A Mandatory 13 Year Sentence
The American public school system, judged by its own merit
Any honest accounting of American public education starts with what it built. Between 1850 and 1950, the United States assembled the largest schooling apparatus in human history. It put a building, a teacher, and a set of books within reach of nearly every child on the continent, including children whose parents could not read, did not speak English, or had arrived the previous month with nothing. It absorbed waves of immigration that would have broken a lesser administrative machine. After 1954, and more forcefully after 1975, it accepted legal obligations that no private institution in the country would take on: educate every child who walks through the door, regardless of disability, behavior, language, or ability to pay. The Individuals with Disabilities Education Act made public schools the only institutions in American life required to serve everyone.
The people inside the system deserve the same accounting. Roughly 3.2 million public school teachers show up every day to do difficult work for pay that has declined in real terms. The Reason Foundation’s analysis of federal data found the average inflation-adjusted teacher salary fell 6.1 percent between 2002 and 2022, from $75,152 to $70,548. Many of these teachers buy supplies out of pocket. Many stay late without compensation. Some are the most stable adult presence in a child’s life. Nothing that follows is an argument against them. They are the most sympathetic figures in this story precisely because the institution they serve spends so much of their effort on things other than teaching.
And the schools do things beyond instruction that communities have come to depend on. They feed roughly 28 million children a day through the National School Lunch Program. They provide de facto daycare that makes the two-income household possible. They screen for vision problems, hearing problems, and abuse. In thousands of small towns, the school is the gymnasium, the auditorium, the polling place, and the largest employer.
All of that is real. It is also beside the central question, which is not whether the system does anything useful but whether a government-funded, government-operated, attendance-compelled monopoly is the right instrument for educating children. Institutions should be judged on their stated purpose. The stated purpose of schools is to teach. On that measure, the evidence has been accumulating for 60 years, and it points in one direction.
The Blueprint
The American public school was not designed by Americans, and it was not designed to produce independent thinkers. Both claims sound polemical. Both are matters of documented record.
The model came from Prussia. After Napoleon destroyed the Prussian army at Jena in 1806, the philosopher Johann Gottlieb Fichte delivered his Addresses to the German Nation, arguing that Prussia’s salvation lay in a new system of state education that would shape the will of the young so thoroughly that they could not choose other than what the state intended. Prussia built it: compulsory attendance, age-graded classrooms, a state-certified teaching corps, a standard national curriculum, and testing to sort children into tracks. The system worked as designed. It produced disciplined soldiers, obedient civil servants, and factory workers who arrived on time.
American reformers went and looked. Horace Mann, secretary of the Massachusetts Board of Education, toured Prussian schools in 1843 and came home an evangelist. His Seventh Annual Report praised the Prussian model at length, and Massachusetts passed the first compulsory attendance law in the nation in 1852. Other states followed over the next 66 years, with Mississippi closing the loop in 1918. The pattern is worth noting. Compulsory schooling did not spread because parents demanded it. It spread because state legislatures, urged on by reformers, industrialists, and nativists anxious about Catholic immigrants, decided that families could not be trusted with the job. Attendance officers, truancy courts, and fines did the rest.
The men who built the system said what it was for, in plain language, in print. Ellwood Cubberley, dean of the Stanford School of Education and the most influential education administrator of his era, wrote in Public School Administration in 1916 that schools were in a sense factories in which children were the raw material to be shaped to the specifications of industrial civilization. He meant it as praise. Frederick T. Gates, who ran John D. Rockefeller’s General Education Board, published the board’s vision for rural schooling in Occasional Papers No. 1 in 1913: the board would not try to make the children of the countryside into philosophers, scientists, authors, or statesmen, because the supply of such people was already ample. The goal, in his words, was to organize children and teach them to do in a perfect way the things their fathers and mothers were doing in an imperfect way. The General Education Board spent the equivalent of billions of today’s dollars making that vision institutional fact across the American South.
None of this required a conspiracy. It required a consensus, and the consensus was open. The early 20th century’s leading education theorists, working under the banner of “social efficiency,” argued explicitly that the purpose of mass schooling was to fit each child to his probable station. The historian Raymond Callahan documented the takeover in Education and the Cult of Efficiency in 1962: school superintendents, desperate for the prestige of businessmen, imported Frederick Taylor’s factory management wholesale. Bells to mark shift changes. Batch processing by date of manufacture, which is what age grading is. Standardized inputs, standardized outputs, and a Carnegie unit, invented in 1906, that defined learning as 120 hours of seat time regardless of whether anything was learned in the seat.
The defenders of the founding era make a fair counterpoint: intentions in 1916 do not determine outcomes in 2026, and plenty of institutions have outgrown ugly origins. The counterpoint would carry weight if the architecture had changed. It has not. The bell schedule, the age-graded cohort, the seat-time credit, the compulsory attendance statute, and the district monopoly on funding are all still in place. A teacher from 1926 dropped into a classroom in 2026 would recognize everything except the screens. When an institution retains every structural feature of its original design, the original design documents remain relevant evidence of what the machine was built to do. The question is whether it still does it. The scoreboard says the machine now fails even at its founders’ modest goal.
