Higher Education Reform After the One Big Beautiful Bill Act

I. The original sin of federal higher-education finance

To understand why American higher education has drifted so far from the interests of the students it serves, it helps to begin with the statute that built the modern system. The Higher Education Act of 1965 (HEA) was one of the last pieces of Great Society legislation, signed by President Lyndon Johnson at his old teachers’ college in San Marcos, Texas, ten weeks after he signed Medicare into law. Johnson described the bill in the language he used for all of his domestic achievements as an act of generosity toward the disadvantaged and a down payment on the promise that a talented student from a poor family should not be kept out of college by the cost of tuition. “Higher education is no longer a luxury,” he told the crowd at San Marcos, “but a necessity.”

The HEA created the first federal grant program for undergraduate tuition, a federal guaranteed-loan program, and the architecture of federal aid to colleges and universities that, in its essential form, still governs the sector today.

What Johnson and his advisers built, however, was not a federal effort to finance the education of students. It was a federal effort to finance the budgets of institutions, delivered through students. That is an important distinction and the failure to grasp it is the source of nearly every higher-education policy problem the country has faced since.

Title IV of the HEA—the title that governs federal student aid, and the reason a term like “Title IV eligibility” has become the central gatekeeping concept in the sector—was drafted to move taxpayer dollars from the federal Treasury to colleges as efficiently as possible, with the student acting as the courier. The student was the nominal beneficiary of the grant or the loan, but the money passed through his or her hands without stopping. Whether the student could repay the loan, whether the degree he or she obtained would be worth anything in the labor market, and whether the college had any stake in the answer to either question were never addressed in the original statute because none of these questions were the ones the bill’s drafters were asking.

The reasons for that design were both philosophical and practical. Philosophically, the drafters of the HEA shared the widespread mid-century assumption that more education was self-evidently good, that the returns to a bachelor’s degree were high and stable across institutions and programs, and that the federal government’s role was to remove cost as a barrier rather than to regulate quality. Practically, there was no informational infrastructure in 1965 that would have allowed Congress to tie federal funding to student outcomes. The College Scorecard was 50 years away. The American Community Survey did not exist in its current form. The federal government did not collect earnings data on former students, could not have linked that data to the institutions they attended, and had no administrative capacity to run the kind of accountability system that outcomes-based funding would have required. The choice, as Congress saw it, was between writing a statute that financed colleges generously and writing no statute at all.

A third reason, rarely acknowledged but more important than the first two, was political. American higher education in 1965 was a collection of state universities, private colleges, and historically Black institutions. Many had deep local constituencies and powerful alumni networks. Others were institutionally disadvantaged. Any federal program that tied funding to outcomes would have meant, sooner or later, withdrawing funding from institutions whose outcomes were bad—and in 1965, as in 2026, there was no appetite in Congress for a bill that might close a regional college in a member’s district.

Johnson’s approach was the politically sustainable one: a federal program that sent money to every accredited institution that wanted it, evaluated only by the crudest quality screens, and left the hard questions of value for some future Congress to sort out. It worked, in the sense that the HEA passed and has been reauthorized nine times since. But the hard questions were never sorted out, and the accumulating costs of that deferral are now larger than the original program.

The central defect of the HEA’s design—the decoupling of federal funding from student success—metastasized in three stages. In the 1970s, Congress expanded the grant and loan programs without creating any mechanism to measure whether the subsidies were translating into earnings gains for the students receiving them. In 1980, Congress created the Parent PLUS program, which extended federal credit to parents on behalf of their dependent children with essentially no underwriting, and the Grad PLUS program followed in 2005 on the same terms.

Both of these programs allowed borrowing up to the full cost of attendance as set by the institution itself, which meant that a college could raise its sticker price and be confident that federally sanctioned lending would expand to meet the new price. In 2010, as part of an accounting gimmick to help fund the Affordable Care Act, the HEA’s loan guarantees were taken over entirely by the federal government. From then on, borrowing became the dominant source of tuition finance for graduate programs and a substantial share of undergraduate programs, and tuition at the institutions most dependent on federal loans rose faster than at any other category of consumer spending in the American economy.

In 2016, the Government Accountability Office estimated the taxpayer losses at $108 billion and growing. By 2024, the total outstanding stock of federal student debt exceeded $1.6 trillion, roughly one-quarter of which FREOPP’s research has estimated is owed by borrowers whose degrees will not generate enough lifetime earnings to justify the cost.

This is the backdrop against which the One Big Beautiful Bill Act (OBBBA)was written. OBBBA is the first serious federal legislative attempt to reconnect federal funding to student success since the HEA was signed sixty years ago. It is not complete. It is not, by the standards of the reform literature, particularly ambitious. But it represents the first occasion in two generations on which Congress has accepted the premise that federal student aid should be conditioned on whether the programs receiving the aid actually work. The purpose of this paper is to describe what OBBBA accomplished, explain what it left undone, and propose a legislative and regulatory agenda for the next federal bill that finishes the job.

II. What Congress finished, and what it left undone

On July 4, 2025, President Trump signed OBBBA into law. In the middle of a reconciliation bill best remembered for its tax and border provisions, Title VIII of OBBBA quietly enacted the largest retrenchment of federal student lending since 1965. To understand why OBBBA’s higher-education provisions matter, it helps first to understand the problems Congress was trying to solve.

Unlimited graduate loans: The principal driver of education inflation

The first problem was the uncapped graduate loan. Under the pre-OBBBA system, two federal loan programs—Grad PLUS for graduate and professional students, and Parent PLUS for the parents of dependent undergraduates—allowed borrowers to take on federal debt up to the full cost of attendance at the institution they chose to attend. That phrase “full cost of attendance” is the critical one. The cost of attendance at a graduate program is set by the institution itself, and it includes tuition, fees, books, and a generous living allowance. In practice, Grad PLUS and Parent PLUS were effectively unlimited: whatever the institution chose to charge, the federal government would lend.

FREOPP’s 2024 analysis of the College Cost Reduction Act documented this dynamic in detail, showing how the uncapped structure of Grad PLUS in particular gave graduate programs a powerful incentive to raise tuition. Any increase was immediately absorbed by expanded federal lending rather than resisted by price-sensitive students. The result was a 30-year tuition spiral in graduate programs, especially in fields like law, medicine, and the master’s-level programs in education and the arts, where the cost of the credential decoupled almost entirely from the earnings it produced.

No quality or accountability tests for federal dollars

The second problem was the absence of any meaningful accountability test for federal dollars. The HEA’s original quality screen for “Title IV eligibility”—the federal umbrella term for the grants, loans, and work-study programs authorized under Title IV of the Higher Education Act, and the category of funding a program must qualify for in order to receive any federal student aid at all—was institutional accreditation.

Accreditation is conducted by private accrediting bodies whose historical focus was on inputs like faculty credentials and library holdings rather than on student outcomes. A succession of administrations tried to add outcomes-based screens on top of accreditation through regulation, most prominently the Obama-era Gainful Employment Rule, which imposed debt-to-earnings ratio tests on for-profit colleges and postsecondary certificate programs (but excluded non-profit colleges).

Gainful Employment was rescinded by the Trump administration, reissued by the Biden administration, and remained tangled in litigation for a decade. No administration had ever succeeded in writing an earnings-based accountability test into the statute itself, which meant that every version of the rule was vulnerable to rescission by the next administration’s Secretary of Education.

Income-driven repayment made debts grow larger over time

The third problem concerned income-driven repayment plans (IDRs). Income-driven repayment plans calculated a borrower’s monthly payment as a percentage of discretionary income. For borrowers with low incomes relative to their loan balances, the required payment was often so low that it was less than the monthly interest accruing on the loan, which meant that the unpaid interest was added to the loan balance, and the borrower’s total debt grew over time even as they made every payment on schedule. This increasing debt despite consistent on-time payment is called negative amortization.

Negative amortization was the single most corrosive element of the federal student loan system from the borrower’s perspective. A borrower who owed $40,000 at graduation and made faithful payments for 10 years could find herself owing $55,000 at the end of the decade, and there was no amount of personal discipline that could prevent the balance from growing. The political backlash against negative amortization was what drove the Obama, Trump, and Biden administrations to attempt some version of IDR reform, and what created the political space for the Biden-era student loan cancellation, Saving on a Valuable Education (SAVE), that was later enjoined in federal court.

Discrimination against vocational training

The fourth problem was the absence of any short-term aid pathway for workforce-oriented students. The Pell Grant—the federal need-based grant program for low-income undergraduates, and the largest source of non-loan aid in the federal system—was restricted to programs of at least 15 weeks in length, which excluded most of the short occupational training programs where labor-market outcomes for low-income students were strongest. Students who wanted to earn a welding certificate, a commercial driver’s license, or a medical assistant credential—programs that routinely produced strong earnings gains within months rather than years—were locked out of Pell eligibility by statute.

OBBBA addressed all four of these problems.

