploads/2022/02/
Charter-School-COVID-Final.pdf
Henderson, M. B., Peterson, P. E., Houston, D., & West, M. R. (2021). What American families experienced
when COVID-19 closed their schools. Education Next, 21(1), 22-31.
Mumma, K. S., & West, M. R. (2018). Charter School Authorizing in California. Technical Report. Getting Down
to Facts II. Policy Analysis for California Education, PACE.
NACSA. (2016). State of Charter Authorizing 2016. Retrieved April 24, 2023 from https://qualitycharters.org/
wp-content/uploads/2018/07/State-of-Charter-School-Authorizing-2016-Findings.pdf
Implications
The charter school policy framework sets the conditions for charter schools’ growing positive outcomes. It
is the fundamental common denominator in every case, and its role is powerful.
The framework oers a divergent approach from the conventional strategy for public schools. The “exibility
for accountability” construct is not just a catchphrase. It is a distinctly dierent mode of operation. The
“loose-tight” parameters of the framework create incentives to which schools and networks respond. The
incentives nd positive support in this study’s ndings and the broader trends. While our study design cannot
make causal claims (because randomly assigning schools to the traditional or charter school approach has yet
to happen), it can deliver a plausible argument of the value of the policy based on available evidence.
On the “loose” side of the approach, the framework establishes a policy of possibility where educators,
leaders and boards of directors have the discretion to build and deliver curriculum and instruction that meets
high standards for learning and is responsive to local needs.
According to this study, there are a lot of positive possibilities. The process has led to many successful schools
nationwide, often with meaningful innovations. The diversity of schools illuminates an important feature of
the framework: success is attainable via many paths. Over time, many have sought and gained permission to
expand and then shown the ability to create strong student learning at scale.
Students in these schools, especially minority students and those in poverty, make larger advances than in
local public schools. Beyond the benets for their students, successful charter schools deliver critical proof
points of ways to improve outcomes for students. In the current regulatory climate, it is dicult to imagine
how similar eorts could become conventional among traditional public schools.
Beyond exibility in school design, school teams have the leeway to tinker with their operations. The results
show that existing charter schools have improved over time. The proportion of charter schools with superior
results is on the rise. The share that lags behind the local TPS alternatives is also shrinking. This means
schools and networks use their discretion and autonomy to foster a standing capacity to adapt over time.
3
Accordingly, the framework also aims to be “tight” at key points as schools open and mature. Authorizers
are expected to behave as governors of quality. They set the bar to receive initial permission to operate,
which exerts quality and safety controls at the outset. Others have documented stronger standards among
authorizers in the review and approval of new applications (Mumma & West, 2018). The ndings of stronger
new schools in this study compared to earlier results attest to the eort and to the CMO replications and new
charter schools that meet the higher bar.
Authorizing is a delicate job that requires resources, expertise and substantial political acumen and courage.
There is growing attention to authorizers adopting rigorous standards and practices and using a variety of
performance data to evaluate schools that apply for renewal (NACSA, 2016).
3 We saw that capacity in stark terms when we examined how charter schools in three states responded to the COVID-instigated school closure orders (CREDO,
2022). Rapid transformation into remote instructional mode; acquisition and distribution of food, technology, or internet access; and strengthening of personal
supports were widespread. Return to in-person instruction in the fall of 2020 was nearly universal. These points rest admittedly on smaller bases of qualitative
evidence, but they provide human dimensions to the point that the present quantitative analysis illuminates nationally. See also: Boast et al. (2020); Henderson
et al. (2021); Childs et al. (2022).
As a Matter of Fact:
The National
Charter School
Study III 2023
As a Matter of Fact:
The National Charter
School Study III 2023
Volume 1
Charter School
Performance in 31 States
Authors
Margaret E. Raymond, Ph.D.
James L. Woodworth, Ph.D., Lead Analyst- 31 State Study
Won Fy Lee, Ph.D., Lead Analyst- CMO Study
Sally Bachofer, Ed.M.
Contributors
Meghan E. Cotter Mazzola, M.S.
William D. Snow
Tzvetelina Sabkova, M.A.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023 21
© 2023 CREDO
Center for Research on Education Outcomes
Stanford University
Stanford, CA
https://credo.stanford.edu
CREDO, the Center for Research on Education Outcomes at Stanford University, was established to improve
empirical evidence about education reform and student performance at the primary and secondary levels.
CREDO at Stanford University supports education organizations and policy makers in using reliable research
and program evaluation to assess the performance of education initiatives. CREDO’s valuable insight helps
educators and policy makers to strengthen their focus on the results of innovative programs, curricula, policies
and accountability practices.
Acknowledgments
CREDO gratefully acknowledges the support of the state education agencies that contributed their data to this
partnership. Our data access partnerships form the foundation of CREDO’s work, without which studies like
this would be impossible. We strive daily to justify the condence placed in us.
The research presented here uses condential data from state departments of education. The views expressed
herein do not necessarily represent the positions or policies of the organizations noted above. No ocial
endorsement of any product, commodity, service or enterprise mentioned in this publication is intended or
should be inferred. In addition:
> The research presented here utilizes SLDS Data from the Idaho State Board of Education (SBOE) and the
Idaho State Department of Education. Any research errors are the sole responsibility of the author(s).
> This research result used data structured and maintained by the MERI-Michigan Education Data Center
(MEDC). MEDC data is modied for analysis purposes using rules governed by MEDC and is not identical
to data collected and maintained by the Michigan Department of Education (MDE) and/or Michigan’s
Center for Educational Performance and Information (CEPI). Results, information and opinions solely
represent the analysis, information and opinions of the author(s) and are not endorsed by, or reect the
views or positions of, grantors, MDE and CEPI or any employee thereof.
> Data for this report was provided by the Missouri Department of Elementary and Secondary Education.
> The conclusions of this research do not necessarily reect the opinions or ocial position of the Texas
Education Agency, the Texas Higher Education Coordinating Board, or the State of Texas.
The analysis and conclusions contained herein are exclusively those of the authors and are not endorsed by
any of CREDO’s supporting organizations, their governing boards, or the state governments, state education
departments or school districts that participated in this study. All errors are attributable to the authors.
CREDO also acknowledges the support of the Walton Family Foundation and The City Fund for supporting
this research.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 2322
1.5 School-Level Impacts ....................................................................... 61
School-Level Academic Growth .............................................................. 61
School-Level Academic Growth by State ......................................................63
1.6 Charter School Academic Growth and Achievement .............................................65
1.7 Gap-Closing Charter Schools .................................................................69
Table of Figures
Figure 1.1: CREDO Virtual Control Record (VCR) Methodology ........................................ 32
Figure 1.2: Average Achievement of All Charter Students by State, Math 2017 ..........................38
Figure 1.3: Average Academic Growth of Charter Students by State, Math 2017 ........................40
Figure 1.4: Percent of Charter School Student Enrollment by Location ................................43
Figure 1.5: Percent of Charter Schools by Grade Level ..............................................44
Figure 1.6: Annual Academic Growth of Charter School Students, Reading and Math ...................45
Figure 1.7: Annual Academic Growth of Charter School Students across Three National Studies. . . . . . . . . . 46
Figure 1.8: Annual Academic Growth in Previously Studied Schools Compared to Current Schools. . . . . . . . 47
Figure 1.9: RECAP: Average Academic Growth for Charter School Students
by Charter School Type, Reading and Math .................................................... 48
Figure 1.10: State Level Average Charter School Student Academic Growth, Reading ....................49
Figure 1.11: State Level Average Charter School Student Academic Growth, Math ......................50
Figure 1.12: Average Reading Growth of Charter School Students by State, 2013 vs 2023 ................ 51
Figure 1.13: Average Math Growth of Charter School Students by State, 2013 vs 2023 ................... 51
Figure 1.14: Days of Learning for Charter School and TPS Students by Race/Ethnicity, Reading and Math ..53
Figure 1.15: Annual Academic Growth for Charter School Students in Special Populations ............... 55
Figure 1.16: Annual Academic Growth for Charter School Students with Compound Designations ........56
Figure 1.17: Annual Academic Growth of Charter School Students by Grade Level ...................... 57
Figure 1.18: Annual Academic Growth for Charter School Students by School Mode, Reading and Math ...58
Figure 1.19: Annual Academic Growth for Charter School Students by Years of Enrollment .............. 59
Figure 1.20: Charter School Student Academic Growth by School Location, Reading and Math ........... 61
Figure 1.21: Academic Growth of Charter Schools Compared to Their Local TPS, Math and Reading ......62
Figure 1.22: Academic Growth of Charter Schools Compared to Their
Local TPS across Studies, Reading and Math ..................................................63
Figure 1.23: Average Academic Growth in Charter Schools versus. Their Local TPS by State: Reading .....64
Figure 1.24: Average Academic Growth in Charter Schools versus. Their Local TPS by State: Math ........64
Figure 1.25: Academic Growth and Achievement, Reading ...........................................66
Figure 1.26: Academic Growth and Achievement, Math .............................................67
Contents
Table of Figures ...............................................................................23
Table of Tables ................................................................................24
1.1 Introduction ...............................................................................25
A Brief Primer on Charter Schools in the United States .........................................26
The Structure of the National Charter School Study III Report ...................................26
Aggregate Charter Student Academic Progress .............................................26
Academic Progress for Student Groups in Charter Schools ...................................27
Student Academic Progress in Dierent School Settings .....................................27
The Role of Charter Management Organizations in Student Academic Progress .................28
1.2 Methods and Data ..........................................................................28
Methodology ..............................................................................28
Consolidating Student Data from Multiple States ..............................................30
Selection of Comparison Observations .......................................................31
Student Match Rates .......................................................................33
School Match Rates ........................................................................33
Fair Analysis of Impacts on Student Academic Progress .........................................33
Basic Analytic Models ......................................................................34
How We Present the Results ................................................................34
1.3 Descriptive Statistics .......................................................................35
Student Characteristics .....................................................................35
Race/Ethnicity Composition of Matched Charter Students ......................................36
Other Student Characteristics ...............................................................37
Perceptions of Charter School Student Advantage ............................................. 41
School Characteristics ......................................................................43
School Location. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
School Level/ Grade Span ................................................................44
1.4 Analytic Findings ...........................................................................45
Academic Growth of Charter School Students .................................................45
RECAP: Academic Growth of Charter School Students by Type of School ..........................48
Charter School Student Academic Growth by State .............................................49
Changes in Charter School Student Academic Growth by State ..................................50
Dierences in Academic Growth by Charter School Student Characteristics ....................... 52
Dierences by Race/Ethnicity ...............................................................52
Academic Growth for Charter School Students in Special Populations .........................54
Student Annual Academic Growth by Charter School Grade Level .............................56
Annual Academic Growth of Online Charter School Students ....................................58
Academic Growth by Continuous Enrollment in Charter School .................................. 59
Charter School Student Academic Growth by Location of their School ............................60
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 2524
1.1 Introduction
The year 1992—the year that City Academy Charter School opened in St. Paul, Minnesota—was over 30 years
ago. The fundamental bargain of public charter schools—“exibility for accountability”—took root in the
school’s rst charter and in the more than 7,800 public charter schools that came after. With over 3.7 million
students currently enrolled in charter schools in 43 states and the District of Columbia, charter schools
represent the largest experiment in public school innovation in the nation’s history.
The current study is the third multistate study of charter school eectiveness—the rst was released in 2009
and the second in 2013. It adds to a large slate of charter school research
1
released by CREDO in 2006. This
study covers the education experience of 2,080,913 unique students enrolled in charter schools in 31 states
from 2014-15 to 2018-19. As our work in this area uses the same peer-reviewed research design and analytic
approaches, the results across studies provide the basis for examining charter school performance trends
since 2006.
CREDO’s work joins a body of research on the subject (Booker et al., 2009; Mead et al., 2015). Our unique
contribution lies in the scope of the eort: CREDO uses longitudinal student-level information derived from
state administrative data from 29 states plus the District of Columbia and New York City.
2
In our research,
we include 94 percent of the nation’s charter school students in tested grades. We use a detailed matching
method to ensure that our analytic comparisons to students in district schools are as precise as the data
allows. Consequently, our ndings carry strong levels of reliability and validity.
A study of the academic impacts of charter schools on their students is timely. Insights about the educational
eectiveness of schools, school operators, K-12 academic programs and education policy are valuable
today more than ever. The 2022 results from the National Assessment of Educational Progress removed any
ambiguity about student learning after the COVID-19 pandemic. As a country, student academic performance
has regressed by two decades in math and fallen steeply in reading, with the most severe performance
declines found among minority, poverty and special needs populations that were already struggling before
the pandemic. The need for evidence-backed approaches to sustained academic success for students
transcends demographic, economic and political divides. As school and district leaders, policy makers,
teachers, families, and philanthropists build and implement plans to address pandemic-accelerated declines
in student learning, they need the analysis of school and system achievement presented here to guide and
support their eorts.
1 Center for Research on Education Outcomes, http://credo.stanford.edu.
2 We refer to these 31 jurisdictions as “states” to maintain consistency with previous studies. New York City data is not included in New York results. The two
groups are mutually exclusive for this study.
Table of Tables
Table 1.1: States Participating in Each CREDO National Charter School Study (NCSS) ....................29
Table 1.2: Match Rates by Race/Ethnic Group ......................................................33
Table 1.3 Demographic Comparison of Students in TPS, Feeders,
and Charter Schools (Brick-and-Mortar and Virtual) in 31 States, 2017–18 .........................35
Table 1.4: Race/Ethnic Proportions for All versus Matched Students ..................................36
Table 1.5: Special Population Proportions for All versus Matched Students ............................37
Table 1.6: Achievement Decile Distribution of Charter Enrollees by State 2017, Math .................... 39
Table 1.7: Percentage Dierences between Entering Charter Students
and Feeder School Students by Decile of Achievement .........................................42
Table 1.8: Charter School Student Academic Growth by Grade Level
across Studies, Reading and Math ...........................................................57
Table 1.9: Charter School Student Academic Growth by Years of Charter
Enrollment across Studies, Reading and Math .................................................60
Table 1.10: Charter Schools with No Learning Gaps and High Achievement ............................69
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 2726
The current results are larger than what we reported in the second national study, which were larger than the
rst national study. The three studies taken together produce a solid positive trend over the 15 school years
between 2004 and 2019. Notably, the upward trend was due to existing charter schools improving over time,
not an inux of higher-performing new schools.
Academic Progress for Student Groups in Charter Schools
The second set of ndings addresses how consistent the results are for all students. We again report these
dierences as marginal days of additional (or fewer) days of learning on a benchmark of 180 days of learning
in a single school year. We found important dierences in the amount of learning for dierent groupings of
students enrolled in charter schools in our study.
Consistent with our earlier studies, we found signicant variations in charter student learning when we exam-
ined results for students in dierent racial/ethnic groups. In math performance, Asian/Pacic Islander students
in charter schools realized more than a year of academic progress in a school year. In contrast, Black, Hispanic,
White and Native American students have academic gains that fall short of a year’s progress in a year. In
reading performance, Asian/Pacic Islander students made gains well above the benchmark 180 days of
learning, while White and Hispanic students were closer to the benchmark of one year of growth in a year.
Black and Native American students fell considerably short of the 180 days of learning mark.
Despite overall low growth, Black and Hispanic students in charter schools fared better when compared with
the learning gains of their TPS peers. White, Native American and multiracial students had smaller learning
gains than their TPS comparisons.
Charter school students in poverty and their TPS counterparts fell short of the learning of their non-poverty
peers. Despite this, charter students in poverty had stronger growth, equal to 17 additional days of learning in
math and 23 additional days of learning in reading, than their TPS peers in poverty. Likewise, English-language
learner (ELL) students who attended charter schools also had stronger growth in math (eight days) and reading
(six days) than their TPS peers but were still left considerably behind non-ELL students. Students receiving
special education services had signicantly weaker growth in both math and reading than their TPS peers.
Specically, they grew 14 fewer days in math and 13 fewer in reading.
Student Academic Progress in Dierent School Settings
As the conversation about public education focuses on schools as units of analysis, the third set of conclusions
refers to the eects of charter school students’ learning when dierent school characteristics are considered.
Across the sample of 6,802 charter schools in math, 36 percent had overall learning gains that were statistically
signicantly larger than the local TPS alternatives. One quarter posted statistically signicantly smaller results,
and 39 percent had gains equivalent to their local peer schools. In reading, the results were stronger: 36
percent had statistically signicantly larger learning results, 47 percent posted gains on par with their TPS
peers, and 17 percent had statistically signicantly smaller results. At both ends of performance, these results
improve on earlier results from the last national study—a greater share of charter schools is stronger than the
local option and a smaller percentage is worse.
A Brief Primer on Charter Schools in the United States
Enabling legislation allows charter school founders and operators to design and tailor organizational
structures, stang and instructional approaches to provide their students with an alternative to local district
schools. They pursue dierent missions such as STEAM, college prep, social justice or new technologies.
They can be small or large; they can operate as single schools or in school networks. Some charter schools
outsource some or all of their operations to outside vendors. Some charter schools mirror traditional public
school (TPS) grade level or grade band congurations, and others serve students K-12 in one school. Some
charter schools own and operate their facilities, and some are tenants of local school districts or rent space
from commercial landlords.
Charter schools operate under governing boards separate from local district school boards. Following the “ex-
ibility for accountability” construct, in exchange for discretion in school design and operation, charter schools
must undergo periodic accountability reviews to remain open and in good standing. These accountability
reviews weigh the schools’ operational and scal health and student academic performance.
Thirty-seven states allow multiple schools to be held and operated under a common management structure
known as charter networks or charter management organizations (CMOs). This option has increased the
number of available charter school seats, yet it raises questions of scalability and quality. This study examines
these questions and the performance of charter schools and charter networks against the legislative and
regulatory incentives in place.
The Structure of the National Charter School Study III Report
We report four sets of ndings, summarized below. The rst three are included in this volume, Charter School
Performance in 31 States (CSP31). The fourth is presented in Volume 2, Charter Management Organizations 2023
(CMO23).
Aggregate Charter Student Academic Progress
The rst set of ndings focuses on student performance in all charter schools included in the study. Looking
at year-to-year academic progress from 2015 to 2019, tested students enrolled in all charter schools in the
31 states had reading and math gains that outpaced their peers in the TPS that charter school students
otherwise would have attended. We report these dierences as marginal days of additional (or fewer) days
of learning on a learning benchmark of 180 days each school year. In math, charter school students, on aver-
age, were found to advance their learning by an additional six days in a year. For reading, on average, their
learning added 16 days of learning.
In the past, a common claim asserted that positive academic results in charter schools arise from advantages
that their students bring to their schooling. In some cases the claim focused on students having more motivat-
ed parents. Another version suggests targeting behavior on the part of the school results in a student body
that is better prepared academically, a practice commonly referred to as “cherry picking” or “cream skimming.”
If true, the students in charter schools would show higher academic achievement at the point of enrollment.
In multiple analyses, we do not see signicant evidence of an undue advantage to charter schools. In fact, we
nd the opposite is true: charter schools enroll students who are disproportionately lower achieving than
the students in their former TPS.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 2928
Table 1.1: States Participating in Each CREDO National Charter School Study (NCSS)
NCSS I - 2009 NCSS II-2013 NCSS III-2023
First School Year of Data 2000-01 2006-07 2014-15
Last School Year of Data 2007-08 2010-11 2018-19
States Included in Each Study
Arkansas Arkansas Arkansas
Arizona Arizona Arizona
California California California
Colorado (Denver) Colorado Colorado
District of Columbia District of Columbia District of Columbia
Florida Florida Florida
Georgia Georgia
Idaho
Illinois (Chicago) Illinois Illinois
Indiana Indiana
Louisiana Louisiana Louisiana
Massachusetts Massachusetts Massachusetts
Maryland
Michigan Michigan
Minnesota Minnesota Minnesota
Missouri Missouri Missouri
Nevada Nevada
New Jersey
New Mexico New Mexico New Mexico
New York New York
New York City New York City
North Carolina North Carolina North Carolina
Ohio Ohio Ohio
Oregon Oregon
Pennsylvania Pennsylvania
Rhode Island Rhode Island
South Carolina
Tennessee Tennessee
Texas Texas Texas
Utah Utah
Washington
Wisconsin
The performance of charter schools in dierent types of communities continues in earlier patterns. As seen in
earlier national studies, students in urban charter schools outpace their TPS peers and post larger gains than
their charter school peers in suburban, town or rural settings.
The academic performance of students enrolled in virtual charter schools compares poorly to the 180-day
learning standard in TPS and the performance of students enrolled in brick-and-mortar charter schools.
Students in virtual schools had 124 fewer days of learning in math and 60 fewer days in reading against our
180-days of learning benchmark. By contrast, students in brick-and-mortar charter schools posted 21 addition-
al days of learning in reading and 14 extra days in math.
The ndings show important dierences for charter schools when grouped by the state in which they operate.
Ten states/regions had learning gains in reading and math that were statistically signicantly larger than the
TPS students: Colorado, Illinois, Massachusetts, Michigan, Missouri, New Jersey, New York City, Upstate New
York, Rhode Island and Tennessee. Seven states posted better gains in reading: Arizona, California, Florida,
Idaho, Minnesota, North Carolina and Texas. Only Oregon saw the reverse: charter school learning was statisti-
cally signicantly smaller in both subjects. Ohio and South Carolina had negative and signicant learning
advances in math.
The Role of Charter Management Organizations in Student Academic Progress
We extensively investigated student progress according to the type of charter school they attended. When the
results were grouped by independently operated charter schools (stand-alone charter schools, or SCS) versus
those in Charter Management Organizations (CMOs or networks), students in schools run by CMOs had
stronger results than their stand-alone student counterparts. While both sets of schools are stronger than
their TPS peers, the CMO learning gains are substantially stronger and carry the overall results of the study
despite having only a third of the schools.
We expanded our typical format for sharing results with this study. We moved all results into a web-based
interactive data set at ncss3.stanford.edu. No individual student data or identiable small group information is
included in the graphics and other data visualizations. All the results from this study on the website mirror the
document’s ndings.
1.2 Methods and Data
Methodology
Since the 2009 study, Multiple Choice: Charter School Performance in 16 States, CREDO has rened our matching
and analysis techniques and expanded our data collection. This chapter provides a nontechnical overview of
the data sources and analytic methods used in the current study. The chapter presents general descriptions of
the data sources used in the recent study and explanations of how the study was organized and executed.
The Technical Appendix to this report and the Technical Appendix of the 2013 National Charter School Study II
(Cremata et al., 2013) includes greater scientic detail on these topics. Table 1.1 represents the states included
in each study and the years of data included in each study.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 3130
Selection of Comparison Observations
As in previous CREDO studies, this study employed the virtual control record (VCR) method of analysis
developed by CREDO (Davis & Raymond, 2012). The VCR approach creates a “virtual twin” for each charter
student who is represented in the data. In theory, this virtual twin would only dier from the charter student
in that the charter student attended a charter school and the twin attended a TPS. The VCR matching
protocol has been assessed against other possible study designs and judged to be reliable and valuable by
peer reviewers (Egalite & Ackerman, 2015).
4
Using the VCR approach, a “virtual twin” was constructed for each charter student by drawing on the available
records of traditional public school (TPS) students with identical traits and aligned prior test scores who were
enrolled in TPS that the charter students would have likely attended if they were not in their charter school.
5
Factors included in the matching criteria were:
> Grade level
> Gender
> Race/ethnicity
> Free or reduced-price lunch eligibility
> English-language learner status
> Special education status
> Prior test score on state achievement tests
Figure 1.1 shows the matching process used by CREDO to create the virtual twins linked to each charter
school student. In the rst step, CREDO identies all TPS students who enrolled in a given charter school.
These schools are referred to as “feeder schools” for that charter. Each charter school has a unique feeder
school list for each year of data. Students attending a charter school that is also a feeder school are
eliminated from the match pool for each charter student to ensure VCRs consist entirely of TPS students. The
feeder school method provides a strong counterfactual as residential school assignment commonly used
to place students in TPS has been shown to group demographically and socioeconomically similar students
into schools. This practice increases the likelihood that students assigned to similar schools have similar
backgrounds, knowledge of school choice programs and school choice options. Once a school is identied as
a feeder school for a particular charter, all the students in that TPS become potential matches for students in
the charter school. All of the student records from all of a charter’s feeder schools were pooled: this became
the source of records for creating the virtual twin match.
4 Details of these assessments of the VCR method are presented in the Technical Appendix of the 2013 National Charter School Study, https://credo.stanford.
edu/wp-content/uploads/2021/08/ncss2013_technical_appendix.pdf
5 The majority of VCRs included only test scores which were exact matches. Non-exact matches must be within 0.1 standard deviations to be included as part
of a VCR.
For this study, CREDO partnered with education departments in 31 jurisdictions to use their student and
school level data. The resulting data set included 81 percent of tested public school students in the
United States, making it one of the largest data sets of student-level observations created to date. We
used this information to create a matched student data set with over 6,500,000 student-level observations
from over 1,853,000 charter students and a matched comparison group.
Our partnerships with the 31 individual states depend on negotiated data-sharing agreements. One common
requirement across all agreements is that the processing, analysis and security of the student-level data must
meet the Federal Education Rights and Privacy Act (FERPA) requirements. This study complies with FERPA
regulations as interpreted by each state providing data.
No single study can provide the denitive analysis on a topic as broad as the eectiveness of charter schools.
A solid body of evidence emerges only by accumulating evidence from multiple studies. With this expansion
and update to CREDO’s earlier works, we add to the growing array of studies about charter schools and their
impact on students’ academic outcomes. In doing so, we strived to create a study that was both as rigorous
and as balanced as possible.
Consolidating Student Data from Multiple States
This study is built on a methodology similar to the one used in the 2009 study. The only change to the
method was to rematch the charter school students to a new set of TPS students each year.
3
The data
collected for this study consisted of student-level demographics, school enrollment and achievement test
scores in reading/English language arts (ELA) and math. Since No Child Left Behind’s implementation, reading
and math tests have been given consistently across grades 3–8. However, testing could be more consistent
across other grades.
Many states had early elementary or high school testing. High school testing often took the form of an end-
of-course (EOC) exam, which was tied to course enrollment rather than a student’s grade. These EOC tests
diered by state in several ways that could impact growth estimates. These variations included the grade in
which the EOC exam was given, the number of times a student is allowed to take the EOC exam, and the time
gap between the EOC tested grade and the previously tested grade. All of these factors had to be considered
when constructing our data set.
Growth is the change in each student’s score from one school year to the next. For each two-year series
of individual student achievement data, we calculated a measure of academic growth. We could compute
complete growth data from the 2013–14 school year through the 2017–18 school year. Two states are missing
one year of data. Nevada is missing growth data from 2016–17 to 2017–18. Tennessee is missing data for
2015–16. Thus, the rst period of growth for Tennessee was measured from 2014–15 to 2016–17.
Additional details about creating the study data set for the 31 states in this study are available in the
Technical Appendix.
3 This change was implemented to meet the new standards of the What Works Clearinghouse at the National Center for Education Evaluation.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 3332
Student Match Rates
CREDO’s VCR matching method resulted in 81.2 percent of the charter students from participating states
being matched with TPS students to create VCRs. This match rate provided a data set with 3,669,446 matched
charter student-by-year records. The match rates vary by the race and ethnicity of the students; smaller race/
ethnic groups had lower match rates.
6
With CREDO’s strict standards to create the VCRs, smaller race/ethnic
groups have fewer identical students to match with the charter students. Table 1.2 provides the match rates
for each race/ethnicity and shows each group’s share of the data set. Racial/ethnic group match rates at the
state level are listed in the Technical Appendix.
Table 1.2: Match Rates by Race/Ethnic Group
Race/Ethnic Group Group Match Rate
Proportion of Student Body in each
Race/Ethnic Group in Study
White 84.4% 32.6%
Black 81.4% 25.3%
Hispanic 83.3% 36.0%
Asian/Pacic Islander 64.0% 3.6%
Native American 38.0% 0.3%
Multiracial 58.1% 2.3%
Students in poverty - commonly measured by those eligible for free or reduced-price lunches—had a slightly
stronger match rate (82.3 percent) than non-poverty students (79.7 percent). Match rates for ELL students
(74.9 percent) were lower than those for non-ELL students (81.7 percent).