The Scoreboard
The National Assessment of Educational Progress is the closest thing the country has to an honest scoreboard. It is federally administered, low-stakes for schools, and not gameable in the way state tests are. The 2024 results, released across 2025, completed the picture for grades 4, 8, and 12.
The numbers, plainly stated: 33 percent of 8th graders scored below NAEP Basic in reading, the largest share in the history of the assessment. Below Basic means a student likely cannot identify the sequence of events or the main idea in a text. Fewer than a third of students at either grade 4 or grade 8 read at the Proficient level. In math, nearly 40 percent of 8th graders sit below Basic, meaning they likely cannot use similarity to find the side of a triangle. Among the class of 2024, tested months before graduation, 45 percent scored below Basic in math and 32 percent below Basic in reading, both the worst results ever recorded. Just 22 percent of 12th graders reached Proficient in math. Federal testing officials estimated that 33 percent of seniors were prepared for entry-level college math, down from 37 percent in 2019, even as more of them reported acceptance to four-year colleges. Reading scores for 4th and 8th graders now sit roughly where they were in the early 1990s.
The standard explanation is the pandemic, and the pandemic did real damage, much of it inflicted by the school closures themselves, which in many districts ran far longer than the epidemiology justified. But the federal government’s own analysts have pointed out that the decline among the lowest-performing students began around 2013, seven years before anyone had heard of COVID-19. Congress then spent roughly $190 billion in emergency relief specifically to repair pandemic learning loss. Five years after the disruption, no grade and no subject has recovered to 2019 levels, and the gap between the highest and lowest performers is the widest ever measured. Whatever the $190 billion bought, it was not recovery.
The long view is worse for the system, not better, because the long view includes the money. In fiscal year 2024, according to the Census Bureau’s Annual Survey of School System Finances, current spending per pupil reached $17,619, and total K-12 spending passed $1 trillion for the first time. Total revenue per pupil, the full tax take, averaged $21,065. New York spends $31,918 per student per year. Adjusted for inflation, per-pupil spending has roughly tripled since 1970 and, per the Reason Foundation’s compilation of Census and federal education data, rose 35.8 percent just between 2002 and 2023, from $14,969 to $20,322 in constant dollars. Over that same half century, the long-term trend NAEP scores for 17-year-olds, the system’s finished product, moved essentially nowhere. Reading scores for 17-year-olds in 2012 were statistically indistinguishable from 1971. The recent movement has been downward.
The output side of the pipeline confirms what the school-age tests suggest. The OECD’s 2023 Survey of Adult Skills found that 28 percent of American adults aged 16 to 65, about 59 million people, read at Level 1 or below, up from 19 percent in 2017. Level 1 means a person can understand short, simple sentences but struggles with multi-step instructions. In numeracy, 34 percent of adults, roughly 72 million, sit at Level 1 or below, one of the worst showings among the 31 participating countries. Average American literacy scores fell 12 points in six years. Nearly all of these adults are graduates of the public system, which enrolls close to 90 percent of American children. Thirteen years of compulsory attendance, at a cumulative cost that now exceeds $200,000 per student in the higher-spending states, delivers a workforce in which one adult in three cannot reliably do the math on a loan document.
Two objections deserve to be met head-on rather than left for a footnote. The first is that American schools face harder demographics than their international peers: more poverty, more language learners, more inequality. This is partly true and entirely damning as a defense, because the American system was sold, from Mann forward, as the great equalizer, and the 2024 NAEP data show it doing the opposite. The declines are concentrated almost entirely among the bottom quartile, whose losses since 2022 grew 70 percent larger, while the top decile held steady or gained. A system that cannot educate poor children after 175 years of promising to do exactly that has not been mugged by demographics. It has failed its own founding pitch.
The second objection is that money does matter, and it does. The Education Recovery Scorecard project at Harvard and Stanford found that each additional $1,000 per pupil in federal relief bought roughly 1 percent of a standard deviation in achievement, with larger effects in poor districts. Take the finding at face value. Real per-pupil spending rose more than $5,000 in constant dollars between 2002 and 2023. At the measured rate of return, that should have produced gains visible from orbit. Scores fell. The honest reading is not that money does nothing but that this particular institutional design converts money into something other than learning at a spectacular rate. The next section covers what it converts it into.
What the Money Buys
Follow the dollars and the mystery of flat scores amid tripled spending resolves quickly. The money did not go to classrooms. It went to headcount and benefits.
Benjamin Scafidi, an economist at Kennesaw State University, assembled the federal staffing data going back to mid-century. Between 1950 and 2015, public school enrollment roughly doubled. The teaching force grew about 243 percent, which cut class sizes substantially, an intervention with modest measured effects. Non-teaching staff, the administrators, coordinators, specialists, aides, and central-office personnel, grew by more than 700 percent. The pattern has continued in the current century: the Reason Foundation’s 2025 spending analysis notes that states keep adding non-teaching staff even in states where enrollment is falling. American schools now employ roughly one non-teacher for every teacher. No other developed country runs its schools this way.