The OBBBA reforms

Beginning July 1, 2026, Grad PLUS ends for graduate and professional students. Under OBBBA, annual Direct Unsubsidized Stafford borrowing for graduate students is capped at $20,500; for professional students, at $50,000. Aggregate graduate borrowing is capped at $100,000, or $200,000 for those pursuing a professional degree. Parent PLUS loans—long the least disciplined corner of the federal lending system—are capped at $20,000 a year and $65,000 over a dependent’s academic career. Lifetime federal borrowing, across all loan types, is capped at $257,500 per student. These caps convert Grad PLUS from an uncapped program into a bounded one and give Parent PLUS its first real statutory ceiling since the program was created in 1980.

For the first time in the history of the Higher Education Act, a federal statute denies Title IV eligibility to any program whose graduates fail to out-earn a state median benchmark for two of three consecutive years: a provision known as Section 84001 after its location in the bill, and the first statutory earnings-based accountability test in the history of federal student aid.

Taken together with the new Repayment Assistance Plan, which eliminates negative amortization for distressed borrowers and introduces a federal “match” on principal paydown of up to $50 a month; and the new Workforce Pell Grants, which tie short-term Pell eligibility to verified completion rates, job placement, and an earnings-based tuition cap; OBBBA accomplishes more federal higher-education reform in a single bill than the last four reauthorizations combined.

What still needs to be fixed

And yet, the essential architecture of federal higher-education finance remains broken. A typical four-year undergraduate at a mid-tier public university can still borrow enough to graduate $30,000 in debt from a program whose median earnings five years out barely exceed what the borrower would have earned with a high school diploma. A Parent PLUS borrower can still sign for $20,000 a year—on top of whatever Stafford loans the student has already taken—on behalf of a child enrolled in a program FREOPP’s research shows will generate a negative lifetime return on investment. The federal government still bears 100 percent of the loss when a student defaults; the institution that enrolled that student, cashed the Pell Grant, and set the tuition loses nothing. And the Section 84001 accountability test that OBBBA created—the most significant new accountability tool in 30 years—sets its failure threshold so low that the median four-year undergraduate program will clear it even if the program adds less than a high school diploma’s worth of earnings power.

These are the gaps that the next wave of federal higher-education reform must close. This paper proposes a legislative and regulatory agenda to close them. The agenda has five pillars:

1. A risk-sharing requirement for institutions, designed with a risk-adjustment layer modeled on the architectures used in Medicare Advantage and the ACA exchanges. Institutions that consume Title IV dollars should bear a share of the losses that their students incur; the share should be calibrated to the economic circumstances of the students enrolled so that the formula does not penalize schools for serving disadvantaged populations.
2. A strengthened Section 84001 earnings test, built on the FREOPP return-on-investment methodology. The new statutory benchmark should target the bottom of the return-on-investment distribution—programs whose graduates actually lose meaningful money by enrolling—rather than attempting to police the vast middle of the sector.
3. Program-level undergraduate loan limits keyed to projected ROI. OBBBA capped graduate borrowing, but left undergraduate Stafford loan limits untouched. Programs whose graduates cannot reliably service their debt should not be able to borrow at the statutory maximum; the loan cap should be a function of the program’s expected earnings.
4. Parent PLUS underwriting with meaningful ability-to-repay review and a co-signer framework. The $20,000 annual and $65,000 lifetime caps are improvements, but Parent PLUS remains the only federal loan program without a serious ability-to-repay standard. Restoring an underwriting requirement—or requiring a co-signer who passes one—would do more to prevent catastrophic household indebtedness than any of OBBBA’s caps.
5. Re-privatization of student lending, modeled on a reformed version of the pre-2010 Federal Family Education Loan Program. The 2010 decision to terminate private origination of federal student loans concentrated 100 percent of the credit risk on the federal balance sheet and removed the last market signal from the pricing of tuition. A restored FFEL, redesigned with the lessons of 2010 in mind, would reintroduce private underwriting discipline while preserving the affordability guarantees that students now depend on.

A federal return-on-investment database, modeled after FREOPP’s but maintained by the Department of Education (ED) and refreshed annually using College Scorecard, Treasury, and American Community Survey data, is the informational backbone all five pillars require. Congress should fund and mandate such a database; in its absence ED can build one administratively.

Each pillar has a legislative path and an executive path. Where Congress is the easier route this paper says so; where the Department of Education has adequate authority under the existing Higher Education Act, we describe the relevant sections. Rulemaking is slow and contested under any administration, but the HEA gives the Secretary more authority than is commonly acknowledged, and the arguments for action are stronger now than they have been in a generation.

III. OBBBA’s achievements, on their merits

Before turning to what is still broken, it is worth walking through what the bill did. Understanding OBBBA’s actual text matters because the commentariat’s characterization of the bill has already drifted from reality: progressive critics describe it as gutting student aid, conservative commentators describe it as a full repeal of income-driven repayment. Neither description survives a careful reading.

The loan caps and the termination of Grad PLUS

The significance of OBBBA’s graduate-loan caps cannot be appreciated without understanding what the pre-OBBBA system looked like. Before July 1, 2026, a graduate or professional student at any accredited institution could borrow from the federal government up to the full cost of attendance as published by the institution, with no statutory ceiling. A student at an elite law school whose institution set a cost of attendance of $110,000 per year could borrow that amount annually through a combination of Direct Unsubsidized Stafford loans ($20,500 per year) and Grad PLUS loans (uncapped). A three-year J.D. program routinely produced graduates with $280,000 or $300,000 in federal student debt. A four-year M.D. program produced, in the most expensive programs, graduates with $400,000 or more.

The institutions setting these costs of attendance faced no market discipline, because their students faced no private lender willing to evaluate whether the debt was serviceable; the federal government lent against the posted price regardless. FREOPP’s 2023 analysis of the return on investment of roughly 14,000 graduate programs found that this unbounded borrowing had produced an extraordinary concentration of negative-ROI programs: more than 40 percent of master’s programs generated a negative lifetime return on investment for the median student, and the concentration was particularly dense in education, the arts, and some humanities fields—precisely the programs whose students tended to max out Grad PLUS because the federal loan had no cap.

Section 81001 of OBBBA amends HEA §455(a) to end the Grad PLUS loan program for graduate and professional students beginning July 1, 2026. After that date, graduate students may still borrow Direct Unsubsidized Stafford loans, but the annual cap is $20,500 and the aggregate cap is $100,000. Professional students—the statutory term captures J.D.’s, M.D.’s, D.D.S.’s, D.V.M.’s, and other programs listed at 34 C.F.R. §668.2—face a higher annual cap of $50,000 and a higher aggregate cap of $200,000. A new overall lifetime cap of $257,500 now applies to everything a student borrows under Title IV, not counting Parent PLUS taken out on the student’s behalf. Institutions retain the authority to lower any of these limits program-by-program, so long as the limit is applied uniformly to all students in that program.

The termination of Grad PLUS closes what FREOPP’s graduate-program research has repeatedly identified as the largest single driver of negative-return federal borrowing. Grad PLUS was not the only reason tuition at high-cost programs escalated to meet the available borrowing, but it was a necessary condition. Taking the cap away ends that dynamic.

The Parent PLUS story is similar but less widely understood. Parent PLUS was created in the 1980 HEA reauthorization as a supplement for families whose students had exhausted Stafford borrowing capacity, and the original statutory design also allowed borrowing up to the full cost of attendance. There was a credit check, but the check was minimal—essentially a screen for active bankruptcy or very recent delinquency—and the Department of Education relaxed it further in 2011. The combination of unlimited borrowing capacity and nominal underwriting produced predictable results. FREOPP’s 2022 analysis of Parent PLUS outcomes documented a cluster of institutions—including selective private colleges and a number of Historically Black Colleges and Universities (HBCUs)—where the median Parent PLUS borrower had accumulated more than $50,000 in debt on behalf of a student who ultimately did not graduate. OBBBA’s $20,000 annual limit per dependent and $65,000 aggregate limit per dependent prevent the most extreme of these cases. They do not fix the underlying underwriting problem, which this paper takes up in Section VII.

The Repayment Assistance Plan

To understand why the Repayment Assistance Plan (RAP) represents a significant improvement over the means-tested repayment architecture that preceded it, it is necessary to begin with the problem RAP was designed to solve.

Prior to OBBBA, the federal student loan system offered a menu of income-driven repayment plans—Income-Based Repayment, Pay As You Earn (PAYE), Revised Pay As You Earn, Income-Contingent Repayment, and most recently the Biden-era SAVE bailout—whose rules differed on nearly every dimension. These rules included an income threshold below which no payment was owed; the percentage of discretionary income above that threshold required as payment; the forgiveness horizon; the treatment of spousal income; and the handling of unpaid interest.

Borrowers routinely chose the wrong plan for their circumstances simply because the choice architecture was too complex to navigate. More corrosively, each of these plans suffered from negative amortization: for borrowers with low incomes relative to their loan balances, the required monthly payment was less than the monthly interest, so the unpaid interest capitalized into the principal balance and the borrower’s total debt grew over time even as she paid on schedule. A borrower who had made 10 years of faithful payments could rationally watch his or her loan balance rise, and the resulting political backlash—visible in every public opinion survey on student debt since 2012—was the single most important driver of progressive pressure for mass forgiveness.

