Graduates per opening: matching IPEDS completions against BLS demand across 17 CIP to SOC pairs

Registered nursing produces 1.39 graduates for every projected annual opening. Paralegal studies produces 0.25. Across the 17 occupations we could honestly map, twelve produce fewer graduates each year than the labor market is projected to absorb, and the spread from top to bottom is 5.6 fold. Sixteen of the 17 are jobs where a specific postsecondary credential is effectively required to be hired. The seventeenth, accountants and auditors, is in the set on purpose, as the pair that shows what a broken mapping looks like.

That is the whole result, and it took an afternoon of querying to produce and a week of arguing with ourselves about whether it means anything. It does, but only with the qualifications in this post attached. The two halves of the ratio come from federal datasets that were never built to sit next to each other and share no key. IPEDS Completions counts awards conferred by institutions in a twelve month window. BLS Employment Projections counts jobs the economy is expected to need to fill each year over a decade. Bridging them requires the NCES CIP to SOC crosswalk, which tells you which fields of study CAN prepare someone for which jobs, not how many graduates actually went there. That crosswalk is published on this site in both directions, all 415 CIP-4 entries with the occupations each one feeds and the reverse view, SOC to CIP, so you can check any pair below without taking our word for it.

We built the bridge anyway, published the pairs we rejected alongside the pairs we kept, and then spent most of the analysis trying to break our own number. The state rankings are where it breaks first. Utah appears to produce 7.09 nursing graduates for every allocated opening, which sounds like a policy story and is really an artifact of where Western Governors University is incorporated.

The numerator is IPEDS Completions, data_year 2025, which is the 2024-25 award year: first major awards at all levels conferred between July 1 2024 and June 30 2025. That numerator counts awards, not people. Someone who earns an associate degree in one award year and a bachelor's in a later one lands in it twice, and nothing in the file connects the two records. The denominator is the BLS average annual openings figure from the 2024-34 projections round, which combines growth, replacement of workers who leave the occupation, and transfers out.

The SOC side does not live in the SQL warehouse. There is no BLS employment, wage, openings or crosswalk table anywhere across the eleven public datasources (ipeds, score_card, campus_crime, eada, pseo, pell, cdr, dapip, ppd, synth_private_data, scc_cohort). Zero tables. So the occupation figures come from the Clema occupations directory, a Cloudflare KV build covering 648 detailed SOC occupations across 22 SOC major groups, 51 states and 21,548 state pages, which carries the OEWS employment and wage block, the projections block, and the NCES crosswalk on every record. You can read the same records on the occupations section of this site and the study side on the programs section.

Every record in that build declares its own vintage, and the declaration is identical on all of them: OEWS release 2025, projections 2024-34, IPEDS year 2025, IPEDS survey year 2024, O*NET 30.3, crosswalk CIP 2020 to SOC 2018, and IPEDS Completions labelled 2024-25 (provisional). We checked the IPEDS half against the warehouse directly rather than trusting the label, and all 17 anchor CIP completion counts reproduced to the unit.

Each occupation was then assigned one anchor CIP: the single program code that supplies the overwhelming majority of its plausible graduates. That is the compromise at the center of this analysis, and it costs a great deal more in some pairs than in others.

Read the last two columns of that table before you read any of the others. The NCES crosswalk is many to many in both directions, and how many CIPs it maps to a given occupation varies enormously: one CIP for dental hygienists, dental assistants, occupational therapists, respiratory therapists, occupational therapy assistants and physical therapist assistants; 21 for registered nurses; 25 for pharmacists and for veterinarians.

Exactly one pair in the set is genuinely one to one. CIP 51.0806 is the only CIP feeding SOC 31-2021 Physical Therapist Assistants, and 31-2021 is the only occupation that 51.0806 feeds. Everywhere else there is leakage in at least one direction. CIP 51.3801 Registered Nursing, for instance, maps to two occupations: registered nurses and postsecondary nursing instructors.

The anchor share column is how much of the leakage we are absorbing. It runs from 100 percent, where the anchor is the whole crosswalk, down to 69.4 percent for paralegals. For the eleven pairs below 100 percent the ratio in the table is a floor, not a point estimate, because it counts only the anchor CIP graduates against all of the occupation openings. If you divide by the anchor share to approximate the all feeder pool version, registered nurses moves from 1.39 to 1.67 (1.39 divided by 0.835) and radiologic technologists from 1.01 to 1.35 (1.01 divided by 0.748). Neither of those is a better number. It is a differently wrong number, because the full feeder pool counts graduates who were never going to sit for that licence. We show the conservative one and tell you the other exists.

This is the same class of problem as any local reporting term carrying several valid definitions at once, which we wrote about in why "enrolled student" has five definitions. The crosswalk is a definitional layer sitting on top of clean data, and it is where the analysis actually gets hard.

Everything from here down is state level, and BLS publishes no state level annual openings in this data. Every state occupation record in the KV build repeats the national outlook block verbatim. The Texas registered nurses record reports 189,100 annual openings, which is the national figure. The Ohio dental hygienists record reports 15,300, which is also the national figure.