The benefits picture is starker. Between 2002 and 2023, inflation-adjusted spending on employee benefits, dominated by pension obligations and health insurance, rose 81.1 percent per student, from $2,221 to $4,022. Salary spending rose 7.7 percent. For every new dollar that reached an employee as salary over those two decades, $3.27 went to benefits, most conspicuously to paying down pension debt for people who no longer work in any school. In New Jersey, benefit costs alone run $8,333 per student per year. Teachers experienced this arithmetic directly: their real salaries fell 6.1 percent over a period when total real spending per student rose by more than a third. The system took in vastly more money and delivered less of it to the person standing in front of the children. That is not a funding shortage. That is an allocation decision, made year after year by identifiable school boards, superintendents, union negotiators, and state legislators, each of whom found it easier to promise deferred compensation and hire another coordinator than to raise a teacher’s paycheck or a test score.
Taxpayers cannot correct what they cannot see, and they do not see it. EdChoice’s national polling has found for years that Americans, including parents and teachers, estimate per-pupil spending at roughly a third of its actual level. Told the real figure, support for spending increases drops sharply. A monopoly funded through property tax assessments and state formulas, rather than through prices paid by willing customers, faces no mechanism that forces this information into the open. The district’s incentive is to describe itself as underfunded in perpetuity, and the description works because the audience has no receipt.
The Skills That Never Arrive
Set aside the test scores and ask a simpler question: after 13 years and 2,300 hours of mathematics instruction, what can the graduate actually do?
The TIAA Institute and the Global Financial Literacy Excellence Center have run the same personal finance assessment on American adults every year since 2017. The result barely moves: adults answer roughly half the questions correctly, and the youngest cohorts, the most recently schooled, score worst. These are not exotic questions. They cover interest, inflation, and risk, the arithmetic of being an adult in a market economy. The Federal Reserve’s household surveys have repeatedly found that a large share of adults could not cover a modest emergency expense without borrowing. The school system requires 4 years of English and typically 3 to 4 of math for a diploma, and in most states, until a recent wave of legislation, required no demonstration that a graduate could read a pay stub, compare two loan offers, or understand what compound interest does to a credit card balance. About half the states have now passed standalone personal finance course mandates, most of them since 2021. It took the institution 170 years to require the single most universally applicable body of quantitative knowledge it could teach, and it acted only after state legislators, responding to public embarrassment, ordered it to.
Civic knowledge follows the same pattern, and here the failure is structural rather than incidental, because civics is the subject the common school was explicitly created to teach. The Annenberg Public Policy Center’s annual Constitution Day survey found in 2017 that 26 percent of American adults could name all three branches of government, and more than a third could not name a single First Amendment right. Recent years have improved to roughly two-thirds naming the branches, which is progress measured against a floor. The NAEP civics assessment for 8th graders, last given in 2022, found 22 percent of students at or above Proficient, and the 2022 results were the first statistically significant decline in the assessment’s history. The institution that justifies its compulsory character on the grounds that democracy requires informed citizens produces a citizenry in which a third of adults cannot pass the test given to immigrants seeking naturalization. The naturalization applicants, studying on their own time with a booklet, pass at rates above 90 percent.
Then there is work itself. For most of the 20th century, American high schools maintained serious vocational programs: machine shops, auto shops, welding bays, drafting rooms, home economics kitchens that taught cooking and budgeting. Beginning in the 1980s, districts dismantled them, partly to chase the college-for-all consensus, partly because shop equipment and liability insurance cost money that coordinators do not. Enrollment in traditional vocational coursework fell steadily for three decades. The results are now legible in the labor market. Employers and trade associations report persistent shortages of electricians, welders, machinists, plumbers, and HVAC technicians, occupations with six-figure ceilings and no college debt, while the system routes a majority of graduates toward universities from which roughly 40 percent will not emerge with a degree in 6 years, though they will emerge with the debt. The 2024 NAEP data close the loop: fewer than half of seniors are academically prepared for the college coursework the system spent 13 years pointing them toward.
The defense here is that schools cannot teach everything and must prioritize academic fundamentals. It would be a strong defense if the fundamentals were being delivered. A system producing record shares of students below Basic in reading cannot claim it sacrificed financial literacy, civics, and trade skills on the altar of reading. It sacrificed them and did not get the reading either. The time went somewhere. The next section is about where.
The Compliance Curriculum
Every school teaches two curricula. The first is written down: fractions, the Constitution, photosynthesis. The second is embedded in the structure of the day, and it is taught far more effectively because it is practiced 7 hours a day for 13 years, with perfect consistency, by every adult in the building.
Consider what the structure itself instructs. A bell rings and 30 children stop what they are doing, whether or not it was worth doing, and move to the next room, where a different adult directs a different activity for exactly 47 minutes. Interest in a subject is irrelevant to how long one may pursue it. Permission is required to speak, to stand, and in most schools to use the bathroom. Work is assigned, not chosen; evaluated by authority, not by results in the world; and rewarded with tokens whose value exists only inside the institution. The child who complies smoothly is called a good student. The child who asks why is called a management problem. None of this is hidden and none of it is accidental. It is the Prussian architecture doing what it was engineered to do, and it operates regardless of the intentions of the teacher standing inside it.
John Taylor Gatto spent 30 years teaching in Manhattan public schools and was named New York State Teacher of the Year in 1991. In his acceptance-era writings, later collected in Dumbing Us Down, he cataloged what he called the real lessons of his profession: confusion, class position, indifference, emotional and intellectual dependency, provisional self-esteem, and constant surveillance. His claim was not that teachers intend these lessons but that the bell schedule, the grading apparatus, and the compulsion teach them no matter what the teacher intends. Gatto was an insider describing the machine from within, and the institution’s response, then and since, was not to dispute his description but to ignore it.
