Completion gaps by field: the higher a bachelor's field pays, the smaller the women's share of its graduates

Sort the 124 bachelor's fields that have at least 2,000 degrees in 2024-25 and a published earnings figure into five groups by what their graduates earn, and the women's share of completions falls from 75.7 percent in the lowest-paying group to 44.8 percent in the highest. That is a 30.9 point gradient, and it is sitting in a file every IR office already has a login for.

Then take Registered Nursing out of the top group, because it is the one large, high-paying, majority-women field in the country. The remaining 23 highest-paying fields are 27.0 percent women, 7.3 percent Black and 15.9 percent Hispanic, against bachelor's baselines of 58.2, 10.5 and 19.0 percent.

None of this is new policy or a new dataset. It is IPEDS Completions joined to PPD 2026, queried one level below the institution line. Institutional equity dashboards stop at the institution total, or at best at the college. The completions file is program-level, and the demographic detail is right there on every CIP-4 row.

This is a numbers release. If you want the policy machinery behind the earnings side, the Earnings Premium math walkthrough covers it and this post does not repeat it. For the volume side of the same file, which fields grew and which stood still, the six-year completions breakout sits alongside this one. The rest of this is baseline: which fields are still 85 percent one gender, where the race gaps are widest, what the CIP-2 families look like, and how little of it has moved in ten years.

Everything here comes from the IPEDS Completions file for the 2024-25 award year: 5,701,913 first-major awards across 403 CIP-4 fields and 273,404 rows, one per institution, program and award level. Two of those 403 codes appear only on rows reporting zero awards, so 401 is the count of fields that actually conferred something. Two filters run on every query. majornum = 1, so a dual major's credential is counted once rather than twice. cipcode <> '99', because the file carries a literal institution grand-total row per award level and summing without dropping it roughly doubles every number. If those column names are unfamiliar, the plain-English IPEDS primer is the place to start.

Every field-level ranking below is restricted to CIP-4 fields with at least 2,000 completions in the year described. The floor matters more than it sounds. Without it, the most women-skewed fields in the country are CIP 30.53 with 19 completions, 30.40 with 3 and 30.43 with 2, all at 100.0 percent women, which is a true statement and a useless one.

Read that table by column, not by row. Race shares use the known-U.S.-race denominator described above, and the earnings figure is the cohort-weighted average of program-level medians four years after exit.

The women's column is the one that moves: 75.7, 64.6, 67.1, 49.8, 44.8. It is not perfectly monotonic, the middle group ticks back up, but the distance between the ends is 30.9 points against a bachelor's baseline of 58.2 percent women.

The Hispanic column moves too, and it does fall monotonically: 20.9, 20.7, 19.4, 18.7, 16.9. That is a real gradient but a modest one, 4.0 points end to end. The Asian column runs the other way, 5.3 to 15.2 percent, a 9.9 point rise.

The Black column has no gradient at all. It runs 10.0, 11.9, 11.0, 10.6, 9.6. Highest-paying minus lowest-paying is 0.4 points and the series is not monotonic. Black under-representation in completions is field-specific rather than earnings-graded: severe in engineering, agriculture, forestry and earth science, mild or absent almost everywhere else. A dashboard that sorts fields by pay will surface the women's gap and miss the Black one entirely.

Across the 124 matched fields, the correlation between field earnings and group share is -0.618 for women, +0.585 for Asian graduates, -0.247 for Black graduates and -0.243 for Hispanic graduates. Those are field-level associations across 124 observations, descriptive and not causal. They say nothing about why any individual student picked a major.

The completions are from 2024-25. The earnings describe students who left in 2017-18 or 2018-19, measured four years later. So the cross-tab reads as today's graduating class sorted by what an earlier class in the same field went on to earn. That vintage gap is real and it has to be named rather than assumed away.

First test: rerun the correlation entirely inside PPD, using PPD's own same-cohort demographic columns and no cross-vintage join at all. This one runs on a cohort-size floor inside PPD rather than the 2,000-completion floor used everywhere else, so the field set is not identical. On 152 fields covering 1,859,335 graduates it comes back at -0.630 for women, -0.375 for Black share and -0.304 for Hispanic share. Same direction, slightly stronger. The pattern is not an artifact of the join.