School Match Rates
The charter school data set contained 7,288 individual schools. Almost all charter schools (98.3 percent) had
at least one school match. In seven states, all the schools had a matching school. The state with the lowest
rate was Washington, at 86.7 percent.
Fair Analysis of Impacts on Student Academic Progress
Most researchers agree that the best method of measuring school eectiveness is to look at schools’ impact
on student academic growth, independent of other possible inuences. The technical term for this is “value-
added” (Betts & Tang, 2008). The central idea is that schools should be judged on their direct contribution to
student academic progress. This necessarily considers the students’ starting scores on standardized tests
and student characteristics that might inuence academic performance. This approach forms the foundation
of our study design.
To conduct a fair analysis, this study followed the approach of the previous CREDO studies: we looked at the
academic growth of individual students as reected in their performance on state achievement tests in both
reading and math. To ensure accurate estimates of charter school enrollment on student academic growth,
6 Due to the variable distribution of students by school type and subgroup across the country, some student subgroups have low match rate in some states. Low
match rates require a degree of caution in interpreting the national pooled ndings as they may not fairly represent the learning of the student groups involved.
Figure 1.1: CREDO Virtual Control Record (VCR) Methodology
The VCR matching method eliminates any of the remaining TPS students whose demographic characteristics
do not match exactly and who did not have an identical or similar prior test score. As part of the match
process, we also drop any students who enrolled in a charter school in subsequent years from the TPS match
pool.
Using the records of TPS students at feeder schools in the year prior to the rst year of growth, CREDO
randomly selects up to seven TPS students with identical values on the matching variables in Figure 1.1,
including aligned prior test scores. Students with similar test scores were used only when there were not
enough TPS students with exact test score matches. The values for the selected TPS students are then
averaged to create values for the virtual twin. As all other observable characteristics are identical, the only
characteristic that diers between the charter student and their VCR is attendance in a charter school.
Thus, we concluded that any dierences in the post-test scores are primarily attributable to charter school
attendance (Unlu et al., 2021). The matching process was conducted separately for reading and math. Table
1.2 below displays the proportion of charter students in each racial/ethnic group for whom CREDO was able
to create a VCR.
PROVIDE LIST OF
MATCH SCHOOLS
FIND MATCHES BASED ON
DEMOGRAPHIC VARIABLES
ELIMINATE MATCHES WHO
ATTEND CHARTER SCHOOLS
MATCH TEST
SCORES
AVERAGE POST
TEST SCORES
Virtual Control Records
Charter School Student Feeder School(s) Students
MATCHING VARIABLES:
✓ Race/Ethnicity
✓ Gender
✓ English proficiency
✓ Poverty status
✓ Special education status
✓ Grade level
MATCHING VARIABLES:
✓ Test scores from prior year
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 3534
1.3 Descriptive Statistics
In this section of the report, we familiarize the reader with the attributes of the research data set and discuss
student and school dierences between charter schools and TPS. Table 1.3 describes and compares our
data. We rst provide information on the TPS sector as a whole. This sector represents all the TPS schools in
the 31 states included in the analysis. We then look at the feeder schools. Feeders are the TPS schools that
charter school students would have attended had they not enrolled in a charter school; they are a subset
of the entire class of TPS. Because charter schools are not evenly scattered across all types of locations and
communities, the feeder schools from which they draw their students have characteristics that are dierent
from the class of all TPS schools (Monarrez et al., 2022).
Student Characteristics
There are also dierences in the characteristics of enrolled students, even within the charter sector. The
students who enroll in virtual charter schools tend to dier demographically from students enrolled in brick-
and-mortar charter schools. It is important to understand how charter school students dier from the larger
body of all TPS students when generalizing charter school outcomes to other student bodies with dierent
demographics. The table below shows the student demographic characteristics for schools in the 31 states
included in the study.
Table 1.3 Demographic Comparison of Students in TPS, Feeders, and Charter Schools (Brick-and-Mortar and
Virtual) in 31 States, 2017–18
All TPS Feeders
All
Charters
Brick-and-
Mortar Charters
Virtual
Charters
Number Schools 69,706 34,792 6,802 6,588 214
Average Enrollment 552 671 463 444 1,565
Total Enrollment 37,369,048 22,658,792 2,963,468 2,755,778 207,690
% In Poverty 51% 57% 55% 56% 44%
% ELL 11% 13% 10% 11% 2%
% SPED 13% 13% 11% 11% 14%
% White 47% 40% 32% 29% 63%
% Native American 1% 1% 1% 1% 1%
% Hispanic 30% 35% 34% 36% 15%
% Black 13% 16% 25% 26% 12%
% Asian/Pacic Islander 6% 5% 4% 4% 2%
% Multiracial 4% 4% 4% 4% 7%
Brick-and-mortar charter schools enroll a larger proportion of students living in poverty than the TPS
schools in our 31-state study. Most states dene a student being in poverty as a student eligible for free
or reduced-price meal programs; however, some states use a state-specic metric to classify a student as
we used statistical methods to neutralize the inuence of student demographics and eligibility for categorical
program support, such as free or reduced-price lunch eligibility and special education. In this way, we
structured the analysis so that dierences in academic growth between the two groups are a function of
which schools they attended.
While we went to great eorts in each state to match the charter students and their virtual twins, it is
important to recognize that states dier in the location of charter schools and the students they serve. These
dierences mean that charter students are not likely to be representative of the state’s full complement
of students. These dierences are described in the Student Characteristics section. Our statistical models
included controls for these dierences between states to consider these dierences when estimating the
overall impact of charter school attendance.
Basic Analytic Models
The purpose of this study is to address multiple questions. All focused around one central question, “How did
the academic growth of charter school students compare to similar students who attended traditional public
schools (TPS)?” By answering this foundational question, we aim to extend the pool of knowledge on charter
school eectiveness and provide reliable information for policy makers.
In CSP31, we analyze charter schools’ eectiveness in the 31 states with which we have data partnerships. We
also discuss the performance change for the states covered in the 2009 and 2013 reports. These cross-study
comparisons are included by research topic when applicable.
How We Present the Results
We present the ndings in units of days of learning to make the results clearer to non-technical readers.
The statistical analysis produces results denominated in standard deviations—an unfamiliar currency to the
general public. The days-of-learning metric takes the statistical ndings of our analysis and transforms them.
It uses a protocol that was developed prior to the study and then applied here.
7
For each growth period, we
identify the one-year learning growth of an exactly average TPS student in each state and grade and set that
learning gain as “180 days of learning in 180 days of schooling.” We then take our results, student by student,
and compare their academic progress to the benchmark learning of 180 days. If a student in our study
has more learning, we award him extra days of learning on top of the 180. If a student learns less than the
benchmark, they are awarded negative days of learning which added to the 180 benchmark result in fewer
days of learning.
8
While transforming the statistical results into days of learning provides a more accessible measure, the
days of learning are estimates and should be used as general guides (Hanushek & Rivkin, 2006). We provide
the dierence in growth in standard deviation units in the outputs of the statistical methods used for each
analysis found in the Technical Appendix.
7 Using nationwide growth data from the National Assessment of Education Progress, the transformation involves multiplying the standard deviation units produced
by our statistical analyses by 578 days. This yields 5.78 days of learning for every 0.01 standard deviation dierence in our analysis. For those wanting to convert
these larger counts into weeks or months: a school week consists of ve days; a school month is 20 days and a quarter or nine-week term is typically 45 days.
8 The expression “additional days of learning” does not mean the students were necessarily in school for more days during the school year. It means that the
additional learning that took place in charter schools during the school year was equivalent to attending school for x additional days in a TPS setting.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 3736
Other Student Characteristics
For other student characteristics, 57.8 percent of students in the study are students in poverty—dened as
eligible to receive free or reduced-price lunches or using their state’s specic economic metric to identify
students in poverty.
10
English-language learning students (ELL) made up 7.1 percent of the data set. Students
receiving special education services made up 7.9 percent of the data set. Just over half (51.5 percent)
of charter school students are female. The proportions of the matched student body are similar to the
proportions of these special populations in the larger sample of all charter students in the 31 states.
Table 1.5: Special Population Proportions for All versus Matched Students
Special Population
Proportion of Student Body in each
Special Population — All Charters
Proportion of Student Body in each Special
Population — Matched Student Data Set
In Poverty 55% 58%
ELL 10% 7%
SPED 11% 8%
One in four students in the data set is a Hispanic student in poverty (26.2 percent), while 20 percent of
students are Black students in poverty. Also, 6.2 percent of students in the data set are Hispanic ELL
students.
11
While the national distribution ts the expected pattern, student achievement decile patterns vary greatly by
state. For example, in 2017, Pennsylvania drew a larger percentage of its charter enrollment from the lower
deciles, as do Michigan and Ohio. The opposite—higher achieving students enrolling in charter schools—is
found in New York City, North Carolina and Arizona. In the gure below, there are 10 boxes in each state, with
the lowest box being the rst decile (lowest achievement) and the highest box representing the 10th decile
(highest achievement).
Figure 1.2 shows the within-state decile of the average achievement level for all students enrolled in charter
schools by state for math. There is a wide variance in the achievement levels of charter students in dierent
states. While some states have average charter student achievement as high as the sixth decile, which means
the average charter student has achievement above the average TPS student in the state, most are in the
third and fourth deciles. The average achievement scores are due to a combination of new charter students’
entry-level achievement and the impact of attending charters for existing charter students.
10 CREDO acknowledges the declining usefulness of free and reduced-price lunch eligibility as an indicator of poverty. We have used a state-specic variable in
states where a better metric is available. For the remaining states, free or reduced-price lunch eligibility was the best indicator available (Fazlul et al., 2023).
11 Hispanic students in poverty and Hispanic ELL students are not mutually exclusive groups. A student could be in both.
being in poverty. We treat these two methods as equally valid for these analyses. The percentage of students
in poverty in charter schools is similar to those in poverty in the feeder schools that students would have
attended if not enrolled in their charter schools. The percentage of charter school students in brick-and-
mortar charter schools identied as English learners and students receiving special education services is
comparable to that of the full set of TPS schools and feeder schools. The brick-and-mortar charter schools
have twice the rate of Black student enrollment as the TPS schools and 10 percentage points higher than their
feeder schools. The enrollment rate for Hispanic students in brick-and-mortar charters is similar to that in the
set of feeder schools, yet lower than the overall rate for all TPS schools. These increased enrollment rates for
Hispanic and Black students were oset by lower rates in brick-and-mortar charters for White students than
in the feeder charters and the complete TPS set of schools.
When it comes to student proles, virtual charter schools have dierent proles from the other forms of
charter schools, traditional public schools and brick-and-mortar charters. Virtual charters have a smaller
percentage of students living in poverty, students identied as English learners, Hispanic students and Black
students. On the other hand, they have a disproportionately high number of White students relative to the
other groupings mentioned in Table 1.3.
Race/Ethnicity Composition of Matched Charter Students
9
The data set was made up of matched charter students with at least two successive test scores who
attended the public charter schools in the years under study in the included states. Therefore, the makeup
of the student body for this study will dier slightly from the student body described in the overall charter
landscape and the 31-state summary (see Table 1.3).
Table 1.4: Race/Ethnic Proportions for All versus Matched Students
Race/Ethnic Group
Proportion of Student Body in each
Race/Ethnic Group — All Charters
Proportion of Student Body in each Race/
Ethnic Group — Matched Student Data Set
White 32% 33%
Black 25% 25%
Hispanic 34% 36%
Asian/Pacic Islander 4% 4%
Native American 1% 0.3%
Multiracial 4% 2%
The largest race/ethnic group included in the study is Hispanic students, who comprise 36 percent of the
matched data set. The next-largest groups are White students (32.6 percent) and Black students (25.3
percent). Asian and Pacic Islander students are 3.6 percent of the data set. Multiracial students, those of two
or more races, are 2.3 percent of the students in the analyses, and Native American students make up the
smallest portion, with only 0.3 percent of students identifying as Native American only.
9 Because the VCR matching protocol produces a single record (the average of up to seven TPS matched students), the demographic proles of charter and VCR
student-year records are identical.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 3938
Table 1.6: Achievement Decile Distribution of Charter Enrollees by State 2017, Math
Charter Student Enrollment by Achievement Decile in 2017 (2016 for MD)
This table contains data that is available in an interactive format on the study website.
State
Decile
1
Decile
2
Decile
3
Decile
4
Decile
5
Decile
6
Decile
7
Decile
8
Decile
9
Decile
10
Total
AR 7% 12% 12% 11% 8% 12% 8% 9% 9% 11% 14,506
AZ 5% 7% 8% 9% 9% 10% 12% 13% 14% 12% 74,868
CA 8% 10% 10% 10% 10% 11% 11% 11% 11% 8% 185,840
CO 7% 9% 10% 10% 10% 10% 11% 11% 11% 10% 50,395
DC 8% 11% 13% 10% 11% 11% 11% 10% 9% 5% 7,486
FL 5% 6% 8% 10% 11% 12% 14% 13% 12% 9% 113,763
ID 6% 7% 8% 8% 9% 11% 12% 13% 15% 11% 8,329
IL 12% 17% 14% 12% 12% 10% 9% 7% 6% 2% 16,210
IN 16% 16% 14% 12% 11% 10% 8% 6% 5% 2% 10,539
LA 11% 13% 13% 12% 10% 11% 10% 8% 7% 4% 30,263
MA 9% 11% 11% 12% 11% 11% 11% 10% 8% 5% 14,962
MD 15% 16% 14% 13% 10% 9% 8% 8% 4% 2% 20,056
MI 17% 17% 14% 11% 9% 9% 8% 6% 6% 4% 44,967
MN 15% 12% 10% 11% 10% 10% 10% 9% 8% 5% 17,674
MO 14% 13% 13% 14% 12% 12% 9% 7% 3% 1% 7,386
NC 7% 7% 7% 9% 11% 11% 13% 12% 13% 10% 33,817
NJ 11% 13% 12% 12% 11% 10% 10% 9% 8% 5% 19,944
NM 7% 10% 10% 11% 10% 10% 10% 10% 10% 11% 9,133
NV 5% 8% 8% 9% 10% 10% 12% 13% 14% 12% 19,153
NY 13% 10% 12% 13% 13% 12% 11% 8% 6% 3% 8,200
NYC 4% 7% 7% 10% 11% 13% 14% 15% 12% 8% 41,627
OH 21% 20% 14% 12% 9% 8% 6% 5% 3% 2% 29,618
OR 5% 9% 10% 10% 11% 12% 12% 12% 12% 7% 7,306
PA 16% 28% 16% 11% 8% 7% 5% 4% 3% 2% 38,985
RI 7% 8% 11% 12% 11% 12% 11% 12% 9% 7% 2,157
SC 5% 14% 13% 11% 11% 10% 9% 9% 7% 10% 11,636
TN 16% 15% 14% 13% 11% 10% 8% 7% 5% 2% 19,924
TX 10% 12% 9% 8% 9% 9% 8% 12% 14% 8% 104,665
UT 11% 9% 9% 9% 10% 11% 11% 12% 11% 6% 26,108
WA 12% 11% 10% 11% 11% 11% 12% 10% 7% 4% 1,244
WI 9% 11% 11% 11% 13% 13% 11% 10% 8% 4% 15,648
9% 11% 10% 10% 10% 10% 10% 11% 10% 7% 1,006,409
To get a better understanding of the achievement of students entering charter schools, Figure 1.2 provides
the percentage of charter students in each state with student math achievement in each decile. If charter
schools drew their students from the same deciles as TPS, we would expect roughly 10 percent of students
to come from each decile.
12
However, the patterns in Figure 1.2 show that is not typical. Some states draw
a disproportionate share of their students from the lower deciles, creating a pyramid-shaped distribution.
Other states invert the pyramid by pulling more high-achieving students into charter schools than the TPS.
Much of the achievement distribution of charter school enrollees has to do with where charter schools are
located. In states where charter schools are located primarily in urban locations, we would expect more
lower decile students to enroll in charter schools. We could expect to see a more even distribution in states
where charter schools are distributed more evenly throughout the state.
Figure 1.2: Average Achievement of All Charter Students by State, Math 2017
13
12 Decile by state percentages for charter school reading achievement are included in the Technical Appendix. The distributions support the insights gleaned from
math achievement.
13 Results for Reading are available in the Technical Appendix
Decile of Achievement
1
2
3
4
5
6
7
8
9
10
TOTAL
WI
WA
UT
TX
TN
SC
RI
PA
OR
OH
NYC
NY
NV
NM
NJ
NC
MO
MN
MI
MD
MA
LA
IN
IL
ID
FL
DC
CO
CA
AZ
AR
TOTAL
WI
WA
UT
TX
TN
SCRI
PA
OR
OH
NYC
NY
NV
NM
NJ
NC
MO
MN
MI
MD
MA
LA
IN
IL
ID
FL
DC
CO
CA
AZ
AR
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 4140
Perceptions of Charter School Student Advantage
In the past, a common claim asserted that positive academic results in charter schools arise from advantages
that their students bring to their schooling. The claim has taken one of two forms: a “push” on the part of
parents or a “pull” on the part of charter schools. The “push” alternative posits that charter school students
have parents that, by the act of enrolling their student in a charter school, reveal they value education more
and/or are more motivated to pursue educational success for their children than other parents. As a result of
parental investments of time and resource, their students are thought to be better prepared academically.
The “pull” version suggests that charter schools signal or sift interested students to enroll more advantaged
students, drawing them away from neighborhood schools. This practice commonly is referred to as “cherry
picking” or “cream-skimming.”
Despite dierent mechanisms, both versions of “charter school students are advantaged” can be tested with
the same analysis. If either or both claims are true, then entering charter school students would present
stronger academic preparation than the students in the feeder TPS schools. With our analysis, we advance
earlier work to examininge achievement distribution for low-end and high-end dierences in starting
achievement (Kho et al., 2022; Zimmer et al., 2009).
We compare students who initially enrolled in a TPS and took at least one achievement test before
transferring to a charter school to their peers who enroll in the TPS. We can observe the distribution of
charter students’ test scores across deciles of achievement and do the same for students in the feeder
TPS. Taking the dierence in the two percentages for each decile illuminates how equal the distributions
of student achievement are in the two school settings. We conduct the analysis by subject for each state,
yielding 62 tests (i.e., 31 states and 2 subjects).
Table 1.6 presents the results. For example, in Michigan, the share of students entering charter schools from
the bottom three deciles of achievement is 24.4 percentage points larger than the share the feeder schools
enrolled. We consider two percentage points dierence for any achievement decile as natural variation.
Table 1.6 presents reveals important results at both ends of the achievement continuum. In 17 states, charter
schools enroll more students from the bottom three deciles of achievement than do their feeder schools. In
many cases, the share is 10 to 20 percent larger than in feeder schools. For eight states, the dierences fall
in the 2-percentage margin of variation. In ve states new charter school student enrollment in the lowest
deciles is smaller by three to six percentage points.
At the upper end of the achievement range, in three states, the share of charter school enrollment from the
top three deciles is three percent larger than their feeder schools. Six states have equivalent enrollment. In 21
states, charter schools enrolled smaller shares of top-decile students than their feeder schools, with smaller
enrollments upwards of 17 percentage points.
To recap the analysis, across the 62 tests the claim charter schools are advantaged by the students they
enroll was unfounded in 54. Where the distributions dierered, the balance of evidence shows larger shares
of students entering charter schools with achievement in the lowest deciles and smaller shares of students
had prior achievement in the highest deciles than in the schools they left. In the handful of tests where the
entering student distribution favors charter schools, the advantage is insubstantial. The evidence dispels
claims that charter schools gain an unfair edge by enrolling “better” students.
Figure 1.3: Average Academic Growth of Charter Students by State, Math 2017
Decile of Achievement
TOTAL
WI
WA
UT
TX
TN
SC
RI
PA
OR
OH
NYC
NY
NV
NM
NJ
NC
MO
MN
MI
MD
MA
LA
IN
IL
ID
FL
DC
CO
CA
AZ
AR
TOTAL
WI
WA
UT
TX
TN
SC
RI
PA
OR
OH
NYC
NY
NV
NM
NJ
NC
MO
MN
MI
MD
MA
LA
IN
IL
ID
FL
DC
CO
CA
AZ
AR
-60
-50
-40
-30
-20
-10
0
10
20
30
40
50
60
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 4342
School Characteristics
School Location
The majority of charter school students in the study attend charter schools located in an urban setting
(53.6percent), according to the locale designations of the National Center for Education Statistics (NCES). The
study’s second largest group of students is those attending a suburban charter school, at 29 percent. Rural
charter school students (8.4 percent) and charter students in towns (3.1 percent) comprise the remainder of
the brick-and-mortar charter school students. The remaining 5.9 percent of charter school students attend
online charter schools.
15
Figure 1.4: Percent of Charter School Student Enrollment by Location
The geographic settings where charter students attend school remain relatively stable between the 2013 and
the 2023 studies. The one exception is the number of rural students dropping by half, from 16 percent of the
matched sample to just eight percent. Between the two studies, CREDO added the virtual category to report
separately for students who attend online charter schools. In the 2013 study, students attending online
schools were categorized by the location of the online schools’ headquarters. Any changes in locale reporting
would impact only the comparisons between locale reporting when comparing outcomes between the 2013
and 2023 studies. The larger overall and state-level comparisons will not be impacted.
15 While online charter schools are assigned an NCES locale based on the locations of their oces, for this study we group students attending an online charter
into a separate “online” locale regardless of where the school’s oces are physically located.
Table 1.7: Percentage Dierences between Entering Charter Students and Feeder School Students by
Decile of Achievement
14
Achievement Group
State Bottom Deciles 1- 3 Middle Deciles 4-7 Top Deciles 8 - 10
AR -0.36 -1.31 1.68
AZ -1.56 1.31 0.26
CA 4.28 0.77 -5.05
CO 3.71 -0.64 -3.07
DC 8.66 0.21 -8.87
FL -5.01 7.36 -2.33
ID -5.13 1.94 3.17
IL 6.20 0.96 -7.16
IN 14.16 -0.50 -13.65
LA 10.29 2.34 -12.63
MA 1.42 -1.13 -0.28
MI 24.44 -7.17 -17.26
MN 13.57 -2.17 -11.42
MO 10.39 2.75 -13.13
NC -3.04 0.49 2.54
NJ 9.60 -0.89 -8.73
NM 0.56 0.75 -1.31
NV -3.32 3.18 0.13
NY 4.78 5.87 -10.67
NYC -1.73 8.86 -7.13
OH 20.56 -4.62 -15.96
OR -1.57 4.73 -3.16
PA 26.03 -8.88 -17.16
RI -6.14 10.08 -3.92
SC 0.53 1.86 -2.40
TN 4.68 6.50 -11.19
TX 3.93 -7.25 3.32
UT -0.69 1.41 -0.74
WA 2.64 2.92 -5.56
WI 6.47 3.06 -9.51
WI 9.00 11.00 11.00
14 Full breakout by decile is included in the Technical Appendix.
Rural
Town
Suburban
Urban
Virtual
53.6%
29.0%
3.1% 8.4%
5.9%
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 4544
1.4 Analytic Findings
Academic Growth of Charter School Students
The typical charter school student in our national sample has statistically signicant positive year-over-year
growth in both math and reading compared to the TPS VCRs.
16
The benet of attending charter schools
during the period of study amounts to additional days of learning equivalent to six days in math (0.011) and
16 days in reading (0.028).
17
Figure 1.6: Annual Academic Growth of Charter School Students, Reading and Math
** Signicant at p ≤ 0.01
This is a step forward for charter school performance over CREDO’s two previous national studies. In the
2009 national study, students attending charter schools had less growth in both math (17 days less) and
reading (six days less) than their TPS VCRs. In the 2013 national study, the growth of charter students was not
signicantly dierent from their TPS VCRs in math but was signicantly stronger in reading (six days more).
16 Throughout this report, numbers referred to as “signicant” are statistically signicant at least at the 0.05 level. In graphics, a single star (*) means statistically
signicant at the 0.05 level; double stars (**) means statistically signicant at the 0.01 level. Dierences that are not statistically signicant are reported as
being similar.
17 As described in the Methodology section of this report, when we transform our analytic growth results from standard deviation units to days of learning, a .01
standard deviation equates to 5.78 days of learning.
School Level/ Grade Span
We also group students into school levels based on the NCES grade-span categories: elementary, middle,
high, and multilevel schools. This gives us a picture of the distribution of charter school enrollment by
school conguration. The majority of charter school students in our study (40.7 percent) are enrolled in K-6
elementary schools; 16.6 percent of charter school students in our study are enrolled in stand-alone middle
schools (grades 6–8); and 5.6 percent are enrolled in charter high schools (grades 9–12). Multilevel schools
serve a combination of grades outside traditional school grade groupings. For example, K-8 schools, 6–12
schools or schools that enroll students in K-12. Students in these schools make up 37.1 percent of charter
school students in this study.
Figure 1.5: Percent of Charter Schools by Grade Level
Between the 2013 and 2023 studies, the only major changes we see in locales are an eight percentage point
decrease in the proportion of charter students attending high schools and a seven percent increase in the
proportion of charter schools classied as multilevel schools.
Multilevel
High School
Middle School
Elementary School
40.8%
37.0%
16.6%
5.6%
0
2
4
6
8
10
12
14
16
18
MathReading
16**
6**
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 4746
Figure 1.8: Annual Academic Growth in Previously Studied Schools Compared to Current Schools
** Signicant at p ≤ 0.01
One mechanism by which the existing charter schools can get stronger over time is by the closure of poorer-
performing charter schools. However, there are reasons other than academic performance that can lead
to school closure. We examined the performance of the 207 charter schools that closed during our study
window. The performance of th ese closed schools was similar in both subjects. Using reading to illustrate,
the majority of these charter schools that closed (58 percent) were those with below-average achievement
and weaker growth than their TPS comparisons. However, 30 percent of the schools that closed had stronger
growth than their TPS comparisons, even if their achievement was below the state’s average. Surprisingly,
seven percent of the closed schools had stronger growth than their TPS comparisons and above-average
achievement for their state.
Figure 1.7: Annual Academic Growth of Charter School Students across Three National Studies
** Signicant at p ≤ 0.01
While these results are the national averages for charter school students, the results vary greatly from
state to state and by student characteristics. Since many charter school policies are set at the state
level, dierences across states are partly a function of variation in charter schools’ legal and regulatory
environments. Below we examine the outcomes by dierent student subpopulations.
To explore the trend of improved performance, we examined the pooled national data to see if schools that
are new to our sample (by being new or having tested grades for the rst time) had dierent results than
schools that were included in earlier national studies. This comparison provides a partial view of the source
of overall improvement over time. The existing charter schools had stronger growth than their TPS peers in
reading (+18 days) and math (+10 days). The new-to-the-study schools had stronger growth in math (+13 days)
and identical growth in reading as their TPS peers. Based on these results, the larger part of the improved
performance of charter schools since the 2013 study stems from the earlier cohort getting stronger.
Interestingly, the new schools in this study had better performance than new schools in the second national
study and outpaced overall growth for all charter schools in both prior studies.
MathReading
18**
13**
10**
1
0
5
10
15
20
25
Schools in NCSS2
Schools not in NCSS2
Schools in NCSS2
Schools not in NCSS2
Days of Learning
-6**
-17**
Days of Learning
-25
-20
-15
-10
-5
0
5
10
15
20
25
3rd Study2nd Study1st Study
-3
6**
6**
16**
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 4948
Charter School Student Academic Growth by State
Charter school students had weaker reading growth than their TPS peers only in Oregon. Signicantly
stronger growth by charter students was seen in 18 states. In the remaining 12 states, growth for students
attending charter schools was similar to that of their TPS peers. The strongest gains were in the Northeast.
Rhode Island (+90 days) and New York State (+75 days) charter students saw the largest gains. New York City
(+42 days) students had strong gains, as did students from Massachusetts (+41 days).
Figure 1.10: State Level Average Charter School Student Academic Growth, Reading
RECAP: Academic Growth of Charter School Students by Type of School
To complement these aggregate analyses, CREDO expanded the analyses of charter school student academic
growth by distinguishing the progress of students attending charter schools associated with charter
management organizations (CMO) from those attending stand-alone charter schools (SCS).
