1,220 programs fail the OBBBA earnings test in PPD:2026. Here is the field and credential breakdown.

1,220 programs in PPD:2026 carry the official OBBBA fail flag. That is the entire list. The denominator underneath it is 44,052, not the 209,321 program rows in the file, and the difference between those two numbers is the whole story.

Divide 1,220 by 209,321 and you get 0.58 percent, a number that measures IRS privacy suppression rather than program quality. 159,461 rows, 76.18 percent of the file, have no published median earnings at all, so no test could run on them. That blank is its own subject, and why your program has no earnings number takes it apart branch by branch. Divide by 44,052, the rows that carry a real pass or fail result, and the official failure rate is 2.77 percent. 42,832 programs pass.

We are not going to re-explain the test. The short version: a program passes if the median earnings of its completers, measured four years after they left, clear a benchmark built from what people with a high school diploma or a bachelor's degree in the same field earn. The statutory math is walked through in the OBBBA workforce Pell earnings premium explainer, and if you are still untangling which acronym does what, STATS vs FVT/GE vs PPD sorts that out.

What the file does not hand you is the breakdown: which fields, which credential levels, which states, and how close to the line the failures actually sit. So we ran it. Twenty-six queries against the PPD:2026 program outcome table, 209,321 rows covering 5,096 institutions, 411 CIP-4 fields and 8 credential levels, using the official flag rather than any analytical proxy. Two validation checks came back clean: the official flag and the analytical master flag disagree on 0 of the 44,052 tested rows, and the GE-aligned benchmark differs from the official benchmark on 0 of them. The mechanics of getting the six PPD files to join correctly are covered separately in six files, one question.

Two boundaries before the numbers. Earnings here describe people who completed in academic years 2017 to 2018 and 2018 to 2019, measured four tax years later. The 2026 in PPD:2026 is a release year and nothing more. And this analysis establishes which programs trip a flag in one file. It establishes nothing about enforcement, timelines, or aid eligibility. No query we ran went anywhere near that question.

Only 49,860 of the 209,321 rows carry any published median earnings figure at all. Of those, 44,052 got an official result and 5,808 did not, because the field they sit in is out of scope for the statute. Total excluded from the test: 165,269 rows, 159,461 suppressed plus 5,808 out of scope.

That 5,808 splits into two groups ED names explicitly. 4,148 are "not covered": out of scope and clearing the bar anyway. 1,660 are "at risk": out of scope and sitting below the bar. Those 1,660 are real programs with real published earnings below a real benchmark, and they receive no official result. They are the most interesting rows in the file, and we come back to them at the end.

41,818 certificate program rows in PPD:2026 receive no official pass or fail result: 38,349 undergraduate certificates and 3,469 post-baccalaureate certificates, 100 percent of them labelled "Not Listed in Section 84001". Graduate certificates are a separate credential level in this file and they are tested, but the two big certificate tiers are not.

Better reporting would not close this gap. The statute is scoped that way. If you read the 1,220 headline and quietly assume certificates were tested and mostly passed, you have the file backwards: the 1,220 says nothing whatsoever about those two tiers, in either direction.

What sits in the gap is large. Among undergraduate certificates that do publish earnings, 1,659 of 5,723 fall below the bar, a rate of 28.99 percent. That is 24 times the bachelor's degree failure rate of 1.20 percent. Post-baccalaureate certificates look nothing like that: 1 of 85 below the bar, 1.18 percent. So when an IR office is asked "how exposed are we", the answer depends on whether the question is about the official test or about program economics. Those two questions have different denominators and different answers.

Among the six credential levels that actually get tested, the associate degree is the worst performer by a clear margin: 504 failures out of 7,642 tested, 6.60 percent. The bachelor's degree, with more than three times the tested base, produces 300 failures at 1.20 percent. That is a 5.5 to 1 gap in rate.