Second test: the denominator. Rerun the cross-tab on the raw all-completions denominator, nonresident and unknown race included, and the Black series is 9.5, 11.2, 10.3, 9.6, 8.6 and the Hispanic series 19.8, 19.6, 18.2, 17.0, 15.2. Same shape. The denominator choice is not doing the work, which is worth knowing because it is not a neutral choice: the nonresident share of bachelor's degrees climbs from 2.5 percent in the lowest-paying group to 6.4 percent in the highest, and unknown race from 2.8 to 3.7 percent.

Third test: the cut itself. Five groups of about 25 fields each gives every field equal weight regardless of size. Cut instead by volume, five buckets each targeting a fifth of the graduates. Fields go in whole, so the realized buckets come out uneven, but the picture holds at the bottom and wobbles at the top: 76.1 percent women in the lowest bucket (29 fields, 384,221 degrees, average 48,791 dollars), then 62.0, then 60.5, then 40.7, then 50.3 percent in the highest bucket (18 fields, 387,179 degrees, average 97,038 dollars). The top bucket bounces back because bachelor's nursing lands in it, which is the next section.

Registered Nursing is 154,529 of the 513,601 degrees in the top earnings group: 89,841 dollars, 86.2 percent women, 14.5 percent Black against a 10.5 percent bachelor's baseline. It is the only field in the top group that is both large and majority women, and on the Black share it is one of the few high-paying fields that sits above baseline rather than below.

Remove it and the remaining 23 fields, still 359,072 degrees, are 27.0 percent women, 7.3 percent Black, 15.9 percent Hispanic and 18.0 percent Asian. Against baselines of 58.2, 10.5, 19.0 and 9.4 percent, that is a deficit of 31.2 points for women, 3.2 points for Black graduates and 3.1 for Hispanic graduates, with Asian graduates 8.6 points above.

For an IR office that is the most useful line in the post, because it tells you what your own high-earning program mix looks like once nursing stops carrying it. If your institution has a large nursing program, your aggregate high-earnings equity numbers are largely nursing's numbers.

Nine of those ten are computing, engineering or quantitative fields, and eight of the nine sit under the bachelor's baseline on Black share, two of them under 3 percent. The exception is Computer and Information Sciences, General at 12.1 percent on 52,157 degrees, above the 10.5 percent baseline. Pharmacy is the only field in the top ten that is majority women, at 67.8 percent, on 2,747 degrees. One note on the baseline row: it pairs the 2024-25 bachelor's completions count with the cohort-weighted average program median across PPD's 1,940,036 covered graduates, which is a different population from the 2,041,806 degrees beside it. The same program-level earnings, institution by institution rather than field by field, are on the programs directory.

The bottom of the distribution does not mirror the top as cleanly. The lowest-paying bachelor's field in the file is Religion and Religious Studies at 38,240 dollars, and it is 19.5 percent women. Just above it sit Dance at 41,390 dollars and 90.8 percent women, Teacher Education Specific Levels and Methods at 48,115 dollars, 90.3 percent women and 50,704 degrees, and Human Development and Family Studies at 49,163 dollars and 92.9 percent women. So the low-paying end is not uniformly female. It spreads wide, with religion sitting alone at the bottom, and the teaching and care fields are what pull the group average up to 75.7 percent.

Above the 2,000 floor, 38 of 220 fields are at least 85 percent one gender: 20 at least 85 percent women, 18 at least 85 percent men. Together they hold 1,430,322 of 5,602,703 above-floor completions, one in four of the graduates above the floor. Widen the line to 75 percent and it is 84 fields and 40.3 percent of graduates. Narrow it to roughly balanced, 45 to 55 percent women, and only 24 fields and 15.4 percent of graduates qualify.

The two extremes are almost exact mirrors. Communication Disorders Sciences and Services is 95.8 percent women on 22,860 completions. Heating, Air Conditioning, Ventilation and Refrigeration Maintenance is 95.8 percent men on 28,286. Same number, opposite sign.

