Nobody keeps a national count of the students lost before census day

The count that gets funded is taken a few weeks into the term

A registrar at a large community college spends the week before census closing one file. Census day, written as the census date in official reporting, is set a few weeks into the term; it is the day a college takes its official enrollment count, and state funding formulas run on the enrollment as of that day. The students in the file the registrar certifies are the students the state will fund. At Dallas College in Texas, which opened its administrative data to RAND researchers, about 30 percent of students registered in the final two weeks before classes. Before their first-semester census, 11 to 16 percent of newly registered students each semester drop or are dropped. The registrar certifies whoever remains.

How much of the pipeline leaks before that file closes? The published answer depends on the sector, because the sectors are measured unevenly.

Four-year colleges that admit from a pool have a federal admissions survey and an annual association chart of yield, the share of admitted students who enroll. Open-admission colleges, the colleges that accept any student who applies, have one 2025 multi-state study and a handful of single-college reports. At the last step, between registering for classes and attending them on census day, no national figure exists in any sector. At community colleges that measure the pre-census window, the loss of newly enrolled students runs 10 to 15 percent.

At four-year colleges, yield fell from 36 percent to 30 percent in eight years

Yield at four-year colleges is measured every year, and the measured series has been falling. The National Association for College Admission Counseling, NACAC, charts yield from federal collections for four-year colleges that enroll first-time freshmen and are not open admission. Average yield fell from 36 percent in 2014 to 30 percent in 2022. Private nonprofits slid from 36.0 percent to 32.6 percent over those eight years. Public four-year colleges fell from 36.9 percent to 25.3 percent, a loss of nearly twelve points, and their line fell in every year of the series. At 25.3 percent, a public four-year college admits roughly four students for every one who enrolls.

The counts underneath the rate come from the federal Digest of Education Statistics, which publishes applications, admissions, and enrollees for each sector. In 2012-13, public four-year colleges received 4.75 million applications, admitted 2.83 million students, and enrolled 971,000 first-year students. A decade later they received 7.6 million applications, admitted 4.9 million, and enrolled 1.1 million. The Digest publishes counts and does not publish a rate, so the following is our computation: applications grew about 60 percent over the decade while first-year enrollment grew about 13 percent, and enrollees per admit fell from 34.3 percent to 22.5 percent at public four-year colleges and from 26.4 percent to 19.0 percent at private four-year colleges, whose first-year enrollment slipped from 532,000 to 519,000.

The computed ratios and the published average rest on different populations and different denominators: each ratio divides a sector's total enrollees by its total admissions, while the association averages college-level yield rates across a different set of colleges. The published average stands at 30 percent.

Open-admission colleges lose students at two points before the count

A community college admits any student who applies and collects no deposit, so no yield rate exists for the sector. Some students apply, are admitted, and never register. Others register and are gone before census. In 2025, RAND published the first multi-state study of the second event surfaced by our search, built on 15 community colleges across Texas, Kentucky, and California and interviews with 130 staff. Community colleges that track the pre-census window lose 10 to 15 percent of their newly enrolled students before the census date.

Dallas College anchored the study, opening eight semesters of administrative data from summer 2021 through fall 2023. Across those semesters, 31 to 38 percent of applicants remained enrolled past the census date. About 65 percent of applicants never enrolled at all. Of the roughly 35 percent who did enroll, 11 percent withdrew between the first day of class and census, and 11 to 16 percent of newly registered students each semester dropped or were dropped before their first-semester census.

Even in the strongest of the eight semesters, the share of applicants still enrolled at census was 38 percent.

Students who register late are 37 percent more likely to be gone by census

Part of the loss is visible before the term begins. At Dallas College, about 30 percent of students register in the final two weeks before classes, and the timing of every registration sits in the college's own file.

Compared with students who registered two to six weeks out, a late registrant's chance of dropping before census ran 37 percent higher in RAND's analysis of fall 2023 registrations.

The published record for this sector is five studies in thirty-four years

In 1991, William Rainey Harper College in Illinois reported that nearly 16 percent of its fall 1990 applicants had not registered by the first week of classes. In 1997, Pennsylvania College of Technology, an open-admission public technical college, reported that 1,619 of its 3,524 accepted students for fall 1995, 46 percent, never enrolled; applicants from families earning $40,000 or more enrolled at 61 percent versus 52 percent for families earning under $30,000. In 2001, Bers and Nyden found that students who enrolled and left before the census date equaled about 10 percent of one community college's total enrollment; nearly two-thirds of them were gone before the semester began, and one-quarter later returned. In 2002, College of the Canyons in California reported that Latino students made up 21 percent of its fall 2000 enrollees and 30 percent of applicants who never enrolled.

After 2002, a search of federal data, ERIC (the federal education research index), association reports, and state sources turns up nothing until RAND's 15-college study in 2025, a gap of 23 years.

The federal admissions survey does not collect data from open-admission colleges

The Integrated Postsecondary Education Data System, IPEDS, is the federal database where colleges report their enrollment, finances, and outcomes. Its admissions survey collects applications, admissions, and enrollment counts only from institutions that are not open admission. A college that admits any student who applies files no admissions data, so the two-year rows in the federal Digest exclude open-admission institutions, which is most of the community college sector.

In fiscal 2025, community colleges collected $2,668 in net tuition revenue per full-time-equivalent student, the standard per-student unit, and enrollment-driven state allocations tied to the census count come on top of that. Public four-year institutions ran on $21,508 in total education revenue per full-time-equivalent student. Both figures come from the State Higher Education Executive Officers association, SHEEO. A student who registers and is gone before census never enters the funded count, and the net tuition a community college forgoes for that student runs roughly $2,668 a year.

Methodology

Every figure in this piece was verified against a fetched primary source on August 21, 2026, in the fact-base prepared for it. Vocabulary follows the structure of the sectors: yield describes four-year colleges that admit from a pool, and the open-admission losses are described as applied-admitted-never-registered and registered-but-gone-before-census, because that sector has no deposit and no admit pool. Ratios computed from NCES Digest table 305.40 are labeled as our computation; the table publishes counts without a yield column. The Dallas College figures describe first-time applicants and registrants across eight semesters, summer 2021 through fall 2023.

Three circulating figures were excluded. A widely quoted 2007 yield figure of 48 percent rests on a different survey methodology and an unfetched primary document. A 2025 PPIC transfer study could not be confirmed against its primary document. Aggregate dollar claims about what these losses cost have no primary source. The student-side summer melt literature measures college-intending high school graduates, a different population from institution-side registrants, and is treated in a companion piece. SHEEO FY2025 per-student revenue figures are reused exactly as verified in the research corpus, presented adjacent to the loss rates and never multiplied into aggregate-loss claims.