18
The complete set
of ndings is available in the second volume: Charter Management Organizations 2023 Students attending CMO-
aliated charter schools have statistically signicant positive learning gains in reading and math compared to
their TPS peers with similar observable characteristics. Students attending SCS had stronger growth in reading
and similar growth in math to their TPS peers. Figure 1.10 shows these dierences to be equivalent to an
additional 27 days of learning in reading and 23 days in math for students attending charter schools associated
with a charter management organization over their comparison group. This is contrasted to 10 additional days
in reading and similar growth in math for students attending SCS as compared to their VCRs.
Figure 1.9: RECAP: Average Academic Growth for Charter School Students by Charter School Type, Reading and
Math
** Signicant at p ≤ 0.01
Figure above originally appears as Figure 2.4 in CMO23.
18 The CMO study does not include Idaho, Maryland, and Ohio.
10**
-3
27**
23**
Days of Learning
SCS CMO
-5
0
5
10
15
20
25
30
MathReading
26.0
7.5
11.0
24.3**
3.5
3.5
39.3**
15.0
36.4
74.6**
33.5**
13.3**
-8.1
8.1**
-1.2
37.0**
DC 4.6
32.9**
RI 90.2**
41.0**
-4.0
21.4**
39.9*
2.9
15.0*
-1.7
19.7**
-18.5*
11.0 **
17.3*
** Signicant at p ≤ 0.01 * Signicant at p≤ 0.05
In 12 of the 31 states in our study, charter school students had signicantly stronger growth in math than
their peers in TPS. Only three states showed weaker growth for charter students compared to their peers.
The remaining 16 states’ math growth was similar between charter students and their TPS peers. Of the
states with signicantly dierent growth for charter students, the largest eects were in Rhode Island, New
York City, and New York State.
19
Charter students in Rhode Island gained the equivalent of attending an extra
88 days of learning per school year over their TPS peers. Charter students gained an additional 80 days in
New York City and 73 days in the rest of New York State.
19 CREDO treats New York City as its own state because the size of New York City would overwhelm the New York State results and because New York City has
several city-level policies that impact education outcomes.
NYC 42**
NY 75**
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 5150
Rhode Island stood out in particular for having high growth in both subjects for both studies, even though
the state’s change in scores from 2013 to 2023 was negligible. Other notable improvements were Missouri
in math and New York state, New York City, and Texas in reading. Even though Texas had a smaller reading
growth score in 2023 than several other states, its change in growth from 2013 was larger.
Figure 1.12: Average Reading Growth of Charter School Students by State, 2013 vs 2023
Figure 1.13: Average Math Growth of Charter School Students by State, 2013 vs 2023
Figure 1.11: State Level Average Charter School Student Academic Growth, Math
Low Growth 2013
High Growth 2023
Low Growth 2013
Low Growth 2023
High Growth 2013
Low Growth 2023
Additional Days of Growth 2013
Additional Days of Growth 2023
High Growth 2013
High Growth 2023
-120
-100
-80
-60
-40
-20
0
20
40
60
80
100
120
100806040200-20-40-60-80-100
NV
PA
RI
NY
TN
DC
LAIN
NJ
MI
MA
NYC
MN
MO
IL
TX
AZ
OH
AR
CO
NC
CA
UT
FL
NM
OR
Low Growth 2013
High Growth 2023
Low Growth 2013
Low Growth 2023
Additional Days of Growth 2013
Additional Days of Growth 2023
NV
PA
RI
NY
TN
DC
LA
IN
NJ
MI
MA
NYC
MO
IL
TX
AZ
OH
AR
CO
CA
MN
NC
UT
FL
NM
OR
High Growth 2013
Low Growth 2023
High Growth 2013
High Growth 2023
-120
-100
-80
-60
-40
-20
0
20
40
60
80
100
120
100806040200-20-40-60-80-100
** Signicant at p ≤ 0.01 * Signicant at p≤ 0.05
Link to full state math and reading results here
Changes in Charter School Student Academic Growth by State
Having longitudinal data over multiple studies allows us to examine the performance of states relative to
each other and each state’s performance over time. This helps us understand the impact of state policies
over time. Of the 25 states in the 2013 and 2023 studies, state charter school academic growth in reading
increased in 2023 for 17 states and decreased for eight states.
20
The dierences for New Jersey, Oregon, and
Michigan were trivial.
In math, 15 states had stronger growth in 2023 than in 2013, and 10 had weaker growth. The dierences for
Ohio, Florida, and Rhode Island were negligible.
The largest decreases in both subjects occurred in Washington, D.C., and Louisiana. The largest increase
for both subjects was in Nevada. Nevada had charter growth in 2023 that was not signicantly dierent
from the state’s VCRs. Still, that modest performance was a vast improvement over the extremely negative
performance of Nevada charter schools in 2013.
20 Due to the mandated destruction of data les from prior studies, dierences for each state between periods could not be tested for statistical signicance.
39.3
3.5
6.9
-4.6
5.8
-1.2
56.1*
16.2
23.7**
72.8**
38.7**
-16.8
-47.4**
-1.2
-21.4
37.0**
DC 32.4**
31.8**
RI 87.9**
41.0**
-37.6**
7.5
48.0**
-26.6
13.3*
-14.5
0.6
-31.8**
4.0
7.5
NYC 80**
NY 73**
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 5352
Figure 1.14: Days of Learning for Charter School and TPS Students by Race/Ethnicity, Reading and Math
Dierences in Academic Growth by Charter School Student Characteristics
Dierences by Race/Ethnicity
As seen in our earlier studies, the impacts of attending charter schools are not uniform for all students. When
looking at student groupings, our analyses reveal varying overall status and growth for dierent subgroups.
Therefore, it is important to examine the outcomes for students by this characteristic to gain a deeper
understanding of the overall results at both the national and state levels. Students from dierent racial and
ethnic groups can have opposite impacts from attending charter schools, which is obscured when looking at
overall student outcomes. For example, in previous CREDO studies, White students attending charter schools
generally have weaker growth than their peers attending TPS. Asian/Pacic Islander and Native American
students tend to have growth similar to their peers. However, previous studies have shown that for Black and
Hispanic students, attending a charter school often leads to signicant academic growth.
We compared the academic growth across student race/ethnicity
groups. Students were grouped into White, Black, Hispanic, Asian/
Pacic Islander, Native American, and Multiracial students.
Compared to their TPS peers, Black students attending charter
schools had 35 days more growth in a school year in reading and
29 days in math. This would be as if the students had attended
an additional 1.5 months of schooling each year. The results were
also positive and signicant for Hispanic students. Relative to their
TPS peers, Hispanic students grew an extra 30 days in reading and
19 additional days in math. Only two subpopulations of charter
school students had weaker growth than their TPS peers in math.
White and Multiracial students attending charter schools grew
24 fewer days per school year than their TPS peers. No racial/
ethnic subpopulations had weaker growth than their TPS peers in
reading.
However, because the TPS peer groups often have growth weaker
than the average 180 days per year that anchors these analyses,
even those subpopulations with positive growth may experience
less than 180 days of growth per school year. The gure below
shows the typical growth in math for each subpopulation of
charter students and their TPS peers.
UNDERSTANDING
SUBPOPULATION
RESULTS
In these analyses, the
growth of subpopulations in
charter schools is compared
to the growth of the same
subpopulations in TPS. This
means learning for Black
charter school students is
compared to their Black
TPS peers. Both TPS and
charter student results are
studied against the 180-day
baseline for White comparison
students.
White
Native American
Multiracial
Hispanic
Black
Asian/Pacific Islander
Race/Ethnicity
260240220200180
180
1601401201008060
Asian/Pacific Islander
Black
Hispanic
Multiracial
Native American
White
TPS
Charter
Reading
Math
Race/Ethnicity Sector
Days of Learning
White
Native American
Multiracial
Hispanic
Black
Asian/Pacific Islander
Race/Ethnicity
260240220200180
180
1601401201008060
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 5554
Figure 1.15: Annual Academic Growth for Charter School Students in Special Populations
** Signicant at p ≤ 0.01 * Signicant at p≤ 0.05
The ndings for separate subgroups detail the growth we observe for all charter students in each group,
all else being equal. Of course, that is not the case; within groups, we know that some students, in addition
to being in a minority group, also experience poverty or are not native English speakers. Students with
compound designations are likely to face more challenges in their education. CREDO studied three such
groups: Black students in poverty, Hispanic students in poverty and Hispanic students who are also English-
language learner students.
In the current study, we nd that Black students in poverty had 37 days stronger growth in reading and
36 days stronger growth in math when compared to their TPS peers. The results were similar for Hispanic
students in poverty: they grew 36 more days in reading and 30 more in math than their TPS peers. There
were also signicant benets for Hispanic students who are English-language learners (ELL). Hispanic ELL
students gained an additional 11 days in reading and an extra eight days of learning in math by attending
charter schools instead of their local TPS option.
Relative to the standard of 180 days of learning per year—the amount of growth that the average White TPS
student in this study makes each year—Figure 1.15 delivers two essential ndings. First, Black and Hispanic
students in charter schools advance more than their TPS peers by large margins in math and reading.
Multiracial, Native American, and White charter students show the reverse in math, lagging behind the
growth of their TPS peers. Reading progress was equivalent for these subpopulations. Asian/Pacic Islander
students in both sectors show similar growth.
The second conclusion is more sobering: neither in the typical charter schools nor in the comparison TPS are
Black, Hispanic, or Native American students posting growth that is close to 180 days a year in either reading
or math. Multiracial students fare better but still don’t reach typical growth. White students in charter
schools are on par in reading and lag in math. Only Asian/Pacic Islander students, a small fraction of the
student population, post better growth than the average growth of White TPS students. The message is clear.
The majority of students in both settings are not learning as much as they need to for their schooling to be
on track. These growth gaps are the building blocks of learning inequality that result in the achievement gaps
that plague the nation.
Beyond the picture of dierent results at the average for dierent groups of students, the insights available
from the distribution of student experience are potentially transformative. There are thousands of minority
and economically disadvantaged students whose progress outpaces or is on par with White students in their
school. We note these gap-busting cases and present more detail in the school-level results below.
Academic Growth for Charter School Students in Special Populations
Many studies have shown persistent disparities between students at the upper and lower ends of the
socioeconomic spectrum (Duncan & Murnane, 2016; Hanushek et al., 2019). In this study, charter school
students in poverty had stronger growth equal to 17 additional days of learning in math and 23 days stronger
growth in reading than their TPS peers. English-language learner students who attended charter schools
also had stronger growth in math (eight days) and reading (six days) than their TPS peers. However, students
receiving special education services had signicantly weaker growth in both math and reading than their TPS
peers. Specically, they grew 13 fewer days in reading and 14 fewer in math. Compared to the 2013 National
Charter School Study, these most recent results represent a slight increase in charter school eectiveness
forstudents in poverty and a slight decrease in eectiveness for English-language learners and special
education students.
23**
17**
6*
8**
-13** -14**
-20
-15
-10
-5
0
5
10
15
20
25
MathReading
Days of Learning
SPEDELLIn poverty
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 5756
Figure 1.17: Annual Academic Growth of Charter School Students by Grade Level
** Signicant at p ≤ 0.01
Examining grade level charter performance against earlier CREDO study results, shown in Table 1.7, we
see a marked improvement in all grade levels in both subjects. Seeing growth in all grade spans helps us
understand that trends in the national aggregate performance are not concentrated in particular grades.
Table 1.8: Charter School Student Academic Growth by Grade Level across Studies, Reading and Math
Reading Math
2009 2013 2023 2009 2013 2023
Elementary 6** 17** 24** 0 12** 17**
Middle 12** 23** 29** 12** 29** 29**
High -29** 0 25** -12** 0 31**
Multilevel -46** -12** 1 -23** -40** -20**
** Signicant at p ≤ 0.01
Figure 1.16: Annual Academic Growth for Charter School Students with Compound Designations
** Signicant at p ≤ 0.01
Student Annual Academic Growth by Charter School Grade Level
Another way CREDO has typically looked at charter school performance has been by examining charters
grouped by the grades served by the school. Typically, there are four levels of schools. These are elementary
(K–5), middle schools (6–8), high schools (9–12) and multilevel schools (a mix of grades that do not fall easily
into one of the other categories, e.g., K–6, 6-12 or K-12).
The results show signicantly positive growth in reading and math for charter schools serving elementary,
middle or high school-age students. In contrast, results for multilevel charter schools were negative in math
and similar to the TPS comparison groups in reading.
In reading, the results for charter schools were stronger. The average increase in growth for elementary
charter school students was 24 additional days of learning. Middle school students saw 29 extra days and
high school students saw 25 extra days on average. Students attending multilevel charter schools had growth
similar to their TPS peers.
The average impact on math growth for charter school students was the same as attending 17 extra days
for elementary students, 29 extra days for middle school students, and 31 additional days for high school
students. Multilevel charter school students, on average, had 20 days less learning per school year.
37**
36** 36**
30**
11**
8**
MathReading
0
5
10
15
20
25
30
35
40
Hispanic ELLHispanic in PovertyBlack in Poverty
Days of Learning
24**
17**
29** 29**
25**
1
-20**
31**
MathReading
Days of Learning
-20
-15
-10
-5
0
5
10
15
20
25
30
35
MultilevelHighMiddleElementary
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 5958
Academic Growth by Continuous Enrollment in Charter School
Students often have their weakest growth in their rst year in a charter school (Cremata et al., 2013). This ts
the known issues around school transitions and decreases in student performance. As seen in Figure 1.20,
the subset of students who enroll in a charter school during our data window follow the pattern.
23
However,
charter students overcome the initial dip associated with a school change, as shown below. By their fourth
year in their charter school, students show 45 days stronger growth in reading than their TPS peers and 39
additional days of learning per year in math. However, it should be noted that the sample size of students
attending a charter school for four years is small, limited by the number of tested grades available for study
and the alignment of the study window with grade patterns in schools.
Figure 1.19: Annual Academic Growth for Charter School Students by Years of Enrollment
** Signicant at p ≤ 0.01
23 This analysis included only those students seen entering charter schools from a TPS. Students already in charter schools in their rst year of the data window
were excluded.
Annual Academic Growth of Online Charter School Students
While enrolling only six percent of the charter school student population, online schooling continues
to grow over time (Lehrer-Small, 2022). CREDO and other partners conducted a study of online charter
schools in 2015, nding signicant underperformance in the online setting compared to brick-and-mortar
charter schools. With time since the previous study and additional focus from a number of charter school
authorizers, we reexamine the growth of students attending online or brick-and-mortar charter schools
compared to their TPS peers.
21
Brick-and-mortar charter school students had signicantly stronger growth in reading (22 more days)
and math (15 more days). Online charter school students had much weaker growth. Online charter school
students grew 58 fewer days in reading and 124 fewer days in math than their TPS peers.
Stated another way, compared to 180 days of learning for their brick-and-mortar TPS peers, the learning for
an average online charter student advanced only 122 days in reading; in math, the progress for online charter
students was 56 days per year. While across the nation, six percent of charter school students attend a virtual
charter school; in Ohio and South Carolina, this rate is as high as 14 percent.
22
It is important to note that examples of equivalent or better academic growth for students in virtual charter
schools exist today, and their numbers have increased. These neutral and positive examples buck the
preponderance of the evidence: of the 214 virtual charter schools in the study, 73 percent had weaker growth
than their comparison group in reading, and 90 percent underperformed their comparison group in math.
Figure 1.18: Annual Academic Growth for Charter School Students by School Mode, Reading and Math
** Signicant at p ≤ 0.01
21 The comparison students for online charter students come from brick-and-mortar TPS. It is not possible to create comparison students from online TPS only.
22 Massachusetts, Maryland, New Jersey, New York, Rhode Island, and Tennessee do not allow online charter schools.
-58**
-124**
22**
15**
MathReading
-150
-120
-90
-60
-30
0
30
Brick-and-mortar charter schoolsOnline Charter Schools
Days of Learning
-17**
17**
31**
45**
39**
12**
27**
-35**
Years in Charter School
-40
-20
0
20
40
60
4 years in3 years in2 years in1 year in
Days of Learning
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 6160
Figure 1.20: Charter School Student Academic Growth by School Location, Reading and Math
** Signicant at p ≤ 0.01
1.5 School-Level Impacts
School-Level Academic Growth
Analyzing school-level performance is another valuable way to assess the eectiveness of charter schools.
Though the overall results of the charter school sector are largely positive in reading and math, it does not
follow that every individual charter school performs better than the alternative. School performance is
important for policy-related decisions such as funding or renewal. Thus, it is helpful to test if charter schools
deliver academic progress that is greater, the same, or smaller than is seen for identical students in the
feeder schools nearby.
In order to determine school-level charter performance, we compute each charter school’s average impact on
student learning over the two most recent growth periods (2017 and 2018). We compare the school average
to the same measure of learning for their TPS VCRs.
25
The average gains of the TPS VCRs serve as a proxy
comparison of what learning would have occurred for a charter schools’ students had they instead enrolled in
the local TPS options. The outcome of interest is the average contribution to student learning gains for each
charter school per year. The measure is expressed relative to the gains the charter school students’ TPS peers
posted. Each charter school is assessed to see if it is performing signicantly stronger, signicantly weaker, or
similar (not statistically signicantly dierent) to its VCR comparison group.
25 We chose to base the school-level analysis on the two most recent growth periods in this analysis for two reasons. First, we wanted to base the result on a
contemporary picture of charter school performance. Second, the two-growth-period time frame made it possible to include the newest schools and still ensure
that performance for all the schools included the same amount of data, thereby creating a fair test for all. The school-level analysis includes only those schools
with at least 30 students to ensure a sucient sample size for the statistical stability of estimates.
Table 1.9: Charter School Student Academic Growth by Years of Charter Enrollment across Studies,
Reading and Math
Reading Math
2009 2013 2023 2009 2013 2023
1 Year in Charter -35** -35** -17** -52** -46** -35**
2 Years in Charter 6** 17** 17** 0 12** 12**
3 Years in Charter 12** 35** 31** 17** 17** 27**
4 Years in Charter n/a 41** 45** n/a 35** 39**
** Signicant at p ≤ 0.01
As with measures of charter growth presented earlier, we see persistent improvement in the charter sector in
reading and math. While students consistently take a large dip in their rst year in charter schools, the size of
the drop has decreased from the 2009 study to the 2023 study. We also see steady or improving performance
for the charter sector in the 2023 study except for a slight drop in reading from 2013 for students in their
third year attending a charter school. These results suggest improved onboarding of new students across the
charter school community.
Charter School Student Academic Growth by Location of their School
In previous studies, CREDO and others have found that charter schools were most eective for students
living in urban communities (Clark et al., 2015; Cremata et al., 2015; Cremata et al., 2013). This remains true
in this latest study. Compared to their TPS peers, urban charter school students had an additional 29 days of
growth per year in reading and 28 additional days in math, both of which were signicant. Suburban charter
school students also had stronger growth in reading (+14 days). However, rural students enrolled in charter
schools tended to have 10 days less growth in math than their TPS peers.
24
24 Analyses of charter performance by school location exclude those students attending virtual charter schools as the location of these students cannot be
determined. The impact on students attending virtual schools was discussed in a previous section.
30**
28**
14**
3
6
-3
5
-10**
MathReading
-20
0
20
40
RuralTownSuburbanUrban
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 6362
Figure 1.22: Academic Growth of Charter Schools Compared to Their Local TPS across Studies,
Reading and Math
School-Level Academic Growth by State
The prior multi-state comparisons can be repeated separately
for each state. Since each state has its policies and practices that
can impact how charter schools operate, these state-specic
school-level comparisons give us a small view of these diering
environments. The data reveals that some states have stronger
charter markets than others. As seen in the gure, New York does
not have any charter schools whose reading growth is signicantly
weaker than their VCRs.
Charter schools with stronger growth comprised 36 percent of the study schools in reading. Forty-seven
percent of charter schools had similar growth to their TPS peers. Charter schools with weaker average growth
in reading than their TPS comparison groups comprised 17 percent of the study.
In math, 36 percent of charter schools had statistically signicantly stronger growth for their students than
the TPS peers. This is compared to 25 percent of charter schools with weaker math growth than their TPS
comparisons and 39 percent with similar growth.
Figure 1.21: Academic Growth of Charter Schools Compared to Their Local TPS, Math and Reading
We can compare these distributions to earlier work. Both prior CREDO studies included local school-level
comparisons for math. The 2013 National Charter School Study presented an analysis for reading, but not
the 2009 report. A consistent pattern appears by examining the results of these analyses over time. Charter
schools have improved performance over time at both ends of the range. Figure 1.22 shows a marked rise
in the number of charter schools with better development and a decrease in those with weaker growth than
their VCR set. This trend amplies the average national charter school eect at the student level, suggesting
that improvements are widespread and not due to concentrated impacts in a subset of schools.
StrongerSimilarWeaker
2023
2013
2009
Results not available
Reading
2023
2013
2009
Math
19% 56% 25%
17% 47% 36%
37% 46% 17%
31% 40% 29%
25% 39% 36%
READER NOTE:
On the interactive website,
the reading and math gures
display the percentages
for each category of
performance.
StrongerSimilarWeaker
Math
Reading
17% 47% 36%
25% 39% 36%
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 6564
1.6 Charter School Academic Growth and Achievement
Student academic growth measures how much students advance their learning in a year, and student
achievement measures the stock of their knowledge at the end of the year. In this section, we integrate the
ndings about growth and achievement to show comprehensively the results that charter schools deliver for
their students.
Both dimensions of student performance are needed to situate charter schools in their local community
contexts and within the larger mission of academically preparing students with knowledge and skills for
future success. Importantly, considering growth and achievement simultaneously also gives us a basis for
making predictive statements about how charter schools will likely support their students in the future.
To ground this presentation, it is useful to consider four basic categories of school performance. This
construct applies to all schools, CMO-aliate charter schools, stand-alone charter schools, district schools
and others.
We can classify any school based on whether and by how much its average academic progress in a year
compares to the other options its students could select. Schools that do not advance student learning as
much as the comparison are considered “low growth.” Those that exceed the local standard are deemed “high
growth.” These dierences can be mapped on a continuum from “very low growth” to “very high growth.” We
use the growth of the local TPS alternative as the standard in this demonstration.
Looking at absolute achievement—the measure of what students know at the end of a school year—we
use the achievement scores that students get on state performance tests as a measure of achievement
and place schools along that distribution based on school-wide averages. Schools that mirror the state
average are designated “50th percentile.”
26
Schools with an average performance at lower (or higher) points
of the achievement range are situated below (above) the average—we use the 25th percentile and the 75th
percentile as additional reference points.
If we map the growth and achievement dimensions together, four groups result:
> High Growth—High Achievement: schools that exceed the growth of their local options and whose
students are above the state average in overall achievement.
> High Growth—Low Achievement: schools that exceed the growth of their local options but with overall
student achievement below the state average.
> Low Growth—High Achievement: schools whose students exceed the state average on achievement
but do not advance as much yearly as their comparisons.
> Low Growth—Low Achievement: schools with lower academic growth than their local alternatives and
whose students’ achievement is lower than the state average at the end of a school year.
26 The 50th percentile is the point value in a range of scores, in this case achievement for each state, that splits all the scores so that 50 percent are above and 50
percent are below the point.
Figure 1.23: Average Academic Growth in Charter Schools versus. Their Local TPS by State: Reading
Figure 1.24: Average Academic Growth in Charter Schools versus. Their Local TPS by State: Math
These results are encouraging, but they require a note of caution. Since the reference point in these
comparisons is the growth that equivalent students in the local TPS realize, this comparison does not reveal
if the dierence is modest or large, nor does it indicate where the dierence occurs in the range of absolute
achievement. Positive dierences at the lowest levels of achievement may not be sucient to move students
ahead fast enough to result ultimately in constructive long-term outcomes such as academic prociency or
post-secondary readiness. Similarly, a charter school may post growth results that are considered outsized for
any school but still lag behind their community schools in achievement. Simultaneous consideration of student
academic growth and achievement is the only way to get the complete picture of charter school performance.
State
AR
AZ
CA
CO
DC
FL
ID
IL
IN
LA
MA
MD
MI
MN
MO
NC
NJ
NM
NV
NY
NYC
OH
OR
PA
RI
SC
TN
TX
UT
WA
WI
State
AR
AZ
CA
CO
DC
FL
ID
IL
IN
LA
MA
MD
MI
MN
MO
NC
NJ
NM
NV
NY
NYC
OH
OR
PA
RI
SC
TN
TX
UT
WA
WI
StrongerSimilarWeaker
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 6766
Schools in the Low Growth—High Achievement quadrant can expect to drift downward in the achievement
ratings if they maintain their current pace of growth since other schools with higher growth rates will
eventually surpass them. Since student achievement in these schools is above state averages, the impact of
lower growth may not be as concerning as for students at lower levels of achievement. Roughly a tenth of
charter schools display this pattern, many of which are close to average in both growth and achievement.
Modest improvements in student learning each year could move those schools into the upper right quadrant.
The remaining charter schools, 57 percent, are situated in the lower two quadrants with achievement that
falls below the state average. This is consistent with the earlier ndings that charter schools enroll both a
larger share of lower-decile students and a smaller share of high-decile achievers. Again, their position and
prospects are distinguished by the amount of growth their students demonstrate.
The High Growth—Low Achievement quadrant displays the results for 31 percent of all charter schools.
These schools serve students with current achievement that is weaker than the average in their states. These
schools have demonstrated success with students of modest or challenged academic backgrounds. With
higher than average growth each year, their students will elevate their achievement over time. In theory,
given enough time, the students in the lower left quadrant would move up to the upper right quadrant.
The 26 percent of schools in the Low Growth—Low Achievement quadrant are of greatest concern. These
schools serve academically challenged students and produce weaker growth than their TPS comparisons.
Should the performance of these schools remain unchanged, their students will drift further behind over time,
even if all the other schools on the map remain stable. Increases in growth are within reach for these schools,
which seem possible for nearly 20 percent, which would migrate them to the lower right area. Especially
concerning at the moment are outcomes for the students attending the four percent of schools in the cell with
the lowest growth and achievement. This group represents charter schools in need of immediate attention.
Figure 1.26: Academic Growth and Achievement, Math
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.2% 2.0% 4.9% 3.8%
70th Percentile
50th Percentile
30th Percentile
1.0% 8.6% 12.0% 7.5%
4.9% 14.3% 13.8% 6.2%
7.1% 7.5% 5.3% 1.3%
Low Growth,
Low Achievement
High Growth,
Low Achievement
Using the last two years of school
performance, we mapped the
charter schools in this study onto
the structure described above.
(For reliability, we only included
schools with 30 tested students.)
We subdivided each quadrant into
four smaller groups, yielding 16
cells within the map. The results
appear in Figure 1.25 for Reading
and Figure 1.26 for Math.
Figure 1.25: Academic Growth and Achievement, Reading
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.1% 1.5% 5.8% 2.8%
70th Percentile
50th Percentile
30th Percentile
0.7% 9.1% 17.0% 6.1%
3.1% 12.3% 17.6% 6.4%
4.1% 6.8% 5.8% 1.1%
Low Growth,
Low Achievement
High Growth,
Low Achievement
As shown in Figure 1.25, summing the percentages in the top quadrants yields 43 percent of schools with
average student achievement above the state average. Currently, these students are better prepared
for future learning than half the students in their respective states. However, their growth performance
signicantly inuences their outlook for the future. Sixty-two percent of charter schools have stronger yearly
growth than the local TPS and 38 percent have weaker growth.
Schools in the High Growth—High Achievement quadrant can expect to remain in that part of the map
if their growth continues at their current pace. Roughly a third of charter schools appear in this quadrant.
At current levels of performance, these schools will likely increase their students’ achievement levels over
time. Of particular interest is the subset of High Growth—High Achievement schools that advance students
of any academic background to high levels of achievement; their operations and practices could help inform
improvements in lower-performing charter and traditional schools.
NOTE TO READERS:
The thumbnail table below presents the total proportion of
students in each major quadrant in Figure 1.25. These values
appear on the study website as a layer of the chart—the user
can see the quadrant totals and then drill down to see the
inner-quadrant values.
11.4 31.7
26.3 30.9
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 1: Charter School Performance in 31 States 6968
1.7 Gap-Closing Charter Schools
Earlier in the ndings, we reported that a signicant share of charter schools deliver gap-busting growth
for their students. We probed this nding further to see if these exceptional schools shared any common
attributes. We found hundreds of schools that satisfy dual criteria: (1) the average achievement of the school
exceeds the state average, and (2) their disadvantaged students (Black, Hispanic, in-poverty, ELL) have growth
as strong or stronger than their non-disadvantaged peers in the same school.