Section 82001 and new HEA §455(q) replace this complex menu with a two-plan system for loans originated after July 1, 2026: a standard plan, with a repayment period that scales from 10 to 25 years depending on balance; and the Repayment Assistance Plan.

Under RAP, a borrower earning less than $10,000 a year pays $120 a year—ten dollars a month, the statutory minimum. A borrower earning $50,000 a year pays four percent of their adjusted gross income (AGI), or about $170 a month before dependent deductions. A borrower earning more than $100,000 a year pays 10 percent of AGI. Each dependent deducts $50 per month from the borrower’s payment.

The most important design features of RAP are the two that critics have paid the least attention to. First, if an on-time monthly payment is insufficient to cover accrued interest, the unpaid interest is simply not charged to the borrower: the federal government pays the difference, eliminating negative amortization.

Second, if an on-time monthly payment reduces principal by less than $50, the federal government pays for an additional principal reduction up to $50 a month, effectively matching small borrower principal contributions from the public fisc. After 360 qualifying payments—30 years—any remaining balance is paid by the government.

RAP is the closest the federal student loan program has ever come to a federally administered income-insured lending product. It is not a particularly generous product: the base payments are higher than under the SAVE plan, and the forgiveness horizon is longer than under Pay As You Earn. But it is internally coherent in a way that neither of its predecessors was, and it eliminates the one feature of IDR that Americans consistently found most objectionable.

Section 84001: The first earnings test in federal statute

Section 84001 amends HEA §454 to add a new subsection (c) that denies Title IV eligibility to any program whose graduates’ median earnings fall below a credential-matched benchmark. For undergraduate programs, the benchmark is the median earnings of working adults aged 25 to 34 in the institution’s state whose highest credential is a high school diploma. For graduate and professional programs, the benchmark is the lowest of three candidate medians: state working adults with only a baccalaureate degree; state working adults in the same two-digit CIP field with only a baccalaureate; or U.S. working adults in the same field with only a baccalaureate. A program fails if its graduates—measured four years after completion, excluding anyone still enrolled in higher education—miss the benchmark for two of three consecutive years. The cohort minimum is 30 individuals, with aggregation rules for smaller programs, and institutions have an appeals process during which the secretary may permit continued participation.

This is the first time Congress has written a direct earnings-based accountability test into federal statute. The Obama administration’s Gainful Employment rule applied a debt-to-earnings ratio test to for-profit programs and certificate programs at public and nonprofit schools, and that rule has been rescinded and reissued repeatedly since 2014. Section 84001 is different from all of those predecessor regulations in two respects: it covers all degree programs, including four-year bachelor’s, master’s, and professional degrees at public and private non-profit institutions; and it is written into statute rather than regulation, so it cannot be unilaterally rescinded by the next administration.

The OBBBA’s accountability provisions are weak

By the standards of accountability proposals that have circulated in the policy community for 15 years, the OBBBA’s accountability provisions are weak. Section 84001 of the OBBBA compares college students’ incomes to a high-school-only working adult in the 25-to-34 age range; at the state level, in 2026 dollars, that is typically somewhere around $35,000 to $42,000. A bachelor’s degree program that generates median earnings four years after graduation of $45,000 passes, even though the student has taken four years out of the labor market, paid tuition, and borrowed to do so. A program barely clears Section 84001 if it adds almost nothing to the student’s earnings trajectory after accounting for the opportunity cost of attendance.

Part IV of this paper picks up where Section 84001 leaves off. The bill has created a framework; the framework needs a more economically relevant threshold.

Workforce Pell expands Pell Grants to short-term occupational programs

Finally, Section 83002 of the OBBBA creates the Workforce Pell Grant under new HEA Sections 401(k) and 481(b)(3). Workforce Pell extends Federal Pell Grant eligibility to short-term occupational programs—programs of at least 150 and fewer than 600 clock hours, lasting between 8 and 15 weeks—for the first time in the history of the program. FREOPP scholars have argued in favor of short-term Pell eligibility for more than a decade, most recently in our 2024 analysis of career and technical education. The statutory design in OBBBA is closer to the FREOPP model than to the loose short-term Pell proposals that had circulated in earlier Congresses. The OBBBA requires:

• That the governor of the state, after consultation with the state workforce board, certify that the program is aligned with high-skill, high-wage or in-demand occupations under Perkins V §122;
• That the program lead to a stackable, portable recognized postsecondary credential, or to one of the narrow category of occupations for which a single recognized credential exists;
• That the credit earned in the Workforce Pell program transfer to related certificate or degree programs at an institution of higher education, including that the institution guarantee the transfer before the program can receive Workforce Pell aid;
• That the program has been offered by the institution for at least one year prior to approval;
• That the program’s verified completion rate be at least 70 percent, measured within 150 percent of normal time;
• That the program’s verified job placement rate be at least 70 percent, measured 180 days after completion; and
• That the published tuition and fees of the program not exceed the “value-added earnings” of the program’s Title IV recipient completers, defined as the median earnings of those completers three years after graduation, adjusted for regional price parities, minus 150 percent of the federal poverty line for a single individual.

Read literally, the last bullet is the most important. The program’s tuition cannot exceed the earnings premium the program generates over a rough subsistence floor. That is the first time a federal financial aid program has imposed an earnings-based tuition cap on a class of institutions, and it deserves to be understood as a precedent for what Congress can eventually do in the four-year sector.

IV. The gap OBBBA left open: Institutional risk sharing

The case for institutional skin-in-the-game

Every federal higher education program in the United States operates on a common architecture. The taxpayer fronts money to the student—as a grant, as a direct loan, as a subsidized interest payment—and the institution captures the money through tuition. If the student repays the loan and earns a reasonable return on the investment, the system works. If the student defaults, drops out, or graduates into a labor market that will not reward the credential, the loss falls somewhere on the balance sheet between the student’s credit file and the federal government’s budget outlays. The institution keeps the tuition payment it received at matriculation, and bears essentially nothing of the loss that its own product created.

This is an unusual way to finance an industry. When the federal government guarantees home mortgages, it requires banks to hold a portion of the credit risk. When the federal government reinsures crop insurance, it requires private insurers to hold a portion of the underwriting risk. When Medicare contracts with managed-care plans, it caps the total amount the government pays and forces the plans to absorb cost overruns. The federal government does not extend a blank credit card to any other sector of the economy and tell the supplier to keep the money regardless of whether the product worked. But that is precisely how federal student aid has operated for 60 years.

The standard defense of this system is that higher education is a public good, that students benefit from the existence of institutions even when they do not benefit from their own enrollment, and that any institutional cost-sharing rule would penalize schools for enrolling the disadvantaged students they are trying to serve.

Take the typical version of that defense: the argument, articulated in almost every House and Senate hearing on risk sharing since 2013, that a default-rate-based penalty would disproportionately affect HBCUs and minority-serving institutions.

It is true that institutions that serve first-generation, Pell-eligible, minority, and academically unprepared students will, all else equal, have higher default rates than institutions that enroll wealthy students. A simplistic rule that penalizes institutions with higher student loan default rates will disproportionately penalize institutions that serve students from lower socioeconomic backgrounds.

But this is no defense of the current system. It is technically possible to risk adjust the accountability mechanism so that it does not penalize institutions for serving disadvantaged students. It is possible to price the risk associated with the student’s entering characteristics separately from the risk associated with the institution’s own choices. That is exactly what modern health insurance regulation does.

What health insurance teaches us about risk adjustment

Private health insurance markets in the United States are highly regulated. A key regulation found in its largest health insurance markets is called community rating. Community rating requires insurers to charge individuals the same premium whether or not they have pre-existing conditions or poor health status. In effect, under community rating, insurers are required to overcharge healthy people and undercharge the sick, relative to their actual consumption of health care services.

In an environment where insurers are forced to charge uniform prices irrespective of individuals’ health status, insurers are theoretically incentivized to only enroll the healthy and avoid the sick. To take a simple example: if insurers charge a uniform $1,000 monthly premium, and the typical healthy person consumes $200 a month in health care, but the typical sick person consumes $2,500 a month, the insurer will make $800 a month on the healthy person but lose $1,500 a month on the sick person.

How do insurers avoid the sick? They can design their plans so that sick people never sign up. Physician and hospital networks can exclude the specialists sicker patients need. Drug formularies can drop the drugs that sicker people require. Insurer marketing can target neighborhoods where the healthy cluster and avoid the ones where the sick cluster.

In order to address this incentive problem, regulators usually pair community rating with risk adjustment. Risk adjustment assigns bonuses or penalties to insurers based on how healthy or sick their enrolled population is. A healthier population will incur a penalty, and a sicker population will incur a bonus payment, in order to neutralize the incentive for insurers to seek out the healthy and avoid the sick.

Instead of allowing the insurer to charge the sick person a higher premium, the regulator measures the expected cost of the sick person based on diagnostic history, and makes a payment to the insurer enrolling the sick person that brings their expected revenue into line with the average. The cream-skimming incentive disappears—or, at least, diminishes—because the insurer’s profit margin on a sick person is now comparable to its margin on a healthy person.