Master's degrees are the surprise in the middle. 377 failures out of 9,373 tested, 4.02 percent, second highest on both count and rate. Doctoral programs almost never fail (11 of 1,082, 1.02 percent) and first professional degrees rarely do (14 of 667, 2.10 percent). Graduate certificates, which are tested when the field is in scope, fail at 3.98 percent on a small base of 352.

Visual and Performing Arts is the one field that leads on both measures, and it is not close. 338 failures out of 1,993 tested, 17.0 percent, which is 27.7 percent of every failure in the file coming from a single CIP-2 family.

After that, count and rate pull apart hard, and publishing either one alone inverts the story. Health Professions is second on count with 295 failures, which sounds alarming until you see the denominator: 7,647 tested, a 3.9 percent rate, barely above the national average. Communications Technologies is ninth on count with 34 failures and first on rate among fields with a usable denominator: 145 tested, 23.4 percent. A big field running close to average, a small field in real trouble, and a table that shows only the count says the opposite of a table that shows only the rate.

Business and Marketing is the cleanest example of scale without exposure: 6,771 tested, 55 failures, 0.8 percent. Computer and Information Sciences matches it at 0.8 percent on 1,717 tested. At the floor sit Mathematics and Statistics (1 failure of 440), Physical Sciences (1 of 545) and Engineering Technologies (2 of 736). Failures do turn up nearly everywhere, though: 34 of the 40 broad CIP-2 families in the file contain at least one failing program.

Field and credential together explain more than either one does alone. Bachelor's programs in Visual and Performing Arts are the single largest failing cell, 184 failures of 1,537 tested, but at 12.0 percent they are not the sharpest. Associate programs in the same field are: 72 failures of 161 tested, 44.7 percent. Master's programs in Visual and Performing Arts sit at 30.6 percent, 81 of 265.

Health Professions splits the other way round. Master's programs there fail at 9.5 percent (173 of 1,822) while associate programs fail at 3.6 percent (91 of 2,521), so in that field the graduate credential carries more of the risk than the entry-level one. The master's credential is not uniformly exposed either: master's programs in Business, on a large tested base of 1,623, fail 18 times, 1.1 percent.

Outside Visual and Performing Arts, the high-rate cells in the top twenty by count are associate degrees in small fields. Family and Consumer Sciences at the associate level fails at 49.4 percent, 43 of 87. Communications Technologies at the associate level fails at 45.1 percent, 23 of 51. Education at the associate level, 55 of 215, 25.6 percent. Psychology at the associate level, 24 of 107, 22.4 percent. In each case the denominator is small enough that a handful of programs moves the rate, which is exactly why the count belongs next to it.

Drop to the CIP-4 level and one program type carries more failures than any other in the file. Master's programs in Mental and Social Health Services and Allied Professions fail at 38.1 percent, 75 of 197 tested: the worst single field by credential cell with a denominator large enough to trust. Rates run higher than that on thinner bases. Master's programs in Music fail 42 times out of 96, 43.8 percent, and associate programs in Human Development and Family Studies fail 42 times out of 83, 50.6 percent. Read those two as counts first.

The table below picks out the cells discussed here rather than ranking every CIP-4 cell in the file. Other cells clear the same 100-tested floor with more failures than the smallest row shown, so do not read it as a top ten.

The benchmark a program is measured against is not one number, and which one gets applied changes what failure means. 699 of the 1,220 failures, 57.3 percent, were measured against the median earnings of high school completers in the institution's own state, an average bar of $34,880. Add the 105 measured against the national high school median and 804 failures, 65.9 percent of the list, come from programs whose completers did not out-earn employed 25 to 34 year olds who are not enrolled and whose highest credential is a high school diploma.

The other third is a different kind of finding. 244 failures, 20.0 percent, were measured against the national median for bachelor's holders in the same field, an average bar of $53,021. Those are graduate programs judged against a considerably higher line. A program that came up short against the $53,021 average bar and a program that came up short against the $34,880 average bar both land in the same 1,220, and they are not the same problem.

