← All investigations

WorthDegree Investigation · October 25, 2026

Colleges where 55% of students graduate report higher median earnings than colleges where 80% do. The graduation rate is measuring what kind of school it is, not how good it is.

I sorted all 4,895 American colleges that report both numbers into graduation bands. Earnings climb, peak at the 55% band, and then fall — all the way through the schools that graduate almost everyone.

The graduation rate is the single most quoted number in American higher education. It powers the rankings, the federal accountability rules, and the sentence every parent says out loud: look at how many of them finish.

So I took the federal College Scorecard data and asked the obvious follow-up question. Do the schools that graduate more people produce graduates who earn more?

At the institution level, the answer is no — and not in the vague, "the correlation is weak" way. The curve actually turns down.

Eight bands, one inverted curve

Group every college by its overall completion rate. For each band, take the median reported earnings of students ten years after they enrolled.

The peak is the 50-60% band, at $42,198. The schools that graduate more than four in five students report a median of $37,777 — 11.7% less. The 70-80% band is lower still, at $37,015, the bottom of the entire curve.

These are not small differences produced by a handful of outliers. Each band holds between 174 and 803 institutions. The 733 colleges that graduate 80%+ of their students collectively report less than the 174 colleges that graduate fewer than one in five.

Before you read that as "drop out and get rich," understand what the two ends of the curve actually contain.

The top of the curve is not better. It is a different institution.

Of the 733 colleges in the 80%+ band, 416 — 57% — are for-profit institutions. Of the 1,878 colleges below 50% completion, 1,148 — 61% — are public, overwhelmingly community colleges.

That single fact explains most of the curve. A community college's completion rate counts only the students who finish a credential there. The student who takes 30 credits and transfers to a state university as a junior is, in the federal formula, a non-completer — a failure recorded against the college that taught them, and a success recorded nowhere. A for-profit career school, by contrast, has almost no transfer outflow: students enter, finish a program, and are counted as graduates. The 57% figure is not a measure of teaching quality. It is a measure of who loses students to someone else's diploma.

The earnings side of the curve has its own composition problem. The schools that graduate nearly everyone are heavily regional public universities and private colleges turning out teachers, social workers, nurses' aides, and counselors — occupations with stable employment and unremarkable starting salaries. The schools in the middle of the curve include the engineering and business institutions.

So the inverted curve is not evidence that finishing college hurts you. It is evidence that the completion rate is a description of institutional type wearing the costume of a quality score.

Now the exception that should not exist

If completion were purely a type label, the two variables would separate cleanly. They mostly do. Of the 250 colleges reporting median earnings above $70,000, the median completion rate is 81%. High pay and high completion travel together — 95% of the time.

The other 5% is thirteen institutions. They report median earnings above $70,000 a decade after entry, with fewer than half their students finishing.

Median earnings of students ten years after entry, in 2023 dollars, and overall completion rate. All thirteen report measured earnings — none of these figures is a federal estimate. Source: U.S. Department of Education College Scorecard.

Nine of the thirteen are nursing and health-sciences schools. That is not a coincidence, and it is the strongest part of the story.

A hospital-based nursing program is one of the few educational products in America where the exit door is narrow on purpose. Clinical placements are capacity-limited, and programs fail students on clinical competency rather than on tuition. Completion rates of 21% to 47% are normal there — and so are ten-year earnings above $70,000, because the people who do finish are licensed RNs, and RN licensure is one of the few credentials that reliably clears $70,000 regardless of where you earned it.

Then there is Embry-Riddle Aeronautical University-Worldwide, the online campus of the aviation university: 9,139 students, a 21% completion rate, and $84,131 in median earnings. Its students are largely working aviation technicians and pilots already employed while they collect certificates and ratings. They are not failing to finish a degree. They are finishing a rating and leaving, because the rating is what the job requires and the degree is not.

That is the trap in the whole exercise. In a nursing school and an online aviation campus, a low completion rate and a high income are not a contradiction to be explained away. They are two measurements of the same deliberate design: a narrow gate into an occupation that pays.

Three honest objections

One: the completion rate has a transfer problem, and I am using it anyway. The federal measure counts students who finish at the institution where they started. Community colleges are structurally penalized because transferring out is a success that the formula scores as a non-completion. This is a real defect, and it is exactly why the 0-20% band sits at $39,156 rather than at the bottom. I report the number because it is the number that rankings and policy use. If the underlying measure is broken, that is an argument against the rankings, not a defense of them.