While risk adjustment is never exactly perfect—insurers always look to optimize their enrollment against it, and regulators play whack-a-mole to identify opportunities for insurer arbitrage—it has proven to be a reasonably effective regulatory instrument that simultaneously protects the sick from premium discrimination and protects the market from adverse selection.

For several decades, the Centers for Medicare and Medicaid Services (CMS) has run two large risk-adjustment systems. The CMS Hierarchical Condition Category model, known as CMS-HCC, prospectively adjusts per-capita payments to Medicare Advantage plans based on enrollees’ age, sex, Medicaid eligibility, and the diagnoses documented in the previous year.

The CMS-HCC model was described by Gregory Pope, John Kautter, and colleagues in a 2004 paper, and has been repeatedly recalibrated since then. Kautter, Pope, and another group also described the HHS-HCC model used in the Affordable Care Act’s exchanges in a 2014 paper.

Because ACA enrollees churn more frequently than do Medicare Advantage enrollees, the HHS-HCC model adjusts risks concurrently rather than prospectively; that is, it uses the current year’s diagnoses to predict current-year cost. But both systems operate on the same intellectual foundation: identify the characteristics of enrollees that predict cost, run a regression to estimate the coefficients, calculate a risk score for each enrollee, and use the score to move money among plans.

Neither system is perfect. A group of Stanford researchers found that when Medicare introduced the HCC model in 2004, Medicare Advantage insurers responded by selecting enrollees whose HCC scores were high but whose actual costs, conditional on score, were low: a finding that undermined claims that the reform would eliminate selection-driven overpayments. Michael Geruso and Timothy Layton estimated in 2020 that Medicare Advantage plans upcode the diagnoses of their enrollees by 6 to 16 percent relative to what fee-for-service Medicare would record, generating billions of dollars a year in excess federal payments.

Similarly, some research suggests that the HHS-HCC model under-predicts cost for the chronically ill and over-predicts cost for the healthy. Even with effective risk adjustment, unsubsidized enrollees in ACA exchange markets tend to migrate toward bronze-tier coverage, because healthy people gravitate to the plans with the lowest premiums.

The lesson is not that risk adjustment works perfectly. The lesson is that risk adjustment works well enough, paired with reinsurance and risk corridors, to sustain community-rated health insurance markets that would otherwise be impossible. It is a better model than any currently circulating student loan risk-sharing proposal, and it is the right starting point for designing an institutional accountability regime that does not penalize mission-driven schools for doing their jobs.

A risk-adjusted risk-sharing formula for student loans

The structure this paper proposes is deliberately analogous to the CMS-HCC framework and deliberately conservative in its parameters. Congress can tighten the parameters later; it is more important to get the architecture right the first time.

Step 1: Build an outcomes database. The Secretary of Education should construct a federal database, similar to FREOPP’s ROI database, that links every Title IV recipient to:

1. Their entering characteristics—Pell status, family income at matriculation, first-generation status, age, dependency status, state of residence, program of study at the four-digit CIP level;
2. Their eventual outcomes—completion status, loan balance at separation, and earnings in years 4, 6, 8, and 10 after separation.

The earnings data already exist inside Treasury, inside the College Scorecard, and, for non-filing and low-earning former students, inside the unemployment insurance wage records held by state workforce agencies. Part VIII of this paper discusses the data infrastructure in more detail.

Step 2: Develop the risk adjustment formula. The Secretary of Education can build a regression analysis, in which the outcome variable is the present value of the student’s loan repayments over 10 years, and the right-hand side variables are the entering characteristics enumerated in step one. The coefficients are the risk-adjustment weights. A Pell-eligible first-generation student from a ZIP code in the bottom decile of median family income receives a high risk score; a non-Pell student from an affluent ZIP code receives a low risk score. The model should be calibrated nationally but refreshed annually, so that macroeconomic and cohort effects are absorbed.

Step 3: Customize a risk-adjusted earnings benchmark for each institution. For each higher education institution, the secretary should then calculate an expected aggregate repayment value by summing the risk-adjusted expectations across that institution’s most recent five cohorts. The institution’s expected repayment is its nationally-calibrated benchmark—the number the institution should hit if its students perform exactly as similar students perform at the average institution.

Step 4: Compare institutions’ performance to its benchmark. The secretary can then compare each institution’s actual aggregate repayment value to its expected value. The difference—positive or negative—is the institution’s risk-adjusted performance. Institutions whose actual repayments exceed their expected repayments by a meaningful margin are rewarded. Institutions whose actual repayments fall short of their expected repayments by a meaningful margin owe ED a fraction of the shortfall.

Step 5: Apply financial rewards and penalties based on performance. The penalty fraction should be capped at a percentage of Title IV revenue—this paper recommends five percent in the first year, rising to 15 percent by year five—so that no institution faces a sudden existential threat. Payments from penalized institutions can fund a reinsurance pool for borrowers whose earnings fall below a catastrophic threshold for reasons uncorrelated with institutional quality. The reinsurance pool could be modeled on the high-cost risk pool used in the ACA exchanges after 2017; its purpose is to absorb the random variation that no regression can predict.

Step 6: Audit the risk adjustment system every year. The model should be audited annually. Institutions should be able to challenge their risk scores; ED must publish the full model specification and a replication dataset with enough detail for outside researchers to verify the calibration. Institutions may not change their reported student characteristics by more than a specified percentage year over year without documentation; this is the higher-education analog to MedPAC’s recommendations on peripheral-diagnosis upcoding in Medicare Advantage, and it is essential if the system is to survive its first decade.

A formula built on this architecture does two things that no default rate-based rule can do. It prices risk rather than punishing schools for enrolling risky students. And it generates an explicit, publicly auditable measure of institutional value-added, which can be used for accreditation decisions, state funding formulas, and consumer disclosure.

The path to accountability reform, with or without Congress

Congress should enact institutional risk sharing by amending HEA §454 to add a new subsection. The statutory text can borrow heavily from the Section 84001 architecture already in place: the same programmatic cohort definitions, the same 30-person minimum, the same appeals process. What it needs to add is the risk-adjustment regression, the reinsurance pool, and the maximum penalty cap.

Absent congressional action, the Department of Education has partial authority to improve accountability under the Higher Education Act’s quality-assurance provisions. HEA §487(a) requires institutions to enter into a program participation agreement as a condition of Title IV participation, and the secretary has historically used that authority to impose a variety of administrative requirements. A narrower version of the risk-sharing proposal—one that conditions continued PPA eligibility on an institution’s agreement to publish risk-adjusted outcomes and absorb a defined share of default losses above a threshold—could be attempted administratively. It would be challenged in court; the Supreme Court’s major questions doctrine jurisprudence after West Virginia v. EPA and Loper Bright would be the principal obstacle.

The major questions doctrine, as the Supreme Court has articulated, requires clear congressional authorization for agency action of vast economic or political significance. The doctrine is designed to prevent agencies from discovering, in dusty corners of ambiguous statutes, the authority to transform entire sectors of the American economy without a legislative mandate. A risk-sharing regulation does not fit that pattern, for four reasons.

First, the statutory hook is not obscure. HEA §454—the provision that governs Title IV program eligibility and that OBBBA itself amended in Section 84001 to create the first statutory earnings test—explicitly tasks the secretary with ensuring that institutions meet “quality assurance” standards. HEA §487(a)(20) gives the secretary authority to include in the program participation agreement “any other provision… necessary to protect the interests of the United States and promote the purposes of this title.” A risk-sharing rule that requires institutions to absorb a portion of Title IV losses is a direct application of both provisions.

Second, the rule is not novel as a matter of federal lending practice. The federal government already requires risk retention of mortgage originators under §941 of the Dodd-Frank Act, of crop insurers under the Federal Crop Insurance Act, and of guaranteed-loan lenders in agriculture and small business under the Farm Credit Act and the Small Business Act. A risk-sharing rule in higher education is a refinement of a regulatory tool Congress has repeatedly endorsed in adjacent domains, not a novel regulatory scheme of first impression.

Third, Congress has effectively ratified the secretary’s quality-assurance role in successive reauthorizations. Every HEA reauthorization since 1972 has expanded the secretary’s authority to condition Title IV eligibility on institutional outcomes, and OBBBA itself—enacted eight months before this paper was written—codified an earnings-based eligibility test in statute. The legislative trajectory is unambiguous: Congress wants the secretary to measure outcomes and condition funding on them. A risk-sharing rule is an incremental step along a path Congress has itself laid down.

Fourth, and most importantly, the architecture of Section 84001 creates its own statutory predicate for the next step. Section 84001 establishes that Title IV eligibility may be withdrawn from programs whose outcomes fail a statutory threshold. It is a short step from “the program loses eligibility if outcomes fail” to “the institution bears a share of the losses when outcomes fail,” and the short step is well inside the discretion Congress granted the secretary in the implementing provisions of the same statute. The major questions doctrine is not a general license for courts to strike down agency actions they disapprove of; it is a rule that demands clarity for agency actions that transform sectors. An incremental rule that extends an explicit statutory framework is not a major questions case.

