A third of colleges cut a team, and the team was small for years first

Between AY 2014-15 and AY 2024-25, 549 of 1,715 continuously filing colleges dropped at least one varsity team. That is 32.01 percent, roughly one institution in three. Over the same eleven years those 1,715 institutions went from 21,206 teams to 23,198, a gain of 9.4 percent, and the average portfolio grew from 12.37 teams to 13.53.

Both sentences are true, and the second is why the first goes unnoticed. Adds outnumbered drops 3,135 to 1,143, and 261 of the 549 cutters finished the decade flat or larger, so their cut never appears in their net number at all.

What goes first is not spread evenly. Tennis and golf together supplied 463 of the 1,143 drops, 40.5 percent of every team a continuously filing college dropped, out of 31 sport families. Men's tennis was dropped by 15.7 percent of the panel institutions that fielded it. Men's basketball, by 0.6 percent.

And the teams that go are small long before they go. Inside the roster distribution of its own sport, sex and year, a team in its final season sits at the 9.4th percentile; a survivor sits at the 46.1st. That gap is open four years out, which makes it the one finding here an IR office can act on. The rest of this post earns the right to state it, because three separate things in this file will hand you a wrong answer if you count naively.

Only institutions that sponsor intercollegiate athletics file under the Equity in Athletics Disclosure Act, so every rate below has an athletics-sponsoring denominator rather than a US-colleges one. Across the eleven-year window 2,220 institutions appear at least once, and only 1,715 of them (77.3 percent) filed in all eleven years.

That distinction carries the analysis. 2,073 institutions filed in AY 2014-15 and 2,037 in AY 2024-25, but only 1,910 filed in both: 163 left and 127 arrived. An institution vanishing from the file is not a college cutting a team, and scoring it as one would have more than doubled the headline while blending closures, shutdowns and filing lapses into one number. So the panel gets built first, and the 163 leavers are reported separately at the end.

A team has to be constructed too, because EADA publishes no team count. One row covers both sexes, so a team is the triple of institution, sport and sex with participants above zero. Roughly half of all sport rows are single sex (3,514 men only and 5,067 women only, of 18,045 rows in AY 2024-25). No row in AY 2014-15, AY 2019-20, AY 2020-21 or AY 2024-25 carried zero participants on both sides, so nothing is lost. The EADA data source page covers the file layout.

Loosen the panel to institutions that merely filed in both endpoint years and you get 1,910 institutions at a 33.87 percent cutter rate, with the same shape underneath: 930 net grew, 625 flat, 355 net shrank. That looser 1,910 set is the panel the sport-level companion post runs on, which is the right choice for measuring how a sport moved and the wrong one for measuring who cut it. Where a figure appears in both posts on different bases, that is usually why.

1,844 institutions filed for AY 2020-21 against 2,073 the year before, with 14,381 sport rows against 17,612. Left alone, that dip manufactures a wave of cuts in 2020 and a matching wave of additions in 2021, neither of which happened.

Two tests settle it. 183 institutions filed in both AY 2019-20 and AY 2021-22 and skipped AY 2020-21 entirely; they fielded 1,822 teams before the gap and 1,762 after it, so their absence is a filing gap. Then the decisive one: 4,095 teams vanished in AY 2020-21 at institutions that filed all three surrounding years, and 3,293 of them (80.4 percent) were back the following season.

Out of any other base year, 448 to 933 teams disappear and 18.4 to 38.0 percent come back within twelve months. Out of AY 2019-20, 3,884 disappear and 80.8 percent come back. The panel top line agrees: 23,210 teams in AY 2019-20, 20,469 in AY 2020-21, 23,705 in AY 2021-22, from the identical 1,715 institutions. Those counts, and the table below, are at the raw sport-label grain, which is why they run higher than the team totals in the opening. The next section explains the gap; every figure after it is collapsed.

AY 2020-21 is excluded from every cut count below. Counting it would have produced roughly 3,300 cuts that never happened. The return-rate test is the reusable part: if you suspect a break year in a panel your office maintains, ask what share of the disappearances come back.

At the raw sport-label level the panel shows 2,121 drops at 783 institutions, a cutter rate of 45.66 percent. Collapse the running family (four sport codes describing one program) and the aquatics family (four more) and it falls to 1,143 drops at 549 institutions, 32.01 percent. Roughly 978 apparent drops and 234 apparent cutters are colleges changing which label they file under: an 86 percent overstatement produced by a filing habit.