The empirical record on classroom practice supports the description. John Goodlad’s A Place Called School, published in 1984 and still the largest observational study of American classrooms ever conducted, examined more than 1,000 classrooms across the country. His researchers found instruction overwhelmingly dominated by teacher talk, worksheets, and recall. Barely 1 percent of instructional time involved questions that required students to reason toward an answer not already supplied. Two decades later, the accountability regime of No Child Left Behind and its successors made the pattern worse by attaching funding and careers to standardized test results, which reward precisely the kind of instruction Goodlad documented. Teachers report, in survey after survey, that test preparation crowds out open-ended work. The system now advertises “critical thinking” in every mission statement while running an instructional regime that Goodlad’s methodology would score about the way it did in 1984. The 2024 NAEP reading results, in which the skills lost fastest are exactly the interpretive and inferential ones, suggest the mission statements are losing.
The standard rejoinder is socialization: whatever the academic record, school is where children learn to function among others. The claim deserves scrutiny rather than deference. The school’s social environment, 25 to 30 same-aged children supervised by one adult, exists nowhere else in human life. No workplace, family, congregation, team, or neighborhood segregates people into single-year age cohorts. It is an artifact of batch processing, not a preparation for society. The research on children educated outside the system, summarized across decades by scholars on both sides of the homeschooling debate, has consistently failed to find the social deficits the objection predicts; measured social skills for home-educated children run equal to or above schooled peers, in part because their daily life includes adults, younger children, and actual community institutions rather than a single-age holding pattern. Meanwhile the socialization the conventional school demonstrably does provide, documented in the federal government’s own school crime and safety reports, includes bullying rates that touch roughly 1 in 5 students. An institution cannot claim credit for teaching children to get along by pointing at the room where a fifth of them are being tormented.
There is one more data series worth reading as testimony, because it comes from the students themselves. Chronic absenteeism, defined as missing 10 percent or more of the school year, ran around 28 percent nationally in the most recent federal counts, roughly double the pre-pandemic rate and improving only slowly. Nearly a third of 12th graders taking the 2024 NAEP reported missing 3 or more days in the prior month. The conventional reading treats this as a discipline problem or a lingering pandemic habit, something to be fixed with robocalls and truancy referrals. There is a plainer reading. A century of compulsion trained everyone to interpret attendance as a given rather than a signal, but attendance is a signal, the only vote a student gets. When 1 child in 4 declines, at real cost to themselves, to show up to a free service they are legally required to attend, they are reporting something about the value of the hours on offer. Institutions that people flee when the enforcement weakens have received their evaluation. The 13 years of the standard sequence contain roughly 14,000 instructional hours, the largest single claim on an American childhood, and the system’s own assessments say what the median child holds at the end of it: reading below the level of 1992, mathematics below every cohort ever measured at grade 12, no trade, no financial competence, and a durable lesson in sitting still. The hours were not free. They were simply billed to someone who could not refuse them.
The Monopoly Problem
The failures cataloged above are usually treated as separate scandals requiring separate reforms: a curriculum fix here, a staffing audit there, a new accountability dashboard everywhere. Treating them separately misses the common cause. They are the predictable outputs of a specific institutional design, and the design has three load-bearing features.
The first is compelled attendance. Every state requires children to be in school, and for the roughly 87 percent of families who cannot pay private tuition on top of the school taxes they already pay, the assigned district school is the school. The customer cannot leave. An institution whose customers cannot leave does not need to satisfy them, and over time it stops organizing itself around them. This is not a slur on the people inside the institution. It is the oldest finding in the study of monopoly, and it applies to school districts exactly as it applies to cable companies and departments of motor vehicles.
The second is funding detached from performance. A district’s revenue arrives through property tax levies and state formulas keyed to enrollment counts. It does not vary with whether children learn to read. A district where 33 percent of 8th graders read below Basic receives its full appropriation; frequently, under compensatory funding formulas, it receives more. Whatever one thinks of the equity logic, the incentive structure is unambiguous: failure is revenue-neutral or revenue-positive, and no institution improves under that arrangement. Sixty years of evidence backs this up. The Coleman Report, commissioned by Congress and published in 1966 as the largest social science study ever conducted to that point, found that school resources explained little of the variation in student achievement. Eric Hanushek at Stanford has spent a career updating the finding: across hundreds of studies, the relationship between spending levels and outcomes within the American institutional structure is weak and inconsistent. The structure, not the sum, is the variable that matters.
The third is governance by capture. School districts are nominally controlled by elected boards, and school board elections are the emptiest ritual in American democracy. Turnout routinely runs from the single digits to the low teens, and the most motivated voters are the system’s employees and vendors. The result is textbook concentrated-benefits, diffuse-costs politics: the people who draw salaries and contracts from the district organize; the taxpayers and children who bear the costs do not. Collective bargaining agreements in most large districts still contain seniority-based layoff rules that federal and state analyses have repeatedly shown remove effective junior teachers ahead of ineffective senior ones, and dismissal procedures under which removing a demonstrably failing teacher takes years and six-figure legal costs, so administrators mostly stop trying. The infamous New York City “rubber rooms,” where teachers deemed unfit for classrooms drew full salaries for years while cases idled, were a symptom the district spent a decade claiming to have fixed and never quite did. Again, the point is not that individuals are wicked. The point is that the board members, union officials, and administrators who negotiated and renewed these arrangements were responding rationally to a design in which the child’s outcome carries no institutional weight.