In reading, seven percent of schools in the study sample (526 schools) met the dual criteria for Black students
compared to their White peers. Comparing Hispanic students to White students, the percentage of charter
schools meeting the dual criteria was 13 percent (912 schools).
Further, 19 percent of charter schools (1,393 schools) met the criteria for students in poverty, compared to
their peers not in poverty. For ELL students compared to non-ELL students, 14 percent of charter schools
(1,015 schools) met the dual criteria.
In math, Black students outpaced their White peers in six percent of charter schools (456 schools). Similar
results for Hispanic students occurred in 10 percent of charter schools (731 schools). Comparing students in
poverty to their peers not in poverty, 16 percent of schools (1,142 schools) met the criteria. For ELL students,
11 percent of schools (809 schools) met the criteria. These charter schools excel at addressing achievement
gaps for their students.
Table 1.10: Charter Schools with No Learning Gaps and High Achievement
Reading Math
Number Percentage Number Percentage
Blacks equal or outperform Whites 526 7.3 456 6.3
Hispanics equal or outperform Whites 912 12.6 731 10.2
Poverty students equal or outperform
non-Poverty students
1,393 19.3 1,142 15.9
ELLs equal or outperform non-ELLs 1,015 14.1 809 11.2
The inferences for math are the same, albeit
with dierent percentages. Above-average
achievement exists in 40 percent of charter
schools, while 60 percent have achievement
lower than their state averages. Compared
to their local TPS, 55 percent of charter
schools had stronger growth, with 45
percent having weaker growth. The data
provides additional evidence that charter
schools tend to serve lower-performing
students but grow them more than is typical.
As with reading, the current and future story
depends on the quadrant in which schools
are located.
The High Growth—High Achievement quadrants contain 28 percent of charter schools, a slightly smaller
share than appeared for reading. Maintaining the current pace of growth would result in these schools
moving higher in the achievement range.
The High Growth—Low Achievement quadrant in the lower right reects schools that deliver stronger
growth to below-average achieving students. This group makes up 26 percent of all charter schools, a smaller
share than in the same reading quadrant. Their students will move higher in the achievement range if these
schools maintain or improve their growth.
Twelve percent of schools land in the Low Growth—High Achievement quadrant in the upper left, with
high average achievement but below average growth. The share of charter schools in this quadrant Is nearly
identical for reading and math. The majority of schools in this quadrant could either move down into the
lower achievement quadrant if they remain static or move to the High Growth—High Achievement area with
improved growth.
The left-hand-side lower quadrant, representing Low Growth—Low Achievement, makes up 34 percent of
charter schools. This is a signicantly larger share of schools than in the analogous reading quadrant. The
greatest worry is that 7 percent of schools are situated in the lowest performing cell. They oer the weakest
growth to students with constantly low achievement levels.
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant
in Figure 1.26. These values appear on the study
website as a layer of the chart—the user can see
the quadrant totals and then drill down to see the
inner-quadrant values.
11.8 28.2
38.8 26.4
As a Matter of Fact:
The National
Charter School
Study III 2023
As a Matter of Fact:
The National Charter
School Study III 2023
Volume 2
Charter Management
Organizations 2023
Authors
Margaret E. Raymond, Ph.D.
James L. Woodworth, Ph.D., Lead Analyst- 31 State Study
Won Fy Lee, Ph.D., Lead Analyst- CMO Study
Sally Bachofer, Ed.M.
Contributors
Meghan E. Cotter Mazzola, M.S.
William D. Snow
Tzvetelina Sabkova, M.A.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023 73
© 2023 CREDO
Center for Research on Education Outcomes
Stanford University
Stanford, CA
https://credo.stanford.edu
CREDO, the Center for Research on Education Outcomes at Stanford University, aims to improve empirical
evidence about education reform and student performance at the primary and secondary levels. CREDO
at Stanford University supports education organizations and policy makers in using reliable research and
program evaluation to assess the performance of education initiatives. CREDO’s valuable insight helps
educators and policy makers strengthen their focus on the results of innovative programs, curricula, policies
and accountability practices.
Acknowledgments
CREDO gratefully acknowledges the support of the state education agencies that contributed their data to
this partnership. Our data access partnerships form the foundation of CREDO’s work, without which studies
like this would be impossible. We strive daily to justify the condence placed in us.
The research presented here uses condential data from state departments of education. The views
expressed herein do not necessarily represent the positions or policies of the organizations noted above.
No ocial endorsement of any product, commodity, service or enterprise mentioned in this publication is
intended or should be inferred. In addition:
> This research result used data structured and maintained by the MERI-Michigan Education Data Center
(MEDC). MEDC data is modied for analysis purposes using rules governed by MEDC and is not identical
to data collected and maintained by the Michigan Department of Education (MDE) and/or Michigan’s
Center for Educational Performance and Information (CEPI). Results, information and opinions solely
represent the analysis, information and opinions of the author(s) and are not endorsed by, or reect the
views or positions of, grantors, MDE and CEPI or any employee thereof.
> Data for this report was provided by the Missouri Department of Elementary and Secondary Education.
> The conclusions of this research do not necessarily reect the opinions or ocial position of the Texas
Education Agency, the Texas Higher Education Coordinating Board, or the State of Texas.
The analysis and conclusions contained herein are exclusively those of the authors and are not endorsed by
any of CREDO’s supporting organizations, their governing boards, or the state governments, state education
departments or school districts that participated in this study. All errors are attributable to the authors.
CREDO also acknowledges the support of the Walton Family Foundation and The City Fund for this portion of
the research.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 7574
Tables of Figures
Figure 2.1. Map of States Included in the 31-State and CMO Studies ..................................80
Figure 2.2. Growth in Number of Schools by Charter School Type, 2014-15 to 2018-19 ..................81
Figure 2.3. RECAP: Annual Academic Growth of Charter School Students, Reading and Math ............83
Figure 2.4. Annual Academic Growth of Charter School Students by
Charter School Type, Reading and Math ......................................................84
Figure 2 5. Distribution of Academic Growth in SCS and CMO Schools - Reading .......................85
Figure 2.6. Distribution of Academic Growth in SCS and CMO Schools - Math ..........................85
Figure 2.7. Academic Growth Trend by Charter School Type, Reading .................................86
Figure 2.8. Academic Growth Trend by Charter School Type, Math ...................................87
Figure 2.9. Academic Growth by Students’ Years of Enrollment by Charter School Type, Reading .........88
Figure 2.10. Academic Growth by Students’ Years of Enrollment by Charter School Type, Math ........... 88
Figure 2.11. Academic Growth by Charter School Grade Span and Charter School Type, Reading .........89
Figure 2.12. Academic Growth by Charter School Grade Span and Charter School Type, Math ...........90
Figure 2.13. Academic Growth by Charter School Locale and Charter School Type, Reading .............. 91
Figure 2.14. Academic Growth by Charter School Locale and Charter School Type, Math ................91
Figure 2.15. Academic Growth by Race/Ethnicity and Charter School Type, Reading ....................95
Figure 2.16. Academic Growth by Race/Ethnicity and Charter School Type, Math .......................95
Figure 2.17. Academic Growth by Poverty Status and Charter School Type, Reading ....................96
Figure 2.18. Academic Growth by Poverty Status and Charter School Type, Math .......................97
Figure 2.19. Academic Growth by ELL Status and Charter School Type, Reading ........................98
Figure 2.20. Academic Growth by ELL Status and Charter School Type, Math ..........................98
Figure 2.21. Academic Growth by Special Education Status and Charter School Type, Reading ...........99
Figure 2.22. Academic Growth by Special Education Status and Charter School Type, Math ............. 100
Figure 2.23. Academic Growth by Race/Ethnicity & Poverty Status and Charter School Type, Reading ....101
Figure 2.24. Academic Growth by Race/Ethnicity & Poverty Status and Charter School Type, Math ......101
Figure 2.25. Academic Growth by Hispanic Students with ELL Status and Charter School Type, Reading ..102
Figure 2.26. Academic Growth by Hispanic Students with ELL Status and Charter School Type, Math ....103
Figure 2.27. Academic Growth in Persisting CMOs and New CMOs ..................................104
Figure 2.28. Academic Growth in Persisting CMO Schools vs. New CMO Schools ......................105
Figure 2.29. Annual Academic Growth in CMOs Operating in Single or Multiple States .................107
Figure 2.30. Student Academic Growth in CMO Schools by Charter School
Growth Fund Support, Reading and Math .................................................... 108
Figure 2.31. Student Academic Growth in CMO Schools, Before and After
Charter School Growth Fund Support, Reading and Math ......................................109
Table of Contents
2.1 Introduction ..............................................................................77
2.2 Methods and Data .........................................................................79
Denition of Network Schools ...............................................................79
Measure of Academic Performance ..........................................................79
Comparison Group and Analytic Model ....................................................... 79
Data ....................................................................................80
2.3 Descriptive Statistics of Students and Schools .................................................81
2.4 Analytic Findings ..........................................................................82
2.4.1. RECAP: Annual Academic Growth of Charter School Students in 31 States ...................83
2.4.2. Academic Growth by Charter School Type ...............................................83
2.4.3. Academic Growth Trend by Charter School Type .........................................86
2.4.4. Academic Growth by Students’ Years of Enrollment in Charter Schools ......................87
2.4.5. Academic Growth by Charter School Grade Span .........................................89
2.4.6. Academic Growth by Charter School Locale ..............................................90
2.4.7. Average Academic Growth of Charter School Students by State ............................92
2.4.8. Academic Growth of Charter School Student Groups .....................................93
2.4.8.1. Academic Growth by Race/Ethnicity ................................................93
2.4.8.2. Academic Growth by Poverty Status ...............................................96
2.4.8.3. Academic Growth by ELL status ...................................................97
2.4.8.4. Academic Growth by Special Education Status ....................................... 99
2.4.8.5. Academic Growth by Race/Ethnicity & Poverty Status ...............................100
2.4.8.6. Academic Growth by Hispanic & ELL Status ........................................100
2.4.9. Operational Analysis of CMOs .........................................................102
2.4.9.1. Does Charter Network Size Matter? ................................................103
2.4.9.2. Annual Academic Growth in New CMO Schools and Networks ........................103
2.4.9.2.1. Annual Academic Growth in New CMO Networks ................................104
2.4.9.2.2. Annual Academic Growth in New Charter Schools versus Continuing Schools ....... 105
2.4.9.2.3. New Charter Schools versus Persisting Schools in the Same Network .............. 107
2.4.9.3. Annual Academic Growth of CMOs Operating in Multiple States ......................107
2.4.9.4. Special Analysis: CMO Growth Accelerator Case Study—Charter School Growth Fund ...107
2.4.9.5. Special Analysis: CMOs and Turnaround Schools .................................... 110
2.4.9.6. Comparison of Average Academic Growth of Charter Schools and Their Local TPS ......113
2.4.9.7. The Relationship of Academic Growth and Achievement ............................. 115
2.4.9.8. Gap-Closing CMOs ..............................................................120
Appendix ...................................................................................122
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 7776
2.1 Introduction
Minnesota’s legislature adopted the rst charter school law in 1991, allowing for the creation of public schools
governed and managed independently from local school boards. City Academy in St. Paul opened in 1992 as
the rst charter public school in the country, serving about 35 students in its rst year of operation. In the
2021-22 school year, over 7,800 charter schools were in operation, serving over 3.7 million students. Forty-
ve states and the District of Columbia permit the operation of charter schools.
Although the majority of charter schools in the United States are single schools, many organize into
formalized entities that pool common governance, operational, nancial and programmatic resources. These
arrangements, called Charter Management Organizations (CMOs) or “networks,” aim to increase operational
eciencies and encourage strong student academic outcomes. Aspire Public Schools created the rst CMO in
the country in the early 1990s for its growing network of schools in Northern California; in the 2020-21 school
year, 432 CMOs operated a total of 2,045 CMO-aliated schools and campuses, serving 955,730 students
(White & Xu, 2022).
For the past two decades, the Center for Research on Education Outcomes (CREDO) at Stanford University
has examined charter schools in general and CMOs as a distinct subset from a nonpartisan and policy-
neutral position. The evolution of charter schools in the United States public school scene is worthy of study.
There is broad interest in their contributions to improving outcomes for the students they serve and, by
extension, to the broader public education group.
In this report, we classify charter schools into two categories.
1
Many denitions exist for Charter Management Organizations (CMO), so it is important to articulate
the one used in this study. We dene a CMO as an organization that is contracted to perform whole-
school services to at least three separate charter schools. A governing board holds the charter for
the school(s) and contracts with the CMO to provide a range of services to the school(s), including, for
example, academic programming, operations and back-oce services. The governing board is ultimately
responsible for scal health, legal compliance and academic performance of the schools it oversees.
Our designation of CMO applies to nonprot or for-prot operators, which are sometimes known as
Education Management Organizations (EMOs). For this study, we include both non-prot and for-prot
organizations in our CMO count.
In this study, we dene stand-alone charter schools (SCS) as any charter school that operates as one or,
at most, two schools.
1 CREDO’s 2017 CMO study categorizes charter schools into four types: 1. CMOs, 2. VOSs, 3. Hybrid, and 4. Independent charters (Woodworth et al., 2017). In the
current study, we break down the charter into two categories. 1. CMOs and 2. Non-CMOs that combine VOSs, Hybrid, and Independent charters.
Figure 2.32. Student Academic Growth in New CMO Schools, Before and After
Charter School Growth Fund Support, Reading and Math ......................................110
Figure 2.33. Academic Growth in Turnaround Schools: All Students vs. Continuously
Enrolled Students, Reading ................................................................ 112
Figure 2.34. Academic Growth in Turnaround Schools: All Students vs. Continuously
Enrolled Students, Math ................................................................... 112
Figure 2.35. Impact of Acquiring Turnaround Schools on Other Schools in CMO Networks ............. 113
Figure 2.36. School Comparisons of Charter School vs. Local TPS Average
Academic Growth by Charter School Type, Reading ........................................... 114
Figure 2.37. School Comparisons of Charter School vs. Local TPS Average Academic
Growth by Charter School Type, Math ....................................................... 114
Figure 2.38. Academic Growth and Achievement in CMO-aliated Charter Schools, Reading ...........116
Figure 2.39. Academic Growth and Achievement in Stand-alone Charter Schools, Reading ............. 117
Figure 2.40. Academic Growth and Achievement in CMO-aliated Charter Schools, Math .............. 119
Figure 2.41. Academic Growth and Achievement in Stand-alone Charter Schools, Math ................ 119
Tables of Tables
Table 2.1. School Characteristics by Charter Charter School Type, Matched Analytic Data ...............82
Table 2.2. Average Academic Growth of Charter School Students by Charter School Type and State ......92
Table 2.3. Student Growth in New Schools Compared to Persisting Schools in Same CMO Network .....106
Table 2.4. CMOs with Above Average Achievement Portfolios and Equitable Learning, Reading .........121
Table 2.5. CMOs with Above Average Achievement Portfolios and Equitable Learning, Math ............121
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 7978
The results provide the most current picture possible of the charter group in the nation.
Section 2 of this report describes methods and data, and Section 3 documents descriptive facts and trends
about the charter groups. The main results from the impact analyses follow in Section 4. We present ndings
disaggregated by student and school characteristics. A market analysis provides evidence of eectiveness by
organizational traits.
Because the National Charter School Study III ndings and this deeper investigation of CMOs and SCS are
intertwined, we prepared a consolidated Summary of Findings, Conclusions and Implications.
2.2 Methods and Data
Denition of Network Schools
Building upon the database created for CREDO’s 2013 and 2017 reports, we identied 368 CMOs operating
in the 28 states between the 2014–15 and 2018–19 school years. There is no national database of CMO
networks. Thus, CREDO used a variety of data sources to identify the CMOs, including data from state
educational agencies, charter school organizations and individual CMOs.
Measure of Academic Performance
For the key outcome variable, we use academic growth at the student level. Academic growth is dened as
the change in learning from one testing period to the next. For readers to understand better the results of
our analyses, academic growth is presented as marginal days of learning compared to a typical student who
obtains 180 days of educational progress in a typical school year of 180 days. Students with above average
growth are said to obtain additional days of learning in the same period and students with lower-than-
average growth are said to have fewer than 180 days of learning.
Comparison Group and Analytic Model
To create a comparison group with similar demographic and academic prole characteristics to that of
students enrolled in the charter schools, we use a combination of matching and statistical analyses to
account for the systematic dierences between students attending dierent types of schools.
In the rst stage of the analysis, we employ the virtual control record (VCR) method, which is a matching
strategy developed by CREDO (Davis & Raymond, 2012) to construct a comparison group including traditional
public school students who exhibit similar socio-demographic and academic characteristics as the students
who attend CMO-aliated and non-CMO-aliated charter schools. The VCR approach creates a “virtual twin”
for each charter student by drawing on the available records of the TPS that the students in a given charter
school would have likely attended if they were not in that charter school. We ensure that all dimensions of
observable characteristics are statistically similar between the students enrolled in the CMO-aliated charter
schools and the comparison group from the TPS.
3
In the second stage, based on the matched sample, we conduct statistical analyses to examine the eect of
CMO-aliated charter school attendance on the student’s academic growth.
3 Due to the variable distribution of students by school type and subgroup across the country, some student subgroups have low match rate in some states. Low
match rates require a degree of caution in interpreting the national pooled ndings as they may not fairly represent the learning of the student groups involved.
Funders and policy makers consider CMOs as an important lever in their aims to provide high-performing
schools. Their assumptions rarely are put to the test. Even when they are, previous research measuring
the impact of CMOs on students’ academic outcomes produced mixed results. Some of the work has been
anecdotal or small-scale, showing improved student outcomes associated with students enrolled in CMO
schools (Angrist et al., 2012; Dobbie & Fryer, 2015; Raudenbush et al., 2011). More generally, the earlier
literature shows CMO impacts on student outcomes to be small. Large variations in CMO quality across the
group have appeared in several studies (Furgeson et al., 2012; Woodworth et al., 2017).
This report presents the results of our third study of CMOs. The rst report from 2013, Charter Growth
and Replication, examined the performance patterns from the opening of schools through the period of
replication and scaling. The second report, Charter Management Organizations, released in 2017, analyzed
the dierent contributions to academic progress by CMOs and SCS (though the nomenclature for this latter
group has changed over time).
This report on CMO performance is part of a more extensive national study of charter schools prepared by
CREDO. As a Matter of Fact: The National Charter School Study III (NCSS3) examines the impact of charter school
enrollment on students’ academic growth. Due to the large scope of the research, the report is sectioned into
two volumes. The rst, Charter School Performance in 31 States 2023 (CSP31) pools all charter students together
to examine sector-wide impacts. This report, Volume 2, explores an important structural and operational
attribute of charter schools; namely, whether students attend a school that is a stand-alone charter
school (SCS, also called independent charter schools) or a member of a Charter Management Organization
(CMO), also called networks in some cases. The nationwide impact of charter schools on student academic
progress over time is, partly, a story of the rise in the number and sizes of CMOs. This report tells that story
empirically.
This study uses anonymized student-level administrative data from 28 states.
2
We treat New York City and
Washington, D.C., as separate jurisdictions to give us 28 “states” included in this study. The data window
spans the school years from 2014-15 to 2018-19, which creates four growth periods. We address the critical
questions on whether systematic dierences in the impact on student learning exist between CMO-aliated
and SCS schools.
Our outcome of interest is the change in students’ knowledge and skills from one year to the next. We use the
terms “growth,” “gains,” and “learning” interchangeably in describing the incremental progress students make
over a school year.
We probe the aggregate results to understand better how students fare in dierent charter school
environments and, in turn, how well dierent charter schools can provide high-quality education to all
their students. Since many students attending charter schools are people of color from educationally and
economically disadvantaged backgrounds, understanding the impact of CMOs and SCS on vulnerable
populations is important. Disparities in academic outcomes are well documented, for example, between
those from high socioeconomic backgrounds and those from underserved communities (Duncan & Murnane,
2016; Hanushek et al., 2019). Here, we seek not only to quantify any dierences between student groups but
also to identify cases where all students benet academically.
2 Idaho, Maryland and Ohio are included in the companion study, CSP31, but not part of the CMO analysis due to restrictions in CREDO’s data use agreements
with each state.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 8180
2.3 Descriptive Statistics of Students and Schools
Figure 2.2 shows the recent nationwide expansion of CMO-aliated and stand-alone charter schools. CMO-
aliated charter schools increased from 2,381 schools in 2014–15 to 2,793 schools in 2018–19, a 17 percent
increase over the ve years. The growth among the stand-alone charter group was about 2.6 percent during
the same period, but there were about two stand-alone charter schools for every CMO-aliated charter
school in 2014–15. The ratio decreased to about 1.8 in 2018–19.
Figure 2.2. Growth in Number of Schools by Charter School Type, 2014–15 to 2018–19
Source: NCES Core of Common Data, 2015–2019. CMO school list identied by authors.
A summary of school characteristics by CMO aliation status is included in the analytic data presented
in Table 2.1. Regardless of the group, many students enrolled in the charter schools are students of color,
and Hispanic students make up the largest minority group in both groups. Most students enrolled in
CMO-aliated charter schools and stand-alone charter schools live in poverty, with 65 and 53 percent,
respectively.
5
Another substantial dierence between the SCS and CMO-aliated charters is the share of
White students: CMO-aliated charters have 21.6 percent White students. In comparison, the share in SCS is
higher at 38.2 percent. The location dierences may contribute to the demographic dierences in the student
bodies between the groups. Approximately 58 percent of CMO-aliated charters are in urban areas, while
46percent of SCS operate in urban settings. The percentage of virtual schools is similar between the groups
at about ve percent.
5 A student in poverty is eligible for free or reduced-price lunches under the National School Lunch Program, is certied as a recipient of public assistance
support or meets state-dened criteria for poverty. Since our study design compares each charter school student to his exact-match VCR from nearby TPS,
both students face the same criteria for poverty designations. The variation in denitions across states does not aect the results.
Data
This study uses student-level administrative data from 29 states.
4
We treat New York City and Washington,
D.C., as separate jurisdictions that give us 31 “states” included in this study. The data window spans school
years from 2014–15 to 2018–19, which creates four growth periods. Under FERPA-compliant data-sharing
agreements, we use anonymized student-level administrative data; this study uses data from ve school
years, from 2014–15 to 2018–19.
Using test scores from Every Student Succeeds Act (ESSA)-mandated achievement tests administered each
spring, we calculate the dierence in a student’s scores.
Figure 2.1. Map of States Included in the 31-State and CMO Studies
4 Figure 2.1 shows the map of states included in the CSP31 and CMO analysis.
States in CSP31 & CMO Study
States in CSP31, but excluded in CMO Study
Excluded Charter States
No Charter States
Number of Schools
SCS CMO
2000
2500
3000
3500
4000
4500
5000
5500
20192018201720162015
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 8382
2.4.1. RECAP: Annual Academic Growth of Charter School Students in 31 States
As mentioned, this study parallels Charter School Performance in 31 States 2023 (CSP31). The primary nding
in CSP31 of positive annual academic gains for charter school students provides the departure point for
this study. As shown in Figure 2.3, CSP31 reported that in a year’s time, students attending charter schools
make an additional 16 days of learning in reading and six days of learning in math, compared to their TPS
comparison peers. Importantly, CSP31 shows steady increases in student academic growth over the years
of the current study and over the 15 years of CREDO’s charter school research. In this report, we probe the
overall charter school results from CSP31 by structural and operational attributes of charter schools.
Figure 2.3. RECAP: Annual Academic Growth of Charter School Students, Reading and Math
Note: The gure above originally appears as Figure 1.7 in CSP31.
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.2. Academic Growth by Charter School Type
The overall impact of attending SCS or CMO charter schools on students’ annual academic growth in
reading and math is shown in Figure 2.4. Compared to their TPS VCR peers, CMO-aliated charter school
students have statistically signicant learning gains in reading and math. Students attending stand-alone
charter schools had stronger growth in reading and similar growth in math compared to their TPS peers.
The students attending CMO schools gain the equivalent of 27 additional days of reading learning and 23
additional days of math learning per 180-day school year. Students attending the stand-alone charter also
make statistically signicant gains in reading (+10 days), but the dierence is not statistically dierent from
their peers. In order to test the dierence in the learning growth in math between the CMO and SCS, we
Table 2.1. School Characteristics by Charter Charter School Type, Matched Analytic Data
SCS CMO
Number of Schools 3,578 1,959
Number of Observations (student-level) 563,224 431,718
Student Demographic Characteristics
Percent Students in Poverty 52.9% 64.5%
Percent ELL 7.6% 10.8%
Percent Students receiving Special Education 7.7% 6.9%
Percent White 38.2% 21.6%
Percent Black 21.2% 27.7%
Percent Hispanic 33.7% 44.8%
Percent Asian/Pacic Islander 3.9% 3.6%
Percent Native American 0.4% 0.2%
Percent Multiracial 2.6% 2.0%
Locale
Urban 45.6% 58.4%
Suburban 31.5% 28.8%
Town 3.9% 1.3%
Rural 10.3% 6.9%
Virtual 5.1% 4.5%
Grade Span
Elementary 42.1% 40.5%
Middle 13.7% 21.2%
High 5.4% 7.0%
Multi-grade 38.8% 31.4%
Note: Values use data for the 2017–18 school year
2.4 Analytic Findings
This section presents the average impact of attending CMO or stand-alone charter schools on a student’s
academic growth. Academic growth is denominated in the days of learning scale, based on an average
student in a TPS who attends school for 180 days and gains 180 days of learning. In each analysis, we
compare the growth of charter school students to the learning of their TPS peers, represented by a virtual
control record as described in the Methods section. We conduct two statistical tests: one to test dierences
between charter learning and TPS learning and the second to examine dierences in results between
students in SCS and those in CMO-aliated charter schools.
0
2
4
6
8
10
12
14
16
18
MathReading
16*
6*
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 8584
Figure 2 5. Distribution of Academic Growth in SCS and CMO Schools - Reading
Figure 2.6. Distribution of Academic Growth in SCS and CMO Schools - Math
conduct a statistical test.
6
For reading and math, the analysis indicates that students attending CMO-aliated
charter schools show stronger growth than students attending stand-alone charter schools in both subjects.
Figure 2.4. Annual Academic Growth of Charter School Students by Charter School Type, Reading and Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
The results in Figure 2.4 reect the average growth based on all tested students in all schools in all the
study years for each type of charter school. It is important to note that around the average, there are wide
variations in academic growth. This is evident in Figures 2.5 and 2.6, showing the school-level distribution of
academic growth by their charter group aliation. In each charter group, the academic growth ranges from
negative 300 days to positive 300 days, suggesting the school quality varies greatly within each group. We use
the variation across students, schools or types of charter schools in the rest of our analysis.
CMOs have multiple schools that, in theory, could have distinctly dierent results. Accordingly, we
disaggregate the distributions from Figures 2.5 and 2.6 to create CMO-specic averages and ranges. The
average academic growth for each CMO is of keen interest to leaders and policy makers; Appendix A
presentsthese results.
6 We conducted a test to determine whether there is a statistical dierence between the academic growth in the two groups.
10**
-3
27**
23**
Days of Learning
SCS CMO
-5
0
5
10
15
20
25
30
MathReading
Fraction
0.00
0.05
0.10
0.15
0.20
4003002001000-100-200-300-400
Mean: 10
Fraction
0.00
0.05
0.10
0.15
0.20
4003002001000-100-200-300-400
Mean: 27
Days of Learning
CMO
SCS
Fraction
0.00
0.05
0.10
0.15
0.20
4003002001000-100-200-300-400
Mean: -3
Fraction
0.00
0.05
0.10
0.15
0.20
4003002001000-100-200-300-400
Mean: 23
CMO
SCS
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 8786
Figure 2.8. Academic Growth Trend by Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.4. Academic Growth by Students’ Years of Enrollment in Charter Schools
Figures 2.9 and 2.10 show changes in learning growth in reading and math with each additional year of
enrollment in a group-specic charter school compared to TPS peers.
7
The academic growth of a student
shows an increase in growth the longer a student is enrolled in either CMO-aliated charter or stand-alone
charter schools. This relationship exists for both reading and math. Students enrolled in stand-alone charter
schools display a comparable rate of improvement, but their growth is smaller than the students in CMO
charter schools. Students in their rst year of a CMO-aliated charter school gain 23 days more of learning
than those in the traditional public school system.