V. Strengthening Section 84001: An earnings test aimed at lower-income households

The weakness of the high school benchmark

As we have discussed, Section 84001 requires that an undergraduate program’s graduates earn at least as much as a state working adult aged 25 to 34 whose highest credential is a high school diploma.

Take the actual numbers this produces. The Census Bureau’s American Community Survey five-year estimates for 2022 put the median annual earnings of working adults aged 25 to 34 with only a high school diploma at $36,200 nationally. State medians range from a low around $28,000 in the poorest states to a high around $45,000 in the wealthiest. A four-year bachelor’s degree program has to produce median graduate earnings of only $36,000 to $45,000—four years after graduation—to clear the Section 84001 threshold.

That is not a demanding test. Consider the comparison to what the student would have earned without enrolling. A 22-year-old high school graduate earning $36,000 a year today is, at age 26, on a wage trajectory that will take her to the mid-$40,000s by her late twenties. The four-year college graduate who clears the Section 84001 threshold at $36,000 at age 26 is actually earning less than her same-age high school counterpart, because the high school graduate has been working for four years and climbing the job ladder while the bachelor’s student was in class. The Section 84001 benchmark implicitly compares bachelor’s graduates at age 26 to high school graduates at age 26, but it should compare bachelor’s graduates at age 26 to high school graduates at age 26 who have been in the labor force since age 18.

FREOPP’s Return on Investment in Higher Education methodology, first published in 2019 and updated annually, calculates the counterfactual correctly. It uses College Scorecard median earnings at six and ten years after entry, adjusts for the time the student spent enrolled rather than working, discounts future earnings to present value, and nets out the full counterfactual stream of earnings a similar worker could expect without the credential. The result is a measure of lifetime net value that can be positive or negative.

In FREOPP’s 2024 edition, roughly a quarter of undergraduate bachelor’s programs in the United States produced negative lifetime ROI, and within that quarter a substantial subset produced lifetime ROI losses in excess of $50,000, $100,000, or more.

Section 84001 as written, in its first year of implementation, would deny Title IV eligibility to perhaps three or four percent of those programs—the ones whose graduates earn less than a high school diploma’s median.

An ROI floor targeted at the worst-performing institutions

The goal of a stronger Section 84001 should not be to police the vast middle of the higher-education sector. Most American bachelor’s programs generate some positive lifetime value for their median student, even if the value is modest, and a statutory test that tried to withdraw Title IV eligibility from programs in the middle of the ROI distribution would set off a political war that reformers will not win.

The right target is the bottom of the distribution: programs whose graduates reliably lose significant money, programs whose negative ROI is so large and so predictable that continued federal subsidies cannot be defended on any plausible theory of the public interest.

Our proposed reform replaces the single high-school benchmark with a two-part floor. Congress can write this into statute by amending HEA §454(c), or the secretary can approximate it through implementing regulations under her existing authority.

The first floor targets the worst-performing programs in absolute dollar terms. A program should lose Title IV eligibility if its median lifetime return on investment—calculated under the FREOPP methodology, net of counterfactual earnings and accounting for the opportunity cost of time enrolled—is worse than negative $50,000. This threshold is not arbitrary. It corresponds roughly to the point at which a program’s median graduate would have been better off putting his or her tuition payment and four years of forgone labor-market earnings into a money-market account and taking a job at the typical starting wage for a high school graduate. A program whose median ROI is worse than negative $50,000 is not merely a bad deal for its graduates; it is a product whose failure is so large and so reliable that the federal taxpayer cannot justify financing it.

The second floor targets the worst-performing programs in relative terms. Under our proposed system, a program would lose Title IV eligibility if its median lifetime ROI falls in the bottom decile of all Title IV-eligible programs of the same credential level (bachelor’s, master’s, professional) nationally. This second floor catches programs whose absolute ROI is not quite catastrophic but whose performance relative to peer programs is consistently the worst in the country: the bottom 10 percent of bachelor’s programs an the bottom 10 percent of master’s programs. A program in the bottom decile is, by construction, outperformed by 90 percent of comparable programs at comparable credential levels, and the burden of proof should fall on the program rather than on the regulator to show that the institution should continue to receive subsidized student loans.

Programs that fail either floor should lose Title IV eligibility for new enrollments. They retain the statutory right to an appeals process identical to the one Section 84001 already provides, and already-enrolled students are protected by the same teach-out provisions. The two floors can be phased-in over three years, starting with the first floor in year one, the first plus the second in year two, and full enforcement in year three, so that institutions have time to adjust.

We believe that this two-floor system is moderate and reasonable. Policing only the bottom decile of programs by ROI, or programs whose negative ROI exceeds $50,000, produces a test that affects perhaps 8 to 12 percent of bachelor’s programs in the United States. These programs are the least defensible to subsidize and the programs whose closure would plausibly improve the earnings trajectories of students who would otherwise have enrolled. It leaves the middle of the sector alone. That is the right place to draw the line, because the statutory tool that Section 84001 created is a blunt instrument, and blunt instruments should be used against clear targets rather than against ambiguous ones.

The graduate program benchmark

Section 84001’s graduate program benchmark is the lowest of three alternative medians: state working adults with only a baccalaureate, state working adults in the same two-digit CIP field with only a baccalaureate, or national working adults in the same field with only a baccalaureate. The “lowest of” rule is meant to be generous—the secretary picks the comparison least favorable to the benchmark. In practice it will mean that most graduate programs pass even if they deliver almost nothing in earnings premium, because the comparison group will be a subset of a subset of working adults, and the smaller the comparison group, the more variable the median.

FREOPP’s graduate-program research identifies the scope of the problem directly. Roughly 40 percent of master’s programs produce a negative lifetime return on investment for the median student; the MBA, which is the single most common master’s credential in the United States, has a negative median ROI; and only 86 percent of doctoral and professional programs produce positive lifetime returns. Applied to graduate programs, the same two-part floor used for undergraduate programs—ROI worse than negative $50,000, or ROI in the bottom decile of comparable graduate programs—would deny Title IV eligibility to the worst-performing master’s programs while leaving the vast majority of graduate education untouched.

VI. Program-level undergraduate loan limits

OBBBA failed to reform undergraduate loan caps

The 2005 Higher Education Reconciliation Act raised Stafford undergraduate loan limits to the current levels: $5,500 for a first-year dependent undergraduate, $6,500 for a second-year dependent undergraduate, $7,500 for each subsequent year, with an aggregate cap of $31,000. Independent undergraduates face higher limits—$9,500, $10,500, $12,500, with an aggregate cap of $57,500. These limits have been unchanged for 20 years. OBBBA left them alone. The bill’s authors made the defensible choice to focus their scarce political capital on the graduate and Parent PLUS programs, where the worst abuses were concentrated. But leaving the undergraduate caps alone means that the same federal dollar is available to a first-year student enrolling in a high-quality nursing program at a state university where the median graduate out-earns his or her counterpart by $20,000 a year, and to a first-year student enrolling in a low-quality bachelor’s program at a regional private college whose median graduate is earning less than that high school counterpart described in Part V.

A federal loan system that extends the same credit line to both students is either subsidizing the second student into a bad deal, or implicitly relying on the second college to cover the resulting losses.

A program-level cap keyed to projected ROI

The ideal fix is to make the annual loan cap a function of the program’s projected return on investment. Specifically, the cap should be the lesser of the current statutory limit ($5,500, $6,500, $7,500 by year) and a program-specific figure equal to eight percent of the program’s projected median earnings at six years after entry. A program whose median graduate earns $75,000 six years after entry could borrow up to $6,000 in the first year of enrollment (eight percent of $75,000). A program whose median graduate earns $35,000 six years after entry could borrow up to $2,800 in the first year. The eight percent figure is the standard mortgage-industry debt-service-to-income ratio, which is the right benchmark because it reflects what is known about sustainable household debt-service burdens.

The program-level cap accomplishes three things simultaneously. It stops the federal government from over-lending to programs whose graduates cannot service the loans. It creates a strong incentive for institutions to hold down tuition in low-ROI programs—if the loan cap is $2,800 and the published tuition is $9,000, the institution has to find a way to close the gap or lose students. And it generates a public, program-level signal of the federal government’s own assessment of the program’s financial value, which has consumer-protection benefits independent of any loan cap.

The statutory authority to do this is HEA §428H for unsubsidized Stafford loans and HEA §425 for subsidized Stafford loans. Both sections currently specify the annual and aggregate limits in dollar terms. A straightforward amendment would replace the current dollar limits with a formula tied to program earnings as maintained in the College Scorecard. Executive action is harder here: the Department of Education has historically treated the loan caps as ministerially fixed in statute and has declined to use its authority under HEA §455(a)(7)(B)—which permits institutions to set lower caps for particular programs—to impose caps itself. A more aggressive reading of §487 would allow the secretary to require institutions to disclose the ROI of their programs and to cap loans at ROI-appropriate levels as a condition of PPA participation, but the legal case would be weaker than for risk sharing because the statute’s dollar amounts are specific and unambiguous.