Two: earnings are institution-wide, not program-specific. The scorecard reports one median per college, pooling every major. A nursing school's $104,403 is a statement about nursing graduates, not about every student who ever enrolled. This is why the thirteen-school table lists net price beside earnings — at $53,020 a year, West Coast University-Texas costs more than most of the public universities with higher completion rates, and a family deciding needs both columns.

Three: the curve is descriptive, not causal. Nothing here says an individual student earns more by dropping out. The bands differ in who attends them, what they study, and what the surrounding labor market pays — all at once. What the curve does establish is narrower and harder to argue with: if you rank colleges by completion rate, you are ranking them by institutional type. That is not the same as ranking them by outcome, and the two lists disagree at both ends.

What the missing 22% should tell you

One more number sits under all of this. Of the 6,273 institutions in the federal file, 1,375 — 22% — report no completion rate at all. Not a zero. An absence. They are excluded from every band above, and from every ranking that claims to cover American higher education.

The graduation rate is not a lie. It is an answer to a question most families never ask: how many people finish here. The question they think they are asking is will this pay off for me. Those two questions produce the same answer for 95% of high-earning colleges and opposite answers for the other 5%, and no ranking in the country shows you which column you are standing in.

So check the specific program, not the institution. The College Scorecard publishes earnings by field of study for thousands of colleges, and those field-level numbers do not always resemble the institution-wide one. If you are considering a nursing or aviation program at a school with a 30% completion rate, that low number may be the gate you want to walk through. If you are considering a liberal arts degree at a school with a 90% completion rate, that high number tells you nothing about what you will earn.

Method

Universe: 6,273 degree-granting institutions in the WorthDegree dataset, built from the U.S. Department of Education College Scorecard. Two filters produce the 4,895 colleges analyzed: a reported median earnings figure for students ten years after entry (6,269 institutions qualify) and a reported overall completion rate (4,898 qualify). Completion here is the Scorecard's integrated measure covering first-time, full-time, part-time, and transfer students within 150% of normal program length.

Earnings are the Scorecard's median for students who are employed and not enrolled ten years after first entry, in 2023 dollars. Every institution in the headline table and the thirteen-school list carries a measured (reported) earnings figure, not a federal estimate — the dataset flags each value's provenance and none of the 13 is an estimate.

Bands are equal-width by completion rate; medians are unweighted across institutions, so a 174-student college counts the same as a 60,000-student one. Institution size is not controlled anywhere in this piece. Net price is average annual cost after grants for first-time, full-time, in-state students. The "95% of the time" framing in the exception section describes the 237 of 250 above-$70,000 colleges whose completion rate is at least 50%; it is a count, not a statistical test.

Reproduce it in four minutes

The whole analysis is one pass over the public dataset. Download the College Scorecard institution file, or use the processed worthdegree_dataset.json in our repository, then:

import json, statistics

rows = json.load(open("worthdegree_dataset.json"))["colleges"]

a = [c for c in rows

     if c["earnings_10yr_median"] and c["completion_rate"] is not None]



# 1. the inverted curve

for lo in range(0, 100, 10):

    band = [c for c in a if lo/100 <= c["completion_rate"] < (lo+10)/100]

    if band:

        print(lo, len(band),

              statistics.median(c["earnings_10yr_median"] for c in band))



# 2. the thirteen outliers

for c in sorted(a, key=lambda x: -x["earnings_10yr_median"]):

    if c["earnings_10yr_median"] >= 70000 and c["completion_rate"] < 0.5:

        print(c["name"], c["state"], c["earnings_10yr_median"],

              round(c["completion_rate"]*100), c["net_price"])

The first loop reproduces all eight bands and both turnarounds. The second reproduces the thirteen rows above, in the same order. If your numbers differ, the dataset version differs — the Scorecard is revised annually and we do not silently update past articles.

The graduation rate is not a measure of quality. It is a description of a business model — and the colleges that score highest on it are, more often than not, the ones that never let a student leave.

Two indicators, both published, both quoted in every ranking, and neither one measures what the family reading them assumes. See the companion analysis on why price does not predict earnings either, and the list of 72 colleges that charge $45,000 a year while their median graduate earns less than a high school graduate.