Do Public Charter Schools Outperform Traditional Public Schools, or 'Cream-Skim' Students and Drain District Budgets? What Stanford's 2-Million-Student CREDO Studies, Admissions Lottery RCTs, and Fiscal Audits Actually Prove

Since Minnesota enacted the nation's first charter school law in 1991, public charter schools have grown from a small bipartisan policy experiment into one of the defining battlegrounds of American K–12 education. By the 2025–2026 academic year, roughly 8,000 public charter schools across 45 states and the District of Columbia enroll more than 3.7 million to 3.8 million students—approximately 7.6% to 8% of all U.S. public school children. Waitlists in major urban centers routinely stretch into the tens of thousands, and in cities such as New Orleans (nearly 100%), Washington, D.C. (~47%), Detroit (~50%), Camden, Newark, and Denver, charter schools educate a massive share of the public student body ^{[1], [6]}.

Unlike private school vouchers or Education Savings Accounts (ESAs), public charter schools are public schools: they are tuition-free, non-sectarian, open to all students without academic entrance exams, required to administer state standardized accountability tests, and governed by performance contracts ("charters") with state or local authorizers that can shut down a school if it fails to meet academic or financial benchmarks ^{[1], [2]}. In exchange for that bottom-line accountability, charter schools are exempted from most traditional school district regulations, geographic attendance-zone assignments, and—in nearly 90% of charter schools nationwide—teachers' union collective-bargaining agreements covering tenure, seniority-based layoffs ("last in, first out"), and rigid length-of-school-day limits ^{[2], [3]}.

In 2025 and 2026, the Trump administration and congressional Republicans—alongside conservative governors and researchers at the Hoover Institution, the Thomas B. Fordham Institute, and the Manhattan Institute—have sought to expand federal Charter Schools Program (CSP) startup grants and tax-credit financing for charter facilities, arguing that public charter schools have decisively disproven the claim that poverty determines academic destiny ^{[1], [9]}. Conversely, the National Education Association (NEA), the American Federation of Teachers (AFT), and progressive urban school boards in cities like Los Angeles, Chicago, and Boston have fought for statutory caps and moratoriums on charter growth—contending that charter schools are an unaccountable privatization scheme that "skims" advantaged students, counsels out costly special-education children, and starves neighborhood public schools of funding ^{[4], [7]}.

"For thirty years, defenders of the traditional public school monopoly claimed that low-income Black and Hispanic children couldn't achieve at grade level unless Washington poured trillions more into district bureaucracies—and that any charter school showing miraculous results must be 'cream-skimming' the best kids. Then Stanford University tracked nearly two million students and MIT economists analyzed randomized admissions lotteries, proving that high-performing urban charter schools eliminate the achievement gap for the exact same neighborhood kids on 25 percent less money per pupil." — Synthesis of Pro-Charter Education Reform Arguments (Hoover Institution, Thomas B. Fordham Institute, and National Alliance for Public Charter Schools) ^{[1], [2], [9]}

Who is telling the truth about charter school performance, student sorting, and school district finances? Because oversubscribed charter schools are required by federal and state law to admit students by random public lottery, and because state longitudinal data systems now track millions of individual students year over year, economists have been able to test every major claim in the charter school debate with randomized and quasi-experimental precision ^{[1], [2], [3]}.

What the Data Shows Part I: Stanford's 1.9-Million-Student CREDO Studies and the Three-Decade Trajectory of Charter Quality

For years, participants in the charter school debate talked past one another by citing different eras of data. Critics of charter schools still routinely cite the 2009 first edition of Stanford University's Center for Research on Education Outcomes (CREDO) study—which examined 16 states from 2001 to 2007 and found that on average, charter students lagged traditional public school students by -7 days of learning in reading and -22 days in math ^[1]. What critics omit is what happened over the subsequent fifteen years as state authorizers shut down hundreds of failing early charters and scaled up proven non-profit charter networks.