Put the three features together and the system’s observed behavior stops being mysterious. Staffing grows fastest in the categories furthest from instruction because hiring coordinators expands administrative careers and satisfies organized constituencies, while raising reading scores satisfies no one with power. Benefits promises balloon because they defer costs past the current board’s tenure. Curricula chase fashions because fashions come with grants and conference invitations, while the phonics-based reading instruction the cognitive science has endorsed since the 1960s took until the 2020s, and a Mississippi embarrassment, to gain ground. The institution is not malfunctioning. It is functioning precisely as its incentives specify. That is why 60 years of reform, every wave of it, from new math to open classrooms to standards to small schools to accountability to Common Core, has washed over the structure and left it unchanged. Reform inside this design amounts to redecorating a building with a cracked foundation. The foundation is the interesting part.
What Came Before
The strongest argument for the current system is an implied history: before government schools, ordinary children went uneducated, and only the state’s intervention brought literacy to the masses. The record does not support this, and the record matters, because it establishes that communities are capable of the thing the state claims only it can do.
Literacy in the early American republic, measured imperfectly but consistently through signature rates on wills, deeds, and military records, was among the highest in the world before a single compulsory attendance law existed. Kenneth Lockridge’s studies of New England put male signature literacy near 90 percent by 1800, and the popular press of the era assumed a mass reading public: Thomas Paine’s Common Sense sold on the order of 100,000 copies within months in a nation of 2.5 million free inhabitants. These figures come with honest caveats. Signature literacy overstates deep reading skill, the record is thinnest for women, and it excludes the enslaved, whom Southern states barred from literacy by statute, a reminder that the government of that era was quite capable of preventing education when it chose to. But the direction of the evidence is clear and largely uncontested among historians of literacy: the compulsory state school arrived in a society that was already substantially literate, having gotten that way through an unplanned weave of dame schools, church schools, subscription schools, tutors, apprenticeships, and parents teaching at the kitchen table.
The institutional forms this weave produced were durable and, in places, spectacular. The one-room district schools of the rural North were funded and governed at the level of the neighborhood; the families who paid for the teacher hired the teacher, boarded the teacher, and fired the teacher. When Catholic immigrants found the ostensibly common schools of the 19th century saturated with Protestant instruction, they built their own system with nickels and dimes collected at parish doors: by 1965, the Catholic school network enrolled 5.2 million children, ran on a fraction of public per-pupil costs, and in later decades produced graduation and achievement results for poor urban children that public districts in the same neighborhoods could not match. The country’s mutual aid infrastructure did the same in every other domain of welfare. David Beito’s history of fraternal societies documents that by 1920 roughly 1 in 3 adult American men belonged to lodges that provided sick pay, funeral benefits, orphan care, and, in the case of societies like the Security Benefit Association, entire hospitals and schools, all financed by member dues from working-class households.
The state school did not out-teach this ecosystem. It out-lawed it, or more precisely, it taxed the ecosystem’s participants to fund a free competitor and then compelled attendance. The clearest episode came in Oregon in 1922, when a ballot initiative backed enthusiastically by the Ku Klux Klan, which viewed Catholic schools as a menace to Americanism, required every child to attend public school. The Supreme Court struck the law down in Pierce v. Society of Sisters in 1925, writing that the child is not the mere creature of the state. The sentence survives as constitutional doctrine. The funding arrangement Pierce left intact, in which every family pays for the government school whether or not any child attends it, accomplished most of what the Oregon initiative intended, just more slowly. An independent school that must charge tuition to families already taxed for its competitor operates at a permanent, deliberate disadvantage. The Catholic system’s long decline from 5.2 million students to under 1.7 million tracks that arithmetic as much as any demographic trend.
None of this is an argument for reconstructing 1850. Dame schools are not coming back, and no one should want the parts of that world that deserved to die. The historical point is narrower and load-bearing: voluntary institutions, funded and governed by the people who use them, demonstrably built mass literacy once, under material conditions far worse than today’s. The claim that education collapses without a government operator is not a finding. It is an assumption, and the one period of American history that tested it points the other way.
What Is Growing Back
The historical ecosystem is not merely a memory. Its modern descendants are the fastest-growing sector of American education, and the growth is measurable.
Start with homeschooling, which is no longer what the word suggests. The Johns Hopkins Homeschool Research Lab, which compiles the state-registered counts, found homeschooling grew 4.9 percent nationally in the 2024-2025 school year, nearly triple the pre-pandemic growth rate, with 80 percent of reporting states showing increases and 36 percent of them, including Georgia, Minnesota, Nebraska, Ohio, and Pennsylvania, recording their highest homeschool enrollments ever, higher than the pandemic peak. South Carolina grew 21.5 percent in a single year. These are minimum counts, capturing only families who file with the state. The pandemic bump did not recede; it compounded. Whatever families saw through the window of remote schooling in 2020, several million of them declined to unsee it.