For orientation, the two national benchmarks in the file are $34,808 for high school only and $60,112 for a bachelor's degree, both ACS 2023 medians for 25 to 34 year olds, CPI-adjusted to 2024. Five benchmark types account for nearly the whole list, and the shares below stop just short of 100 percent.

The median failing program earns $34,006 against a median benchmark of $36,082. The median shortfall is $3,142. For half the list, that is the entire distance between passing and failing, and it is smaller than most people picture.

455 of the 1,220, 37.3 percent, fail by less than $2,000. 252 of them, 20.7 percent, fail by less than $1,000. 828, 67.9 percent, fail by less than $5,000. Only 125 programs, 10.2 percent, miss by $10,000 or more. The interquartile range runs from $5,961 below the bar at the 25th percentile to $1,252 below at the 75th.

The line cuts both ways, and that is the part that matters for planning. 606 programs pass by less than $2,000, more than the 455 that fail by that margin, and 289 pass by less than $1,000. Put the two sides together and 1,061 of 44,052 tested programs, 2.4 percent, sit within $2,000 of the line in one direction or the other. A cohort composition shift, a benchmark update, or a different pooling year moves a meaningful share of them across.

One program makes the point better than the distribution does. Walden University in Minnesota runs a master's in Mental and Social Health Services whose earnings cohort is 630 people, with median earnings of $60,034 against a $60,112 benchmark. It fails by $78. At the other end of the size distribution, the largest failing program in the file is Ultimate Medical Academy in Florida, an associate in Health and Medical Administrative Services with a cohort of 6,973, $30,765 against a $34,808 benchmark, missing by $4,043. Across all 1,220 failing programs, those cohorts total 64,300 people. Cohort here means completers who were working and not enrolled four years after exit, the denominator of each earnings median rather than a count of everyone who finished the program.

Debt data thins out fast: only 754 of the 1,220 failing programs publish a median debt figure at all. Among those, the median debt is $27,000 and 177 programs, 23.5 percent, carry median debt above median earnings. Treat that comparison carefully. The debt figures are CPI-adjusted to 2019 while the earnings are adjusted to 2024, so the ratio is real but it is not like for like.

PPD reports institutional control as a text label (Public, Non-Profit, For-Profit) rather than an IPEDS sector integer, so control is the right word here and sector is not.

For-profit institutions have the highest official failure rate by a wide margin: 143 failures of 1,596 tested, 8.96 percent, which is 3.6 times the public rate of 2.47 percent. Private non-profits sit at 2.66 percent, 389 of 14,644. But rate is not volume. Public institutions supply 688 of the 1,220 failures, 56.4 percent of the list, on much the largest tested base of 27,812. Neither number is the story on its own: for-profits are not most of the list, and publics are not the worst performers.

Then include the out-of-scope certificates and the picture changes shape entirely, because for-profit institutions concentrate there. Counting every program with published earnings that sits below the bar, whether or not it is in scope, 1,268 of 3,613 for-profit programs are below the line, 35.10 percent. Public institutions come to 1,146 of 31,309, 3.66 percent, and non-profits to 466 of 14,938, 3.12 percent. The official test sees a small slice of the for-profit picture.

Read that widened gap with its denominators in mind. All three rates count only programs that publish earnings, and the share of a sector's catalog that clears IRS suppression is not the same across control: a much larger fraction of for-profit rows survives into the comparison than of public ones. So those three percentages describe differently selected slices, not three full portfolios, and that selection is part of why the gap widens once the out-of-scope rows are counted.

Geography splits the same way. California leads on count with 134 failures of 3,409 tested, 3.9 percent, and New York follows with 122 of 3,395, 3.6 percent. Both are ordinary rates on very large bases. Mississippi has the highest failure rate of any state with 300 or more tested programs: 28 failures of 435, 6.4 percent. Washington comes next at 4.1 percent, and then a cluster that rounds to between 3.7 and 3.9 percent: California, Colorado, Utah and Oregon.