In comparison, students enrolled in stand-alone charters only make three additional days of progress. The
number of additional days of learning grows as the students’ years of enrollment in the school increase. In
their fourth year, CMO students gain 40 additional days of learning, while stand-alone charter students gain
27 more than their TPS peers. The statistical tests indicate that the dierence in the academic performance
between the two charter groups is statistically signicant in all years in the data window.
7 This analysis included only those students seen entering the charter schools from a TPS. Students already in charter schools in their rst year of the data
window were excluded.
2.4.3. Academic Growth Trend by Charter School Type
Figures 2.7 and 2.8 show the academic growth by charter group estimated in CREDO’s series of CMO reports
(Woodworth et al., 2017; Woodworth & Raymond, 2013). CMO-aliated charter schools have seen a marked
improvement in student academic growth in reading and math, adding approximately 10 additional days of
learning in each study. In reading, students’ progress in stand-alone charter schools is positive in two of the
three studies and equivalent to the learning of TPS peers in the third. For math, learning gains for students
in stand-alone charter schools lagged that of their TPS VCR peers by seven days of learning in the 2013 study.
Growth improved over time to show six days of additional learning in the 2017 study and has no signicant
dierence from growth in TPS students in the current study. Examining the graphs also reveals a widening
gap between SCS and CMO-aliated charter schools in the magnitude of student academic growth for
reading and mathematics.
Figure 2.7. Academic Growth Trend by Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Days of learning
SCS CMO
-20
-10
0
10
20
30
40
50
3rd Study2nd Study1st Study
3**
4** 0
17**
27**
10**
Days of learning
SCS CMO
-20
-10
0
10
20
30
40
50
3rd Study2nd Study1st Study
-3**
-7**
6**
17**
23**
-3
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 8988
2.4.5. Academic Growth by Charter School Grade Span
Students enrolled in all grades K through 12 in CMO-aliated charter schools show statistically signicant positive
academic growth compared to their TPS VCRs. Consistent with previous CREDO ndings, students in CMO-
aliated middle schools exhibit the most sizable academic growth at 40 additional days of learning for reading
and math. Figures 2.11 and 2.12 show adverse eects only for the students enrolled in multilevel stand-alone
charter schools. The statistical test shows that the dierence in the academic performance between the CMO and
stand-alone schools is statistically signicant for students in all grade bands except for high schools (grades 9–12).
For high schools, the dierence in the size of the academic growth between the CMO and stand-alone schools is
minimal, especially for reading. The results show no meaningful dierences between the two groups in terms of
reading and math scores.
Figure 2.11. Academic Growth by Charter School Grade Span and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.9. Academic Growth by Students’ Years of Enrollment by Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.10. Academic Growth by Students’ Years of Enrollment by Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Days of learning
SCS CMO
-20
-10
0
10
20
30
40
50
23**
3
Years in
Charter=4
Years in
Charter=3
Years in
Charter=2
Years in
Charter=1
18**
28**
27**
24**
40** 40**
Days of learning
SCS CMO
-20
-10
0
10
20
30
40
50
-12**
17**
Years in
Charter=4
Years in
Charter=3
Years in
Charter=2
Years in
Charter=1
28**
6**
38**
18**
35**
20**
20**
18**
25**
-3
32**
40**
25**
15**
Days of Learning
-10
-5
0
5
10
15
20
25
30
35
40
45
MultilevelHighMiddleElementary
CMOSCS
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 9190
Figure 2.13. Academic Growth by Charter School Locale and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.14. Academic Growth by Charter School Locale and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.12. Academic Growth by Charter School Grade Span and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.6. Academic Growth by Charter School Locale
Figures 2.13 and 2.14 reect the academic growth of charter school students by the physical locale of charter
schools aliated with CMOs and those that are stand-alone. Because virtual charter schools can enroll
students from larger geographic areas than brick-and-mortar charter schools, they appear as a separate
category in these analyses. As shown earlier, CMO charters are more likely to be in urban areas (58 percent,
vs. 46 percent for SCS). The gures demonstrate that students in CMO-aliated charters in urban areas
experience 40 more days of reading instruction and 46 more days of math instruction compared to the TPS
VCRs. While the dierence in student learning is still noticeable in suburban CMO schools, the dierence is
less dramatic. Students attending urban stand-alone charter schools make 20 additional days of learning in
reading and 12 additional days of learning in math. Urban and suburban stand-alone charters make up more
than 80 percent of the total stand-alone charter groups, and students attending these stand-alone charters
show growth on par with their TPS peers. When comparing the academic performance between the CMO-
aliated and stand-alone charters, statistical tests point to the fact that CMO-aliated charters located in
urban and suburban areas provide better results than stand-alone charters.
A troubling result is virtual schools’ dramatically sizeable negative impact on academic growth. The students
in the virtual CMO schools trail behind their TPS peers by 107 days in reading and 155 days in math. The
results for stand-alone virtual charters is similar at 77 days less learning in reading and 142 days less learning
in math than their TPS peers.
8
This nding is consistent with previous CREDO studies that found substantially
lower academic growth in virtual charter schools across the group (Woodworth et al., 2015, 2017).
8 The results for CMO and SCS mirror but do not precisely align with the ndings in CSP31 because three states are omitted from the CMO/SCS analysis.
12**
20** 20*
-25**
32**
40**
37**
1
Days of Learning
MultilevelHighMiddleElementary
CMOSCS
-30
-20
-10
0
10
20
30
40
50
20**
7** 8
1
-77**
40**
31**
8
15*
Days of Learning
Rural VirtualTownSuburbanUrban
CMOSCS
-120
-100
-80
-60
-40
-20
0
20
40
60
-107**
12**
-6
5
-12*
-142**
46**
28**
6
-1
Days of Learning
Rural VirtualTownSuburbanUrban
CMOSCS
-155**
-200
-150
-100
-50
0
50
100
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 9392
Reading Math
CMO SCS CMO SCS
Days of
Learning
Signicance
Days of
Learning
Signicance
Days of
Learning
Signicance
Days of
Learning
Signicance
NYC 62 ** 21 ** 114 ** 45 **
OR -33 -17 * -72 * -27 *
PA 14 -8 -1 -31 *
RI 134 ** 75 ** 169 ** 60 **
SC -44 ** -2 -91 ** -40 *
TN 24 ** 44 ** 32 * 46 **
TX 34 ** 2 16 ** -49 **
UT -2 -2 -8 -15
WA -71 * 63 -9 58
WI -2 18 * 10 17
Signicant
Positive Total
14 15 11 7
Signicant
Negative Total
2 1 2 6
Not Signicantly
Dierent
11 11 14 14
Note: NM has been excluded from the list due to the small number of CMO-aliated charter schools in the state.
Numbers appearing in bold signify statistically signicant dierences between CMOs and SCS.
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.8. Academic Growth of Charter School Student Groups
2.4.8.1.Academic Growth by Race/Ethnicity
Beyond the overall learning impacts of attending CMO schools or stand-alone charter schools, we are
interested in knowing if all students share the gains. We rst examine the gains for dierent race/ethnicity
groups. This is one way to track if schools are fullling their role as builders of opportunity for every enrolled
student.
As shown in Figures 2.15 and 2.16, the academic growth of students in CMOs and stand-alone charter schools
can be arranged by student groups. For each type of charter school, we compare students to their TPS peers
of the same race/ethnicity, whose performance is benchmarked on the zero line. (For instance, we assess the
educational improvement of Black CMO and SCS students relative to their Black TPS peers, likewise Hispanic
students in comparison to TPS Hispanic learners, etc.) The impact of attending dierent groups of charter
schools is nearly null on reading and 23 days weaker in math for White students compared to their White
peers in the TPS. On the other hand, Black and Hispanic students in charter schools display substantially
2.4.7. Average Academic Growth of Charter School Students by State
Table 2.2 shows the academic growth for students in CMOs and SCS in each state included in the study.
Across the states in both charter school settings, statistically signicant positive growth in reading was more
prevalent than in math. CMO-aliated charters showed statistically signicant growth in 14 states in reading
and 11 in math. For SCS, students in 15 states had signicantly positive reading gains but signicant math
gains appeared only in seven states.
The state results also revealed a few cases where charter school students had statistically signicantly
smaller learning gains than their TPS peers: CMO-based learning lagged TPS in two states in reading and two
states in math. SCS learning signicantly lagged TPS comparisons in only one state in reading but was found
in six states for math learning.
The remaining comparisons to TPS were statistically insignicant.
The bolded text in each column indicates the contrast between student academic growth in various types of
charter schools in each state. If a particular group has larger growth with a statistically signicant dierence
within the same state, it is highlighted.
Table 2.2. Average Academic Growth of Charter School Students by Charter School Type and State
Reading Math
CMO SCS CMO SCS
Days of
Learning
Signicance
Days of
Learning
Signicance
Days of
Learning
Signicance
Days of
Learning
Signicance
AR 14 -3 -5 1
AZ 24 ** 14 ** 5 -5
CA 19 ** 7 * 10 1
CO 14 16 * 34 ** 5
DC 12 -6 50 ** 6
FL 21 ** -1 13 -12 *
IL 46 ** 32 ** 66 ** 27 **
IN 7 -1 -11 -42
LA -6 10 13 1
MA 51 * 40 ** 72 * 38 **
MI 54 ** 21 ** 45 ** 6
MN 35 ** 19 * 22 5
MO 24 56 ** 34 79 *
NC 19 ** 12 ** 15 -22 *
NJ 55 ** 20 * 63 ** 14
NV 15 -2 16 -11
NY 110 ** 65 ** 124 ** 60 **
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 9594
Figure 2.15. Academic Growth by Race/Ethnicity and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.16. Academic Growth by Race/Ethnicity and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
higher growth when compared to the TPS students of the same racial/ethnic group enrolled in TPS. For
example, Black students enrolled in CMO-aliated charter schools make an additional 41 days of learning in
reading and 47 days in math relative to the Black students in TPS. For Black students attending a stand-alone
charter, the impact is smaller with 25 additional days in reading and 17 days in math. The data shows that the
dierences between the types of charter schools of 16 days for reading and 30 days for math are statistically
signicant.
The story is quite similar for Hispanic students. Hispanic students attending either the CMO-aliated or
stand-alone charters perform substantially better than their peers in TPS. However, Hispanic students
attending CMO-aliated charter schools had 22
9
days more reading gain than Hispanic students attending
SCS. The dierence in math for Hispanic students was even larger, with CMO-aliated Hispanic students
gaining 30 days more learning than those in stand-alone charter schools. Black and Hispanic students
comprise many of the student bodies in schools in urban cities across the United States. The statistical
analysis results indicate that the dierences in academic performance between the CMO and stand-alone
schools for Black and Hispanic students were statistically signicant. These ndings indicate that both stand-
alone and CMO-aliated charters, on average, may contribute to narrowing the racial achievement gaps, but
CMO-aliated charter schools give the stronger boost.
The eects on Asian/Pacic Islander students are not as strong as those on Black and Hispanic students.
However, those enrolled in charters associated with CMOs increased their learning by 17 days in reading,
while no statistically signicant impact was found for math. Meanwhile, Asian/Pacic Islanders in stand-alone
charters show similar growth to their TPS peers in reading but are lagging in math by 11 days. This dierence
between CMO charters and stand-alone charters is statistically signicant, signifying that CMOs have a more
positive impact on Asian/Pacic Islanders over stand-alone charters.
According to our analysis, the academic performance of Native American students does not improve when
they attend charter schools. In addition, multiracial students enrolled in charter schools do not perform as
well in reading as their counterparts in traditional public schools and have similar performance in math.
The estimates in this section align with the previous ndings. CREDO’s previous reports show that the
impact of charter schools on academic growth was positively signicant for Black and Hispanic students:
the 2017 CMO study reported that Black students attending CMO-aliated charter schools made, on
average, 40 additional days of learning in reading and 29 additional days of learning in math compared to
the Black students attending TPS. Similarly, Hispanic students attending CMO-aliated charter schools
made 34 additional days in reading and 29 additional days in math compared to the Hispanic students in TPS
(Woodworth et al., 2017).
9 The 22 days of learning dierence is derived by subtracting days of learning of SCS (16 days) from the days of learning of CMO (38 days).
-3 -3
1
-2
1
25**
16**
-3 -4
41**
38**
17**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
MultiracialNative
American
Asian/PIHispanicBlackWhite
-22**
-23**
-10
-18
-14**
17**
6*
-11**
-25**
47**
36**
9
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
MultiracialNative
American
Asian/PIHispanicBlackWhite
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 9796
Figure 2.18. Academic Growth by Poverty Status and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.8.3. Academic Growth by ELL Status
Public schools commit to educating students whose rst language is not English. This requires additional
expertise and resources. The learning outcomes of English-language learner (ELL) students is a continuing
interest in public education. In the context of this study, serving ELL students is also an area where a CMO’s
scale of multiple schools potentially could provide advantages over independent charter schools.
CMO-aliated charter school ELL students outperform their TPS ELL peers as well as the ELL peers in the
stand-alone schools: ELL students enrolled in CMO-aliated charter schools make 18 additional days of
learning in reading and 24 additional days of learning in math relative to the ELL TPS peers (Figures 2.19 and
2.20). The academic growth is slightly larger in reading and similar in math for non-ELL students enrolled in
the CMO-aliated charter schools. For reading, they make 27 additional days of learning while exhibiting
23 additional days of learning in math. Stand-alone charter students trail behind CMO students in academic
growth in all categories, and the dierences are statistically signicant.
2.4.8.2. Academic Growth by Poverty Status
Education is a critical factor in improving life outcomes for students in poverty. The role of charter schools in
opening future options for students has been a strong interest of policy makers, funders and educators for
much of the 30 years of charter school operations.
Our analysis indicates that students enrolled in CMO-aliated charter schools show 35 days of additional
learning in reading and 36 days in math compared to their VCR TPS peers who are also in poverty (Figures
2.17 and 2.18). Students in poverty attending stand-alone charter schools show positive learning gains in
reading and similar gains in math compared to their VCR TPS peers. The academic gains of CMO-aliated
students are signicantly larger than those attending stand-alone charters, yielding a 22-day dierence in
reading and a 32-day gap in mathematics.
For non-poverty students, the magnitude of the eect is signicant but smaller in reading. Non-poverty
students in CMOs had similar growth to their peers in math. Non-poverty students attending stand-alone
charter schools had negative growth compared to their TPS peers. CMO students not in poverty made
greater learning gains for both subjects than those students in stand-alone charter schools.
Figure 2.17. Academic Growth by Poverty Status and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
13**
35**
6**
12**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-poverty studentsStudents in Poverty
4
36**
-11**
0
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-povertyPoverty
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 9998
2.4.8.4. Academic Growth by Special Education Status
One of the largest federal funding sources for K–12 education is IDEA
10
, which serves more than 7.5 million
eligible children and students with disabilities. State spending policies allocate additional investment for
students with disabilities eligible for specialized education services. It is important to understand how the
learning of this vulnerable population fares in either type of charter school.
Approximately 11 percent of students attending charter schools receive special education services. As shown
in Figures 2.21 and 2.22, when it comes to the academic growth of special education students, CMO special
education students gain equivalent learning as their TPS counterparts in reading and math. In this case, a
“no dierent” nding reects an improvement over earlier periods. However, special education students
attending stand-alone charter schools exhibit signicantly smaller learning gains than their TPS peers, on the
order of 18 fewer days of learning in reading and 23 fewer days in math.
11
The dierence was even larger in
math at 22 days. The relative dierences between the CMO and stand-alone charter schools are statistically
signicant for special and non-special education students in reading and math.
Figure 2.21. Academic Growth by Special Education Status and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
10 See Individuals with Disabilities education Act (IDEA) at https://sites.ed.gov/idea/
11 The dierence between the learning of SPED students in CMO and SCS was 15 days. The 15 days of learning dierence is derived by subtracting days of learning
of SCS (-18 days) from the days of learning of CMO (-3 days).
Figure 2.19. Academic Growth by ELL Status and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.20. Academic Growth by ELL Status and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
-5
18**
11**
27**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-ELLELL
-6
24**
-2
23**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-ELLELL
-18**
-3
12**
29**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-SPEDSPED
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 101100
Figure 2.23. Academic Growth by Race/Ethnicity & Poverty Status and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.24. Academic Growth by Race/Ethnicity & Poverty Status and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.22. Academic Growth by Special Education Status and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.8.5. Academic Growth by Race/Ethnicity & Poverty Status
As shown in Table 2.1, students served by the CMO-aliated charter schools are predominantly low-income
minority students. In this section, we examine how student learning diers for student groups in dierent
types of charter schools by race/ethnicity and poverty status. Compared separately for CMOs and stand-
alone charter schools, we estimated the growth of each student group against its TPS peers. Although the
learning gains of attending stand-alone charter schools are smaller than that of CMO-aliated charter
schools, Black and Hispanic students, regardless of the poverty status in both settings, make statistically
signicant positive academic growth compared to their TPS VCRs in both subjects.
12
As shown in Figures 2.23 and 2.24, CMO-aliated charters appear to show more positive impacts for Black
students and Hispanic students in both subjects. In addition, the amount of growth is larger for the students
in poverty than those not in poverty. For reading, Black students in poverty enrolled in CMO-aliated
charter schools make, on average, 42 additional days of learning compared to their TPS peers, while the Black
students in poverty enrolled in stand-alone charter schools make 24 additional days of learning than their
TPS peers. While the results demonstrate a positive and robust impact for Black and Hispanic students, it is
notable that for white students in poverty underperform by 15 days in CMO and 13 days in SCS compared
to the white students in poverty in TPS schools. This research implies that CMO-aliated charters are more
successful in the academic development of children from minority backgrounds and low-income households.
12 With one exception for the Hispanic non-poverty group in math, where students attending stand-alone charters grow on par with TPS peers.
-23**
-1
-1
25**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Non-SPEDSPED
-13**
-15**
24**
17**
42** 42**
0
1
29**
36**
12**
25**
CMOSCS
Days of Learning
-40
-30
-20
-10
0
10
20
30
40
50
60
Hispanic
non-Poverty
Black
non-Poverty
White
non-Poverty
Hispanic
Poverty
Black
Poverty
White
Poverty
-32**
-34**
18**
10**
49**
43**
-18**
-17**
13**
39**
-4
17**
CMOSCS
Days of Learning
-40
-30
-20
-10
0
10
20
30
40
50
60
Hispanic
non-Poverty
Black
non-Poverty
White
non-Poverty
Hispanic
Poverty
Black
Poverty
White
Poverty
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 103102
Figure 2.26. Academic Growth by Hispanic Students with ELL Status and Charter School Type, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.9. Operational Analysis of CMOs
In this section of the report, we focus exclusively on CMOs with analyses targeted to their particular
operating attributes. A critical interest about CMO networks is how well they maintain academic gains for
their students as they grow. This question cuts both ways: Do CMO-aliated new schools demonstrate
equivalent learning gains as the rest of the CMO portfolio? In addition, does adding new schools aect the
rest of the schools in the CMO? After presenting the full sample results, we focus on exceptional cases.
2.4.9.1. Does Charter Network Size Matter?
Network size reects the number of schools for which a charter organization holds the charter and
responsibility for operations and performance. We exclude any schools with operating contract arrangements
with other educational institutions. The average number of schools managed by CMOnetworks is 6.96,
ranging from three to 73 schools.
We examined the relationship between size and student learning with several measures and proxies for
portfolio size. We found a weak correlation between portfolio size and student academic progress. At every
increment of size (and similarly of age), we saw roughly the same shares of positive, negative and equal
growth CMOs relative to their TPS counterparts, but since the larger portfolios enroll more students, the
balance shifts slightly in favor of larger scale.
Earlier CREDO work pointed out that CMOs can only replicate schools at the quality level they already
produce. That might explain how some larger CMOs have smaller gains than others. Authorizers need to
explain fully how operators with low performance receive permission to expand.
2.4.8.6. Academic Growth by Hispanic & ELL Status
In recent years about 30 percent of Hispanic students identied as English-language learners (ELL), and
Hispanic students make up three-quarters of total ELL students in the United States (De Brey et al., 2019).
Given the high proportion of Hispanic students in charter schools and the signicant share of ELL, we
examine the impacts of dierent types of charter schools on the academic success of Hispanic students with
and without ELL status.
We found a marked dierence in the learning impacts for Hispanic ELL students across the two types of
charter schools. Figures 2.25 and 2.26 show that CMO-aliated charters promote higher academic growth
for Hispanic students in both subjects, independent of their ELL status. Hispanic ELL students benet if
enrolled in CMO schools; they gain 20 extra days of learning in reading and 25 additional in math. This was
not the case if students enrolled in SCS, where their learning was on par with their TPS peers. The magnitude
of learning impacts was greater for non-ELL Hispanic students; they made an average of 42 days of learning
in reading and 39 days in math more than the TPS peers. Non-ELL Hispanic students attending stand-alone
charter schools saw an increase in reading and math learning of 19 and eight days, respectively, relative to
those in traditional public schools. The gap between ELL students attending CMO and stand-alone charter
schools was statistically signicant.
Figure 2.25. Academic Growth by Hispanic Students with ELL Status and Charter School Type, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
20**
-3
19**
42**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Hispanic non-ELLHispanic ELL
25**
-6
8**
39**
CMOSCS
Days of Learning
-30
-20
-10
0
10
20
30
40
50
60
Hispanic non-ELLHispanic ELL
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 105104
2.4.9.2.2. Annual Academic Growth in New Charter Schools versus Continuing Schools
Another way to examine growth among CMOs is to see how well they can replicate and grow new schools.
Adding new schools to a CMO portfolio entails two dierent forms of risk. One is that the school will not
edge successfully, and students will not have strong academic results. The second risk is that launching new
schools burdens the CMO and its existing schools to the point that its results suer.
We regard CMO charter schools established after 2014–15 as new entrants. Sixteen percent of CMO-aliated
schools in our data are new, pointing to signicant eorts to grow networks over the years of this study.
Persisting schools are those in operation before 2014–15.
Figure 2.28 shows that new and persisting CMO schools have a positive and statistically signicant inuence
on student academic growth on average compared to traditional public school peers.
The academic growth observed in persisting CMO schools was stronger in both subjects than in newer ones.
The impacts were 29 additional days of learning for reading and 27 for math in persisting CMO schools
versus new school learning of 21 more days in reading and 13 additional days in math. While the dierences
between persisting and new charter schools are statistically signicant for both subjects, newer schools
retain a considerable share of their CMO DNA even in their early years.
Figure 2.28. Academic Growth in Persisting CMO Schools vs. New CMO Schools
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.9.2. Annual Academic Growth in New CMO Schools and Networks
Questions of scale and eectiveness accompany growth in the CMO community. Policy makers and funders
have targeted CMO expansion to increase education options for families or shift the proportion of high-
quality seats in high-needs areas. Authorizers have faced a degree of scrutiny in their treatment of CMO
applications for new schools. The underlying assumption is that CMOs oer better odds of creating strong
schools than alternative approaches. This study has a unique vantage point to examine that idea empirically.
2.4.9.2.1. Annual Academic Growth in New CMO Networks
One facet of CMO growth is the emergence of new networks. Recall that we dene CMO networks as
operating three or more schools. Eighty CMOS, roughly 20 percent of the CMO networks in the study, opened
their third (or more) school during our study window. The increase in the number of CMOs allows us to see if
newer CMOs “come out of the gates” with student academic learning that supports backing CMOs as strong
education instruments.
We compare learning gains for students in newly emerged CMOs to those enrolled in previously existing
CMOs. As demonstrated in Figure 2.27, new and existing CMOs had a signicantly positive impact on student
academic growth compared to their TPS counterparts. New CMOs contribute less to academic gains than
older CMOs, but still aid in delivering improved education for their students.
Figure 2.27. Academic Growth in Persisting CMOs and New CMOs
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
28**
24**
20**
14**
MathReading
Days of Learning
0
5
10
15
20
25
30
New CMOPersisting CMO
29**
27**
21**
13**
MathReading
Days of Learning
0
5
10
15
20
25
30
New SchoolsPersisting Schools
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 107106
Figure 2.29. Annual Academic Growth in CMOs Operating in Single or Multiple States
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.9.4. Special Analysis: CMO Growth Accelerator Case Study—Charter School Growth Fund
Expanding the number of high-quality schools and seats in the United States has been a target of
considerable interest. One strategy is to foster the growth of successful CMOs. Several CMO growth
accelerators operate nationwide, ranging from supporters of single CMOs or networks to public and private
programs that support dozens of CMOs. Our broader study of the eectiveness of CMOs provided the
opportunity to conduct a case study of one such entity, the Charter School Growth Fund (CSGF). CSGF shared
its list of 72 funded CMOs for this analysis.
13
CSGF is a nonprot organization that makes multiyear investments in charter school networks to grow into
multi-school networks. The CSGF is known for selecting high-quality charter schools to receive expansion
funds. A related expectation is that the entire portfolio will grow its impact on students. We can test whether
student academic performance improves after a CMO receives support from CSGF.
Approximately eight percent of charter schools in this study belong to CSGF-aliated CMOs. We examined
the impact of the Charter School Growth Fund aliation of CMOs on student academic growth. The
estimates of the impact of CSGF appear in Figure 2.30.
13 It bears noting that the Charter School Growth Fund has other strands of work that focus on leaders and organizations at earlier points in their history. This
analysis does not assess the results of those endeavors.
2.4.9.2.3. New Charter Schools versus Persisting Schools in the Same Network
Pushing the inquiry about new CMO schools further, we probe the relationship between old and new schools
within individual CMOs to discern if CMOs are launching schools of equivalent quality. We took the 383 new
schools we examined earlier and related their performance to the other schools in the same portfolio. The
relative performance of the new school appears in Table 2.3.
Table 2.3. Student Growth in New Schools Compared to Persisting Schools in Same CMO Network
Percentages of CMOs (with new schools)
Compared to CMO portfolio, student learning in new school is: Reading Math
Better by 13 days or more 32 % 31 %
About the same (+/- 12 days) 23 % 13 %
Smaller by 13 days or more 45 % 56 %
Total 100 % 100 %
Almost a third of CMOs start schools that are noticeably stronger than the average of their existing schools.
Using an arbitrary cut of plus-or-minus 12 days of student learning in the rest of the CMOs schools, 23
percent of CMOs replicate the new school at about the same performance in reading and 13 percent do so in
math. The share of CMOs that started new schools with notably weaker student learning (by a shortfall of 13
days or more) was 45 percent in reading and 56 percent in math. That about half of new CMO schools dilute
the overall performance of their portfolio with weaker student gains suggests an area for future attention by
replicating CMOs.
2.4.9.3. Annual Academic Growth of CMOs Operating in Multiple States
A third facet of CMO growth concerns the geographic concentration of networks. The number of CMOs that
extend their school networks across state lines has grown since our last study. Managing a CMO portfolio
across states might provide diversication of policy and scal risks for the better long-term sustainability
of the network. On the other hand, dispersed schools might present leadership, operations and reporting
challenges that highly localized networks don’t need to face. Committing resources to buer these eects
might play out in the student learning experience.
Our test examines whether there are dierences between CMOs operating in multiple states and those
conning operations to a single state. Our denition of the CMO network used in this analysis is region
specic. Some large national CMOs include multiple regional networks that operate in a single state. For
example, KIPP New Orleans or KIPP New York City is included in our work as a separate entity that operates in
a single state.
Figure 2.29 suggests that students learning in CMOs operating in multiple states have weaker growth than
students in single-state CMOs. Single-state CMOs support additional learning of 30 extra days in reading and
29 more days in math. This compares to 19 days of additional reading in multistate CMOs and on-par learning
in math. The dierences between the two groups of CMOs are large and statistically signicant. Assuming
that new school start-up is a challenge wherever it occurs, the ndings suggest that more tightly clustered
CMOs have a better time of it.
30**
29**
19**
9
MathReading
Days of Learning
0
5
10
15
20
25
30
35
CMOs operating in multiple statesCMOs operating in one state
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 109108
A corollary question is whether the CMOs aliated with CSGF maintain their high performance levels after
selection. The nal columns of Figure 2.31 display their post-selection student learning, which covers up to
three additional years of operation, depending on when the CSGF selected the CMO in our study window.