The proof is direct. 209 panel institutions stopped reporting women's cross country. 148 of them started reporting the combined track and cross country label inside the same window, and 156 still run some form of track. Only 53 have no running program left. Of 67 institutions that stopped reporting women's swimming, 42 moved to the combined swimming and diving label and only 25 left the pool.

So the naive drop table, the one that puts indoor track, cross country and swimming on top at 28 to 36 percent, answers a filing question rather than an athletics question. Every figure below collapses the families first. And what survives is not a one-year gap landing on the endpoint: 864 of the 1,143 drops (75.6 percent) were absent three consecutive years, AY 2022-23 through AY 2024-25, and 794 were absent all four years from AY 2021-22.

The question only works with the right denominator. Not how many colleges dropped tennis, but what share of the colleges that fielded tennis dropped it.

Of the 870 panel institutions fielding men's tennis in AY 2014-15, 137 had dropped it by AY 2024-25: 15.7 percent. Women's tennis, 131 of 1,028, 12.7 percent. Men's golf, 116 of 1,095, 10.6 percent. Women's golf, 79 of 816, 9.7 percent.

Then the floor. Among sports fielded by more than 800 panel institutions, men's basketball was dropped 0.6 percent of the time, baseball 1.2 percent, women's basketball 1.6 percent, women's volleyball 1.9 percent, women's soccer 2.0 percent, men's soccer 2.4 percent, softball 2.6 percent and football 2.7 percent. The two running families sit in between, at 4.5 percent for men and 4.8 percent for women. Men's tennis goes roughly 29 times as often as men's basketball, on a comparable base.

In absolute terms the concentration is sharper still. Tennis supplied 268 of the 1,143 drops (23.4 percent) and golf 195 (17.1 percent). Two sports out of 31 families account for two fifths of every team a continuously filing college let go.

Read the added column before the dropped column or the table misleads in both directions. Men's bowling was dropped by 26.3 percent of the colleges that fielded it and still went from 57 teams to 100. Women's bowling, 23.1 percent dropped, 108 to 182. Women's lacrosse, 10.2 percent dropped, 453 to 521. Men's wrestling, 5.7 percent dropped, 298 to 399. A high drop rate on a fast-growing sport is churn, not retreat. The one small sport in the table whose adds do not cover its drops is equestrian, on both sides: women's from 65 teams to 59 against only 10 adds, men's from 19 to 13. Read that as a fact about this panel and this window rather than a sector trend, because nationally equestrian has wandered between 57 and 74 institutions for twenty years and AY 2014-15 sits near the top of that band.

Tennis is the only sport fielded by more than 800 panel institutions that shrank on net on both sides, men's from 870 teams to 795 and women's from 1,028 to 951. Those are eleven-year-panel counts, so they run below the wider-panel and national counts for the same sport elsewhere. The decline is already 44 men's teams and 16 women's teams deep by AY 2019-20, so it predates the pandemic. What is institution-level about it, and therefore this post's to report, is that it inverts by resource level: NCAA Division I-FBS dropped women's tennis at 1.7 percent and men's at 6.7 percent, against 18.0 percent and 18.6 percent in NCAA Division III without football. The better funded the division, the likelier the sport survives.

That is as far into sport-level territory as this post goes. Whether tennis is shedding whole teams or thinning rosters, which sports actually grew, and why equestrian's ten-year dip sits inside its normal range are all settled in the companion sport-level analysis.

Sector is the variable most people reach for, and here it barely does anything. The three sectors holding 1,698 of the 1,715 panel institutions sit inside a 12.7 point band on cutter rate: public 2-year at 39.1 percent, private nonprofit 4-year at 31.0 percent, public 4-year at 26.4 percent. Two sectors post extreme rates, private nonprofit 2-year at 71.4 percent and private for-profit 4-year at 70.0 percent, on 17 institutions between them. Nobody should quote those.

Athletic classification spreads from 19.8 percent to 65.7 percent, and it inverts the usual intuition. NJCAA Division III cuts most: 46 of its 70 panel institutions dropped at least one team, and 20.8 percent of its AY 2014-15 teams are gone. NCAA Division I-FBS cuts least, 25 of 126 institutions, with 2.3 percent of its teams gone. The money divisions are not the ones cutting.

The subtler Division I finding sits in the net column. Only two classifications shrank on net and both are Division I: Division I without football at minus 9 teams, Division I-FBS at minus 4. Every other classification grew. Division I does not cut often; when it does, it does not backfill. Classification is pinned to AY 2014-15 here, because the vocabulary is not stable across the window.