So we built the state denominator ourselves, and you should treat it accordingly. Allocated openings equal national annual openings multiplied by the state's share of national OEWS employment in that occupation. Texas registered nurses: 189,100 times 271,380 divided by 3,379,720, which is 15,184 allocated openings against 20,080 Texas graduates, a ratio of 1.32.

That allocation assumes every state has the same separation and transfer rate, which is an assumption, not a measurement. States with older workforces retire more nurses per employed nurse than states with younger ones, and the allocation cannot see that. Every state ratio in this post inherits that assumption. We are putting this here, in the body, rather than in a footnote, because the state rankings are the part people will screenshot.

Utah reports 7.09 graduates per allocated opening. Alaska reports 0.48. That is a 14.8 fold spread across 50 states and DC for a single occupation with a single anchor CIP, one national licensure exam and broadly similar accreditation. A spread that wide says something structural is wrong with the measurement. Utah has not solved nursing supply.

The bottom of the ranking is just as suspect. Massachusetts at 0.85, Hawaii at 0.84 and Oregon at 0.81 all have large nursing programs. They look thin here because the denominator is scaled to their large employed nurse workforces, not because their colleges stopped producing nurses.

The rank column runs 1 to 51 across 50 states plus DC. The bottom block is ranks 45 to 51 with Alaska last. Territories are excluded: Puerto Rico alone reports 2,769 awards in CIP 51.3801, but PR, VI, MP, MH and FM carry no OEWS employment in these records, so there is no denominator to divide by. Dropping them is why the national ratio in the state table arithmetic is 1.38 rather than the 1.39 in the pairs table.

IPEDS books a completion to the institution's home state, not the student's residence. In a sector with large online providers that is not a rounding issue. It is the single largest distortion in these rankings.

Two online universities headquartered in Minneapolis account for three quarters of Minnesota's nursing pipeline: Capella University with 9,210 awards, 51.5 percent of the state, and Walden University with 4,224, another 23.6 percent. Together 13,434 of the state's 17,869 awards, 75.2 percent. Neither awards a single associate degree. Western Governors University is 4,940 awards, 45.4 percent of Utah's 10,871, all at bachelor's level; add Joyce University (1,474, 13.6 percent) and Nightingale College (1,270, 11.7 percent) and three institutions are 70.7 percent of the state. Chamberlain University-Illinois is 58.6 percent of Illinois. Grand Canyon University is 41.9 percent of Arizona.

So any sentence of the form "Utah trains 7.09 nurses for every Utah opening" is false. Utah hosts the mailing address of an institution that awards nursing degrees to students in all 50 states. The same is true of Minnesota, and to a lesser extent Illinois and Arizona. Read the top of that ranking as a map of where distance education providers are incorporated, because that is what it is.

There is a second problem underneath the geography one, and it is worse because it also affects the national number.

CIP 51.3801 awarded 263,325 credentials in 2024-25: 148,864 bachelor's degrees (56.5 percent), 88,840 associate degrees (33.7 percent), 19,622 master's, 4,191 sub baccalaureate certificates, 1,044 doctoral and 764 post baccalaureate certificates. Those are six digit figures and they will not match a four digit one. The whole 51.38 family, which is what the field-level completions breakout reports, comes to 317,986 awards and 154,529 bachelor's degrees, because it picks up nurse anesthesia, nurse midwifery and the other 51.38xx codes that the crosswalk does not route to SOC 29-1141. An unknown share of those bachelor's degrees are RN to BSN completions earned by people who are already licensed and already working as nurses. They are not new entrants to the occupation. They are existing nurses upgrading a credential, and counting them in the numerator of a supply ratio double counts the workforce.

IPEDS cannot separate the two. There is no field on the completions table that distinguishes a pre licensure BSN from a degree completion BSN. Both are award level 5 in CIP 51.3801. We looked; the column inventory has nothing, and the table details call returns all 71 columns with null descriptions.

The robustness check is to rebuild the ratio using only associate degrees, which are almost entirely pre licensure. Across the 50 states and DC, the ratio falls from 1.38 to 0.47, and associate degrees are 33.8 percent of that pipeline: 88,103 awards against 260,493 at all levels. The state ranking does not survive at all: Utah falls from 7.09 to 1.30, Minnesota from 4.56 to 0.42, Illinois from 1.90 to 0.33.

If the nursing ratio is contaminated by credential upgrading, the obvious move is to look at an occupation where upgrading barely happens. Radiologic technology is that occupation. CIP 51.0911 awarded 13,035 credentials in 2024-25: 8,896 associate (68.2 percent), 2,315 sub baccalaureate certificates (17.8 percent), 1,688 bachelor's, 134 master's and 2 doctoral. That is 86.0 percent at associate level or below, against a BLS typical entry requirement of an associate degree. There is very little room for a post licensure degree completion market to distort it.

The national ratio is 1.01, which is about as close to balance as this method ever gets. The state spread is still wide, from Utah at 3.61 to South Carolina at 0.15, and Utah is again at the top, but the rest of the ordering looks much more like a story about program capacity: Texas at 1.94, Idaho 1.85, Indiana 1.81 at one end; Maryland 0.17, New Jersey 0.29, Wisconsin 0.37 at the other.