Three new schools started by the newly funded CMOs are included. After receiving CSGF support, student
learning remained signicantly stronger than TPS, with 57 additional days of learning in reading and 80 extra
days in math. The dierences between pre- and post-CSGF support are not statistically signicant, showing
that the CMOs remain strong but do not quickly improve student learning.
Figure 2.31. Student Academic Growth in CMO Schools, Before and After Charter School Growth Fund Support,
Reading and Math
Since the major purpose of CMO growth accelerators is to launch new schools, the most important question
is how eective the new schools are. During our years of study, 43 CSGF-aliated CMOs opened 96 new
schools. We compare the newly opened schools’ performance to the existing schools in all the CMOs that
CSGF has supported. Figure 2.32 shows the comparison. Students enrolled in the new schools in the CSGF
sphere produced large gains in reading (37 additional days of learning) and math (55 extra days) compared
to their TPS peers. These new starts were dramatically stronger than the performance of the complete set of
new schools (13 additional days in reading and one more day in math) reported in the NCSS3. These results,
however, were signicantly lower than the gains students in the continuing CMO schools had, which were 62
additional days of learning in reading and 69 additional days in math compared to their TPS peers.
CMOs have student progress that outpaces the peers’ learning in TPS independent of CSGF designation.
This is consistent with CREDO’s 2013 and 2017 CMO studies. That said, the strength of CSGF student results
cannot be ignored. The advantage of attending CSGF-aliated schools is quite large for reading (an additional
61 days) and math (an additional 69 days) compared to their TPS peers. It suggests that schools funded by
CSGF provide very large academic benets to student quality. The benet is also outsized compared to the
CMOs that never received funding, despite the non-CSGF CMO schools showing meaningful positive impact
in reading (a margin of 18 days) and math (12 more days). There is a statistically signicant dierence in
academic gains between the two groups of CMOs.
Figure 2.30. Student Academic Growth in CMO Schools by Charter School Growth Fund Support, Reading and
Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Accelerator programs of all types receive the scrutiny of their results. Curiosity revolves around the relative
weights of selecting already high-performing organizations and the lift the program provides from that point
forward. Our data can test the relative contributions of these elements.
To address the question, we restrict our analysis to 29 CMOs that received support from CSGF for the rst
time between 2015–16 and 2017–18. We estimate the average academic growth before and after CSGF
aliation. As shown in Figure 2.31, students attending CSGF-supported schools exhibit much larger academic
growth prior to aliation than students in CMOs that never received funding. In both subjects, CSGF-selected
CMOs have student learning 75 days greater than their TPS peers in reading and 76 in math. The striking
dierence illustrates the CSGF’s focus on choosing strong CMOs for investment.
61**
69**
16**
9**
MathReading
Days of Learning
0
10
20
30
40
50
60
70
80
Never CSGF CMOCSGF
16**
9**
75** 76**
57**
80**
MathReading
Days of Learning
0
10
20
30
40
50
60
70
80
90
Post CSGFPrior to CSGFNever CSGF CMO
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 111110
Figures 2.33 and 2.34 compare turnaround schools’ pre- and post turn-around student academic growth.
Growth is measured at two points: the academic growth period before CMO takeover and the subsequent
period. Before the transfer, students in turnaround schools had 21 fewer days of growth in reading and 35
fewer days in math than their TPS peers. The small number of cases helps to explain why these results were
not statistically dierent from the experience of TPS students. After joining their respective CMOs, average
student performance improved compared to their TPS peers: students enrolled after the turn-around were
observed to have 21 more days of learning in reading and 38 more days in math. The fact that learning was
on par with TPS—i.e., that the dierence was not statistically signicant—can be viewed as a positive. Even if
the learning only rose to equal TPS progress, movement occurred in the right direction.
However, these comparisons at each point only tell part of the story. The change in growth for the
turnaround schools over time was statistically signicant: student learning increased by 42 days for reading
and 73 days for math.
15
These changes appear for “all students” in Figures 2.33 and 2.34.
To thoroughly test the strength of improvement, however, we need to consider whether the observed
positive academic growth stems from churn in student enrollment after the transfer. Some families may not
have supported the newly reconstituted school and moved to other public schools. Some may have read the
transfer as a signal of serious failure and left the system entirely. Alternatively, the CMO might have had a
waitlist of students wanting to enroll who joined the school after the turn-around. Any of these factors could
elevate the post-turnaround results.
As a robustness check, we redo the analysis, only including the students enrolled in the same turnaround
school before and after the transfer. These are the students most in need of turnaround eorts. In Figures
2.33 and 2.34, we contrast the academic growth of the continuously enrolled students to the full set of
enrolled students in the turnaround before and after the transfer. For students who remained enrolled (that
is, continuously enrolled) before and after the transition, we can see 42 days of learning gains in reading
between the two periods of transition and 113 days of learning gains in math. The growth we observe for “all
students” in the pre-and post-turnaround periods occurs for dierent sets of students. In the “pre” period,
the value includes students who left the school before the CMO took over; the “post” period value includes
students who were newly enrolled in the school.
The question of the spillover impact of adding a turnaround school to a CMO’s portfolio is more
straightforward. Looking only at the CMO schools that existed before the transfer, Figure 2.35 shows that
compared to their TPS peers, the academic growth for students prior to the addition of the turnaround
school is positive and statistically signicant at 39 additional days of learning in reading and 28 more
days in math. After the turnaround school joined the CMO, academic growth in the pre-existing portfolio
declined by 12 days of learning in reading but remains positive and statistically signicant at 27 more days
of learning compared to their TPS peers. In math, student academic learning increases by three days to 31
days of learning. Between the two periods, neither the change in reading gains nor the change in math gains
is statistically signicant. These results indicate that adopting turnaround schools is not injurious to the
performance of the rest of the CMO portfolio.
15 The pre-, post-turn dierence was statistically signicant at the 5 percent level in reading and the 10 percent level in math.
Figure 2.32. Student Academic Growth in New CMO Schools, Before and After Charter School Growth Fund
Support, Reading and Math
2.4.9.5. Special Analysis: CMOs and Turnaround Schools
Turnaround schools are schools that intentionally change leadership and governance in an eort to improve
their eectiveness. Since 2007, billions of dollars from the federal government were funneled through Race
to the Top and School Improvement Grant (SIG) programs to divert the learning trajectory of chronically
low-performing schools (Corbett, 2015; Legislation, Regulations, and Guidance—School Improvement Fund,
2010). The turnaround typically takes the form of restarting the schools with a new management system
(Zimmer et al., 2017). We examine the impact on student learning from a hando of school operations in a
low-performing school (either charter school or TPS) to an existing multi-school charter operator.
Two questions frame this special analysis. Where turnaround schools became part of CMOs, what is the
subsequent evidence on students’ academic growth? Additionally, what eect, if any, did the CMO’s choice to
accept a turnaround school have on the other schools in the CMO portfolio?
We are grateful to Public Impact for sharing its extensive data repository on turnaround schools across the
country. From its list, we identied 12 underperforming schools with tested students who migrated to CMOs
between the 2015–16 and 2017–18 school years.
14
Many others occurred prior to our data window, so their
transition is not visible with our available data. With the small set of schools with timely turnarounds, we
measure student performance before and after the school is moved to management by a CMO.
14 Two of the 12 schools became the third school operated by their new organization, meeting the minimum criteria for CMO inclusion in this study of three
schools.
62**
69**
37**
55**
MathReading
Days of Learning
0
10
20
30
40
50
60
70
80
90
New CSGF SchoolExisting CSGF School
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 113112
Figure 2.35. Impact of Acquiring Turnaround Schools on Other Schools in CMO Networks
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
2.4.9.6. Comparison of Average Academic Growth of Charter Schools and Their Local TPS
In this section, we examine school-level performance to assess the eectiveness of schools by charter group.
The evidence presented in the prior sections showing a positive impact on student academic performance
displays the average growth, which is the correct way to gauge the impact of each type of school.
However, this does not mean that all CMO-aliated or stand-alone charter schools perform better than
their TPS counterparts. For each type of charter school, we identify the proportion of schools that perform
better, the same and worse than their TPS comparison group. The approach mirrors prior studies and
the companion CSP31. However, the reader should be aware that the values for CMO schools and SCS will
not necessarily sum to the totals in the CSP31 report due to the exclusion of several states from this CMO
analysis.
Figure 2.36 presents the comparisons for reading. The analysis shows that 42 percent of CMO-aliated
charter schools have statistically signicantly greater reading gains than their TPS peers. In comparison, 15
percent have statistically signicantly smaller academic growth than their TPS peers. Forty-three percent
of the remaining schools advance their students in reading similarly to their TPS counterparts. When
considering the relative performance of stand-alone charter schools, the results in Figure 2.36 show that 32
percent of these schools have statistically signicantly greater gains in reading than their TPS alternatives.
We nd that 18 percent of stand-alone charter schools have reading gains that are statistically signicantly
smaller than their local TPS. The remaining 50 percent of stand-alone charter schools have no dierence in
Figure 2.33. Academic Growth in Turnaround Schools: All Students vs. Continuously Enrolled Students, Reading
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
Figure 2.34. Academic Growth in Turnaround Schools: All Students vs. Continuously Enrolled Students, Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
-21
-3
21
39**
Continuously Enrolled StudentsAll Students
Days of Learning
-30
-20
-10
0
10
20
30
40
50
Post turn-aroundPrior to turn-around
-35
38
69**
Days of Learning
-60
-40
-20
0
20
40
60
80
Post turn-aroundPrior to turn-around
-44
Continuously Enrolled StudentsAll Students
39**
27**
31**
MathReading
Days of Learning
28*
0
5
10
15
20
25
30
35
40
45
Post turn-aroundPrior to turn-around
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 115114
One nal note about these results: For CMO schools and stand-alone alike, the share of schools with stronger
learning impacts is larger, and the share of schools with less academic progress is smaller than seen before in
any of CREDO’s studies. In both spheres of charter schools, the record of performance is improved.
These results are encouraging but require a note of caution in interpretation. Since the reference point
in these comparisons is the growth that equivalent students in the local TPS realize, this comparison
does not reveal where in the range of absolute achievement the dierence occurs. Positive dierences
at the lowest levels of achievement may not be sucient to move students ahead fast enough to result
ultimately in constructive long-term outcomes such as academic prociency or post-secondary readiness.
Similarly, a charter school may post growth results that are considered outsized for any school but still lag
their community schools in achievement. Simultaneous consideration of student academic growth and
achievement is the only way to get the full picture of charter school performance.
2.4.9.7. The Relationship of Academic Growth and Achievement
Student academic growth measures how much students advance their learning in a year, and student
achievement measures the stock of their knowledge at the end of the year. In this section, we integrate the
ndings about growth and achievement to show comprehensively the results that charter schools deliver for
their students.
We need both dimensions of student performance to situate charter schools both in their local community
contexts and within the larger K-12 mission of preparing students with knowledge and skills for future
success. Importantly, considering growth and achievement simultaneously also gives us a basis for making
predictive statements about how charter schools are likely to support their students in the future.
To ground this presentation, it is useful to consider four basic categories of school performance. This
construct applies to all schools: CMO-aliate charter schools, stand-alone charter schools, district schools
and others.
We can classify any school based on whether and by how much its average academic progress in a year
compares to the other TPS options. Schools that do not advance student learning as much as the comparison
are considered “low growth.” Those that exceed the local standard are deemed “high growth.” These
dierences can be mapped on a continuum from “very low growth” to “very high growth.” We use the growth
of the local TPS alternative as the standard in this demonstration.
Looking at absolute achievement—the measure of what students know at the end of a school year—we use
the achievement scores that students get on state performance tests as a measure of achievement and
place schools along that distribution based on school-wide averages. Schools that mirror the state average
are designated “50th percentile.”
16
Schools with an average performance at lower (or higher) points of the
achievement range are situated below (above) the average; we use the 25th percentile and the 75th percentile
as additional reference points.
17
16 The 50th percentile is the point value in a range of scores—in this case, achievement for each state—that splits all the scores so that 50 percent are above and
50 percent are below the point.
17 The measures of achievement show student learning after enrollment in a charter school.
reading gains compared to local TPS. The graphs make clear that for reading, the CMO advantage compared
to stand-alone charter schools applies top to bottom: larger shares of CMO schools are stronger performing
than their local TPS and smaller shares are on par or posting smaller gains.
Figure 2.36. School Comparisons of Charter School vs. Local TPS Average Academic Growth by Charter School
Type, Reading
In terms of math results, the dierence between CMO schools and independent charter schools is much
more notable. As Figure 2.37 displays, 44 percent of CMO schools have statistically signicantly larger
academic gains in math, 22 percent have statistically signicantly smaller learning gains and 34 percent are
not markedly dierent from the TPS alternatives.
The results for stand-alone charter schools in math run parallel to their reading results. Figure 2.37 shows
that 31 percent of stand-alone charter schools have statistically better gains than TPS. The proportion with
statistically signicantly smaller math gains than TPS is 27 percent in math. Of the rest of the stand-alone
charter schools, 42 percent demonstrate equivalent academic gains as their local TPS.
Figure 2.37. School Comparisons of Charter School vs. Local TPS Average Academic Growth by Charter School
Type, Math
Same BetterWorse
18% 50% 32%
15% 42% 43%
CMO
SCS
Same BetterWorse
27% 42% 31%
22% 34% 44%
CMO
SCS
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 117116
Figure 2.39. Academic Growth and Achievement in Stand-alone Charter Schools, Reading
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.1% 1.6% 6.3% 2.7%
70th Percentile
50th Percentile
30th Percentile
0.9% 11.0% 17.0% 5.0%
3.3% 13.5% 16.5% 5.7%
4.3% 6.2% 4.9% 1.0%
Low Growth,
Low Achievement
High Growth,
Low Achievement
The stronger growth of CMO-aliated schools nds a parallel in achievement patterns. As illustrated
in Figure 2.38, 68 percent of CMO-aliated charter schools have average reading growth above their
comparison groups (sum of two right columns) and 32 percent below. For reading achievement, 44 percent of
charter schools have average student achievement above their state’s average (sum of top two rows) and 56
percent below. In Figure 2.39, the chart shows 59 percent of students enrolled in stand-alone charter schools
show stronger growth than their TPS comparisons in reading, with 41 percent of schools having weaker
growth. Fifty-ve percent of students in SCS had average student achievement below their state’s average,
and 45 percent of charter schools had an above-average performance.
Schools in the High Growth—High Achievement quadrant of Figure 2.38 can expect to remain in that part
of the map if their reading growth continues at the current pace. Thirty-ve percent of CMO-aliated charter
schools and 31 percent of stand-alone charters appear in this quadrant. There is no meaningful dierence
between the two types of charters in creating outstanding academic results. At current levels of performance,
these schools will likely increase their students’ achievement levels over time. The gap-busting schools and
networks reside in this quadrant. Of particular interest is the subset of High Growth—High Achievement
schools that advance students of any academic background to high levels of achievement; their operations
and practices could help inform improvements in lower-performing charter and traditional schools.
If we map the growth and achievement dimensions together, four groups result:
High Growth—High Achievement: schools with larger growth than their local alternative and whose
students are above the state average in overall achievement
High Growth—Low Achievement: schools that exceed the growth of their local options but with overall
student achievement below the state average
Low Growth—High Achievement: schools whose students exceed the state average on achievement
but do not advance as much yearly as their comparisons
Low Growth—Low Achievement: schools with lower academic growth than their local alternatives and
whose students’ achievement is lower than the state average at the end of a school year
We mapped the charter schools in this
study onto the structure described
above using the last two years of
school. (For reliability, we included only
schools with 30 tested students.) We
subdivided each quadrant into four
smaller groups, yielding 16 cells within
the map. The results appear in Figures
2.38 and 2.39 for reading and Figures
2.40 and 2.41 for math.
Figure 2.38. Academic Growth and Achievement in CMO-aliated Charter Schools, Reading
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.2% 1.5% 5.4% 3.1%
70th Percentile
50th Percentile
30th Percentile
0.4% 6.7% 18.5% 7.9%
2.8% 11.0% 19.6% 7.6%
3.4% 6.2% 5.1% 0.5%
Low Growth,
Low Achievement
High Growth,
Low Achievement
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in Figure
2.38. These values appear on the study website as a
layer of the chart—the user can see the quadrant totals
and then drill down to see the inner-quadrant values.
8.8 34.9
23.4 32.8
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in Figure
2.39. These values appear on the study website as a
layer of the chart—the user can see the quadrant totals
and then drill down to see the inner-quadrant values.
13.6 31.0
27. 3 28.1
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 119118
Figure 2.40. Academic Growth and Achievement in CMO-aliated Charter Schools, Math
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.1% 2.2% 5.2% 5.7%
70th Percentile
50th Percentile
30th Percentile
0.8% 7.3% 12 .1% 10.8%
4.0% 12.0% 14.2% 7.9%
6.1% 6.3% 4.5% 0.8%
Low Growth,
Low Achievement
High Growth,
Low Achievement
Figure 2.41. Academic Growth and Achievement in Stand-alone Charter Schools, Math
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.2% 2.0% 4.8% 2.8%
70th Percentile
50th Percentile
30th Percentile
1.1% 9.9% 12.5% 5.9%
5.7% 16.4% 13.7% 5.1%
6.8% 7.2% 5.0% 1.2%
Low Growth,
Low Achievement
High Growth,
Low Achievement
The inferences for math are the same as for reading, albeit with dierent percentages (Figures 2.40 and 2.41).
In 61 percent of CMO schools, their growth outpaced their TPS comparisons, with 39 percent having weaker
growth. Forty-four percent of CMO-aliated charter schools had average student achievement larger than
their state’s average. Fifty-six percent of CMO charter schools had a below-average performance.
Schools in the Low Growth–High Achievement quadrant can expect to drift downward in the achievement
ratings if they maintain their current pace of growth since other schools with higher growth rates will
eventually surpass them. Nine percent of CMO charter schools and 14 percent of stand-alone charter schools
sit in this quadrant. Since student achievement in these schools is above state averages, the impact of lower
growth may not be as concerning as for students at lower levels of achievement. Since many of the schools
in this quadrant are close to average in both growth and achievement, modest improvements in student
learning each year could move those schools into the upper right quadrant.
The remaining charter schools are situated in the lower two quadrants with achievement below the state
average. For CMO charters, this amounts to 56 percent; for stand-alone charters, 55 percent are below
the state average. This is consistent with the earlier ndings that charter schools enroll both a larger share
of lower-decile students and a smaller share of high-decile achievers Their position and prospects are
distinguished by their students’ growth.
The High Growth—Low Achievement quadrant displays the results for 33 percent of all CMO charter
schools and 28 percent of stand-alone charter schools. Though these schools serve students with current
achievement weaker than the average in their states, they have demonstrated success with students of
modest or challenged academic backgrounds. With higher-than-average yearly growth, their students will
elevate their achievement over time. In theory, given enough time, the students in the lower right quadrant
would move up to the upper right quadrant.
The share of schools in the Low
Growth—Low Achievement quadrant
is of greatest concern. These schools
serve academically challenged students
and produce weaker growth than their
TPS comparisons. The proportions of
schools in this quadrant are similar for
the two types of charter schools. For
CMO charter schools, the performance
of 23 percent of schools maps to this
quadrant. For stand-alone charter
schools, the share is 27 percent. Given
the substantial dierence in average growth in reading between CMO-aliated and stand-alone charter
schools, it is surprising to see the proportions in this quadrant be so similar. Should the performance of these
schools remain unchanged, their students will drift further behind over time, even if all the other schools on
the map remain stable. Increases in growth are within reach for many of these schools, which would migrate
them to the lower right area. Especially concerning at the moment are outcomes for the students attending
schools in the cell with the lowest growth and achievement. This group represents charter schools in need of
immediate attention.
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in Figure
2.40. These values appear on the study website as a
layer of the chart—the user can see the quadrant totals
and then drill down to see the inner-quadrant values.
10.4 33.8
28.4 27.4
NOTE TO READERS:
The thumbnail table below presents the total
proportion of students in each major quadrant in Figure
2.41. These values appear on the study website as a
layer of the chart—the user can see the quadrant totals
and then drill down to see the inner-quadrant values.
13.2 26.0
36.1 25.0
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 121120
How do we know these results are not simply the fortunate alignment of events at these individual schools? Is
there evidence that the practice can be systematic? We looked at CMOs’ impacts on growth for minorities and
students in poverty compared to their White student counterparts.
Table 2.4. CMOs with Above Average Achievement Portfolios and Equitable Learning, Reading
CMOs where: Number Percentage
Black outperforms White 65 18%
Hispanic outperforms White 95 26%
Lunch outperforms no-lunch 122 33%
ELL outperforms non-ELL 128 35%
Total 368
q
Note: q Percentages do not sum to 100% since a CMO could be included in multiple rows.
Table 2.5. CMOs with Above Average Achievement Portfolios and Equitable Learning, Math
CMOs where: Number Percentage
Black outperforms White 51 14%
Hispanic outperforms White 72 20%
Lunch outperforms no-lunch 97 26%
ELL outperforms non-ELL 115 31%
Total 368
q
Note: q Percentages do not sum to 100% since a CMO could be included in multiple rows.
Tables 2.4 and 2.5 present the numbers of CMOs with student achievement that exceeded the state average
(“High Achievement”) and in whose schools Black and Hispanic students had learning gains on par or better
than the White students. The tables also present the number of CMOs with students in poverty making larger
gains than their non-poverty peers or English-language learners who outpace their non-ELL classmates.
18
The importance of these ndings is obvious: when dozens of schools and networks can prevent dierences in
learning across student groups while also delivering learning above their state averages, they are forestalling
and even reversing the achievement gap that has persisted for decades in our country. The discovery that this
is prevalent in numerous CMOs suggests that these entities have found a way to implement and disseminate
this transformative knowledge on a large scale.
18 CMOs that are included in the results of Tables 2.4 and 2.5 are agged in Appendix A.
Regarding math performance in stand-alone charter schools, about 51 percent of schools show stronger
growth than their TPS comparisons, with 49 percent having weaker growth. Thirty-nine percent of stand-
alone charter schools had average student achievement above their state’s average; 61 percent of stand-
alone charter schools had average achievement below their state averages. The data indicates that, similar to
the CMO charters, stand-alone charters tend to serve lower-performing students but grow them more than
their TPS peers.
The High Growth—High Achievement quadrants contain 34 percent of CMO charter schools, a slightly
smaller share than appeared for reading. Among stand-alone charters, the share was 26 percent. Maintaining
the current pace of growth would result in these schools moving higher in the achievement range.
The High Growth—Low Achievement quadrant in the lower right reects schools that deliver stronger
growth to below average achieving students. This quadrant contains 26 percent of all CMO charter schools
and 25 percent of all stand-alone charter schools. Both proportions are smaller than occurred in the same
reading quadrant. Their students will move higher in the achievement range if these schools maintain or
improve their growth.
Ten percent of CMO-aliated charter schools land in the Low Growth—High Achievement quadrant in the
upper left, schools with high average achievement but below average growth. Thirteen percent of stand-
alone charter schools appear in the same quadrant. The majority of schools in this quadrant could either
move down into the lower achievement quadrant if they remain static or move to the High Growth—High
Achievement area with improved growth.
The left-hand-side lower quadrant, representing Low Growth—Low Achievement, contains 28 percent
of CMO charter schools and 36 percent of stand-alone charter schools. The CMO-aliated percentage is
substantially smaller than for stand-alone charter schools. This is a noticeably larger share of CMO and stand-
alone schools than in the analogous quadrant for reading. The greatest worry is the schools situated in the
lowest performing cell. They oer the weakest growth to students with constantly low achievement levels.
2.4.9.8. Gap-Closing CMOs
In the companion report, CSP31, we highlight the dramatic performance of thousands of charter schools with
outstanding progress for minority and poverty students. These “gap-busting schools” show that disparate
student outcomes are not a foregone conclusion: people and resources can be organized to eliminate these
disparities. The fact that thousands of schools have done so removes any doubt.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 123122
Appendix
Appendix A. Average Annual Academic Growth of CMOs and Networks, Reading and Math
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
A+ Charter Schools, Inc. -0.063 ** -0.040
Academics Plus 0.040 0.023
Academy of Academic Excellence -0.347 ** -0.485 **
Academy of Mathematics and Science, Inc. 0.058 * 0.085 **
3
Academy of Tucson 0.059 **
3
-0.052
ACCEL Schools -0.006 -0.014
Accelerated Intermediate Academy 0.129 ** 0.205 **
Accelerated School, The 0.062
3
-0.011
ACE public charter schools 0.001
3
0.102 **
3
Acero schools -0.025 0.031
Achievement First NY 0.114 **
3
0.253 **
3
Achievement First RI 0.189
3
0.270
3
Albert Einstein Academies -0.101 ** -0.056
Algiers Charter School Assoc. -0.145 ** -0.054 *
Alliance for College-Ready Public Schools 0.185 **
3
0.167 **
3
Alpha Public Schools 0.055 **
3
0.108 **
3
Alta Public Schools -0.181 ** -0.178 **
Altus Institute Network of Charter Schools -0.044 -0.032
America CAN! -0.229 ** 0.036
3
American Indian Public Charter School 0.124 **
3
0.189 **
3
American Leadership Academy Inc. -0.030 -0.001
American Paradigm 0.013
3
0.038
3
American Preparatory schools 0.040 **
3
0.060 **
3
American Promise Schools (now known as
Promise Schools)
0.041
3
0.014
3
American Quality Schools 0.011
3
-0.049
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
AmeriSchools (Ideabanc, Inc.) (The Charter
Foundation, Inc.)
0.085 **
3
0.112 **
3
Amethod Public Schools 0.050
3
0.103 **
3
Archimedean Academy 0.157 **
3
0.242 **
3
Arizona Agribusiness & Equine Center 0.102 **
3
0.084 **
Arizona Community Development Corporation -0.062 ** 0.016
Arlington Classics Academy 0.039 ** -0.032
Arrow Academy, Inc. 0.071
3
0.035
Ascend Learning 0.077 **
3
0.209 **
3
Ascent Academies of Utah -0.017 -0.026
Aspira Inc. of Illinois -0.078 ** -0.104 **
Aspira Inc. of Pennsylvania -0.074 *
3
-0.027
3
ASPIRA of Florida, Inc. -0.028
3
-0.039
3
Aspire Public Schools 0.052 **
3
0.073 **
3
ASU Preparatory Academy 0.047
3
0.135 **
3
Athlos Charter Schools 0.031
3
-0.053
BakerRipley-TX -0.006
3
0.048
Ball Charter Schools 0.073 *
3
0.111
3
BASIS Schools, Inc. 0.104 ** 0.094 **
Bay Haven Charter Academy Inc. -0.011 0.063
3
Beginning with Children Foundation 0.007
3
-0.019
Ben Gamla Charter School Foundation 0.073 **
3
0.034
3
Benjamin Franklin Charter Schools 0.016 * 0.022
Betty Shabazz International Charter School 0.092 ** -
Blackstone Valley Prep Mayoral Academy 0.171 ** 0.269 **
3
Blueprint Education -0.160 ** -0.193 **
Bob Hope School 0.118 **
3
0.217 **
3
Brazos School for Inquiry & Creativity (BSIC) -
Democratic Schools Research Inc.
