Which Programs Don't Pay Off? Inside the 2026 Federal Earnings Test Data
The Department of Education's 2026 Program Performance Data identifies 2,880 college programs whose graduates earn less than the regional benchmark for high school grads. Here's where they are, what they teach, and what the data leaves out.
Update, July 8, 2026: This post was first published on April 25, 2026, before the final rule was published. It has been updated to reflect the STATS Earnings Accountability final rule (July 1, 2026), including the graduated consequence ladder, the corrected institution-level trigger, and the new exemptions. The underlying data analysis is unchanged.
Under the STATS Earnings Accountability framework, enacted through the Working Families Tax Cuts Act (originally the One Big Beautiful Bill Act), the Department of Education now compares the median earnings of program graduates to a regional benchmark of high school earnings. Programs that fall short face sanctions, and if enough programs at an institution fail, the entire school can lose federal financial aid. (For a full walkthrough of the rule, see How the New Federal Earnings Test for Colleges Works.)
We analyzed the Department’s own data, the 2026 Program Performance Data covering 209,321 programs at more than 5,000 institutions, to see which programs are on the wrong side of that line.
The Numbers
Of the roughly 50,000 programs with enough graduates to measure, 2,880 fail the earnings test, a 5.8% failure rate. Their graduates earn, on average, $31,600 per year. Passing programs average $68,200, more than double.
The failing programs aren’t small: they collectively enroll an estimated 644,000 students in their most recent year.
Most failing programs fall short by a relatively modest amount: the median shortfall is about $3,100 below the benchmark. But some miss by $20,000 or more.
A Universe of Programs
The clearest way to see the test is to plot every testable program at once. The chart below shows each program as a dot, with the diagonal line representing the federal benchmark for that program’s credential and field. Programs below the line fail. Programs above it pass. Dots are colored by credential type and sized by enrollment, so a single dot can represent anything from a small certificate to a four-year degree with thousands of students.
The cluster of amber dots below the line is the largest single feature of the chart. Those are undergraduate certificates, which fail at a much higher rate than degree programs of any level. The blue dots (bachelor’s and associate degrees) and teal dots (graduate degrees) sit mostly above the line, with the exceptions concentrated in a few specific fields we’ll come to next.
Where the Failures Concentrate, by State
Failure rates aren’t evenly distributed across the country. Louisiana has the highest at 11%, followed by Mississippi (10%), Utah (9%), California (9%), and Florida (9%). Several of those states have large for-profit sectors or extensive community college certificate offerings, both of which are overrepresented in the failing set.
State rates reflect the mix of institutions and programs in that state, not policy quality. A state with a heavy concentration of short-term workforce certificates will have a different failure profile than one whose graduate population skews toward four-year degrees, even if individual programs are performing comparably.
Which Fields Don’t Pay Off
The chart below ranks fields by the number of failing programs, and personal and culinary services (cosmetology, barbering, culinary arts) leads with 893. Its failure rate is in a class of its own: 79% of testable programs in that field fail, the highest of any field. By raw count, though, health professions is nearly tied at 807, followed by visual and performing arts (363) and business (117). One caveat sits behind that top bar: many personal and culinary services programs, cosmetology, barbering, and bartending among them, fall under the final rule’s tipped-occupation delay, so they won’t face pass/fail determinations until the earnings test uses post-2026 tax data, a reprieve of at least one year.
Click into any bar to see the sub-fields driving the totals. Personal and culinary services breaks down into cosmetology (the largest source of failing programs in the entire dataset), barbering, and a long tail of related certificates. The “health professions” category is more heterogeneous: most of its failing programs are short-term certificates like nursing assistant or medical assistant, not the longer health-degree programs.
Who Fails: Sector and Credential
The other slicing dimension is who offers the program. For-profit institutions account for 44% of all failures despite representing a much smaller share of total programs. And undergraduate certificates are the epicenter, with 58% of all failing programs sitting in that single credential category.
The combination matters more than either dimension alone. A for-profit undergraduate certificate is the single most common shape of a failing program in this dataset. The sanctions in the new rule, which scale with how many of a school’s programs fail, are most likely to bite at institutions whose business model concentrates in this corner.
How Far Below the Benchmark
Failing isn’t binary in a practical sense. A program that misses by $500 looks very different from one that misses by $20,000. The histogram below shows the distribution of gaps, with negative values to the left (failing) and positive values to the right (passing).
The bulk of failures clusters within a few thousand dollars of the line. Those are programs where graduates earn near the benchmark, and small changes in the cohort, the benchmark itself, or the regional comparison group could push them across in either direction. The long tail of severe failures (programs missing by $10,000 or more) is much smaller in count, but those are the cases where the test most clearly identifies programs whose graduates would be better off without the credential.
What You Can’t See
There’s an important caveat: 76% of programs have no earnings data at all. The IRS suppresses earnings when a program has 15 or fewer graduates in the measurement cohort. That means the testable universe skews toward larger, more established programs. Small programs, which may be more likely to have poor outcomes, are invisible in this data.
This doesn’t invalidate the findings, but it means the 2,880 failing programs are likely a lower bound. The true number of programs where graduates earn less than the benchmark is almost certainly higher.
The Backward-Looking Problem
There’s an uncomfortable reality embedded in this test, one that Phil Hill has called a fundamental flaw: the data being used to judge programs today reflects students who graduated years ago.