148 of the 220 above-floor fields sit below the 12.84 percent Black baseline, and 60 of them sit below half of it. On the Hispanic side, 156 of 220 sit below the 22.66 percent baseline but only 14 fall below half. Hispanic under-representation is broad and shallow. Black under-representation is narrower and much deeper.

The deepest is Wildlife and Wildlands Science and Management at 0.8 percent Black on 2,509 completions, 12.0 points below baseline. Nine of the ten lowest sit in agriculture, forestry, environmental science, earth science or engineering; the tenth is zoology. That is a tight enough cluster to be a finding rather than a coincidence. Just outside the ten sit the two fields in this group that carry real volume: Natural Resources Conservation and Research at 3.3 percent on 23,942 completions and Physics at 3.4 percent on 13,129.

At the other end, Ground Transportation is 32.1 percent Black on 19,489 completions, 19.3 points above baseline. Health and Medical Administrative Services is 27.1 percent on 88,893, Human Services General 26.0 percent on 12,186, Practical and Vocational Nursing 23.8 percent on 110,132, and HVAC 22.3 percent on 28,286: the same HVAC that is 95.8 percent men. Fields are rarely skewed on one axis only, and the axes do not line up.

Low pay does not predict high Black representation either. Among bachelor's graduates, Ecology and Evolutionary Biology is 2.3 percent Black and Zoology and Animal Biology 2.9 percent, and both sit among the lowest-paying bachelor's fields in the country.

The Hispanic extremes are wider than the Black ones in both directions. Health Aides, Attendants and Orderlies is 71.2 percent Hispanic, 48.5 points above baseline. Religion and Religious Studies is 3.6 percent, 19.1 points below, and the three lowest Hispanic shares above the floor are all theology fields. The low-Hispanic fields with real volume behind them sit a little further up the list: Teacher Education, Specific Subject Areas at 11.3 percent on 39,764 completions, Medicine at 11.1 percent on 30,803 and Data Science at 11.0 percent on 11,111.

That is the table worth keeping. When a dean asks whether their program's demographics are unusual, the useful comparison is not to the institution and not to the nation, it is to the field's own CIP-2 family. One caution on reading it: the completions column is a raw total, while the race columns are computed on the known-U.S.-race denominator, so those percentages are not shares of the number to their left.

Two families sit at the bottom on Black share: Agricultural and Veterinary Sciences at 4.7 percent and Engineering at 4.9 percent, both about a third of the 12.84 percent national baseline. Engineering is also 26.7 percent women, so it is under baseline on both axes at once. Computer and Information Sciences is close to Engineering on gender at 27.9 percent women, but its Black share is 14.7 percent, above baseline. The two families that get talked about together do not have the same problem.

At the other end, Family and Consumer Sciences is 91.7 percent women, Culinary and Personal Services 85.3, Health Professions 81.8, Education 80.9 and Public Administration and Social Service 80.8. Construction Trades is 7.1 percent women and Mechanic and Repair Technologies 8.6.

Apply the same 2,000 floor inside 2014-15 and 43 fields were at least 85 percent one gender, 20 women and 23 men, covering 28.5 percent of above-floor completions. In 2024-25 it is 38 fields, 20 women and 18 men, covering 25.5 percent. The whole net decline came off the men's side. The overall women's share of above-floor completions is 58.8 percent in both years, so the aggregate did not move at all.

Of the 23 fields that were at least 85 percent men in 2014-15, 22 still clear the volume floor. Twenty-one of those 22 moved toward women, and seven crossed below the 85 percent line: Computer Engineering Technologies 87.8 to 79.5 percent men, Computer Systems Networking 86.1 to 79.9, Aerospace Engineering 86.7 to 80.2, Mechanical Engineering 86.5 to 81.5, Engineering Technologies General 87.1 to 82.5, Construction Engineering Technology 90.2 to 83.1, and Industrial Production Technologies 87.1 to 83.6. Every one of the seven is an engineering or technology field.

The trades barely moved. HVAC went from 97.6 to 95.8 percent men, Electrical and Power Transmission Installers 96.6 to 94.9, Heavy and Industrial Equipment Maintenance 95.2 to 93.3. Environmental Control Technologies went from 93.1 to 93.9, the only one of the 22 that became more male.