-0.145 -0.137 **
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 125124
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
Bright Star Schools 0.083
3
0.085 **
3
Brighter Choice Charter Schools 0.740 ** -
Brooke Charter Schools 0.096
3
0.126
Burnham Wood Charter Schools 0.057 * 0.078 *
3
CAFA, Inc. -0.016
3
-0.066
California Montessori Project -0.025 * -0.019
Calvin Nelms Charter Schools -0.019 0.051
3
Camden's Charter School Network 0.008
3
-0.031
Camino Nuevo 0.069 **
3
0.078 **
3
Capital City Public Charter School 0.002
3
0.034
3
Capstone Education Group 0.023
3
0.055
3
Career Success School District -0.149 -0.106
Carl C. Icahn Charter Schools 0.109 ** 0.256 **
Carmen Schools of Science & Technology -0.055 * 0.056 **
Carpe Diem (IN) -0.123 ** -0.315 **
Catalyst Schools -0.002 0.015
Celerity Educational Group 0.046
3
0.095 **
3
Celerity Schools Louisiana, Inc. 0.039 ** 0.294 **
Center City Public Charter Schools 0.027
3
0.052
3
Center for Academic Success 0.004
3
0.046
3
Cesar Chavez Academy -0.181 ** -0.100 **
Cesar Chavez PCS for Public Policy 0.005
3
-0.049
Champion Schools 0.120 **
3
0.074 **
3
Championship Academy of Distinction -0.058 **
3
-0.119
Chandler Park Academy -0.018 0.007
Chicago International Charter Schools -0.044 -0.010
Choice Foundation 0.083 -0.036
Christel House Academy 0.028
3
0.049
Citizens of the World 0.092 **
3
0.116 **
3
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
City Center for Collaborative Learning 0.002 -0.049
City University-TN -0.061 0.159 **
Civitas Schools -0.059 ** 0.059 **
Classical Academies (Colorado) 0.024 0.047
Classical Charter Schools 0.136 ** 0.291 **
3
College Achieve Public Schools -0.091 -0.084
Collegiate Academies -0.138 ** 0.113 **
Colorado Early College 0.045 * 0.099 **
Community Day 0.230 **
3
0.265 **
3
Community School for Apprenticeship Learning -0.068 -0.028
Compass Charter Schools -0.124 ** -0.291 **
Concept Schools 0.047
3
0.075 *
3
Conuence Academies -0.054 -0.047
Coral Education Corporation -0.013 0.034
CORE Butte -0.092 -0.078
Cornerstone Charter Schools 0.081 * 0.097 *
3
Crescent City Schools 0.071 **
3
0.050
3
Cumberland Academy Schools -0.031 ** -0.032
3
Da Vinci Charter Schools 0.062 0.165 **
3
Daisy Education Corporation (DEC) (now
Sonoran Schools)
0.076 **
3
0.100 **
3
DC Prep Charter Schools 0.073 ** 0.228 **
Delta Charter Schools -0.133 ** -0.040
3
Democracy Prep Public Schools 0.045
3
0.147 **
3
Denver School of Science and Technology Public
Schools
0.083 *
3
0.170 **
Distinctive Schools -0.008 -0.016
Doral Academy 0.104 **
3
0.122 **
3
Downtown College Prep Charter Schools -0.165 ** -0.189 **
E.L. Haynes Public Charter Schools -0.019 0.058
3
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 127126
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
e-Institute -0.119 **
3
0.008
ECI Academy -0.027 -0.072
Edkey Schools -0.058 ** -0.071 **
Education for Change 0.099
3
0.172 **
3
Einstein Schools (New Orleans) 0.041 *
3
-0.022
Energized for Excellence 0.114 *
3
0.357 **
3
Environmental Charter Schools 0.084
3
0.079
3
Envision Schools 0.115 **
3
0.034
3
Equitas Academy 0.063
3
0.156 **
3
Espiritu Community Development Corp. 0.011 -0.076
eStem Public Charter Schools 0.107 **
3
0.059
3
Evolution Academy -0.430 ** -0.348 **
Excel Academy (TX) -0.335 ** -0.452 **
Excellence Community Schools Inc. 0.020 ** 0.178 **
3
Explore Schools Inc. 0.037 *
3
0.136 **
3
Faith Family Academy Charters -0.185 -0.172 **
Fenton Charter Public Schools 0.062 *
3
0.116 **
3
FirstLine Schools (formerly Middle School
Advocates, Inc.)
0.033 * 0.109 **
3
Five Keys Public Schools -0.055 ** -
Family Life Academy Charter Schools (FLACS) 0.028 0.110 **
3
Foundation for Behavioral Resources 0.012
3
-0.007
Founders Classical Academy 0.023 ** -0.046 *
Franklin Academies 0.016
3
0.028
3
Freedom Preparatory Academy 0.065
3
0.154 **
Freire Schools 0.185 **
3
0.282 **
3
Friendship Schools -0.001 0.134 **
Frontier Schools 0.049
3
0.104 **
3
Gateway Community Charters -0.020 -0.043
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
GEO Foundation 0.023
3
0.058 *
3
Gestalt Community Schools 0.008 -0.016
Global Education Collaborative 0.056 ** 0.076
3
Golden Rule Charter Schools 0.088 ** 0.165 **
Goodwill Education Initiatives (Goodwill Excel
Center)
-0.132 -0.074
Great Hearts Academies 0.029 ** 0.043 **
Great Oaks Foundation 0.062
3
0.123 *
3
Green Apple School Management, LLC -0.009
3
0.048
3
Green Dot Public Schools 0.012
3
0.011
3
Guadalupe Centers -0.028 -0.029
Gulf Coast Council of Raza 0.044 ** -0.236 **
Haas Hall Academy 0.209 **
3
0.346 **
3
Harmony Schools (Cosmos Foundation, Inc.) 0.061 **
3
0.126 **
3
Harvest Network of Schools 0.065 **
3
0.019
3
Harvest Power Community Development -0.042 **
3
-0.013
3
Hebrew Public 0.077
3
0.059 **
3
Heritage Academy 0.106 **
3
0.160
3
Heritage Academy AZ 0.008 -0.167 **
Hiawatha Academies 0.014
3
0.052
3
Hickman Community Charter District 0.037 *
3
0.052
3
High Tech High CA -0.012 -0.022
Hogan Preparatory Schools -0.020 -0.037
Honors Academy -0.091 ** -
Hope Online -0.116 **
3
-0.077
3
Houston Gateway Academy 0.150 **
3
0.364 **
3
Humanities and Sciences Academy of the
United States, Inc.
0.085 *
3
0.047
I CAN Schools -0.160 ** -0.368 **
IDEA Public Schools 0.145 **
3
0.130 **
3
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 129128
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
iLEAD Charter Schools -0.047
3
-0.095 **
iLearn Schools -0.034 **
3
-0.003
Imagine Schools -0.016
3
0.001
Inuence 1 Foundation -0.009 -0.097 **
Ingenium Schools -0.112 ** -0.054
Inner City Education Foundation (ICEF) -0.004
3
-0.016
3
Innovative Education Management -0.079 * -0.101 **
Innovative Teaching Solutions -0.057 -0.004
Inspire charter schools -0.146 ** -0.245 **
InspireNOLA Charter Schools -0.001 0.166 **
3
IntelliSchool Charter High Schools -0.177 * 0.183 **
International Leadership of Texas (ILT) 0.005 -0.033
iSchool High -0.030 -0.318 **
James Irwin Charter Schools (CO) -0.016 0.055
Jeerson Chamber Foundation Academy (JCFA) -0.116 ** -
John Adams Academies 0.002 -0.010
John H. Wood Jr. Public Charter District -0.116 -0.371 **
Jubilee Academic Center, Inc. -0.068 ** -0.156 **
K12 curriculum only (Virtual) -0.067 -0.119 *
K12, Inc. -0.138 ** -0.201 **
Kaleidoscope Charter Schools 0.055
3
0.105
Kid’s Community College -0.067 * -0.163 **
King-Chavez -0.022
3
0.010
3
Kingman Academy of Learning -0.010 0.024 *
Kingsburg Elementary Charter School District 0.043 **
3
-0.011 **
3
KIPP Austin 0.110 **
3
0.044
3
KIPP Bay Area 0.122 **
3
0.137 **
3
KIPP Chicago 0.132 **
3
0.203 **
3
KIPP Colorado 0.061
3
0.084
3
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
KIPP Dallas-Fort Worth -0.005
3
0.029
3
KIPP DC 0.063 **
3
0.144 **
KIPP Delta 0.003 -0.038
KIPP Eastern North Carolina 0.026 *
3
0.005
3
KIPP Houston 0.095 **
3
0.050 *
KIPP Memphis -0.056 * -0.036
KIPP Nashville 0.143 **
3
0.321 **
3
KIPP National 0.083 **
3
0.111 **
3
KIPP New Orleans 0.074 **
3
0.051 **
KIPP New York City 0.124 **
3
0.238 **
3
KIPP Philadelphia 0.023
3
0.064
KIPP San Antonio 0.037 *
3
-0.016
3
KIPP SoCal 0.110 **
3
0.151 **
3
KIPP St. Louis 0.092 ** 0.180 **
La Amistad Love & Learning Academy
(L Lowell Byrd Memorial Education and
Community Dev. Corp.)
-0.040 ** -
LEAD Public Schools 0.055 *
3
0.092 *
3
Leadership Public Schools 0.309 **
3
0.313 **
3
Leading Edge Academy -0.014 0.014
3
Learn Charter School 0.094 **
3
0.122 **
3
Legacy Preparatory -0.104 * -0.199 **
Legacy Traditional Schools 0.095 **
3
0.092 **
3
Leman Academy of Excellence, Inc. 0.069 ** 0.019
3
Life Schools 0.013
3
-0.051
3
Life Skills Centers -0.293 * -0.147
Lighthouse Academies 0.016
3
0.036
3
Lighthouse Academy (Michigan) -0.308 ** -
Lincoln-Marti management services, LLC 0.148 **
3
0.259 *
3
Lionsgate Academy 0.044 ** -0.076 **
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 131130
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
LISA Academies 0.094 ** 0.131 **
Magnolia Science Academy (Magnolia
Foundation)
0.032 *
3
0.042
3
Manara Academy, Inc. 0.047
3
-0.169 **
Mastery Charter Schools 0.080 **
3
0.088 **
3
Mastery Learning Institute (Arthur Academy) 0.184 **
3
0.095
3
Match Charter Public School 0.093 ** 0.214 **
Matchbox Learning -0.072
3
-0.114 **
Mater Academy of Nevada, Inc. 0.215 ** 0.243 **
Mater Academy, Inc. 0.055 **
3
0.062
3
Mavericks in Education, LLC -0.122 **
3
-0.244 **
McKeel Academies -0.004
3
0.024
3
Memphis Business Academy -0.025 -0.038
Memphis Scholars -0.090 ** -0.129 *
Milwaukee College Prep 0.189 ** 0.184 **
Minnesota Internship Center -0.213 -
Minnesota Transition Schools (MTS) -0.038
3
-0.009
3
Muskegon Heights Public School Academy -0.202 ** -0.218 **
MYcroSchool -0.185 ** -0.424 **
National Heritage Academies 0.079 **
3
0.120 **
3
National University Academy 0.054 ** -0.013 **
3
Natomas Pacic Pathways Prep 0.027
3
0.015
3
New America Schools -0.269 ** -0.181 **
New Beginnings Schools Foundation -0.071 -0.022
New Orleans College Prep Academies -0.122 ** -0.086
New Paradigm for Education 0.199 **
3
0.187 **
New Tech Network -0.020 0.021
New Technology Foundation -0.017 0.036
New Visions for Public School 0.226 **
3
0.021
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
Newman International Academy -0.002 -0.067
Noble Network of Charter Schools 0.148 **
3
0.291 **
3
North Texas Collegiate Academy -0.039 -0.074
NorthStar Academies -0.527 ** -
Nova Academy 0.087 **
3
0.052 *
3
Oasis Charter Schools -0.017
3
0.000
3
Ombudsman Educational Services, Ltd., a
subsidiary of Educational Services of America
-0.397 ** -0.360 **
Open Sky Education -0.018 *
3
-0.006
3
Opportunities for Learning -0.108 ** -0.165 **
Options for Youth -0.119 ** -0.184 **
Orenda Education-TX 0.014 -0.019
Oxford Preparatory Academies 0.134 **
3
0.212 **
3
Pacic Charter Institute -0.098 ** -0.100 **
Panola Schools -0.125 ** -0.337 **
Para Los Ninos -0.061
3
-0.090
Parnassus Preparatory 0.028 0.067 *
Partnerships for Uplifting Communities (PUC) 0.041
3
0.090 **
3
Performance Academies (formerly EdVantages
Academies)
-0.056
3
-0.074
Perspectives Charter Schools 0.031 *
3
-0.021
Phalen Leadership Academy - IN Inc. 0.006 -0.003
Pinecrest Academy 0.072 **
3
0.097 **
3
Pineywoods Community Academy 0.027 **
3
0.020
3
Pinnacle Charter Academies (SC) -0.206 -0.125
Pinnacle Charter School (CO) -0.076 ** 0.001
3
Pinnacle Education, Inc. -0.343 ** -0.358 **
Pivot Charter School (Roads Education
Organization)
-0.215 ** -0.312 **
Plato Academy Schools -0.021
3
-0.007
3
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 133132
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
Pointe Schools -0.070 ** -0.148 **
Pontiac Academy for Excellence -0.065 * -0.047
Por Vida, Inc. -0.172 ** -0.114
Portable Practical Educational Preparation
Training for Employment Centers (PPEP §
Aliates)
-0.297 ** -0.065
Prairie Seeds Academy -0.087
3
-0.144 **
3
Premier High Schools -0.108 ** -0.238 **
PrepNet LLC -0.176 ** -0.102 **
Priority Charter Schools -0.058 ** -0.037 *
3
Propel Schools 0.073 **
3
0.049 *
Public Preparatory Network, Inc. 0.100 **
3
0.116 **
3
Quest Middle Schools 0.006 -0.029
Rapoport Academy Public School (East Waco
Innovative School Development Inc.)
-0.046 -0.060
Raul Yzaguirre School for Success 0.050
3
0.129
3
ReGeneration Schools 0.174 ** 0.150 **
3
Renaissance Charter School, Inc. 0.023
3
0.009
3
ReNew Schools -0.036 -0.015
RePublic Charter Schools 0.064
3
0.096
3
ResponsiveEd Classical Academies 0.058 **
3
0.053
3
Richard Milburn Academies -0.403 ** -0.475 **
River City Science Academy 0.011
3
0.048 **
3
Rocketship Education 0.166 **
3
0.239 **
3
Rocklin Academies 0.024 ** 0.051 **
Rocky Mountain Prep 0.075 * 0.331 **
3
Roger Bacon 0.051 **
3
0.048 **
3
Rose Management Group -0.134 0.126
Scholar Academies 0.121 *
3
0.027
School of Excellence in Education (SEE) -0.027
3
-0.044
3
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
School of Science and Technology 0.054 *
3
0.104 *
3
Seeds of Health Inc. -0.040 0.039
SER-Ninos, Inc. 0.023
3
0.160 **
Shekinah Learning Institute, Inc. -0.037
3
0.017
3
Sherman Thomas Public Charter Schools -0.090 0.096
Skyline Schools, Inc. -0.198 ** -0.120 **
Somerset Academy 0.021
3
0.033
3
South Texas Education Technologies, Inc. 0.044 *
3
0.029
3
Southwest Schools (Educational
Leadership Inc.)
-0.092 ** -0.049
Southwest Winners Foundation, Inc. -0.115 ** -0.147 **
3
Springs Charter Schools (SCS) -0.012
3
-0.035 *
3
St. Croix Preparatory Academy 0.135 **
3
0.107 **
St. Hope Public Schools 0.149 **
3
0.193 **
3
Strive Prep Charter Schools -0.003
3
0.031
Student Alternatives Program Incorporated -0.241 ** 0.110
3
Success Charter Network 0.185 **
3
0.357 **
3
Summit Academies Utah -0.059 ** 0.035
3
Summit Academy of Schools 0.016 0.027
Summit Public Schools 0.055 0.083 *
3
Superior Schools Corporation 0.048 **
3
0.054
3
Synergy Academies 0.008 0.052
3
TeamCFA 0.013
3
0.000
3
Tekoa Academy of Accelerated Studies 0.157 0.363 **
Texas Boys Choir 0.062 ** -0.006
Texas Education Centers (Salvaging Teens
at Risk)
0.016
3
-0.127
Texas Leadership Public Schools -0.091 ** -0.193 **
The Charter Schools of Excellence 0.064
3
0.066
3
The Classical Academies -0.024 0.035
3
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 2: Charter Management Organizations 2023 135134
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
The Foundation for Hispanic Education -0.063 * 0.086 **
3
The Odyssey Preparatory Academy Inc. 0.035
3
-0.022
The Rhodes School -0.024 * -0.023
UT Tyler University Academy 0.017 -0.093
The W.E.B. Du Bois Consortium of Charter
Schools, Inc.
-0.038 -0.136 **
Tindley Accelerated Schools 0.126 **
3
0.223 **
3
Tracy Learning Center -0.084 -0.059
Tri-Valley Learning Corporation -0.056 -0.149 *
Trinity Charter Schools -0.085
3
-0.172
Tucson International Academy -0.035
3
-0.043
Two Dimensions Preparatory Charter 0.140 ** 0.042 **
UCP Charter Schools -0.248 ** -0.166 **
Uncommon Schools New York City 0.034 ** 0.115 **
Uncommon Schools Newark 0.169 **
3
0.220 **
3
Uncommon Schools Rochester 0.138 **
3
0.188 *
United Schools of Indianapolis 0.012 0.046
Universal Education Management Company 0.037
3
0.058
3
University Academy Missouri 0.099 ** 0.160 *
University of Chicago Charter School
Corporation
-0.059 ** 0.095 **
University of Texas - University Charter School -0.186 * -0.295 **
University Preparatory Academy -0.011 0.045
UP Education Network -0.048 **
3
-0.028
3
Uplift Education 0.049 **
3
0.046
3
Urban Prep Academies 0.032
3
0.014
Value Schools 0.181 **
3
0.176 **
Vanguard Academy, Inc. 0.112 ** 0.091 **
3
Vanguard CO 0.072 0.109 **
Reading Math
Estimate Signicance Gap Buster Estimate Signicance Gap Buster
The Varnett Public Schools -0.006 0.049
3
Vista Academies -0.089 ** -0.140 **
Voices College-Bound Language Academies 0.076 ** 0.135 **
Wayside Schools -0.080 ** -0.129 *
Widening Advancements for Youth -0.531 **
Winfree Academy Charter School -0.341 ** -0.653
YES Prep Public Schools 0.089 ** 0.175
3
Youth Connections Charter Schools -0.197 ** -0.279
Zoe Learning Academy, Inc. -0.043 -0.039
* Signicant at the 0.05 level, ** Signicant at the 0.01 level; The & symbols in GB column indicates the “gap-busting” CMOs described in section 2.4.9.8.
As a Matter of Fact:
The National
Charter School
Study III 2023
As a Matter of Fact:
The National Charter
School Study III 2023
Volume 3
Summary of Findings,
Conclusions and Implications
Authors
Margaret E. Raymond, Ph.D.
James L. Woodworth, Ph.D., Lead Analyst- 31 State Study
Won Fy Lee, Ph.D., Lead Analyst- CMO Study
Sally Bachofer, Ed.M.
Contributors
Meghan E. Cotter Mazzola, M.S.
William D. Snow
Tzvetelina Sabkova, M.A.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023 139
© 2023 CREDO
Center for Research on Education Outcomes
Stanford University
Stanford, CA
https://credo.stanford.edu
CREDO, the Center for Research on Education Outcomes at Stanford University, aims to improve empirical
evidence about education reform and student performance at the primary and secondary levels. CREDO
at Stanford University supports education organizations and policy makers in using reliable research and
program evaluation to assess the performance of education initiatives. CREDO’s valuable insight helps
educators and policy makers strengthen their focus on the results of innovative programs, curricula, policies
and accountability practices.
Acknowledgments
CREDO gratefully acknowledges the support of the state education agencies that contributed their data to
this partnership. Our data access partnerships form the foundation of CREDO’s work, without which studies
like this would be impossible. We strive daily to justify the condence placed in us.
The research presented here uses condential data from state departments of education. The views
expressed herein do not necessarily represent the positions or policies of the organizations noted above.
No ocial endorsement of any product, commodity, service or enterprise mentioned in this publication is
intended or should be inferred. In addition:
> The research presented here utilizes SLDS Data from the Idaho State Board of Education (SBOE) and the
Idaho State Department of Education. Any research errors are the sole responsibility of the author(s).
> This research result used data structured and maintained by the MERI-Michigan Education Data Center
(MEDC). MEDC data is modied for analysis purposes using rules governed by MEDC and is not identical
to data collected and maintained by the Michigan Department of Education (MDE) and/or Michigan’s
Center for Educational Performance and Information (CEPI). Results, information and opinions solely
represent the analysis, information and opinions of the author(s) and are not endorsed by, or reect the
views or positions of, grantors, MDE and CEPI or any employee thereof.
> Data for this report was provided by the Missouri Department of Elementary and Secondary Education.
> The conclusions of this research do not necessarily reect the opinions or ocial position of the Texas
Education Agency, the Texas Higher Education Coordinating Board, or the State of Texas.
The analysis and conclusions contained herein are exclusively those of the authors and are not endorsed by
any of CREDO’s supporting organizations, their governing boards, or the state governments, state education
departments or school districts that participated in this study. All errors are attributable to the authors.
CREDO also acknowledges the support of the Walton Family Foundation and The City Fund for supporting
this research.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 141140
3.1 Summary of Findings
As a Matter of Fact: The National Charter School Study III 2023 (NCSSIII) is the third national study by CREDO
evaluating the academic progress of students enrolled in charter schools in the United States. The current
report presents ndings from 2014 to 2019, which yields four periods of year-to-year student growth as
measured by state achievement tests. It includes data from 29 states plus Washington, D.C., and New
York City, which for convenience we report as 31 states. In addition, because we have used a common
methodology across the three studies, we can combine results into trends to support insights of the
performance of students enrolled in charter schools over the past 15 years.
To organize the extensive body of this current research eort, CREDO separated the analysis into two parts
and produced two reports: (1) Charter School Performance in 31 States (CSP31) and (2) Charter Management
Organization 2023 (CMO23). CSP31 examines the performance of the full set of charter school students and
schools, while CMO23 analyzes the dierence in academic growth between students attending charter
schools associated with charter management organizations (CMOs) and those attending stand-alone charter
schools (SCS).
1
In this volume, we integrate the Summary of Findings, Conclusions and Implications sections
from both reports to ensure we present the fullest picture of performance in charter schools.
Our work deliberately focuses on a specic outcome: the annual progress that students make over an
academic year. In this report, we look at students in charter schools compared to the experience they would
have had in the traditional public schools (TPS) they would otherwise have attended. One notable limitation
of this approach is that we have limited line of sight “under the hood” and into the role that localized
environmental, regulatory and organizational factors play in individual school performance. Our contribution
to the K–12 education research and practice landscape is to test fundamental questions of the eectiveness
of charter schools and highlight outcomes and trends rooted in academic progress.
Looking at year-to-year academic progress from 2015 to 2019, the typical charter school student in our
national sample had reading and math gains that outpaced their peers in the traditional public
schools (TPS) they otherwise would have attended. We report these dierences as marginal days of
additional (or fewer) days of learning on a learning benchmark of 180 days of learning each school year for
matched TPS students. In math, charter school students, on average, advanced their learning by an additional
six days in a year’s time, and in reading added 16 days of learning.
1 The CMO study does not include Idaho, Maryland, and Ohio.
Table of Contents
3.1 Summary of Findings ......................................................................141
3.1.1 Do All Students Benet? ............................................................... 142
3.1.2 Where Is Positive Academic Growth Happening? .........................................143
3.1.3 What Can We Learn from CMOs? .......................................................144
3.1.4 Variations in Charter School Performance ...............................................146
3.1.5 Charter School Growth and Achievement ...............................................147
3.1.6 Exceptional Performance in Charter Schools ............................................. 148
3.1.7 Evidence of Improvement over Time .................................................... 149
3.2 Conclusions ...............................................................................149
3.3 Implications ..............................................................................153
References ...................................................................................155
Table of Figures
Figure 3.1: RECAP – Annual Academic Growth of Charter School Students, Reading and Math ...........142
Figure 3.2: RECAP – Annual Academic Growth of Charter School Students
by Charter School Type, Reading and Math ...................................................144
Figure 3.3: RECAP – Academic Growth of Charter Schools Compared to Their Local TPS, Reading .......146
Figure 3.4: RECAP – Academic Growth of Charter Schools Compared to Their Local TPS, Math ..........146
Figure 3.5: RECAP – Academic Growth and Achievement 2015 to 2018, Reading .......................147
Figure 3.6: RECAP – Academic Growth and Achievement 2015 to 2018, Math .........................148
Figure 3.7: RECAP – Annual Academic Growth of Charter School Students
across Three National Studies .............................................................149
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 143142
When we examined academic growth for special populations of students, we found that, compared with their
TPS peers:
> Charter school students in poverty had stronger growth
> English-language learner students attending charter schools had stronger growth
> Students receiving special education services had signicantly weaker growth in both math and
reading on average, though CMO-aliated students with Special Education needs have learning on par
with their TPS Special Education peers.
In the past, a common claim asserted that positive academic results in charter schools arise from advantages
that their students bring to their schooling. In some cases the claim focus on students having more motivated
parents. Another version suggests targeting behavior on the part of the school results in a student body that
is better prepared academically, a practice commonly referred to as “cherry picking” or “cream skimming”. If
true, the students in charter schools would show higher academic achievement at the point of enrollment. In
multiple analyses, we do not see signicant evidence of an undue advantage to charter schools. In fact, we
nd the opposite is true: charter schools enroll students who are disproportionately lower achieving than the
students in their former TPS.
3.1.2 Where Is Positive Academic Growth Happening?
Deeper into our analysis, we examine where student learning gains are occurring, and nd that positive and
strong eects exist in charter schools that vary widely by location and conguration.
> States – 18 states in the NCSS3 study produced signicantly stronger growth for students enrolled in
their charter schools when compared with their TPS peers; in 12 states, growth was similar to TPS peers.
Students attending charter schools had weaker reading growth than their TPS peers in only one state,
Oregon. In 12 states, charter school students had signicantly stronger growth in math than their peers
in TPS. In 16 states, math growth was similar between charter students and their TPS peers. Only three
states showed weaker growth for charter students compared to their peers.
> Locale – compared to their TPS peers, urban charter school students had 29 additional days of growth
per year in reading and 28 additional days of growth in math, both of which were signicant. Suburban
charter school students also had stronger growth in reading (+14 days) and in math (+3 days). Rural
students enrolled in charter schools had the equivalent of ve additional days of learning in reading,
but 10 days less growth in math than their TPS peers. These results are strongly hampered by the
performance of virtual charter schools; despite having only six percent of charter school students
enrolled, their impact on student progress of 58 fewer days of learning in reading and 124 fewer days in
math has damaging consequences for students and exerts a outsized drag on overall national results.
> Grade conguration – charter schools serving elementary, middle, and high school students had
statistically positive growth in both reading and math. Results for multilevel charter schools were
negative in math and similar to the TPS comparison groups in reading. Seeing growth in all grade
spans helps us understand that trends in the national aggregate performance are not concentrated in
particular grades.
Figure 3.1: RECAP – Annual Academic Growth of Charter School Students, Reading and Math
* Signicant at the 0.05 level, ** Signicant at the 0.01 level
This gure originally appears as Figure 1.7 in CSP31.
These average eects are across all students, all schools, for all time periods. There is considerable variation
around these averages and this variation forms the foundation for additional analyses and ndings in our
two papers.
This growth represents accelerated learning gains for tens of thousands of students across the country. Each
student and each school is a proof point that shows that it is possible to change the trajectory of learning
for students at scale, and it is possible to dramatically accelerate growth additional students who have
traditionally been underserved by traditional school systems.
3.1.1 Do All Students Benet?
When we probe these results to determine if all students benet, we nd positive results are not only present
in the aggregate, but also across student race/ethnicity groups:
> Black and Hispanic students in charter schools advance more than their TPS peers by large margins in
both math and reading.
> Multiracial, Native American, and White students in charter schools show equivalent progress to
their TPS peers in reading, but had weaker growth than their TPS peers in math.
> Asian students in charter schools showed similar growth to their TPS peers.
0
2
4
6
8
10
12
14
16
18
MathReading
16**
6**
Days of Learning
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 145144
Our analysis uncovered additional ways that CMOs are returning more positive, and often gap-busting,
results:
> New CMOs and new schools in existing CMOs open with strong results, in both cases delivering
stronger average gains for their students than their local TPS. The student gains in new CMOs are not as
strong initially as that of their older CMO peers. New schools started by mature CMOs deliver positive
gains in their early years that were none the less smaller than the older CMO schools.
> Size or age of a CMO does not relate to their quality—which means some CMOs are growing poorly
performing networks of schools.
> Clustering of CMOs’ schools within a single state returns signicantly more days of learning for their
students than in CMOs that operate schools in more than one state.
> CMOs that took on “turn-around” schools, absorbing those schools into their portfolios, positively
impacted results for students who remained enrolled in the turn-around school. In addition, the balance
of the CMO portfolio did not experience a downturn in student learning.