VII. Parent PLUS underwriting and co-signer reform

Parent PLUS is the only federal student loan program that has ever lacked a meaningful ability-to-repay standard. The program was created in the 1980 reauthorization of the HEA as a supplement for families whose students had exhausted their Stafford borrowing capacity. The original statutory design required a credit check, but the credit check was minimal—essentially a screen for active bankruptcy or recent delinquency—and ED relaxed it further in 2011 in response to complaints about declining approval rates. The 2011 standards, which remain in force, permit a Parent PLUS loan to any applicant who does not have an adverse credit history as defined by 34 C.F.R. §685.200(c). Applicants with thin credit files, limited income, or existing debt loads that would disqualify them from any private loan product are routinely approved for Parent PLUS loans in the tens of thousands of dollars.

OBBBA’s $20,000 annual and $65,000 lifetime caps are historic improvements, but they do not address the underwriting problem; that is, the fact that federal loans are not preceded by an assessment of the borrower’s ability to repay.

Lending caps prevent the most catastrophic cases—the family that accumulates $150,000 of Parent PLUS debt on behalf of a single child—but they do not prevent the more common cases in which a family with $40,000 in annual income borrows $20,000 a year for four years and ends up unable to service a loan the federal government had no business underwriting in the first place.

Reforming this flaw is straightforward. The federal government should restore a meaningful credit check to student lending: income-and-assets verification, a debt-to-income ratio test, and a credit-score minimum that borrows from the underwriting standards the Veterans Administration uses for home loan guarantees.

Families who fail the credit check should have the option to bring a co-signer with adequate credit, which is what private student loan lenders have done for decades. A co-signer framework would preserve access for the significant share of Parent PLUS borrowers for whom the program remains a reasonable financing choice while disqualifying the much smaller share for whom it has become a trap.

The statutory authority to impose underwriting requirements on Parent PLUS is HEA §428B, which governs the origination of Federal Family Education Loan Program (FFELP) and Direct PLUS loans. Subsection (a)(1)(A) currently provides that “the secretary may establish” underwriting criteria, language that the Obama administration relied on to tighten the adverse-credit standard in 2011 and the Trump administration relied on to loosen it in 2019. The same authority supports the broader underwriting framework this paper proposes; the question is whether any administration will use it. A statutory amendment that codifies a specific debt-to-income threshold would take the question out of the political cycle.

VIII. Re-privatizing the federal student loan portfolio

Nationalizing the student loan sector was a mistake

Before 2010, the federal student loan system operated on a public-private hybrid model known as the Federal Family Education Loan Program. Under FFELP, private lenders—banks, state loan agencies, and nonprofit guarantee corporations—originated federally guaranteed student loans to students and their parents. The federal government guaranteed the loans against default, and paid lenders a subsidy to ensure capital remained available. However, the initial underwriting decision, the origination process, and the servicing relationship were all private.

FFELP coexisted with a smaller Direct Loan program, created in 1993, in which the federal government lent directly to students without a private intermediary. For nearly two decades, the two programs operated side by side, and the presence of competition between them was the single most important discipline on both.

In 2010, as part of the Health Care and Education Reconciliation Act—the budget-reconciliation companion to the Affordable Care Act—Congress terminated FFELP entirely. The claimed justification was a Congressional Budget Office score suggesting that eliminating the federal subsidies to private lenders would “save” money, which Congress then applied toward the ACA’s coverage expansion. These were phantom savings, however, because they counted the value of the loans as revenue, and undercounted the risk of defaults that would occur outside of the CBO’s 10-year scoring window.

As we noted earlier, in 2016, the Government Accountability Office estimated the taxpayer losses at $108 billion and growing from the nationalization of student lending. By 2024, the total outstanding stock of federal student debt exceeded $1.6 trillion, roughly one-quarter of which FREOPP’s research has estimated is owed by borrowers whose degrees will not generate enough lifetime earnings to justify the cost.

The substantive consequence was that, from July 2010 forward, every new federal student loan in the United States has been originated directly by the federal government, and every new federal student loan has been held on the federal balance sheet for the life of the loan.

The federal takeover of student loans was not only a fiscal loss, but a policy mistake. Eliminating private origination removed the last market signal from the pricing of federal student loans. Under FFELP, private lenders still had a modest incentive to look at the borrower’s likelihood of repayment, because the federal guarantee did not cover 100 percent of losses, and because servicing costs varied with default rates.

Under the post-2010 Direct Loan system, the federal government began absorbing 100 percent of the default risk, and the only administrative check on the flow of capital to low-quality programs was the private-sector accreditation process. Private accreditors were not designed for this role; their historical focus has been on inputs, like how many buildings or professors a campus has, rather than student outcomes. The post-2010 system is, in effect, a federally administered blank check, bounded only by the dollar caps in statute and whatever sparse accountability regulations happen to be in force at a given moment.

It is not a coincidence that tuition inflation in the federal student loan-financed sector accelerated sharply after 2010. It is also not a coincidence that the political conversation around student loan debt—mass forgiveness, income-driven repayment reform, the SAVE plan—began to dominate federal higher-education policy in the decade that followed.

A system in which the federal government is the only lender, the only underwriter, and the only party holding credit risk is a system in which every bad loan is a federal political problem rather than a private credit loss, and it is a system in which the reform conversation inevitably migrates from “how do we underwrite better” and “how do we hold institutions accountable for excessive prices” to “how do we forgive more.”

What FFELP did well before 2010, and what it did badly

FFELP was not a well-designed program. Its principal virtue was the presence of private capital in the origination process; its principal vices were the generous federal subsidies to private lenders that made the program politically indefensible, the opacity of the servicing relationships that made borrower protections weak, and the moral hazard created by federal guarantees that covered nearly all of the losses. Any reform that aims to re-privatize student lending has to preserve the virtue—market discipline in origination—while fixing the vices.

A reformed FFELP, designed with the lessons of 2010 in mind, would reintroduce private origination under four conditions.

First, risk retention. Private lenders should be required to retain a meaningful portion of the credit risk on every loan they originate. The appropriate threshold is the same five percent risk retention that §941 of the Dodd-Frank Act imposes on mortgage-backed securities issuers. A lender who keeps five percent of the credit risk has a direct financial stake in whether the borrower can actually repay, which is the core market signal the Direct Loan program has eliminated. The federal guarantee would cover the remaining 95 percent of the loss, which is enough to preserve broad access to federal credit while introducing genuine underwriting discipline.

Second, transparent and capped lender compensation. The pre-2010 FFELP program paid lenders through an opaque combination of interest subsidies, special allowance payments, and consolidation fees that generated windfall returns and were the proximate justification for the program’s elimination. A reformed FFELP should pay lenders a single, capped, transparently calculated origination fee—a flat percentage of loan principal, set by statute and indexed to the cost of capital—and nothing else. Lenders would compete for borrower business on the strength of their servicing, their underwriting accuracy, and their willingness to work with borrowers in distress. They would not compete for subsidy dollars.

Third, statutory consumer protections that travel with the loan. One of the genuine failures of the pre-2010 FFELP program was that borrower rights—access to income-driven repayment, deferment options, forgiveness under Public Service Loan Forgiveness—were inconsistently honored by private servicers, and the servicing relationship moved among lenders in ways that made it hard for borrowers to track. A reformed FFELP would codify every borrower protection available under the Direct Loan program as a statutory condition of the federal guarantee, so that a loan originated by a private lender carries identical Repayment Assistance Plan access, identical deferment rights, and identical Public Service Loan Forgiveness qualifications as a loan originated by ED. Any lender that fails to honor those protections forfeits the guarantee on the affected loan.

Fourth, an ROI screen at origination. A reformed FFELP would permit private lenders to originate federally guaranteed loans only for programs that clear a published Department of Education ROI floor—the same floor proposed in Section V of this paper. Programs below the floor would not be eligible for guaranteed private origination; students wishing to attend such programs could still borrow directly from ED under a smaller, residual Direct Loan program, but the federal guarantee and the private capital flow would be restricted to programs that pass the ROI test. This design uses the private lending channel as an additional mechanism for directing federal credit toward programs that work, without requiring ED to build a new regulatory apparatus.

What a re-privatization bill could accomplish

A reformed FFELP, designed along these lines, would accomplish three things the current Direct Loan program cannot. It would reintroduce market discipline into origination, which will slow or even reverse the tuition inflation that the post-2010 system has fueled. Reform would move a meaningful share of the credit risk off the federal balance sheet and back onto private lenders, which would reduce the political pressure for mass forgiveness by converting the problem from a federal budget question into a private credit question. And it would create a second, market-based check on the quality of the programs receiving federal subsidies—a check that operates alongside risk sharing, the Section 84001 earnings test, and the program-level loan caps this paper has already described.

