Not long ago, a child who preferred reading to recess, memorized train schedules instead of baseball stats, and found eye contact uncomfortable and small talk pointless was considered quirky. Perhaps shy. Perhaps a little odd. His parents might have worried, or they might not have. His teachers might have flagged him for extra attention, or simply seated him near the window and let him be. What almost no one would have done was send him to a specialist who would diagnose him with a neurological disorder and refer him for 30 hours a week of publicly funded behavioral therapy.

Autism was once a rare and severe diagnosis reserved for people who could not speak, could not care for themselves, and would never live independently. The American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM) originally reflected that severity. Over time, clinicians rightly recognized that milder presentations, such as high-functioning individuals with Asperger’s syndrome, had been overlooked, and the criteria expanded to include them. But in 2013, the DSM’s fifth edition went further, merging several previously distinct categories into a single “autism spectrum disorder.” The revision made clinical sense for researchers seeking a unified framework. Its practical consequence was to stretch the word “autism” until it could describe both the nonverbal child who needs lifelong care and the socially awkward teenager whose difficulties amount to a preference for solitude.

With the boundary between disorder and ordinary personality blurred, it became increasingly common for children with social difficulties, particularly those who experienced social isolation during the Covid-19 pandemic, to be labeled “on the spectrum.” The Centers for Disease Control and Prevention now reports that one in 31 American children carries the diagnosis, nearly five times the rate recorded in 2000.

This trend looks like an epidemic, and both left and right have treated it as one. But autism, one of the most genetically determined conditions in psychiatry, is up to 90 percent heritable. A heritable trait does not quintuple in a generation. Meanwhile, according to the same CDC data, the rate of moderate and severe autism has actually declined thanks to early intervention and applied behavior analysis (ABA) therapy, which teaches children social skills. What has exploded is the mild end of the spectrum — the diagnoses given to children who speak fluently, attend regular classrooms, and would, a generation ago, have simply been called introverted or eccentric.

We are not discovering more disabled children. We are sweeping ordinary variation into a clinical category that was never designed to hold it.

If overdiagnosis were merely a labeling problem, it would be a concern for clinicians and parents. But it is far more than that. In the American health-care system, an autism diagnosis is a key that unlocks substantial public funding, and once that door opens, it is extraordinarily difficult to close.

Federal law requires states to cover all medically necessary services for children on Medicaid, and a 2014 federal clarification made ABA therapy, originally developed for children with severe developmental delays, effectively mandatory for any child with an autism diagnosis. ABA is billed by the hour, with no hard cap on how many hours a provider can prescribe. Because the federal government reimburses states for 50 to 83 cents of every Medicaid dollar, a state approving a new claim spends as little as 17 cents of its own money. Meanwhile, the workforce needed to convert diagnoses into billable services has expanded dramatically. The credentialed professionals who supervise ABA programs and enroll as Medicaid providers numbered roughly 12,000 nationwide in 2014; by 2025, that figure had reached 81,000, a nearly sevenfold increase that far outpaces the doubling of diagnoses over the same period. The consequences show up in state budgets. In “The Autism Therapy Gold Rush,” our recent Cato Institute analysis, we found that across eight states with publicly available data, combined Medicaid spending on autism therapy grew from $347 million to over $2.2 billion in recent years. The national total is certainly much larger, but most states do not report autism therapy as a separate line item. Where autism-specific spending data were available, spending far outpaced the already suspect national increase in autism prevalence reported by the CDC, to the tune of hundreds of millions of dollars per state.

Minnesota’s spending alone went from $671,000 in 2018 to $342.8 million by 2024. During those years, the number of Minnesota Medicaid autism-therapy recipients rose from 400 to more than 5,700, but per-recipient spending rose even more dramatically, from less than $1,700 to more than $60,000 annually. That surge is part of a pattern of social-services fraud that engulfed the state under Governor Tim Walz, from the $250 million Feeding Our Future scheme — in which providers fabricated meal counts to collect child-nutrition funds during the pandemic — to fraudulent daycare centers and sham housing programs. Federal prosecutors traced the autism-therapy fraud directly to the same networks: at least a dozen Feeding Our Future defendants also owned or were connected to autism centers.