1. Inside CREDO's "Virtual Twin" Methodology

Led by Stanford economist Margaret E. ("Macke") Raymond, CREDO developed the Virtual Control Record ("virtual twin") methodology to evaluate charter schools across states where lottery data are unavailable ^[1]. For every student enrolled in a charter school, CREDO searches the state's administrative database inside the exact traditional public "feeder" schools from which the charter draws its students, matching each charter student to up to seven traditional public school students with identical baseline prior-year test scores, race/ethnicity, gender, grade level, free/reduced-price lunch (poverty) status, English Learner (ELL) status, and Special Education (IEP) designation ^[1]. Standardizing annual test-score growth where 0.01 standard deviations equals roughly 5.78 days of learning in a standard 180-day school year, CREDO released three landmark national studies tracking the evolution of the sector:

• CREDO I (2009 — 16 States): Charter students trailed matched traditional public school (TPS) peers by -7 days of learning per year in reading and -22 days in math ^[1].
• CREDO II (2013 — 27 States): As states closed low-performing early charters, charter students pulled ahead in reading (+8 additional days of learning per year) and reached virtual parity in math (-3 days) ^[1].
• CREDO III (June 2023 — "As a Matter of Fact" — 29 States, NYC, & D.C., N = 1,895,286 Matched Charter Students): Tracking nearly two million charter students across more than 6,200 charter schools from 2014–2015 through 2018–2019 (representing over 81% of all U.S. charter students), CREDO III documented a decisive nationwide turnaround: on average, charter students gained +16 additional days of learning per year in reading (+0.028 SD) and +6 additional days in math (+0.008 SD) over their matched traditional public school peers ^[1].

2. Where Charter Schools Triumph: Urban Centers, Low-Income Minority Students, and Non-Profit CMOs

When CREDO III disaggregated the 1.9-million-student national dataset by student demographics, geography, and operating model, the results strongly validated the conservative and school-reform case that urban public charter schools serve as an equalizer for disadvantaged children ^[1]:

• Urban Charter Dominance (+29 Days Reading / +28 Days Math): Students attending charter schools in urban districts gained an average of +29 extra days of learning per year in reading and +28 extra days in math compared to their virtual twins in urban traditional public schools—equivalent to adding more than a full extra month of academic progress every single year ^[1].
• Outsized Gains for Black, Hispanic, and Low-Income Students: Black students in charter schools gained +35 additional days of learning per year in reading and +29 additional days in math over Black peers in traditional public schools; Hispanic charter students gained +30 extra days in reading and +19 extra days in math; and students living in poverty gained +23 extra days in reading and +17 extra days in math (with Black and Hispanic students in poverty posting vastly stronger growth in charters than in TPS) ^[1].
• Non-Profit Charter Management Organizations (CMOs): Charter schools affiliated with multi-school non-profit networks—such as KIPP, Success Academy, Uncommon Schools, Achievement First, IDEA Public Schools, and DSST—outperformed both traditional public schools and standalone independent charters, delivering +27 additional days of learning in reading and +23 additional days in math per year ^[1]. Furthermore, CREDO III proved that the sector's decade-long improvement came from two sources: existing charters improving their instruction over time, and authorizers actively closing underperforming charter schools ^[1].

What the Data Shows Part II: Testing the "Cream-Skimming" and "Push-Out" Myths via Randomized Admissions Lotteries

Confronted with the CREDO data, union-affiliated critics such as the National Education Policy Center (NEPC) have long argued that even when researchers match students on prior test scores, poverty, race, and IEP status, a hidden "selection bias" remains: perhaps families who bother to fill out a charter school application are simply more motivated, read to their children more at home, or possess unobserved advantages ("cream-skimming"), and perhaps charters quietly counsel out struggling or disruptive students before springtime state testing ("push-out") ^{[2], [4]}.