The legislative vehicle is straightforward. Congress would repeal the 2010 termination of FFEL and replace it with a new Title IV, Part B framework that reflects the four design principles above. Existing Direct Loans would continue to be serviced by ED on their current terms; new originations beginning on a specified date—July 1, 2028 is a reasonable target, allowing time for private lenders to rebuild origination capacity—would be routed through the reformed FFEL program, with a residual Direct Loan program remaining in place for borrowers who cannot access private origination.

The political obstacle is real. Progressive critics of pre-2010 FFELP correctly pointed out that the subsidies paid to private lenders were excessive and the servicing relationships were inconsistent, and they will read any re-privatization proposal as an attempt to reopen the subsidy windows they closed. The reformed FFELP described here is not that. It eliminates the interest subsidies, caps lender compensation at a transparent origination fee, requires risk retention, codifies borrower protections, and screens originations against program quality. It borrows the best feature of the old program—private underwriting discipline—while fixing every major flaw of the old program. Whether it can be passed in the current Congress is a question of political will rather than of design.

IX. The federal ROI database as public infrastructure

Every reform in this paper depends on a shared informational infrastructure—a federally maintained database that links entering student characteristics, program of study, completion, and earnings outcomes at meaningful intervals after completion. Some of that infrastructure already exists. The College Scorecard, launched in 2015 and expanded repeatedly since, publishes median earnings by program at two, six, and ten years after entry, drawn from Treasury tax filings linked to the National Student Loan Data System. The Scorecard data have been the backbone of FREOPP’s ROI methodology since 2019 and of ED’s own 2024 Financial Value Transparency regulations.

But the Scorecard has two deep limitations. First, it reports only median earnings, not the underlying distribution, which means an analyst cannot measure within-program variation or identify subgroups that a program serves particularly well or poorly.

Second, the Scorecard has incomplete coverage of former students who do not file federal tax returns—a population that includes many of the students whose outcomes matter most for the accountability questions this paper has been discussing. Both limitations can be fixed.

On distributional data. The Treasury’s Statistics of Income program already produces percentile distributions for every IRS data product the public sees; the technical obstacle to publishing earnings at the 25th, 50th, and 75th percentile for each program on the Scorecard is small. The legal obstacle—§6103 of the Internal Revenue Code, which protects the confidentiality of individual tax data—is already managed in the Scorecard’s current form through cell suppression for programs with fewer than 30 students. The same suppression rules work for distributional statistics. Congress can amend §6103(l)(13) to explicitly authorize distributional statistics at the program level; ED can issue the specifications and Treasury can produce the data. Nothing about this requires new bureaucracy.

On non-filer coverage. The state unemployment insurance wage records cover employment for roughly 90 percent of the workforce, and the Workforce Innovation and Opportunity Act already permits limited federal-state data sharing through the State Longitudinal Data System grants. A modest expansion of that framework, authorizing ED to receive anonymized wage records from state workforce agencies for Title IV recipients, would close most of the coverage gap. It would also let ED see earnings of former students in years three and four after completion—the years that matter most for the Section 84001 earnings test—rather than waiting until the Scorecard’s six-year look-back window.

Congress should fund the database expansion explicitly as an appropriation to the Department of Education’s Institute of Education Sciences, which has the technical staff and the statutory authority to operate it. The annual cost will be modest—in the range of $15 million to $25 million once the initial data-sharing agreements are in place—and the return in better-targeted federal aid is large. A fully-built federal ROI database is also the foundation for the institutional risk-sharing formula proposed in Section IV; without it, the regression that produces the risk-adjustment weights cannot be run.

A second virtue of the database, less discussed in the policy literature, is that it creates a consumer-protection signal that is difficult to obtain any other way. A high school senior choosing among colleges has no reliable way to compare the earnings trajectories of graduates of particular programs at particular institutions. The Scorecard is the closest thing that exists, and it is already heavily used by the organizations that serve first-generation college applicants. An expanded Scorecard with distributional data and complete earnings coverage would move the needle on the kind of information asymmetry that underlies the worst enrollment decisions.

X. Answering the objections

“This is just the Gainful Employment rule in a different form.”

It is not, and the difference matters. The Gainful Employment rule of 2014, reissued in 2023, applied a debt-to-earnings ratio test to certificate programs at public and nonprofit institutions and all programs at for-profits. Programs whose graduates spent more than eight percent of gross earnings or 20 percent of discretionary earnings on loan service for two of three years lost Title IV eligibility. The rule generated years of litigation and was rescinded twice and reissued twice. Its basic analytical problem was that it measured debt service as a share of actual earnings, which meant that a program whose graduates could have serviced the debt but happened to be in a recession year failed the test, and a program that had driven its graduates’ earnings up only modestly but had also succeeded in pushing its tuition down even faster passed it.

Section 84001, as enacted, measures earnings against a credential-matched benchmark rather than against debt service. The proposals in this paper strengthen that test by targeting the bottom of the ROI distribution and by adding program-level loan caps and institutional risk-sharing layered on top. None of the resulting structure is a debt-to-earnings ratio test. The closest analog in the existing statutory landscape is the Workforce Pell value-added-earnings cap in Section 83002, which Congress has already blessed for short-term programs.

“Institutional risk sharing will destroy mission-driven schools.”

This is the most common objection to accountability measures, and it deserves a serious answer. The answer is that a well-designed risk-adjustment formula does not penalize mission-driven schools because it prices the risk associated with their students’ entering characteristics before measuring their institutional performance. An HBCU that enrolls 90 percent first-generation Pell students and generates median earnings of $38,000 four years after graduation is—under the formula proposed in Section IV, compared to the national average performance of institutions serving similar students—not to an institution serving the median American eighteen-year-old. If the HBCU’s outcomes exceed the risk-adjusted benchmark, it receives a bonus, not a penalty. If its outcomes fall short, the penalty is capped, the school has appeal rights, and the reinsurance pool absorbs catastrophic losses that are not the school’s fault.

The crude default-rate tests that circulated in earlier Congresses would have penalized HBCUs and other minority-serving institutions. A risk-adjusted test will not, and the best evidence for that claim is that the health insurance literature has spent twenty years refining exactly this distinction. CMS does not penalize Medicare Advantage plans that enroll sicker beneficiaries; it adjusts their capitation payments upward so that the profit margin on a sick beneficiary is comparable to the margin on a healthy one. A higher-education version of the same principle would work the same way.

“Institutions will game the risk score.”

They will try. Geruso and Layton’s upcoding work in Medicare Advantage showed that plans captured 6 to 16 percent of the risk score through coding intensity rather than through genuine differences in beneficiary health. The higher-education analog—institutions reclassifying students to increase their reported risk scores—is real, and regulators should anticipate it.

The antidote is the same as in health insurance: cap the rate at which an institution’s reported risk profile can change year-over-year; require documentation for any material change; audit a random sample of enrollment files annually; and publish the model’s fit, power, and balance statistics so that outside researchers can independently verify the calibration. The Centers for Medicare and Medicaid Services runs the Risk Adjustment Data Validation program for precisely this purpose, and the higher-education version would borrow its structure.

“This will lead to thousands of program closures.”

The version of accountability proposed in this paper will lead to several hundred closures. The reform proposed in Section V—a Section 84001 floor set at the bottom decile of programs by ROI or at absolute ROI losses exceeding $50,000—would affect roughly 8 percent to 12 percent of bachelor’s programs in the United States. That is a meaningful number of programs, and some institutions will lose multiple offerings, if they are unwilling to lower the prices of those programs to improve their return on investment.

Our proposed narrow targeting is deliberate. The programs in the bottom decile are the ones whose graduates end up financially worse-off than if they had not enrolled, and it is difficult to defend continued taxpayer subsidies for an extremely costly product that reliably harms its customers in the $50,000-loss range.

The transition matters. A well-designed rule gives programs meaningful notice, an appeals process, a period of probation before loss of eligibility, and an off-ramp that protects already-enrolled students from stranding. Section 84001 already incorporates a version of this sequence—one year of notice, two years of failing before eligibility loss, a reapplication process after two years of ineligibility—and the stronger proposals in this paper should preserve the same protections.

“The data aren’t ready.”

They largely are. The College Scorecard has six years of matched earnings data for most Title IV recipients. The National Student Loan Data System has entering characteristics for every federal loan recipient going back to the late 1990s. Treasury has been willing to produce the tax-return linked earnings file for the Scorecard since 2015, and the computational infrastructure for running the risk-adjustment regressions exists at the Institute of Education Sciences. What is missing is the non-filer coverage and the distributional data, and both are tractable within the next two years if Congress funds the expansion.

“This is federal overreach into higher education.”

The federal government already spends approximately $150 billion a year on student aid, including grants, loans, tax benefits, and interest subsidies. The question is not whether the federal government has a role; it plainly does. The question is whether that role should continue to be one of writing blank checks without regard to outcomes, or whether it should evolve into the kind of risk-managed, outcomes-conditioned financing that characterizes every other federally backed lending market. The proposals in this paper move federal student aid toward the second model. They do not federalize the curriculum, the hiring of faculty, or the admissions process at any institution. They condition continued federal funding on outcomes the federal government is already measuring, using data the federal government already holds.