The demographic composition has changed as fast as the numbers. Survey research compiled by EdChoice and federal household data show the homeschooling population now includes substantial Black, Hispanic, and Asian participation and spans income levels and politics; urban Black families running co-ops in Birmingham are as characteristic of the sector as anyone’s stereotype from 1995. The motivations have shifted from predominantly religious to a broad mix: safety, special needs, academic customization, and plain dissatisfaction with the assigned district.
Around the homeschool core, an institutional layer is assembling itself, and it looks strikingly like the 19th-century weave rebuilt with modern logistics. Microschools, typically 5 to 15 students in a home, church annex, or storefront, now number in the tens of thousands nationally; the National Microschooling Center’s 2025 analysis found more than half operate as homeschool cooperatives and about 30 percent as small private schools. Hybrid academies run 2 or 3 classroom days a week with home study between. Co-ops share a chemistry teacher across 12 families. Churches that sat empty on weekdays are filling with pods. Retired machinists are teaching shop again, for pay, to teenagers whose district sold the lathes decades ago. None of this required a master plan, which is rather the point. It required demand, legality, and adults willing to take direct responsibility for children in their own community.
Public money is now following children into this ecosystem, which introduces both fuel and hazard. As of 2025, 18 to 19 states operate education savings account programs serving more than 1.2 million students with over $13 billion annually, 13 of them universal. ESA participation jumped from roughly 40,000 students in 2022 to nearly 500,000 by 2025 before passing the million mark. In Arkansas, nearly 10 percent of all K-12 students joined the state’s program within 3 years, about 40 percent of them homeschoolers, and more than 100 microschools registered with the state. Iowa saw new private school openings jump from 2 in the year before its ESA to 24 in year two. The supply side, in other words, responds within months when the funding monopoly cracks.
Honesty requires the hazards be stated as plainly as the growth. The academic evidence on voucher-style programs is genuinely mixed: several state programs, notably in Louisiana and early Ohio, produced negative test-score effects for participants, though a 2025 Urban Institute study found Ohio voucher students markedly more likely to enroll in and complete college, which suggests test scores capture the effects incompletely. Some ESA programs have suffered visible spending abuses, and Utah’s legislature had to tighten allowable expenses after year one. Tennessee’s comptroller found most early ESA users were leaving average-or-better public schools, not failing ones. And a structural risk shadows the whole mechanism: public money has never traveled anywhere without regulation following it, and a homeschool sector remade into a state-administered voucher clientele could find itself re-absorbed into the compliance apparatus it left. Watson’s research at Johns Hopkins has so far found no evidence that choice funding has increased homeschool regulation, and finds regulation decreasing over time, but the incentive is permanent and the vigilance must be too. ESAs are best understood not as the destination but as a transitional mechanism: a way of returning tax money already extracted from families while the deeper questions, who funds, who governs, who is responsible, get settled where they belong.
The apprenticeship model supplies the other half of the picture, and here the useful evidence is foreign. Switzerland routes roughly two-thirds of its teenagers into vocational education and training, 3-to-4-year paid apprenticeships in some 230 occupations, from precision machining to banking, run jointly by employers and industry associations with the state providing certification. Swiss youth unemployment is persistently among the lowest in the developed world, the country’s median wages are among the highest, and the apprenticeship track carries no stigma because it demonstrably leads somewhere. The system’s core feature is the one that matters for this argument: the training is designed and largely delivered by the people who will employ its graduates, which means the feedback loop between instruction and usefulness is direct. American employers spent the last 40 years complaining that schools send them graduates who cannot do anything, while accepting an institutional arrangement in which employers have no role in instruction. The Swiss arrangement, and the fast-growing American registered apprenticeship programs now imitating it, closes the loop by giving the work of preparation back to the people who understand the work.
Who Does What
Strip away the branding and every education system answers three questions: who pays, who governs, and who is accountable when a child cannot read. The current American answer is: everyone pays, a captured bureaucracy governs, and no one is accountable. The alternative answer is older than the country and simpler: the people closest to the child pay what they can, govern what they fund, and answer for the results to neighbors who know their names.
The division of labor follows from asking what actually requires the state’s distinctive tool, which is coercion. Courts require it; someone must be able to compel testimony and enforce judgments. Physical security requires it. Contract enforcement requires it, and an education ecosystem of tutors, microschools, co-ops, and apprenticeships runs on contracts, so the state’s honest role in education is the unglamorous one it already performs everywhere else in the economy: enforce agreements, prosecute fraud, and prosecute abuse and neglect under laws that exist regardless of where a child learns. Teaching a child fractions requires no coercion at all. It requires an adult who knows fractions, a child, and time. An activity that requires no coercion is an activity the coercive institution has no comparative advantage in running, and 60 years of scoreboard confirm the theory.
What replaces the district is not a void; it is the layered structure already visible in the growth data, thickened. Families as the first governors of a child’s education, because they are the only party whose incentives are permanently aligned with the child. Cooperatives and congregations pooling to hire specialist teachers, as they already do. Employers and trade associations running credentialed apprenticeships, as the Swiss have proven at national scale. Mutual aid societies, the modern heirs of the lodges, underwriting tuition for members’ children and for the children of the neighborhood, as the parishes did for a century at a per-pupil cost the districts never approached. Philanthropy doing what the General Education Board did with its billions, ideally with more humility about molding anyone. And the existing physical plant, the buildings taxpayers already bought, transferring to community ownership as libraries, learning centers, and co-op campuses rather than sitting as monuments to an enrollment that is leaving anyway.