The women's side is the opposite story. Of the 20 fields that were at least 85 percent women, 19 still clear the floor, exactly one crossed below the line (Specialized Sales, Merchandising and Marketing Operations, 85.5 to 82.8 percent), and ten of the 19 became more female, among them Dance 87.5 to 89.4, Health Aides 90.0 to 91.1, Teaching Assistants 92.7 to 93.8, Communication Disorders 94.9 to 95.8, Teacher Education Specific Levels and Methods 86.6 to 87.7, Cosmetology 90.0 to 90.5 and Practical and Vocational Nursing 88.1 to 88.7.

None of that is a CIP reclassification artifact. Volumes across the five snapshots are continuous, with no step change that would signal a code split or merge: nursing 268,139 to 317,986, computer science 27,671 to 88,326, mechanical engineering 35,505 to 43,624, vehicle maintenance 70,094 to 81,616, cosmetology 119,696 to 133,218, teacher education 102,770 to 107,537, social work 50,894 to 57,709, HVAC 18,219 to 28,286, against national totals of 4,952,795, 5,124,287, 5,316,848, 5,299,203 and 5,701,913.

Two datasets. IPEDS Completions at data_year 2025, which is the 2024-25 award year, source file C2025_A with no revision suffix. Every trend point comes from a revised file: C2015_A_RV, C2018_A_RV, C2021_A_RV, C2023_A_RV. So the latest year is provisional, and small revisions to the 2024-25 figures are normal. CIP titles come from a static lookup table, not from the completions file itself.

Earnings are PPD release_year 2026, variable md_earn_wne_p4: median earnings of working completers pooled across the 2017-18 and 2018-19 exit cohorts, observed four tax years later, adjusted to 2024 dollars. The completions side has ten years of history behind it; the earnings side has one snapshot and nothing to compare it against, so every gradient here is a cross-section.

The bachelor's earnings universe is 24,936 programs at 1,721 institutions covering 1,940,036 graduates, with a cohort-weighted average program median of 67,925 dollars. Coverage is uneven by credential level, which is why the cross-tab is bachelor's only: post-baccalaureate certificate 2.5 percent, graduate certificate 4.2, doctoral 9.1, undergraduate certificate 14.9, associate 17.8, master's 27.9, bachelor's 36.2, first professional 40.5. Pooling levels would average a certificate median against a doctoral median inside the same field, so we did not.

Joining the two: 2,041,806 bachelor's degrees were conferred in 2024-25 across 373 CIP-4 fields. 126 of those fields clear the 2,000 floor, 1,933,475 degrees. 124 of the 126 also carry a PPD earnings figure, 1,927,936 degrees, 94.4 percent of the national bachelor's total. The two that do not match are Data Science (3,174 degrees) and Data Analytics (2,365), both too new to have a cohort measured four years after a 2018-19 exit. They are excluded from every quintile figure in this post, and they are exactly the two fields you would most want in it.

Imputation is not a material source of error here. 273,377 of 273,404 first-major program rows in 2024-25 carry flag R, covering 5,701,632 of 5,701,913 awards. Everything else is 27 rows and 281 awards: 22 carried forward from a prior year, 5 analyst-corrected, 0.005 percent of awards between them. On the demographic columns the only other frequent flag is Z, which marks a structural zero rather than an imputation. xcbkaat carries Z on 78,103 rows, xchispt on 63,011, xctotalw on 23,278. One honest caveat: the x-prefixed flag columns are not documented in the IPEDS data dictionary, so the R, Z, P and C reading is our interpretation rather than a published definition.

The full query log lives in this repository at docs/data-evidence/completion-gaps-by-field-who-finishes-what/queries.md, with the SQL and returned rows for every figure above, and the schema notes and standing filters in schema.md alongside it. If you would rather not write the SQL at all, here is what asking a question like this in plain English looks like.

IPEDS, use the data. IPEDS survey components and documentation, NCES. Classification of Instructional Programs. Financial Value Transparency and Gainful Employment knowledge center, Federal Student Aid. College Scorecard data documentation.

Reproducible query log for this post: docs/data-evidence/completion-gaps-by-field-who-finishes-what/queries.md and schema.md.