Every college on the list gets a letter from A+ to F. The letter answers one question: does this college choose its students, or does it take whoever comes? It is built from six numbers that colleges report to the federal government each fall, and it never looks at money.
Fall 2024 data · 1,767 colleges graded
A college has command of its market when more students want a seat than it has seats, when the students it admits pick it over their other offers, and when they stay once they arrive. That command is a cushion. The number of high-school graduates in the U.S. peaked with the class of 2025 and falls from here. A college that turns students away can absorb a smaller applicant pool by admitting a little deeper. A college that already admits everyone has no cushion: every lost applicant is a lost student.
The grade is deliberately not a measure of academic quality or of finances. A wealthy college whose applications are collapsing can earn a D; a college with thin reserves and a waiting list can earn an A. The odds that a college closes are a separate number, from a separate model, shown beside the grade on the overview page. The two agree more often than not (rank correlation −0.57 among private colleges), because weak demand is one of the main roads to closure. But they answer different questions.
Who is graded. Private nonprofit four-year colleges, and public four-year colleges that mainly award bachelor’s degrees, that have enrolled at least 200 undergraduates in some year: 1,191 private and 576 public colleges in fall 2024. Public community-college systems that award a handful of bachelor’s degrees are left out. Nearly all of them admit everyone and report no admissions data, so they would sit at the bottom by construction.
The measures fall into three groups. Market position is where the college stands today: how many students it turns away, how many compete for each seat, and how often admitted students choose it. Momentum is the direction of travel over three years. Holding on is first-year retention, the earliest sign that a class isn’t holding, a year before it shows up in enrollment.
The weights are a judgment, not a fit. We tried a range of alternatives, and how well the score sorted colleges that went on to lose students from those that didn’t barely moved (AUC between 0.68 and 0.69 across them). So the split follows what the grade is meant to describe: half on position, because “command of the market” is a statement about where a college stands, and the rest on whether that position is holding.
The six measures come in different units: percentages, ratios, changes. Each one is converted to points from 0 to 100 on a fixed curve, a short list of anchor points joined by straight lines. A college that falls between two anchors gets its proportional share of the distance between them:
x is the college’s value; (xk, yk) and (xk+1, yk+1) are the anchors on either side of it. Beyond the last anchor the points stop at 0 or 100.
Two choices sit behind the curves.
Students apply to far more colleges than they used to. The typical graded college received 3.7 applications for every first-year who enrolled in 2004 and 7.8 in 2024, and its yield fell from 40% to 19%. Demand didn’t collapse everywhere; every admitted student simply holds more offers. Three of the measures are counted in applications, so each is measured against that year’s typical college, the same way prices are adjusted for inflation:
g is the college’s three-year change in applications. A ratio of 1.0, or a trend of 0%, means typical for that year, and scores about 55 points.
The median is taken across all graded colleges, public and private together, so a real shift of students toward one sector still shows. Enrollment and retention are counted in students, which didn’t inflate, so they are left as they are.
A high yield usually means a college wins the students it wants. But it can also mean the applicants had decided before they applied: commuter campuses where admission is a formality, religious colleges whose applicants come from one community, nursing programs that admit late in the process. Those colleges post yields of 75–98% from a thin pool of applicants, and they go on to lose students as often as colleges with ordinary yields, or more often. Among colleges with at least twice the typical number of applicants per seat, by contrast, almost none lost students, whatever their yield.
So yield only counts above typical when there are applicants behind it:
Harvard (yield 4.4 times typical, with 4.2 times the typical applicants per seat) keeps full credit. Florida (2.2 times typical yield, 1.3 times the pressure) is credited at 1.9. Brigham Young (4.1 times typical yield, but a quarter of the typical applicants per seat) is credited as typical. That is the cost of the rule: BYU’s applicants self-select, a genuinely loyal market that the data cannot tell apart from a processing funnel.
The score is the weighted average of the points, over the measures a college reports:
si is the points on measure i, wi its weight, and R the measures the college reported. A missing measure’s weight is shared among the rest.
A college reporting less than half the weight is NR, not rated. The admissions measures are only used when a college receives at least 150 applications, because below that a handful of students swing the ratios, so very small colleges (yeshivas, bible colleges, nursing and online units) are usually NR. Open-admission colleges get 0 points for admit rate and applicant pressure, and yield is not used.
Syracuse admits a bit under half its applicants, has more applicants per seat than typical, yields about the typical share, grew applications more slowly than the typical college, added undergraduates, and keeps 90% of its first-years.
That clears the A line at 66. Syracuse also passes the A test (it admits under 70% of applicants) and raises no red flag, so it earns an A. It stops short of A+ on yield: an A+ college must yield at least 1.6 times the typical share, and Syracuse yields about the typical share.
The score is cut into letters at fixed lines. They were placed by reading the data: each band matches a recognizable kind of college (below), and each step down comes with a clear jump in how often colleges went on to lose students. Because the lines are fixed, the share of colleges in each letter moves from year to year with the sector.
Any one of these caps the grade at D, whatever the score. Each was checked against what later happened to colleges that raised it. A red flag also grades a college that couldn’t otherwise be rated: losing a quarter of its students is evidence enough.
Below each description is the typical college in that letter, fall 2024. Yield and applications are relative to the typical college that year.
| Boundary | What decides it |
|---|---|
| A+ / A | Yield: does the college win the students it admits? |
| A / B | Selectivity: does it turn away at least 30% of applicants? |
| B / C | Momentum: are applications still growing? |
| C / D | Decline: are applications and enrollment shrinking? |
| D / F | Retention: can it keep the students it takes? |
The grade describes the present, so there is no single outcome to fit it to. But a grade about demand ought to sort colleges by what happens to their enrollment next, and it does. Of colleges graded A+ between 2010 and 2021, 2.6% lost a tenth or more of their undergraduates over the next three years; of colleges graded F, 52% did. Private colleges graded D or F closed within four years at 5.1% and 8.8%, against 0.5% or less for B and above.
Put another way: take one college that went on to lose a tenth of its students and one that didn’t, and the first had the lower score 70% of the time (AUC 0.70). For private closures, which the grade was never built to predict, the figure is 0.80. Grades are stable at the top: an A+ college was still A+ the next year 88% of the time.