> The Charter School Growth Fund serves as a case study of charter school growth accelerators. CMOs
that the Growth Fund chooses to support have dramatically larger pre-funding learning gains than other
CMOs. The schools that existed at the time of selection remain strong. New CMO schools also open with
dramatically larger learning gains in both subjects judged against their TPS comparisons.
> Excellence at Scale puts dozens of CMOs at the forefront of eorts to provide education that is both
equitable and eective in moving student achievement to give their students full preparation for their
next steps.
3.1.4 Variations in Charter School Performance
In our reports, we analyze school-level performance, in addition to student-level performance, continuing to
report on growth as the outcome variable. Not every charter school provides quality academic programming
or an eective learning environment for students. Across all charter schools in our study, 36 percent have
greater growth, 47 percent have equivalent growth and 17 percent have lower growth relative to their local
TPS. CMO-aliated charter schools display stronger performance, with 43 percent having greater growth,
42 percent having equivalent growth, and 15 percent having lower growth in comparison to their local TPS.
Stand-alone charter schools have slightly more moderate results.
> Continuous Enrollment – charter students overcome an initial learning dip associated with a school
change, and by their fourth year in their charter school, they show 45 days stronger growth in reading
than their TPS peers and 39 additional days of learning per year in math. The longer a student stays
enrolled in a charter school, the better the student’s academic outcomes are.
> School Management – students who attend a charter school that is part of a charter management
organization (CMO) experience signicantly accelerated growth compared to students enrolled in stand-
alone charter schools (SCS). Even so, CMO schools and SCS provide stronger learning than TPS in reading,
and CMOs do so in math. CMO-aliated students advanced by 27 additional days in reading and 23
more days in math over TPS, both of which are statistically signicant. Stand-alone charter schools still
grew signicantly more than TPS in reading by 10 additional days of learning, but were no dierent in
math. Given that SCS serve two-thirds of all students enrolled in charter schools, soft math performance
in these schools taints the otherwise decisive results in other parts of the study.
3.1.3 What Can We Learn from CMOs?
Comprising one-quarter of the schools, but serving 37 percent of students in our national data set, Charter
Management Organizations (CMOs) are producing much of the learning gains we observed for charter school
students.
As with our national top-line results, we nd robust results for CMOs when we grouped their students by
race/ethnicity, special populations, where the CMOs are located, grade spans of the schools in the network
and how long a student enrolls in the school. As with all schools, there is a range of performance for CMOs,
and we share their student impacts in Appendix A.
Figure 3.2: RECAP – Annual Academic Growth of Charter School Students by Charter School Type,
Reading and Math
** Signicant at p ≤ 0.01
This gure originally appears as Figure 2.3 in CMO23.
10**
-3
27**
23**
Days of Learning
SCS CMO
-5
0
5
10
15
20
25
30
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 147146
These encouraging results require a note of caution. Since the reference point in these comparisons is the
growth that equivalent students in the local TPS realize, this comparison does not reveal if the dierence
is modest or large, nor does it indicate where in the range of absolute achievement the dierence occurs.
Positive dierences at the lowest levels of achievement may not be sucient to move students ahead fast
enough to reach long-term outcomes such as academic prociency or post-secondary readiness. Similarly,
a charter school may post growth results that are considered outsized for any school but still lag behind
the community schools in achievement. Simultaneous consideration of student academic growth and
achievement is the only way to get the complete picture of charter school performance.
3.1.5 Charter School Growth and Achievement
Student academic growth measures how much students advance their learning in a year’s time, and student
achievement measures the stock of their knowledge at the end of the year. We believe it is critical to examine
both growth and achievement in order to understand how well schools prepare students for next steps
in school and life. We map each school’s average growth and average achievement against the growth of
matched TPS students and average state performance. Examining both measurements for all schools in our
national data set during the most recent growth period, we present ndings in four basic categories of school
performance:
> High Growth—High Achievement: schools that exceed the growth of their local options and whose
students are above the state average in overall achievement
> High Growth—Low Achievement: schools that exceed the growth of their local options but with overall
student achievement below the state average
> Low Growth—High Achievement: schools whose students exceed the state average on achievement
but do not advance as much yearly as their comparisons
> Low Growth—Low Achievement: schools with lower academic growth than their local alternatives and
whose students’ achievement is lower than the state average at the end of a school year.
Figure 3.5: RECAP – Academic Growth and Achievement 2015 to 2018, Reading
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.1% 1.5% 5.8% 2.8%
70th Percentile
50th Percentile
30th Percentile
0.7% 9.1% 17.0% 6.1%
3.1% 12.3% 17.6% 6.4%
4.1% 6.8% 5.8% 1.1%
Low Growth,
Low Achievement
High Growth,
Low Achievement
This gure originally appears as Figure 1.25 in CSP31.
Figure 3.3: RECAP – Academic Growth of Charter Schools Compared to Their Local TPS, Reading
This gure combines ndings that originally appear as Figures 1.22 in CSP31 and Figure 2.36 in CMO23.
In math, more charter schools have weaker results than they do in reading, as presented in the gure below.
As the share of charter schools with growth greater than their TPS peers is comparable with the same
growth in reading across all categories, the driver of the overall weaker performance in math is the greater
percentage of charter schools (all, CMO-aliated and stand-alone charter schools) that perform worse than
their TPS peers. Stand-alone charter schools have the largest share of schools with lower growth in math in
comparison to their local TPS.
Figure 3.4: RECAP – Academic Growth of Charter Schools Compared to Their Local TPS, Math
This gure combines ndings that originally appear as Figures 1.22 in CSP31 and Figure 2.37 in CMO23.
READING
BetterSameWorse
STAND-ALONE CHARTERS
CMO CHARTER SCHOOLS
ALL CHARTER SCHOOLS 17% 47% 36%
15% 42% 43%
18% 50% 32%
MATH
BetterSameWorse
STAND-ALONE CHARTERS
CMO CHARTER SCHOOLS
ALL CHARTER SCHOOLS 25% 39% 36%
22% 34% 44%
27% 42% 31%
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 149148
3.1.7 Evidence of Improvement over Time
Findings from this study take on even more weight when considered in the historical context of the 15 years
of CREDO studies on student academic progress in charter schools. Between the 2009 and 2023 studies,
against a backdrop of at performance for the nation as a whole, the trend of learning gains for students
enrolled in charter schools is both large and positive.
Figure 3.7: RECAP – Annual Academic Growth of Charter School Students across Three National Studies
** Signicant at p ≤ 0.01
This gure originally appears as Figure 1.8 in CSP31.
3.2 Conclusions
The outcomes of these studies are largely positive and support several conclusions about the current
landscape of charter schools across America. Perhaps more importantly, the opportunity to position these
ndings in the larger body of research leads to a number of implications about the fundamental policies and
practices of charter schooling at a more global level.
1. In both reading and math, charter schools provide students with stronger learning compared with the
learning in the traditional public schools that are otherwise available to them.
Across the broad range of charter schools, the evidence suggests that they are a robust education option
under many conditions. Whether stand-alone or networked, charter schools operate by law mainly on
their own, making decisions they expect will serve their students well. According to our latest ndings, the
autonomy given to them usually yields positive results. The majority of charter schools provide better year-
to-year outcomes for students compared to their traditional public-school options. Most of these schools
perform better to such a degree that the dierence is statistically signicant.
Schools that have average student achievement above the state average (above the 50th percentile) are
presented in the top half of the gure. In reading, 43 percent of all schools have average performance in the
upper half in their respective states, with a majority of those high achievement schools also having stronger
growth than their local TPS. Zeroing in on the low-growth/low-achievement quadrant, 207 schools (4.1
percent) in our study have lower academic growth than their local alternatives and have student achievement
that is below the 30th percentile of state achievement at the end of the school year.
Figure 3.6: RECAP – Academic Growth and Achievement 2015 to 2018, Math
Low Growth,
High Achievement
High Growth,
High Achievement
Growth (in Days of Learning)
-87 0 87
0.2% 2.0% 4.9% 3.8%
70th Percentile
50th Percentile
30th Percentile
1.0% 8.6% 12.0% 7.5%
4.9% 14.3% 13.8% 6.2%
7.1% 7.5% 5.3% 1.3%
Low Growth,
Low Achievement
High Growth,
Low Achievement
This gure originally appears as Figure 1.26 in CSP31.
In math, above average achievement exists in 40 percent of charter schools, while 60 percent of schools
have achievement that is lower than their state averages. Twenty-eight percent of schools in the data set are
high-growth/high-achievement schools, returning great gains for their students. Zeroing in again on the low-
growth/low-achievement quadrant, 348 schools (7.1 percent) have lower academic growth than their local
alternatives and have student achievement that is below the 30th percentile of state achievement at the end
of the school year.
The number of schools in the low-growth/low-achievement quadrant, though smaller in reading than in math,
remains a key concern.
3.1.6 Exceptional Performance in Charter Schools
Perhaps the most revealing nding of our study is that more than 1,000 schools have eliminated learning
disparities for their students and moved their achievement ahead of their respective state’s average
performance. We refer to these schools as “gap-busting” charter schools. They provide strong empirical
proof that high-quality, high-equality education is possible anywhere. More critically, we found that dozens
of CMOs have created these results across their portfolios, demonstrating the ability to scale equitable
education that can change lives.
-6**
-17**
Days of Learning
-25
-20
-15
-10
-5
0
5
10
15
20
25
3rd Study2nd Study1st Study
-3
6**
6**
16**
MathReading
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 151150
The real surprise of the study is the number of charter schools that have achieved educational equity for
their students: we call them “gap-busting” schools. Ensuring equivalent yearly growth across student groups
has two critical consequences. First, ensuring minority and poverty students learn on par with or better
than their White peers interrupts or reduces the achievement gap. It happens regularly in a large swath of
charter schools. More critically, there is strong evidence that these gap-busting schools can be scaled. Added
to the traditional district schools that achieve similar results, this is the life-transforming education that so
many students need. Second, these schools deliver hundreds of independent proof points that learning gaps
between student groups are not structural or inevitable; better results are possible.
Charter schools function as a portfolio, and their varied impacts on student learning are expected. Charter
school boards and authorizers are responsible for ensuring students perform well. Evidence shows that the
charter school enterprise benets students, and its positive outliers (e.g., gap busters) can pressure the rest
of the system.
The near-term implication for charter school boards and authorizers is two-pronged. Addressing chronic and/or
severe underperformance is necessary and imperative in the current education climate. Identifying high-impact
exemplars for probationary charter schools to study and emulate is possible. Transfer of sub-par schools to higher-
performing operators could be part of a larger incentive for growth and replication. At the same time, authorizers
might consider longer charter terms for charter schools that consistently demonstrate outstanding student learning
success.
Education leaders and policy makers need to understand that in eorts to improve, some failure is inevitable. Any
subsequent failure to address the poor performance compounds the damage. It also blocks constructive learning for
the future. Strong examples of authorizing exist and should be emulated.
Leadership and responsibility demand embracing practices and policies that lead to better results for students, not
maintaining the status quo.
3. The larger scale of Charter Management Organizations does not guarantee high performance—but on
balance, it helps.
When taken as a whole, schools managed by Charter Management Organizations and charter networks bring
a greater learning benet to students compared to stand-alone charter schools. Despite the dierences, both
groups of charter schools have had larger student success than traditional public schools with respect to
reading. We note, however, that math gains in stand-alone charter schools were equivalent to TPS learning.
Our analysis highlights attributes of higher-performing CMOs and networks that could be useful in future
discussions. Size or age of the CMO does not relate to student learning: at every increment of CMO age
or portfolio size, we see high- and low-impact CMOs and networks. This further supports earlier CREDO
research that showed that CMOs only replicate the quality they already have. The implications of replicating
schools with weak results is clear. The big upside is the ability of dozens of CMOs to scale their gap-busting
performance. Additionally, CMOs that concentrate their operations within a single state have stronger gains
than multistate CMOs, though both groups do well by their students.
The results stand up to deeper investigation. Charter schools produce superior student gains despite
enrolling a more challenging student population than their adjacent TPS. They move Black and Hispanic
students and students in poverty ahead in their learning faster than if they enrolled in their local TPS. They
are more successful than the local public school alternatives across most grade spans and community
settings. These results show that charter schools use their exibility to be responsive to the local needs of
their communities.
These ndings generalize into lessons for policy leaders, educators, and funders. Knowing that the average student
in the average charter school can outperform their TPS peers raises important questions about the priority placed
on student outcomes in education decisions in many communities.
2. Some charter schools provide less student learning than their local district schools, although a larger
proportion delivers better learning outcomes. The latter group includes over 1,000 charter schools
managing stang and resources to deliver superior academic results that eliminate the learning gap
across student groups.
Vital lessons also come from the distribution of school performance around the average. Over the past 30
years, small, large, urban, rural, networked or stand-alone charter schools, autonomous and independent of
each other, have arrived at their own solutions for giving their students stronger learning experiences. The
discretion that charter schools enjoy does not guarantee that each school or every charter network realizes
strong student outcomes. Our study illuminated the range of learning across schools.
Despite declining shares, there remain a concerning number of charter schools with weaker student
outcomes. While lower-performing schools make up a larger share of stand-alone charter schools, CMOs and
networks also have a substantial share that produces low gains for their students. This study has profound
implications for charter schools and charter networks that do not support student learning. Charter boards
and authorizers are the accountability side of the charter school equation. They evaluate school performance
and, if necessary, dictate remedies. As our analysis shows, disturbing numbers of charter schools and
networks have low learning levels. There are brick-and-mortar, online, networked, and stand-alone charter
schools with sub-par results.
The number of school closures we observed in the years of this study was small compared to the counts of
schools with the lowest student growth and academic achievement. Since primary and secondary education
is essential to the social contract, providing a foundation for future opportunities, the claim of “choice” cannot
justify derailing students’ preparation. Especially in the post-COVID era, the need for charter boards and
authorizers to address under-performance in their schools has never been more critical.
Closure is not the sole remedy. As we learned from our special investigation, the “takeover“ of
underperforming schools by strong CMOs led to improved student learning for the students who remained
enrolled before and after the transfer. The gains did not adversely aect student academic progress in the
rest of the CMOs’ schools. This policy tool may have broader utility than previously realized.
At the high end of the performance range, good news exists in the growing share of schools outpacing
learning in their local TPS. In both subjects and for both CMO and stand-alone schools, larger shares are
“better than” and a smaller share is “weaker than” compared to earlier work.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 153152
3.3 Implications
The charter school policy framework sets the conditions for charter schools’ growing positive outcomes. It
is the fundamental common denominator in every case, and its role is powerful.
The framework oers a divergent approach from the conventional strategy for public schools. The “exibility
for accountability” construct is not just a catchphrase. It is a distinctly dierent mode of operation. The
“loose-tight” parameters of the framework create incentives to which schools and networks respond. The
incentives nd positive support in this study’s ndings and the broader trends. While our study design cannot
make causal claims (because randomly assigning schools to the traditional or charter school approach has yet
to happen), it can deliver a plausible argument of the value of the policy based on available evidence.
On the “loose” side of the approach, the framework establishes a policy of possibility where educators,
leaders and boards of directors have the discretion to build and deliver curriculum and instruction that meets
high standards for learning and is responsive to local needs.
According to this study, there are a lot of positive possibilities. The process has led to many successful schools
nationwide, often with meaningful innovations. The diversity of schools illuminates an important feature of
the framework: success is attainable via many paths. Over time, many have sought and gained permission to
expand and then shown the ability to create strong student learning at scale.
Students in these schools, especially minority students and those in poverty, make larger advances than in
local public schools. Beyond the benets for their students, successful charter schools deliver critical proof
points of ways to improve outcomes for students. In the current regulatory climate, it is dicult to imagine
how similar eorts could become conventional among traditional public schools.
Beyond exibility in school design, school teams have the leeway to tinker with their operations. The results
show that existing charter schools have improved over time. The proportion of charter schools with superior
results is on the rise. The share that lags behind the local TPS alternatives is also shrinking. This means
schools and networks use their discretion and autonomy to foster a standing capacity to adapt over time.
2
Accordingly, the framework also aims to be “tight” at key points as schools open and mature. Authorizers
are expected to behave as governors of quality. They set the bar to receive initial permission to operate,
which exerts quality and safety controls at the outset. Others have documented stronger standards among
authorizers in the review and approval of new applications (Mumma & West, 2018). The ndings of stronger
new schools in this study compared to earlier results attest to the eort and to the CMO replications and new
charter schools that meet the higher bar.
2 We saw that capacity in stark terms when we examined how charter schools in three states responded to the COVID-instigated school closure orders (CREDO,
2022). Rapid transformation into remote instructional mode; acquisition and distribution of food, technology, or internet access; and strengthening of personal
supports were widespread. Return to in-person instruction in the fall of 2020 was nearly universal. These points rest admittedly on smaller bases of qualitative
evidence, but they provide human dimensions to the point that the present quantitative analysis illuminates nationally. See also: Boast et al. (2020); Henderson
et al. (2021); Childs et al. (2022).
Programs of external funding and support to CMOs to grow their networks, represented here by the Charter
School Growth Fund, focus on some of the stronger CMOs and networks in our study. After high-performing
CMOs receive endorsement, the learning of students in those CMO schools rises in reading but holds steady
in math.
The majority of new CMO schools are no better or worse than the parent organization has already produced, so
decisions to approve applications by CMOs to open new schools must consider the contributions to student learning
of schools in the existing portfolio.
CMO growth accelerators help augment board and authorizer reviews through their extensive selection process; the
growth of their grant-receiving CMOs maintains the strong student learning that led to their selection. The expansion
of these high-quality schools and networks benets more students and communities.
4. Charter schools and networks improve over time, as do the systems that oversee them.
Insights about improvement in schools and networks stem from this study and CREDO’s prior multistate
studies.
In the years of this study, student growth in charter schools was the strongest observed in any of CREDO’s
multistate studies. Added to the results from the previous two studies, a strong trend of improvement
becomes clear. We see substantial increases in student learning in CMOs in both tested subjects and in
reading for stand-alone charter schools. Even the nding of no dierence in math learning in stand-alone
charter schools vis a vis TPS, a decline from the 2017 study results, still marks an improvement from the
statistically signicant negative results in the rst CMO vs. stand-alone comparisons in 2013.
A better understanding of the improvement in the sector comes from two dierent ndings. The rst is
that the largest share of improvement comes from existing charter schools. Compared to the National
Assessment of Education Progress (NAEP) trend, evidence of schools getting better over time is welcome
news.
Second, new schools opened with stronger results than at any time in the past. Growth in the number of
CMOs since the last study plays a role. Many stand-alone charter schools also pushed their results upward.
Strengthening authorizer standards and practices, a drive that took root in the 2010s, also sets a higher bar
that resulted in better schools opening.
Finding ways to improve student academic outcomes is an ambition shared by policy and community leaders,
educators, funders and parents. Charter school results show that change for the better is possible in the larger
education system. The key to improvement lies outside any particular school or network model, though many are
worthy of emulation. It is simply not possible to drive single solutions through the diverse landscape that is U.S.
public education. Lessons from the charter school experience and results may be helpful in charting a future course
in public education.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 155154
References
Angrist, J. D., Dynarski, S. M., Kane, T. J., Pathak, P. A., & Walters, C. R. (2012). Who benets from KIPP? Journal
of Policy Analysis and Management, 31(4), 837–860.
Betts, J. R., & Tang, Y. E. (2008). Value-added and experimental studies of the eect of charter schools on
student achievement. Center on Reinventing Public Education, Seattle, WA. Retrieved February 24, 2023,
https://crpe.org/wp-content/uploads/pub_ncsrp_bettstang_dec08_0.pdf
Boast, L., Cliord, B., & Doyle, D. (2020). Learning in real time: How charter schools served students during
COVID-19 closures. National Alliance for Public Charter Schools. Retrieved from https://www.publiccharters.
org/our-work/publications/how-charter-schools-served-students-during-covid-19-closures
Booker, K., Gill, B., Zimmer, R., & Sass, T. (2009). Achievement and attainment in Chicago charter schools.
RAND Corporation. Retrieved March 10, 2023, from https://www.rand.org/pubs/technical_reports/TR585-1.
html
Childs, J., Grooms, A., & Mozley, M. P. (2022). Hidden in (virtual) plain sight: A charter district’s focus on
attendance during COVID-19. Education and Urban Society.
Clark, M., Gleason P., Tuttle, C., & Silverberg, M. (2015). Do charter schools improve student achievement?
Educational Evaluation and Policy Analysis 37(4), 419–436. Retrieved March 13, 2023, from https://www.jstor.
org/stable/43773520
Corbett, J. (2015). Chartering turnaround: Leveraging public charter school autonomy to address failure.
National Alliance for Public Charter Schools.
CREDO. (2022). Charter Schools’ Response to the Pandemic in California, New York and Washington State.
The Center for Research on Education Outcomes. https://credo.stanford.edu/wp-content/uploads/2022/02/
Charter-School-COVID-Final.pdf
Cremata, E., Woodworth, J., & Raymond, M. (2015). Urban Charter School Study. https://credo.stanford.edu/
reports/item/urban-charter-school-study/
Cremata, E., Davis, D., Dickey, K., Lawyer, K., Negassi, Y., Raymond, M., & Woodworth, J. (2013). National
Charter School Study. Center for Research on Education Outcomes. https://credo.stanford.edu/reports/item/
national-charter-school-study/
Davis, D. H., & Raymond, M. E. (2012). Choices for studying choice: Assessing charter school eectiveness
using two quasi-experimental methods. Economics of Education Review, 31(2), 225–236.
De Brey, C., Musu, L., McFarland, J., Wilkinson-Flicker, S., Diliberti, M., Zhang, A., Branstetter, C., & Wang, X.
(2019). Status and trends in the education of racial and ethnic groups 2018 (NCES 2019-038). U.S. Department
of Education, National Center for Education Statistics. https://nces.ed.gov/pubs2019/2019038.pdf
Authorizing is a delicate job that requires resources, expertise and substantial political acumen and courage.
There is growing attention to authorizers adopting rigorous standards and practices and using a variety of
performance data to evaluate schools that apply for renewal (NACSA, 2016).
Poorly performing charter schools are often ignored. A number of these schools were observed during this
study window. There is data to assess policy leaders and authorizers to hold them accountable for protecting
children’s futures. As tough as closing schools is, the disservice of not closing poorly performing schools has
large and lingering ripple eects.
Our results show that the framework of charter schools helps current students and strengthens public
education overall. We contend these incentives have broader applicability in public schools and see signs
of their spread. Collaborations between charter schools and local district schools have grown over time.
Some states, including Kentucky and Maine, have adopted policies to give educators freedom in adjusting
instruction and boosting performance. However, uptake is slow.
In the year 2023, the importance of strong academic achievement among America’s students has never
been greater. The students hit hardest by school closures during the coronavirus pandemic are precisely
those whom this research illuminates as being able to benet the most from charter schools. In this study
thousands of charter schools have proved that we can do better for our students. The current number of
students beneting from these schools is 3.7 million, but the number could drastically increase if more
schools agreed to the same arrangement. Whether it be termed “charter school” or something else, the
deduction from this data is that when both sides of the equation—exibility and accountability—are working
together for more schools, more students’ academic results will improve.
Executive Summary Volume 1
Charter School Performance
in 31 States
Volume 2
Charter Management
Organizations 2023
Volume 3
Summary of Findings,
Conclusions and Implications
As a Matter of Fact: The National Charter School Study III 2023Volume 3: Summary of Findings Conclusions Implications 157156
Mumma, K. S., & West, M. R. (2018). Charter School Authorizing in California. Getting Down to Facts, September
2019.
NACSA. (2016). State of Charter Authorizing 2016. Retrieved April 24, 2023 from https://qualitycharters.org/
wp-content/uploads/2018/07/State-of-Charter-School-Authorizing-2016-Findings.pdf.
Raudenbush, S. W., Jean, M., & Art, E. (2011). Year-by-year and cumulative impacts of attending a high-mobility
elementary school on children’s mathematics achievement in Chicago, 1995 to 2005. In G. J. Duncan and R. J.
Murname (Eds.) Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances (pp. 359–375).
Russell Sage Foundation.
Unlu, F., Lauen, D. L., Fuller, S. C., Berglund, T., & Estrera, E. (2021). Can quasi-experimental evaluations
that rely on state longitudinal data systems replicate experimental results? Journal of Policy Analysis and
Management, 40(2), 572–613.
White, J., & Xu, Y. (2022, December 6). Who Manages Charter Schools? Retrieved August 22, 2022, from
https://data.publiccharters.org/digest/charter-school-data-digest/who-manages-charter-schools/
Woodworth, J. L., Raymond, M., Chirbas, K., Gonzalez, M., Negassi, Y., Snow, W., & Van Donge, C. (2015). Online
charter school study 2015. Center for Research on Educational Outcomes.
Woodworth, J. L., & Raymond, M. E. (2013). Charter School Growth and Replication, Vol. 2. Center for Research on
Education Outcomes.
Woodworth, J. L., Raymond, M. E., Han, C., Negassi, Y., Richardson, W. P., & Snow, W. (2017). Charter
management organizations. Center for Research on Education Outcomes.
Zimmer, R., Gill, B., Booker, K., Lavertu, S., & Witte, J. (2009). Do charter schools “cream skim” students and
increase racial-ethnic segregation? Prepared for School Choice and School Improvement: Research in State,
District and Community Contexts, Vanderbilt University, October 25–27.
Zimmer, R., Henry, G. T., & Kho, A. (2017). The eects of school turnaround in Tennessee’s achievement school
district and innovation zones. Educational Evaluation and Policy Analysis, 39(4), 670–696.
Dobbie, W., & Fryer Jr., R. G. (2015). The medium-term impacts of high-achieving charter schools. Journal of
Political Economy, 123(5), 985–1037.
Duncan, G. J., & Murnane, R. J. (2016). Rising inequality in family incomes and children’s educational outcomes.
RSF: The Russell Sage Foundation Journal of the Social Sciences, 2(2), 142–158.
Egalite, A., & Ackerman, M. (2015). Rethinking charter school evaluations when the gold standard is o the
table. Education Next. Retrieved March 10, 2023, from https://www.educationnext.org/rethinking-charter-
school-evaluations-when-the-gold-standard-is-o-the-table/
Fazlul, I., Koedel, C., & Parsons, E. (2023). Free and reduced-price meal enrollment does not measure student
poverty: Evidence and policy signicance. Economics of Education Review, 94. Retrieved March 13, 2023, from
https://doi.org/10.1016/j.econedurev.2023.102374
Furgeson, J., Gill, B., Haimson, J., Killewald, A., McCullough, M., Nichols-Barrer, I., ... & Lake, R. (2012). Charter-
school management organizations: Diverse strategies and diverse student impacts. Mathematica Policy
Research, Inc.
Hanushek, E. A., Peterson, P. E., Talpey, L. M., & Woessmann, L. (2019). The achievement gap fails to close.
Education Next, 19(3), 8–17.
Hanushek, E. A., & Rivkin, S. G. (2006). Teacher quality. In E. A. Hanushek and F. Welch (Eds.), Handbook of the
Economics of Education: Vol. 2 (pp. 1051–1078). North Holland. Retrieved March 22, 2012, from http://hanushek.
stanford.edu/sites/default/les/publications/Hanushek%2BRivkin%202006%20HbEEdu%202.pdf
Henderson, M. B., Peterson, P. E., Houston, D., & West, M. R. (2021). What American families experienced
when COVID-19 closed their schools. Education Next, 21(1), 22–31.
Kho, A., Zimmer, R., & McEachin, A. (2022). A descriptive analysis of cream skimming and pushout in choice
versus traditional public schools. Education Finance and Policy, 17(1), 160–187.
Legislation, Regulations, and Guidance—School Improvement Fund. (2010). Oce of Elementary and
Secondary Education. https://oese.ed.gov/legislation-regulations-and-guidance-school-improvement-fund/
Lehrer-Small, A. (2022). Virtual school enrollment kept climbing even as COVID receded, new data reveal.
The74. https://www.the74million.org/article/virtual-school-enrollment-kept-climbing-even-as-covid-receded-
new-data-reveal/
Mead, S., Mitchel, A. L., & Rotherham, A. J. (2015). The state of the charter school movement. Bellwether
Education Partners.
Monarrez, T., Kisida, B., & Chingos, M. (2022). The eect of charter schools on school segregation. American
Economic Journal: Economic Policy, 14(1), 301–340.