But the problems of fraud and overspending extend far beyond one state. The federal inspector general has now audited Medicaid ABA payments in Colorado, Indiana, Maine, and Wisconsin. In each state, 100 percent of sampled claims contained at least one improper or potentially improper payment. Confirmed improper payments totaled $198 million, or 31 percent, of the $638 million reviewed. In multiple states, staff without credentials delivered sessions. Providers billed for children napping, eating snacks, watching movies, and riding water slides. One Colorado recipient was reimbursed for 200 hours of therapy in a single month, or more than eight hours per day, assuming a five-day workweek, the vast majority of which were undocumented. In some facilities, notes were signed before the session had even concluded. Session notes were photocopied across dates with identical text and signatures, and some files contained the wrong child’s name.

None of the four audited states had ever systematically reviewed its own autism-therapy claims before the federal government showed up. Billions of dollars had been moving through a system that operated entirely on trust.

Even where no fraud is present, the spending does not align with clinical evidence. The research supporting ABA therapy is strongest for young children with significant deficits in multiple areas of development, and industry guidelines recommend reserving intensive treatment for that population. Yet Indiana found that half of its Medicaid-funded ABA recipients were getting more than 25 hours of therapy per week, with a third exceeding 30. North Carolina reported that rising intensity of services bore no connection to the severity of children’s needs or to any change in clinical guidelines. Providers prescribed essentially the same regimen for every child who came through the door. The therapy built for the most profoundly disabled children was being dispensed at full dosage to a population that increasingly did not match that description.

No single villain explains that. The system’s architecture does. Providers deliver services and collect hourly payments. Families receive therapy with no out-of-pocket cost. States approve claims for which federal taxpayers cover most of the tab. Diagnosticians apply criteria broad enough to describe a substantial share of the child population. Everyone is acting rationally within the rules. The rules are the problem.

Indiana’s rate reform offers a proof of concept. Indiana’s autism-therapy budget climbed from $21 million in 2017 to $611 million in just six years, consuming 3.4 percent of the entire state Medicaid budget at its peak. But when the state finally replaced open-ended billing with standardized payments in 2024, its spending fell 27 percent without a corresponding reduction in mental health care. Distributing behavioral-health funding to states via block grants, or imposing per capita spending caps with flexibility in program design, would force officials to weigh autism therapy against competing priorities using their own revenues rather than a bottomless federal credit line. Requiring that the clinician who diagnoses a child be independent from the provider who bills for therapy would eliminate the most glaring conflict of interest in the system. Mandating periodic reassessments of medical necessity would prevent the ongoing allocation of hundreds of thousands of dollars on a child’s decade-old referral.

These are not radical proposals. They are the kind of basic accountability that any private insurer would have demanded years ago.

When the government subsidizes something, it gets more of it. But the autism spending crisis is also a story about what happens when a culture medicalizes ordinary human variation and then builds a payment system around the new label. The casual expansion of “on the spectrum” as a clinical diagnosis activates a financial machinery that nobody is tending. The families of genuinely disabled children, the ones for whom Medicaid was built, are the ones paying the price. Their resources are diluted, they face longer waitlists, and their children compete for clinical attention with a growing population whose needs bear little resemblance to theirs.

And the children on the other end of the misdiagnosis bear a cost, too. A child who is merely introverted or socially awkward but is told by clinicians that he has a neurological disorder comes to understand his own personality as a deficiency. Instead of learning that he can navigate the social world on his own terms, he learns that he is broken and needs professional management. That message, absorbed early enough, may do more lasting damage than the traits it was meant to correct.

Reversing the real autism crisis will require not just fiscal reform but a more honest conversation about what autism is, what it is not, and whom the system ought to serve. That conversation is overdue.