Whether the cuts fell harder on men's or women's teams is the first question anyone asks, and the answer is neither. Men's teams were dropped at 5.54 percent (556 of 10,028) and women's at 5.25 percent (587 of 11,178). Both sides grew on net, women's teams by 1,146 and men's by 846.

The mechanism is that cuts come in pairs. Where a college fielded both a men's and a women's team in the same sport in AY 2014-15 and cut one, it cut both 60.4 percent of the time: 302 both-sides cuts against 105 men's-only and 93 women's-only, out of 6,923 offerings carrying both sexes. At the institution level, 305 of the 549 cutters (55.6 percent) cut on both sides, 132 cut only women's teams and 112 only men's.

The second null runs mildly backwards. Institutions that cut only men's teams had a median AY 2014-15 proportionality gap of 13.83 points; institutions that cut only women's teams had 14.36, slightly wider. Institutions that cut nothing had the narrowest gap of the three, 11.69 points. Cutters were a little more exposed than non-cutters. But the direction of the cut does not track the direction of the exposure, so nothing here supports reading the decade's men's-team cuts as compliance moves. What the national gap is, and what it does not imply, is in 55 percent of the students, 42 percent of the athletes.

Institutions that went on to shrink their sport portfolio had a median AY 2014-15 athletics expense of $2,313,206. Institutions that went on to grow it: $2,532,555. Median expense per athlete, $8,765 against $9,099. Median athlete count, 285 against 283. Money at the start of the decade does not sort colleges into cutters and non-cutters. The group that looks different is the one that ended the decade flat, at $4,352,960 and $12,688 per athlete. It is the best funded, and it holds the largest institutions: a median of 393 athletes.

Expense appears here only as a size measure. It cannot do more than that, because EADA revenue equals expense by construction for most filers, so no figure in this file can tell you whether a program pays for itself. Athletics revenue is not revenue has the test behind that, across 40,829 filings.

Enrollment is where the separation shows up. Institutions that net shrank their portfolio saw a median undergraduate enrollment decline of 23.6 percent, close to double the 12.9 percent at portfolio growers and the 10.8 percent at flat institutions, and 77.4 percent of shrinkers lost undergraduates. In levels: shrinkers went from a median 2,862 undergraduates to 2,110, growers from 1,802 to 1,534, flat institutions from 3,446 to 2,867.

All three groups lost enrollment at the median, so this is a difference of degree. But the reframe earns an IR office's attention: portfolio cutting travels with an enrollment problem you already track.

870 panel institutions grew their portfolio over the decade, against 288 that shrank it. At the median the growers went from 13.7 athletes per 100 undergraduates to 24.8, with median athlete count climbing from 283 to 420 while median enrollment fell 12.9 percent. A quarter of the student body at the median portfolio grower is now on a roster.

Two corrections before that number travels. Roster spots are counted, not people, so a two-sport athlete counts twice, and the ratio rises partly because enrollment fell in all three groups. Even so, at small colleges athletics looks like an enrollment channel. That is the context the cutting story sits inside.

Our analysis of which academic programs colleges are quietly closing found that dying programs were tiny for years before they closed: a median of 2 credentials in the final year against 16 for survivors. Teams behave the same way.

Control for sport first, because a 12-athlete golf team and a 112-athlete football roster are not comparable. Inside the roster distribution of its own sport, sex and year, a team in its final season sits at the 9.4th percentile; a team still fielded in AY 2024-25 sits at the 46.1st. Median roster, 9 against 18. And 72.0 percent of final-year teams sit in the bottom roster quartile of their own sport, against 29.0 percent of surviving team-years. That rests on 1,037 termination events and 142,033 surviving team-years.

The useful part is how early the flag is up. Teams that would be gone were already at the 14th to 18th percentile of their own sport four and three years before their final season, sliding to the 9.4th percentile at the end, while the median roster moves only from 10 to 9.

So the intuitive heuristic, watch for a decline, is the wrong one. Doomed teams do not collapse. They sit small for years and then stop. The median team in its final year had lost 1 athlete over the previous two years against 0 for survivors, and 51.9 percent had shrunk against 41.1 percent of survivors, a gap of only 10.8 points.

One trend measure does separate: a sharp collapse. 20.7 percent of teams in their final year had lost 30 percent or more of their roster in two years, against 5.2 percent of survivors. Watch the level and the collapses. Ignore the drift.