The equity objection arrives here, and it is the serious one: markets serve the median, and the poor, the disabled, and the unlucky get left. Three answers, in ascending order of importance. First, the objection describes the present system, not a hypothetical: the 2024 NAEP collapse is concentrated entirely among poor children, the current financing model ties school quality to neighborhood property wealth, which is the most regressive education funding design ever implemented at scale, and the affluent already have full school choice, exercised through the real estate market. Second, the historical institutions of voluntary provision were not boutiques for the rich; the lodges and parish schools were built by and for the working poor, because the working poor are the people for whom institutional failure is not survivable and who therefore build institutions that work. Third, to the extent a transition needs a financial bridge for families whose taxes built the old system, that is what the ESA mechanism, whatever its imperfections, already is: 1.2 million students wide and growing. The disabled-student question is the hardest, and the early evidence runs opposite to the fear: families of students with disabilities have proven among the heaviest and most satisfied users of choice programs, largely because the legally guaranteed services of the district system so often exist on paper, enforced only by parents with the resources to litigate.
The rural objection is nearly as common and weaker. The claim is that thin populations cannot support alternatives. Rural America is where the one-room community school was invented, where homeschool networks are currently densest, and where the consolidated district, with its 45-minute bus rides to a distant campus run from a county office, replaced a school the town actually owned. Low density is an argument against the big-box district model, not for it.
The Wind-Down
Institutions that fail slowly should be replaced deliberately, and the case against the government school is not a case for chaos on a Tuesday. Roughly 45 million children are inside the system. The transition has to be honest about them, about the 3.2 million teachers, and about the towns where the district is the largest employer. A serious sequence looks less like demolition and more like a controlled transfer of responsibility, and most of its components are already in motion in one state or another.
The first phase is already underway and requires only that legislatures keep doing what 19 of them have done: let the funding follow the child, universally, with the money flowing to families rather than to institutions. The design details matter more than the slogan. Programs that pay tuition invoices directly to schools recreate the provider-capture problem one layer out; programs that fund family accounts with audited but broad allowable uses, the Arkansas structure, feed the co-op and microschool layer where the genuine institutional novelty lives. The fraud that surfaces, and some will, should be prosecuted as fraud, by prosecutors, rather than answered with a compliance regime that slowly rebuilds the district’s central office inside the state treasurer’s software. Utah’s course correction after year one, tightening allowable uses without strangling the program, is roughly the right reflex.
The second phase is legal rather than fiscal: dismantle the compulsory attendance statutes’ equation of education with enrollment. Every state already recognizes homeschooling, which concedes the principle that instruction, not seat time in a licensed building, is the obligation. The remaining work is to finish the thought: the state’s legitimate interest, to the extent it has one, is that children not be neglected, and neglect is already illegal under laws that have nothing to do with school. A child reading three grades ahead in a co-op that meets in a fire hall is not a truant, and the statutes treating her as one are the last enforcement teeth of the 1852 settlement.
The third phase is the slow one: the orderly retirement of the district as an operator. Enrollment is doing part of this on its own; public school enrollment has been falling since 2020 through some combination of demographics and exit, and districts from San Francisco to rural Kansas are already closing buildings. The choice is whether the wind-down happens the way public institutions usually shrink, with the cuts landing on classrooms while the central office survives to the last invoice, or whether it is managed with the priorities reversed. Managed well, buildings transfer to community trusts and co-op networks at their depreciated value rather than selling to developers; pension obligations, which are contracts, get honored out of general revenue rather than loaded onto a shrinking pupil count; and the district’s genuine social functions, the meals, the screenings, the gym that is also the town’s civic hall, transfer to the churches, granges, clubs, and mutual societies that ran such things before and are, in thousands of towns, still standing next door.
The teachers deserve a specific word, because the standard framing casts them as the losers of any transition and the standard framing has it backwards. The 3.2 million teachers are the ecosystem’s scarcest asset, and the emerging sector is already bidding for them: microschool networks recruit experienced teachers as founders, not employees, and a teacher running a 12-student microschool at typical rates can match or beat district pay while shedding the compliance load that surveys consistently identify as the top driver of the profession’s attrition. The current system pays a teacher like a mid-grade clerk, supervises her like a suspect, and consumes a growing share of her compensation in benefits she may never collect. A community-funded ecosystem pays her like what she is, which is a skilled professional with a client list. Some teachers will prefer the district until the last bell. Many are already leaving for the other side of the ledger, and the 2024-2025 microschool founding data suggest the flow is accelerating.
What must be watched, throughout, is the oldest failure mode of American reform: the new institutions petitioning to have their competitors regulated away, and the state obliging. The parochial schools of 1900 did not lobby to close the common schools; the common schools’ partisans lobbied, in Oregon successfully for 3 years, to close the parochial schools. The temptation will run the other way this time, and it should be refused on the same principle. The argument of this piece is not that families must exit the district school. It is that the district school must earn its enrollment the way every other institution in a free society earns its members, and that the tax apparatus should stop pre-deciding the question on the district’s behalf. A district school that can fill its classrooms with willing families paying with portable funds has proven something no district can currently prove. Some will. The Catholic schools’ partisans in 1920 would have settled for exactly this deal.