1. Nobel-Winning Randomized Lottery Experiments (Angrist, Pathak, & Walters; Dobbie & Fryer)

Economists solved the "motivated parent" objection completely by exploiting a legal mandate built into state charter laws: whenever more families apply to a charter school than there are open seats, the school must admit students by a blind, random public lottery ^{[2], [3]}. In an admissions lottery, both the lottery winners (treatment group) and the lottery losers (control group) had parents who were identically motivated to apply to the exact same charter school on the exact same day—creating a true randomized controlled trial (RCT) that eliminates all unobserved family or student selection bias ^[2].

Across more than a dozen landmark peer-reviewed lottery studies, the findings not only confirmed CREDO's observational estimates—they found even larger causal gains in urban charter schools:

• Massachusetts Urban vs. Non-Urban Lotteries (Angrist, Pathak, & Walters, AEJ: Applied Economics, 2013): Nobel Prize-winning MIT economist Joshua D. Angrist, MIT economist Parag A. Pathak, and UC Berkeley economist Christopher R. Walters analyzed admissions lotteries across 36 Massachusetts charter middle and high schools ^[2]. They found that winning a random lottery to attend an urban Massachusetts charter school increased student achievement by a massive 0.36 standard deviations per year in math and 0.15 standard deviations per year in English Language Arts (ELA)—roughly equivalent to closing the entire Black-White math achievement gap in three years of middle school ^[2]. Moreover, when the authors directly compared lottery-based RCT estimates against observational matching estimates on the exact same schools, the results were nearly identical—proving that unobserved parental motivation ("cream-skimming") does not explain charter school outperformance ^[2].
• The Harlem Children's Zone, NYC, and KIPP Lotteries (Dobbie & Fryer, AEJ: Applied Economics, 2011; QJE, 2013; Hoxby, Murarka, & Kang, 2009): Harvard economist Roland G. Fryer Jr. and Princeton/Harvard economist Will Dobbie analyzed randomized admissions lotteries at the Promise Academy charter schools in New York City's Harlem Children's Zone and across 39 NYC charter schools ^[3]. Winning the lottery boosted math achievement by 0.23 to 0.28 standard deviations per year, eliminated the Harlem-to-suburban white math gap, reduced teenage pregnancy among female lottery winners by 59% (from 10.1% to 4.2%), and reduced male incarceration rates to virtually zero ^[3].
• Four-Year College Enrollment and Persistence (Angrist, Cohodes, Dynarski, Pathak, & Walters, Journal of Labor Economics, 2016): Critics next claimed that "No Excuses" charters merely "teach to the test," inflating standardized scores without producing real higher-education success. Tracking Boston charter high school lottery winners into adulthood, Angrist et al. (2016) proved that winning a charter lottery doubled the pass rate on Advanced Placement (AP) Calculus exams, boosted SAT scores by 51 points, and increased immediate four-year college enrollment by 18 percentage points (from 41% to 59%), shifting students out of two-year community colleges into four-year bachelor's degree programs ^[2].

2. What Explains Urban Charter Success—and What About the "Push-Out" Accusation?

Why do high-performing urban charter schools produce such large causal gains? In their 2013 Quarterly Journal of Economics paper ("Getting Beneath the Veil of Effective Schools") and Angrist et al.'s (2013) mechanism analysis, Dobbie, Fryer, and Angrist tested whether traditional input metrics—smaller class sizes, higher per-pupil spending, or formal teacher master's degrees/certification—explained charter effectiveness ^{[2], [3]}. They found that traditional inputs explained zero percent of the variance. Instead, five core "No Excuses" practices accounted for roughly 45% to 50%+ of charter school gains: (1) 25% more instructional time (longer school days and school years), (2) high-dosage small-group tutoring, (3) frequent teacher feedback and classroom coaching (10+ times per semester), (4) data-driven instruction using interim assessments to regroup students, and (5) a relentless school-wide culture of high academic and behavioral expectations ^{[2], [3]}. Notably, when Fryer (Quarterly Journal of Economics, 2014) transplanted those exact five charter management practices into traditional public schools in Houston, Texas via a randomized trial, student math scores in the traditional public schools jumped by 0.15 to 0.18 standard deviations per year—proving that the operational model, not student sorting, drives the results ^[3].