A team in the bottom roster quartile of its own sport, sex and year disappears within four years 5.79 percent of the time, against 1.00 percent in the top quartile, and that ratio holds separately within each of AY 2016-17, AY 2017-18 and AY 2018-19.

As a raw headcount band it is blunter and much easier to run against your own roster file. Teams with 6 or fewer athletes are gone within four years 14.56 percent of the time. 7 to 10 athletes, 6.28 percent. 11 to 20, 1.93 percent. More than 20, 1.04 percent. Fourteenfold between the ends.

Ship the rule with its error rate or do not ship it: 85.4 percent of teams at 6 or fewer athletes are still there four years later. This is a watch list, not a prediction. What makes it worth maintaining is that the flag is rare on the other side of the matrix. Only 3.5 percent of surviving team-years sit at 6 or fewer athletes, while 33.8 percent of final-year teams do and 61.9 percent had 10 or fewer.

It is sport specific too. A women's golf team in its final year has a median of 3 athletes and 88.3 percent are at 6 or fewer, against a median of 8 and 29.3 percent for survivors. A men's tennis team in its final year has a median of 8 against 10 for survivors, with 30.0 percent at 6 or fewer against 5.6 percent. Read the flag against the sport's own distribution.

The largest net retreat in the panel is Suffolk County Community College (NY, NJCAA Division III), 17 teams down to 8, nine drops and zero adds, while undergraduate enrollment fell from 14,293 to 9,273. Saint Cloud State University (MN) went from 19 teams to 14 while enrollment fell from 8,980 to 4,008. That is the common shape: the colleges that shrank their portfolios are also the ones that lost enrollment fastest, though nothing in EADA dates one against the other.

The second shape has no enrollment pressure behind it. George Washington University (DC, NCAA Division I without football) went from 24 teams to 17, seven drops and zero adds, while undergraduate enrollment rose from 9,738 to 10,686. The seven were men's rowing, men's sailing, men's squash, men's tennis, women's sailing, women's squash and women's water polo, hand verified against the raw rows. Brown University went from 34 teams to 30 while enrollment rose from 6,241 to 7,185. Florida Institute of Technology dropped 7 teams with enrollment essentially flat, 3,225 against 3,282.

Whatever drove those, it was not a shrinking student body. If your model of athletics cuts is purely financial or purely demographic, this second group is where it fails.

The biggest limitation is a number, and it is larger than the headline. 163 institutions filed in AY 2014-15 and not in AY 2024-25. On the same family grain the panel uses, those institutions carried 1,428 teams and 31,006 athletes in their final filing year, an average of 8.8 teams each. That is more than the 1,143 drops the panel records across the whole decade, and no panel design can see any of it. These are colleges leaving the file, not programs being trimmed. EADA has no status column, so it cannot say how much of that 1,428 is genuine team loss and how much is a filing lapse at a department that is still operating. Treat the headline as a floor on total team loss rather than a measure of it.

Second, the recent years are right censored. AY 2019-20 is the largest single year of terminations at 246 teams across 158 institutions, and AY 2023-24 is next at 239 across 172 institutions, the highest institution count in the window. But a team last seen in AY 2023-24 can return in AY 2025-26, which is not in the file, and the normal one-year return rate runs 18.4 to 38.0 percent. Those counts are upper bounds, and AY 2023-24 is not a peak until another file says so.

Third, participants are roster spots rather than people. Fourth, only institutions sponsoring intercollegiate athletics file at all. Fifth, no revenue figure appears anywhere here, and the single expense figure is a size measure only.

Dataset: U.S. Department of Education, Equity in Athletics Disclosure Act data, academic years 2014-15 through 2024-25, queried through Clema's public data layer. Undergraduate enrollment on the sport file comes from IPEDS and repeats on every sport row of an institution-year, so it is aggregated with a maximum rather than a sum. Background on the file is on the EADA data source page.

Reproducible query log: docs/data-evidence/which-colleges-are-cutting-teams-and-what-goes-first/queries.md (36 numbered queries with their returned rows, plus the thirteen discarded analyses and why each was dropped) and schema.md alongside it (table grains, the constructed measures, join paths and eleven landmines). Twenty-five of the load-bearing queries were re-executed independently against the live source before publication, and the post-publication fact check that corrected the sector cut, the roster horizon and the leaver count is logged at the top of the same file.

Related reading on this site: which college sports are growing and which are vanishing for the sector-wide sport-by-sport side of this file, the Title IX proportionality gap, measured, why EADA athletics revenue is not revenue, and the federal data release calendar for when the next EADA vintage lands.