The PPD:2026 earnings data is based on students who completed their programs in the 2017-18 and 2018-19 award years. Their earnings are measured four years after exit, in calendar years 2022 and 2023. For a student in a four-year bachelor’s program, that means they may have enrolled as early as fall 2014. The first formal notifications to institutions won’t arrive until early 2027, and the first sanctions don’t take effect until the 2028-29 academic year. That is a span of up to 14 years from enrollment to consequence.
Institutions cannot retroactively change those outcomes. A cosmetology certificate program flagged as failing in 2027 is being judged on graduates who entered the workforce in 2018 and 2019, before the pandemic reshaped labor markets, before generative AI began altering entry-level job prospects, and before anyone knew this test was coming. The benchmark itself raises questions: the “high school earnings” figure comes from self-reported Census data, which includes workers with significant postsecondary training like apprenticeships and skilled trades, potentially inflating the bar that programs must clear.
For the small handful of institutions where half or more of their programs fail, that creates a genuinely existential bind: they face provisional certification based on outcomes they had no opportunity to influence under these rules. For everyone else, the more pressing question may be whether the threat of a failing program reshapes how schools talk about, price, and recruit into their lowest-earning offerings, before the sanctions ever arrive.
What Happens Next
Programs that fail the earnings test don’t lose federal funding immediately. The consequences are graduated. After a first failure, the institution must warn current and prospective students, and it then has 120 days to choose one of three paths: voluntarily block Direct Loans for the program for at least five years, which shields it from institution-level sanctions while preserving Pell eligibility; begin an orderly teach-out and close the program; or take no action and face the test again. Once a program fails a second time, the first two options are off the table.
A program that fails in two of any three consecutive years is designated a “low-earning outcome program” and loses eligibility for federal Direct Loans. Appeals are limited to calculation errors, not program quality or alternative data. Because a program must fail twice first, the earliest Direct Loan losses cannot occur before Academic Year 2028-29, which gives institutions a window to improve outcomes or wind programs down.
As Robert Kelchen has noted, the earnings test doesn’t operate in isolation. Starting July 2026, new borrowing caps limit graduate students to $100,000 in total federal loans (down from unlimited Grad PLUS borrowing), while professional programs are capped at $200,000. For institutions already facing earnings test failures, the simultaneous loss of loan flexibility compounds the financial pressure on the very programs most likely to be affected.
The final rule also adds an institution-level trigger. If an institution’s low-earning outcome programs account for more than half of its Title IV recipients, or more than half of its Title IV dollars, in two of three consecutive years, the institution is placed on provisional certification and those programs lose eligibility for all Title IV aid, including Pell Grants. This is the framework’s most severe consequence, aimed at institutions where low-earning programs are the core of the portfolio rather than isolated outliers.
Finally, the final rule carved out several exemptions. Programs for tipped occupations, a list of 20 specific six-digit CIP codes spanning cosmetology and barbering, bartending, casino gaming, and massage, get a delayed timeline before determinations begin. Institutions that have not used Direct Loans for five years are exempt from the institution-level trigger. Institutions that exclusively serve students with specific learning disabilities or autism are exempt from eligibility consequences. And approximately 2,650 graduate programs where the benchmark data is too thin to be reliable default to a $1 earnings threshold, which effectively guarantees a pass.
The reach of the framework over certificate programs is contested. The American Association of Cosmetology Schools has fought earnings-based accountability in court for years; its challenge to the Department’s 2023 Gainful Employment rule, built on the argument that federal earnings data undercounts the cash and tip income of beauty-school graduates, is now before the Fifth Circuit (No. 25-11303) after a district court upheld the rule in October 2025. AACS makes a second argument against the new framework: that Congress excluded undergraduate certificates from the WFTCA’s earnings-accountability provisions, and that the Department cannot re-impose them under its older Gainful Employment authority. The stakes are large. Undergraduate certificates make up 58% of all failing programs in this data, and the Department’s own projections show more than 90% of cosmetology and barbering programs would fail. If certificates are ultimately carved out, the failure count in this analysis roughly halves, to around 1,200. The final rule’s severability provision means the framework for degree programs would stand regardless.
Methodology
This analysis uses the 2026 Program Performance Data (PPD:2026), released by the U.S. Department of Education for the AHEAD negotiated rulemaking process. Earnings are IRS-sourced median earnings measured four years after program exit for the pooled AY 2017-18 and 2018-19 completer cohorts, adjusted to 2024 dollars using CPI-U. Benchmarks are derived from 2023 American Community Survey 5-year estimates. The “master” fail flag (mstr_obbb_fail_cip2_wageb) from the corrected January 2, 2026 release is used. Programs with suppressed earnings data (15 or fewer tax filers) or fewer than 30 completers are excluded from pass/fail calculations.
The PPD:2026 data is informational and uses a pre-COVID completer cohort (AY 2017-18 and 2018-19). The first operational earnings test, expected in early 2027, will use AY 2020-21 and 2021-22 completers with CY 2025 earnings, a different cohort that completed during the COVID-19 pandemic. The operational test will also run at CIP6 granularity (vs. CIP4 in the PPD) and apply multi-year cohort aggregation for small programs. Pass/fail outcomes may differ from the PPD results shown here.
Data source: U.S. Department of Education, Office of the Chief Economist and Office of Federal Student Aid. (2026). Program Performance Data (PPD:2026) for AHEAD Negotiated Rulemaking.