What about the claim that charter schools boost their averages by "pushing out" or expelling their lowest-performing students back to traditional public schools? In a definitive 2013 study published in Educational Evaluation and Policy Analysis ("Is There Empirical Evidence That Charter Schools 'Push Out' Low-Performing Students?"), economists Ron W. Zimmer and Cassandra M. Guarino tracked longitudinal student-level transfer records across a major urban school district ^[4]. They found zero empirical evidence that charter schools systematically push out low-scoring students: low-achieving students exited charter schools at no higher rate than they exited traditional public schools, and students transferring out of charters had prior test scores that were statistically indistinguishable from transfer patterns in traditional public schools—findings independently replicated in New York City and Denver by Marcus A. Winters (Economics of Education Review, 2015) ^[4].

The Full Picture: Where Charter Critics Have Real Empirical Merit—Virtual Charters, Non-Urban Schools, Disability Gaps, and District Fixed Costs

While the empirical record thoroughly refutes the "cream-skimming" and "push-out" myths against urban brick-and-mortar charters, an honest, non-partisan analysis also shows that overbroad conservative claims that "charter schools always outperform traditional public schools without fiscal downsides" fail on four crucial empirical points ^{[1], [2], [5], [7]}.

1. The Catastrophic Failure of Full-Time Virtual (Online) Charter Schools (-60 Days Reading / -124 Days Math)

The single greatest Achilles' heel of the charter school movement is the proliferation of full-time virtual (cyber) charter schools—many operated by large, publicly traded or for-profit Education Management Organizations (EMOs) that collect state per-pupil tax dollars while educating students entirely online from home ^[1]. In CREDO's 2023 National Charter School Study III (as well as CREDO's dedicated Online Charter School Study), the performance gap between brick-and-mortar charters and virtual charters was staggering:

• While students in brick-and-mortar charter schools gained +21 extra days of learning per year in reading and +14 extra days in math, students enrolled in full-time virtual charter schools lost -60 days of learning per year in reading (-0.10 SD) and -124 days of learning per year in math (-0.21 SD) compared to matched peers in traditional public schools ^[1].
• Losing 124 days out of a 180-day school year in math means that the average full-time online charter student learned in an entire academic year what a traditional public school peer learned in barely eleven weeks ^[1].
• Similarly, while non-profit CMOs posted huge gains (+27 days reading, +23 days math), charters operated by for-profit Education Management Organizations (EMOs)—which heavily populate the virtual and low-regulation charter subsector in states like Ohio, Pennsylvania, Michigan, and Arizona—failed to outperform traditional public schools overall and exhibited far higher rates of financial self-dealing and related-party real-estate leases ^[1].

2. The Non-Urban and Texas Earnings Nuance (Angrist et al., 2013; Dobbie & Fryer, 2020)

Second, the academic magic of urban "No Excuses" charters does not automatically generalize to suburban or rural charters, or to progressive/non-rigorous charter designs. In their 2013 Massachusetts lottery study, Angrist, Pathak, and Walters found that while urban Massachusetts charters boosted math by +0.36 SD/year, non-urban (suburban and town) Massachusetts charters actually had a statistically significant negative effect on test scores (-0.14 SD in ELA and -0.12 SD in math) ^[2]. Why? First, non-urban charter lottery applicants were leaving high-performing, well-funded suburban traditional public schools rather than struggling urban districts; second, non-urban charters rarely used the intensive "No Excuses" instructional model (extended time, tutoring, and data-driven pedagogy) ^[2].