The market those graduates enter is 230,490 radiologic technologists employed nationally at a median wage of 80,110 dollars. Registered nursing is 3,379,720 employed at a median of 97,550 dollars.

Dental hygiene and physical therapist assisting are the two places where nearly every state points the same way as the national number. That consistency is what makes them worth reporting.

Dental hygienists: 44 of the 49 states with both sides populated fall below 1.00. The five that clear it are Wyoming at 1.83, Kentucky 1.14, West Virginia 1.12, New Mexico 1.05 and Utah 1.01. The bottom is Alaska at 0.18, Vermont 0.26, Alabama 0.26, New Hampshire 0.32, Delaware 0.33 and New Jersey 0.33. CIP 51.0602 is 8,920 awards across 366 institutions, 6,051 associate (67.8 percent), 2,471 bachelor's, 297 certificates, 100 master's, against a BLS typical entry of an associate degree. Nationally 222,740 dental hygienists are employed at a median of 98,100 dollars.

Physical therapist assistants: not one of the 48 states with both sides populated reaches 1.00. The highest is Idaho at 0.68 and the lowest New Hampshire at 0.06. This is also the single cleanest pair in the set, the only genuinely one to one crosswalk, and its award mix is 5,137 associate degrees out of 5,489 (93.6 percent), 274 certificates and 78 bachelor's, from 399 institutions. National employment 112,430, median 68,380 dollars.

Of everything in this post, the physical therapist assistant number is the one we would defend hardest. One CIP, one SOC, no leakage, no credential upgrade market, an associate degree entry requirement matched by an associate degree pipeline, and a ratio of 0.28.

A supply ratio only means something if the graduates in the numerator turn up in the occupation in the denominator. Census PSEO gives a partial answer by tracking where completers are employed one year out, by industry sector.

Health professions graduates do turn up where you would expect. One year after graduation, 85.6 percent of CIP 51 associate completers (678,769 of 793,159) and 77.5 percent of bachelor's completers (693,666 of 895,508) are employed in NAICS 62, Health Care and Social Assistance. Certificate completers are at 57.7 percent and master's completers at 67.1 percent. Architecture concentrates too: 52.9 percent of CIP 04 bachelor's completers and 60.6 percent of master's completers enter NAICS 54, Professional, Scientific and Technical Services. Engineering sits in between, with 31.3 percent of CIP 14 bachelor's completers in manufacturing and 28.5 percent in professional and technical services, 59.8 percent across the two.

The check has a hard limit. PSEO flows in this database carry no 4 digit CIP. The cip_level values are only "2" (1,324,631 rows) and "A" (1,125,750 rows), and a query filtering on cipcode 51.38 returns zero rows. So this is not "do nursing graduates become nurses". It is "do health professions graduates work in health care", which is a weaker claim about a wider group. PSEO also covers 32 states and 950 institutions, not the nation, and its industry detail exists only for the pooled all cohorts record, so it has no single year attached.

Accountants and auditors is in the table on purpose, as the demonstration of what a broken pair looks like. Its ratio is 0.49, and you should not repeat that number as a fact about accounting.

Two things kill it. First, CIP 52.0301 feeds eight different SOC occupations, more than any other anchor in the set, so the graduates in the numerator are spread across jobs the denominator does not include. Second, PSEO shows business bachelor's graduates scattering more widely than any other field measured: the largest single destination sector takes only 18.3 percent of them (340,768 of 1,861,445 into professional and technical services), with finance and insurance at 16.3 percent, retail trade at 9.3 percent and manufacturing at 7.0 percent. There is no second sector above 16.3 percent. Compare that with health professions associate completers at 85.6 percent into a single sector.

Concentration in the destination is the precondition for this ratio to mean anything. Where a field disperses, "graduates per opening" is measuring a numerator and a denominator that describe different populations, and the arithmetic still returns a tidy number. That is the trap. Check destination concentration before you compute the ratio, not after.

Numerator: IPEDS Completions, data year 2025, award year 2024-25, file C2025_A, awards conferred July 1 2024 to June 30 2025, labelled provisional in the occupation build. Institution geography from the IPEDS HD directory, data year 2025.

Denominator: BLS Employment Projections, 2024-34 round, base year 2024, average annual openings. Employment and wages from BLS Occupational Employment and Wage Statistics, May 2025 release.

Bridge: the NCES CIP to SOC crosswalk, CIP 2020 to SOC 2018, published on this site as 415 CIP-4 entries and as the SOC to CIP reverse view, read from the Clema occupations records at clema.ai/occupations and matched to the study side at clema.ai/programs.

Destination check: Census Post-Secondary Employment Outcomes, pooled cohort, 32 states and 950 institutions.

Reproducible query log: docs/data-evidence/graduates-per-opening-cip-completions-vs-soc-demand/queries.md and schema.md in this repository, covering all 15 numbered query sections and the eight discarded lines of inquiry.