Mississippi is worth a second look because it cuts against the obvious explanation. The default undergraduate benchmark is the high school wage in the institution's own state, so low-wage states get a lower bar. Of the fifty states, Mississippi has the lowest same-state high school benchmark at $30,408, against Alaska's highest at $41,356, a spread of $10,948. New Mexico and Alabama sit at $30,927, Montana at $31,564, Arkansas at $31,625. At the other end, New Hampshire is $40,791, Massachusetts $39,449, Colorado $39,174, Washington $39,073, Rhode Island $38,865. Mississippi is measured against the lowest state bar in the file and still posts the highest state failure rate. Territories are excluded here and everywhere else in this section: all six carry the same benchmark value, well below any state, which makes them a different comparison rather than a lower rung on this one.

2,974 institutions have at least one tested program. 822 of them, 27.6 percent, have at least one officially failing program. So roughly one in four institutions with a tested program is on the list somewhere, which is a much larger number of institutions than the 2.77 percent program-level rate suggests.

The shape underneath that is thin, though. 578 institutions have exactly one failing program, and those single failures account for 47.4 percent of the entire list. Another 230 institutions have 2 to 4, accounting for 553 failures. Only 14 institutions have 5 or more, and together they account for 89 failures, 7.3 percent. 2,152 institutions have a tested program and no failures at all.

For an IR office, that distribution is the operational finding. A handful of bad actors would be easier to plan around. What the file actually holds is one program at a time, at hundreds of otherwise unremarkable institutions, and that program most likely missed by less than $5,000. For scale, the earnings cohorts behind the tested universe total 3,424,109 people, and 64,300 of them sit in officially failing programs.

Back to the rows with no official result. 1,660 programs publish earnings, sit below their benchmark, and are out of scope: 1,659 undergraduate certificates and 1 post-baccalaureate certificate. ED's own vocabulary calls these "at risk", and that is the phrase to use. They are not failures. No official test was applied to them.

They are also bigger than the official list. The earnings cohorts in at-risk out-of-scope certificate programs total 175,476 people, 2.7 times the 64,300 in officially failing ones. So the file's largest below-benchmark population is the one the test does not cover.

One field dominates it. Personal and Culinary Services certificates: 868 of 980 with published earnings sit below the bar, 88.6 percent, and not one of them receives an official result. Health Professions certificates add 512 of 2,251, 22.7 percent. Business certificates, 62 of 358, 17.3 percent. The trades run far lower: Construction Trades 11 of 190 (5.8 percent), Mechanic and Repair Technologies 17 of 527 (3.2 percent), Precision Production 10 of 329 (3.0 percent), Security and Protective Services 6 of 208 (2.9 percent). One note on the table below: the field rows count undergraduate and post-baccalaureate certificates together, while the total row is undergraduate certificates only, so it will not add up down the column.

If your institution runs certificates in culinary or allied health, the official test says nothing about them and the underlying economics say quite a lot. Keep those two conversations apart, even when they land in the same board deck.

Every figure above comes from PPD:2026, release year 2026, published January 2026, queried through Clema's public data layer. Earnings are IRS-derived medians for working completers measured four tax years after exit, pooled across the academic year 2017 to 2018 and 2018 to 2019 completer cohorts, CPI-adjusted to 2024. Benchmarks are ACS 2023 medians for 25 to 34 year olds, also CPI-adjusted to 2024. Median debt is CPI-adjusted to 2019, a different base year. The full query log and schema notes are linked in Sources.

Five things to hold onto if you quote any of these numbers.

Analysis of the U.S. Department of Education's Program Performance Data (PPD:2026) release, program outcome file, release year 2026. Full reproducible query log: docs/data-evidence/programs-failing-the-earnings-test-ppd-2026-numbers/queries.md, with column definitions and the analytical-versus-official flag resolution in schema.md alongside it. Statutory context: Title VIII, Section 84001 of the One Big Beautiful Bill Act. Benchmark construction: American Community Survey 2023 median earnings. Reporting background: Federal Student Aid, Financial Value Transparency and Gainful Employment knowledge center.