Even more revealingly, in a landmark 2020 Journal of Political Economy study linking Texas administrative school records to state unemployment-insurance wage records ("Charter Schools and Labor Market Outcomes"), Will Dobbie and Roland G. Fryer Jr. examined all Texas charter schools (not just oversubscribed lotteries) ^[3]. They found a stark divide: "No Excuses" charters in Texas significantly increased test scores and boosted four-year college enrollment, whereas regular ("non-No Excuses") Texas charters—many of which were credit-recovery or alternative schools—actually reduced test scores, had zero effect on college graduation, and produced slightly negative or null effects on early-adult earnings at ages 24–28 ^[3].

3. Students with Disabilities: The GAO Enrollment Gap vs. Setren's (2021) Declassification Discovery

Third, what does the evidence show regarding students with disabilities (SWD) and English Learners (ELLs)? Here, empirical research uncovers a fascinating two-part reality that partially validates both critics and defenders ^{[5], [6]}:

• The Federal Enrollment & Severe-Disability Gap (GAO-12-543 & NCES): A landmark audit by the U.S. Government Accountability Office (GAO-12-543) and ongoing National Center for Education Statistics (NCES) data confirm that nationwide, public charter schools enroll a lower share of students with Individualized Education Programs (IEPs) than traditional public schools (historically 8.2% to ~11.2% in charters vs. 11.2% to ~15.0% in TPS) ^[5]. Crucially, the gap is most pronounced among students with severe, high-cost, low-incidence disabilities (such as severe autism, multiple disabilities, traumatic brain injury, or deaf-blindness requiring self-contained therapeutic classrooms), both because traditional districts maintain centralized special-education infrastructure and because some standalone charters lack specialized staff or informally discourage parents during pre-enrollment tours ^[5].
• The Boston Lottery Breakthrough (Setren, Journal of Human Resources, 2021): Does the lower IEP percentage in charters mean charters refuse to admit or fail to teach special-needs kids? In a brilliant 2021 lottery study in the Journal of Human Resources, Tufts economist Elizabeth Setren tracked students with IEPs and English Learners who entered random admissions lotteries in Boston charter schools ^[6]. Setren discovered that Boston charters admitted special-education and ELL applicants at similar baseline rates, generated massive causal test-score and college-prep gains for them (equal to or larger than general-education gains), and increased the probability that an IEP student was successfully declassified out of special education into general education by 12 percentage points (and increased English-learner reclassification by 21 percentage points) ^[6]. Thus, a substantial fraction of the charter special-education enrollment gap is actually caused by charters moving mild-to-moderate special-needs students into mainstream general-education proficiency rather than keeping them permanently labeled ^[6].

4. School District Finances: "Stranded Fixed Costs" (Ladd & Singleton, 2020) vs. Positive Competitive Spillovers (Cordes, 2018)

Finally, do charter schools "drain" traditional public school budgets and hurt the students left behind? Economists distinguish sharply between fiscal externalities on district budgets and academic spillover effects on remaining public school students ^{[7], [8]}:

• Why Charters Are Funded Less Per Pupil—Yet Still Strain District Budgets (Ladd & Singleton, Education Finance and Policy, 2020): Charter advocates accurately point out that charter schools are not "overfunded": national finance audits by the University of Arkansas Department of Education Reform (Patrick J. Wolf et al.) show that public charter schools actually receive roughly 20% to 30% less total funding per pupil (~$13,000 vs. ~$18,000+ in urban areas) than traditional public schools because charters rarely receive local property-tax capital bond revenues for school buildings ^[9]. However, Duke University economist Helen F. Ladd and University of Rochester economist John D. Singleton (Education Finance and Policy, 2020) audited North Carolina school districts after the state lifted its charter cap in 2011 and demonstrated why traditional districts still experience genuine budgetary pain ^[7]. When state and local per-pupil operating funds follow 500 students out of a district into a new charter school, those 500 students do not leave from a single classroom or a single building—they leave 2 or 3 students at a time across 30 different schools and 13 grades. As a result, the traditional district loses 100% of the state per-pupil allotment immediately, but cannot immediately reduce its "stranded fixed costs" (heating and maintaining half-empty buildings, central district administration, debt service, and disproportionate severe special-education costs) ^[7]. Ladd and Singleton calculated that this mismatch between average-cost funding formulas and marginal-cost reductions imposed a net fiscal burden of more than $500 per remaining pupil in Durham Public Schools (or ~$3,600 per charter-enrolled student) and $300 to $700 per pupil in non-urban districts unless districts take the politically painful step of closing and consolidating underutilized school buildings ^[7].
• Positive Competitive Academic Spillovers on Remaining Public School Students (Cordes, 2018; Ridley & Terrier, 2024): Does that fiscal strain cause test scores to collapse for the children who remain in traditional public schools? The preponderance of econometric evidence proves the exact opposite. In a landmark 2018 study in Education Finance and Policy ("In Pursuit of the Common Good: The Spillover Effects of Charter Schools on Public School Students in New York City"), Temple University economist Sarah A. Cordes analyzed longitudinal data on over one million NYC traditional public school students ^[8]. Cordes found that when a charter school opened nearby—or even co-located inside the exact same public school building—test scores in the neighboring traditional public school significantly increased (+0.02 to +0.09 SD in math and ELA) and grade retention dropped, as competition spurred traditional principals and teachers to raise expectations and improve instructional focus ^[8]. Similar positive or neutral academic spillovers on traditional public school students were confirmed in Massachusetts after a state fiscal-reimbursement reform by Matthew Ridley and Camille Terrier (Journal of Political Economy Microeconomics, 2024) and across multiple states by Marcus Winters (2012) ^[8].

Conclusion: The Triumph of Accountable Urban Charters—and the Case for Smart Authorizing and Fiscal Reform

When three decades of randomized admissions lotteries, two-million-student longitudinal datasets, and school district financial audits are evaluated without ideological blinders, the evidence dismantles the most extreme talking points of both teachers' unions and doctrinaire market fundamentalists.

On the central question of whether public charter schools can transform the educational trajectory of disadvantaged children, conservative and bipartisan charter reformers have won the empirical argument ^{[1], [2], [3]}. The union-backed narrative that charter schools only succeed by "cream-skimming" privileged families or "pushing out" struggling students has been thoroughly falsified by Nobel Prize-winning randomized admissions-lottery studies (Angrist, Pathak, and Walters; Dobbie and Fryer), longitudinal student-exit panels (Zimmer and Guarino), and Stanford University's 1.9-million-student CREDO III evaluation ^{[1], [2], [3], [4]}. In America's urban centers, brick-and-mortar public charter schools—especially non-profit Charter Management Organizations deploying extended instructional time, high-dosage tutoring, frequent teacher feedback, and high behavioral standards—deliver nearly an extra month and a half of learning every year for low-income Black (+35 days reading, +29 days math) and Hispanic (+30 days reading, +19 days math) students, double AP exam pass rates, boost four-year college enrollment by 18 percentage points, and generate positive competitive spillovers that raise test scores in neighboring traditional public schools ^{[1], [2], [8]}.

At the same time, the data prove that a charter label alone is not a magic wand—governance, accountability, and fiscal design matter immensely ^{[1], [2], [7]}. Full-time virtual charter schools have been an empirical disaster, costing students -60 days of learning in reading and -124 days in math per year ^[1]. And as Ladd and Singleton (2020) demonstrate, when state funding formulas transfer average per-pupil revenue to charter schools without accounting for the stranded fixed overhead costs of underutilized district buildings and severe special-education services, traditional districts suffer real transitional budget deficits of $500+ per pupil ^{[5], [7]}. For policymakers in 2026, the evidence points toward a clear, data-driven synthesis: lifting arbitrary caps on proven, non-profit urban brick-and-mortar charter networks; shutting down chronically failing full-time virtual and for-profit cyber charters; weighting special-education funding by actual severity of student disability rather than a flat rate; and pairing charter growth with transitional state facilities aid and district building consolidation so that both charter and neighborhood public school students thrive ^{[1], [6], [7]}.