XI. A legislative package

The proposals in this paper can be enacted through a single higher-education bill. The package should include:

1. An institutional risk-sharing provision, amending HEA §454 to add a new subsection (d) that requires the secretary to construct and maintain a risk-adjustment model meeting the specifications described in Section IV of this paper, to calculate each institution’s risk-adjusted expected repayment performance annually, and to impose a risk-sharing payment on institutions whose actual performance falls short of the risk-adjusted benchmark. The payment should be capped at five percent of the institution’s Title IV revenue in the first three years and at 15 percent thereafter. Proceeds should fund a catastrophic reinsurance pool for borrowers who experience qualifying economic shocks.
2. A strengthened Section 84001, amending the new HEA §454(c) to replace the single high-school benchmark with a two-part floor that denies Title IV eligibility to programs whose median lifetime ROI is worse than negative $50,000 or in the bottom decile of comparable programs. The new test should be phased in over three years.
3. Program-level undergraduate loan limits, amending HEA §428H and §425 to cap annual borrowing at the lesser of the current statutory dollar limits and eight percent of the program’s projected median earnings at six years after entry. Programs with insufficient data for a projected-earnings calculation should be permitted to borrow at the current statutory limit for a grace period of three years.
4. Parent PLUS underwriting, amending HEA §428B to impose a debt-to-income threshold, an income-and-assets verification requirement, and a credit-score floor for all Parent PLUS originations, with a co-signer option for families who fail the primary underwriting test.
5. A reformed FFELP program, repealing the 2010 elimination of the Federal Family Education Loan Program and replacing it with a new Title IV, Part B framework that restores private origination under four conditions: five percent risk retention by the lender; capped and transparent lender compensation in the form of a flat origination fee; statutory borrower protections that travel with the loan; and an ROI screen at origination that limits guaranteed private origination to programs clearing the Section V floor. Existing Direct Loans would continue to be serviced by ED; new originations would be routed through the reformed FFELP beginning July 1, 2028, with a residual Direct Loan program remaining in place.
6. A federal ROI database authorization, amending HEA §132 to require the Department of Education to maintain a public database of program-level earnings outcomes by entering student characteristics, and amending IRC §6103(l)(13) to explicitly authorize distributional statistics at the program level. The authorization should include an annual appropriation to the Institute of Education Sciences.
7. Workforce Pell expansion, building on OBBBA Section 83002 by extending Workforce Pell eligibility to programs between 600 and 1,500 clock hours, provided they meet the same completion, placement, and value-added-earnings requirements. The current 150-to-599 clock hour cap is too restrictive to capture many of the career and technical programs with the strongest outcomes.

Taken together, these provisions finish the work that OBBBA began. They do not require Congress to raise new revenue, they do not expand the federal bureaucracy beyond the Institute of Education Sciences, and they do not federalize curriculum or admissions. They do require Congress to accept the premise that the federal taxpayer is entitled to ask whether the programs receiving federal dollars are working, and to withdraw the subsidy when the answer is very much no. That is a premise we have long accepted in every other venue in which the federal government subsidizes the private sector. It is past time to apply it to higher education.

XII. What ED can do without Congress

If Congress does not act, the Department of Education has meaningful authority under the existing Higher Education Act to move parts of this agenda forward on its own. The authority is not unlimited, and the post-Loper Bright legal landscape will constrain any aggressive reading, but several of the proposals in this paper can be attempted administratively. FREOPP scholars have previously described the contours of the secretary’s existing quality-assurance authority, and the framework below builds on that analysis.

Institutional risk sharing. HEA §487(a) conditions Title IV eligibility on an institution’s execution of a Program Participation Agreement with the Secretary, and paragraph (20) of that subsection requires the PPA to include “any other provision… necessary to protect the interests of the United States and promote the purposes of this title.” ED has used this authority to impose administrative requirements ranging from 90/10 compliance to borrower-defense claim processing. A narrow risk-sharing requirement—one that requires institutions to agree to absorb a fraction of Title IV losses above a threshold, with the fraction calibrated by a published risk-adjustment model—would fit within the existing PPA framework and would be defensible under §487(a)(20). The legal vulnerability is the major questions doctrine, but as the analysis in Section IV explains, the secretary has the better of the argument: the statutory hook is explicit, the regulatory tool is common in adjacent federal lending domains, Congress has repeatedly ratified the quality-assurance role, and the Section 84001 framework itself creates the statutory predicate.

Strengthened Section 84001 implementation. HEA §454(c)(3)(B) gives the secretary discretion to determine the source of data used to calculate the working-adult benchmark. An implementing regulation that selects data sources and adjustments carefully—for example, by requiring that the benchmark be calculated for same-age working adults who have been in the labor force for the same number of years as the program’s graduates, or by incorporating a decile-based floor—can tighten the test considerably without amending the statute. The secretary also has discretion to define “programmatic cohort” in ways that capture the full cost of a program’s financial product, including by requiring aggregation of closely related CIP codes within an institution.

Program-level loan limits. The statutory dollar caps in HEA §428H are specific and not easily overridden by regulation. But HEA §455(a)(7)(B), as added by OBBBA, authorizes institutions to impose lower limits on particular programs at the discretion of a financial aid administrator. ED can use HEA §487 to require institutions to exercise that authority consistent with ROI thresholds published by the department. This is a weaker instrument than a statutory cap, but it moves the ball.

Federal ROI database. ED already maintains the College Scorecard under the authority of HEA §132. Expanding the Scorecard to include distributional earnings data, within the bounds of §6103 confidentiality, requires only a memorandum of understanding with Treasury and an appropriation for additional data processing. Expanding coverage to non-filers requires a data-sharing agreement with state workforce agencies, which ED has attempted in the past and can attempt again under the existing SLDS framework.

Parent PLUS underwriting. The secretary’s authority to set underwriting criteria under HEA §428B is established; the question is how the secretary exercises it. A rule requiring more robust adverse-credit review, including income verification and a debt-to-income ratio, would be defensible as an exercise of the same authority ED used in 2011 and 2019.

None of these administrative routes is as clean as legislation. All of them invite litigation. But we believe they are defensible on the legal merits, and in combination they could accomplish perhaps half of what a legislative package would accomplish, on a timeline that the legislative calendar is unlikely to match.

XIII. Conclusion: Finishing what OBBBA began

Federal higher-education policy has been stuck for twenty years in a political equilibrium that satisfies no one. Progressives want loan forgiveness and expanded grant aid; conservatives want loan caps and accountability; higher-education institutions want continued federal funding with minimal strings attached; and the students whose interests the federal taxpayer is supposedly protecting end up with a system that extends them credit without regard to whether the credit is a good deal. OBBBA broke part of the equilibrium. It capped graduate borrowing, eliminated Grad PLUS, tightened Parent PLUS, created Workforce Pell with a value-added-earnings cap, and wrote the first statutory earnings test into federal accountability law. But the bill left the four-year undergraduate sector almost entirely untouched, and that is where the largest share of federal Title IV spending goes and where the largest share of negative-ROI programs sit.

The proposals in this paper finish what OBBBA started. They condition federal subsidy on outcomes, price risk rather than punishing schools for serving risky students, restore private underwriting discipline to student lending, and give both Congress and the Department of Education the data infrastructure to know what programs are working and which are not.

The intellectual foundation of our concept comes from the health insurance risk adjustment systems that have managed similar selection and information problems for decades. The empirical foundation is the College Scorecard data the Department of Education already publishes and FREOPP has already put to work in its annual ROI database.

While applying risk adjustment to student lending is a new concept, other ideas we have discussed have been around for some time. Risk sharing has been on the reform menu since 2013. Program-level loan caps have been on the menu longer than that. Parent PLUS underwriting was tightened in 2011 and loosened in 2019. A federal ROI database has been discussed in every higher-education reauthorization debate since 2017. Private origination of federal student loans existed for four decades before it was eliminated in 2010.

What is new is that the OBBBA has provided a foundation and a catalyst for further reform. The federal government has, for the first time in a generation, enacted comprehensive reforms to the financing architecture of higher education. The institutions that built the old architecture are paying attention in a way they have not in decades. And the next piece of federal higher-education legislation—whether a HEA reauthorization or a stand-alone accountability bill—will be written against the backdrop of OBBBA rather than against the backdrop of the pre-OBBBA status quo.

That is the moment to finish the job. A federal student aid system that prices risk, measures outcomes, and withdraws subsidies from programs that reliably harm their students is not a radical proposal. It is what every other federally backed lending market in the United States already does. It is what the CMS Hierarchical Condition Category model does for Medicare Advantage and what the HHS-HCC model does for the ACA exchanges. And it is what the federal taxpayer is entitled to, after sixty years of financing an education industry whose product is increasingly failing to deliver on its promise.

Higher education institutions will fight hard to preserve the blank check status quo. But the case for further reform is compelling. The federal government should not continue to extend $150 billion a year in credit to this multi-trillion-dollar industry without asking whether the credit is being repaid, whether the product it financed worked, and whether the institution cashing the check has any stake in the answer. OBBBA began the process of answering those questions. The next bill should finish it.