The Affirmative Case
It is worth ending where the founders of the system ended, with a statement of what education is for, because the disagreement was never really about test scores.
The Prussian answer, imported in 1843 and never amended, is that education exists to produce the citizens the state requires: punctual, sorted, credentialed, and governable. The system built on that answer has performed to specification for a century and is now failing even the specification, at $1 trillion a year, with a third of its 8th graders unable to follow the plot of a story and half its graduating seniors below Basic in mathematics. The people responsible for this are not abstract forces. They are the specific legislatures that renewed the funding formulas, the specific boards that signed the staffing plans and the benefit deferrals, and the specific administrative class that grew 7-fold while reading scores went nowhere. They were responding rationally to a design. The design is the defect.
The older answer, the one legible in the signature rates of 1800, the parish schools of 1900, the lodge halls of 1920, and the co-op schedules of 2026, is that education is a duty owed by particular adults to particular children, discharged in communities small enough that failure has a face. That answer built mass literacy without a single attendance officer. Its descendants are currently the only part of American education whose enrollment, institution-formation, and parent satisfaction are all rising at once, and they are doing it with a fraction of the money, in church basements and storefronts, faster than anyone is planning it.
The constructive project, then, is not to burn anything down. Enrollment trends are handling the demolition without assistance. The project is to build the receiving structure: join or found the co-op, fund the neighborhood scholarship, take the apprentice, put the trade back in the fire hall, and treat the education of the town’s children as the town’s work rather than as a service procured from a distant monopoly and complained about at intervals. Communities did this once under harder conditions with fewer tools. The evidence that they are doing it again is in the state enrollment files, growing at 3 times the old rate, in 36 states at record highs. What remains is for the law and the money to get out of the way of people who have already started.
References
National Assessment Governing Board, “The Nation’s Report Card Shows Declines in Reading, Some Progress in 4th Grade Math,” January 2025, and “Nation’s Report Card Shows Declines in 8th-Grade Science and 12th-Grade Math and Reading,” September 2025 (nagb.gov)
National Center for Education Statistics, NAEP Long-Term Trend Assessments; NAEP Civics 2022; Digest of Education Statistics, Tables 236.55-236.75
U.S. Census Bureau, Annual Survey of School System Finances, FY2024 release, May 2026
Reason Foundation, “K-12 Education Spending Spotlight 2025,” November 2025
EdChoice, “How Much Do Public Schools Spend?”, May 2026; EdChoice Share and national polling on spending perception
Hechinger Report, “A dismal report card in math and reading,” 2025 (analysis of bottom-decile NAEP declines and ESSER spending)
Education Recovery Scorecard (Harvard/Stanford), Dewey et al., 2025
OECD / National Center for Education Statistics, Program for the International Assessment of Adult Competencies (PIAAC), 2023 U.S. National Results, December 2024
TIAA Institute-GFLEC Personal Finance (P-Fin) Index, annual reports 2017-2025
Annenberg Public Policy Center, Constitution Day Civics Survey, 2017-2024
James S. Coleman et al., Equality of Educational Opportunity, U.S. Office of Education, 1966
Eric Hanushek, research on school expenditure and student performance, Hoover Institution / Stanford
Benjamin Scafidi, The School Staffing Surge, EdChoice (analysis of NCES staffing data, 1950-2015)
Raymond Callahan, Education and the Cult of Efficiency, University of Chicago Press, 1962
Ellwood P. Cubberley, Public School Administration, Houghton Mifflin, 1916
Frederick T. Gates, “The Country School of To-Morrow,” General Education Board Occasional Papers No. 1, 1913
Horace Mann, Seventh Annual Report to the Massachusetts Board of Education, 1843
Johann Gottlieb Fichte, Addresses to the German Nation, 1808
John I. Goodlad, A Place Called School, McGraw-Hill, 1984
John Taylor Gatto, Dumbing Us Down, New Society Publishers, 1992
Kenneth Lockridge, Literacy in Colonial New England, Norton, 1974
David Beito, From Mutual Aid to the Welfare State: Fraternal Societies and Social Services, 1890-1967, University of North Carolina Press, 2000
National Catholic Educational Association, historical enrollment data
Pierce v. Society of Sisters, 268 U.S. 510 (1925)
Johns Hopkins Institute for Education Policy, Homeschool Research Lab, “Homeschool Growth: 2024-2025,” Angela R. Watson, November 2025
National Microschooling Center, 2025 sector analysis
FutureEd, “Directional Signals: A New Analysis of the Evolving Private School Choice Landscape,” November 2025
Ballotpedia, “School choice in the United States,” 2025-2026 updates
Tennessee Comptroller of the Treasury, OREA evaluation of the ESA program, January 2026
Urban Institute, Ohio voucher college enrollment study, 2025
Swiss State Secretariat for Education, Research and Innovation, vocational education and training (VET) system data
National Center for Education Statistics, Report on Indicators of School Crime and Safety