US college enrolment falls 30%

Where things stand

• Baseline (B). The Research Center's Final Fall Enrollment Trends report (15 Jan 2026) puts fall 2025 total enrolment at just over 19.4 million, up 1.0% on the year. The threshold is therefore about 0.70 × 19.4M ≈ 13.6 million. https://nscresearchcenter.org/final-fall-enrollment-trends/
• Latest final figure. Final Spring Enrollment Trends 2026 (4 Jun 2026) reported 18.6 million, also up 1.0%. Dual enrolment of high-school students grew about 10%, and community colleges grew about 3%. For a spring term to qualify, spring enrolment would have to fall about 27% from the spring 2026 level. https://nscresearchcenter.org/enrollment-insights/ ; https://www.highereddive.com/news/spring-enrollment-ticks-up-1-but-graduate-headcounts-take-a-hit/821957/ ; https://bryanalexander.org/enrollment/higher-education-enrollment-new-data-for-spring-2026/
• Fall 2026 so far (term under way, colleges' own figures, Sept–Oct 2026): results are mixed but small. International students are down sharply at some public systems, which pulled totals down at Colorado (−1.4%), Wisconsin (−1%) and North Texas (−3.6%). Many community colleges and public universities are setting records. Nothing points to a collapse. https://www.highereddive.com/news/major-public-universities-report-hits-to-international-enrollment/831336/ ; https://www.ccdaily.com/2026/10/enrollments-continue-to-increase-at-many-community-colleges/
• Calendar. Only final reports count. Final fall reports come out in January (first quarter) and final spring reports in late May or June (second quarter). The November preliminary report is excluded. So 2026 cannot resolve YES. Nine further reports fall inside the window: fall 2026 through fall 2030, and spring 2027 through spring 2031. The fall 2031 final report is due January 2032, after the window closes.

Base rate

I downloaded the federal education statistics agency's historical fall-enrolment table (Digest Table 303.10, https://nces.ed.gov/programs/digest/d23/tables/dt23_303.10.asp) and computed the worst declines over 1947–2022:

• worst 1-year fall: −7.9% (1950→51);
• worst 2-year fall: −14.0% (1949→51, as the post-war GI Bill surge unwound during the Korean War);
• worst drop from a previous peak: −14.0%;
• 2011→2021: −11.2% over ten years.

The worst shock on record is World War II: 1,494,203 students in 1939–40 fell to 1,155,272 in 1943–44, a drop of 22.7% (Biennial Survey, cited in ERIC ED172458). A 30% fall within about five years has never happened, not even with wartime mobilisation. COVID cut enrolment by only about 5%.

Paths to YES

1. Transformative AI. AI would have to empty out the value of degrees fast enough that freshman and graduate enrolment collapse for several years in a row by 2029–31. So far the evidence points the other way. Entry-level hiring is weak, but college-going rates ticked up, and workers worried about AI are going back to school for trades and health programmes. Recessions have historically raised enrolment. I put this path at about 0.5%.
2. Other catastrophes. A pandemic far worse than COVID, a war with a draft, or the abolition of federal student aid. Today's warnings of a Pell Grant funding shortfall and the new graduate loan caps imply single-digit effects at most. About 0.2–0.3%.
3. Data or methodology quirks. Coverage changes are handled by the restated-baseline and coverage-adjusted rules. A gross published error is very unlikely. Below 0.1%.

Hazard construction

Per-report hazards rise from 0.01% (fall 2026, already under way and roughly flat) and 0.03% (spring 2027) to 0.14–0.15% for the fall 2030 and spring 2031 reports. Spring reports carry slightly more because spring headcount runs about 4% below fall. The third and fourth quarters get tiny allowances for a delayed report or an official correction. Accumulated (1 − product of (1 − h)), the series reaches about 0.94% by the end of 2031.

Strongest case against

Forecasters with short AI timelines could argue for several percent. A Metaculus question ("Will US undergraduate enrollment decline by more than 10% from 2024 to 2030?") shows 29%, which implies fat tails. However, it has only 19 forecasters, and its visible comments are low quality. A 10% drop in undergraduates is also far milder than a 30% drop in total enrolment that includes growing dual enrolment and graduate students. Against that, enrolment is still growing, and the history above sets a ceiling. I hold the total at about 1%.

Question. By the last day of each quarter, has the National Student Clearinghouse reported total US postsecondary enrolment at least 30% below its fall 2025 level?

Definitions.

• Total US postsecondary enrolment: the National Student Clearinghouse Research Center's coverage-adjusted national total postsecondary enrollment estimate, including undergraduate and graduate enrollment across all institutional sectors in its reporting universe. Use the published overall total, not full-time-equivalent enrollment, first-year enrollment, a major-field subtotal, or the raw count of records submitted by participating institutions. Retain the series' treatment of international students, dual enrollment, and students enrolled at multiple institutions.
• Fall 2025 level (B): the full-precision national total for fall 2025 in the Research Center's Final Fall Enrollment Trends report, published January 15, 2026, including its dashboard and downloadable data. Its rounded headline value is 19.4 million; that rounded value is not the numerical baseline. Use official corrections to this baseline available by the qualifying publication date. If the Research Center changes the series' coverage or methodology, use its officially restated fall 2025 value on the new basis; do not splice incompatible series or construct an independent adjustment.
• Comparison enrollment (E): a final national total for a subsequent fall or spring term. Either term qualifies; the comparison is always with fall 2025, not with the corresponding season in 2025. The observation period must end after the opening date under the Observation period convention.
• At least 30% below: E ≤ 0.70 × B, calculated from the most precise published counts. A rounded percentage or rounded headline total alone does not establish that the threshold is crossed when the underlying figures leave it uncertain.

Resolves YES if a qualifying final report or official correction published within the Resolution window establishes E ≤ 0.70 × B. A correction or revision of a qualifying post-opening observation can establish YES on its publication date. No attribution of the decline to AI is required.

Resolves NO for a horizon if no qualifying publication has established the threshold by that horizon. Preliminary fall reports, fixed-panel early estimates, forecasts, undergraduate-only declines, and declines confined to a sector, state, or subject do not qualify. A publication that only repeats or revises pre-opening observations does not qualify. A reduction in reporting coverage alone, without a comparable coverage-adjusted national estimate, does not qualify. Later revisions do not undo an established YES under the Resolution window convention.

Resolving source. National Student Clearinghouse Research Center publications only: Final Fall Enrollment Trends and Final Spring Enrollment Trends, their dashboards, data downloads, methodological notes, and official corrections, within the Clearinghouse Enrollment Insights series. These are the successors to the fall and spring Current Term Enrollment Estimates reports. A renamed successor qualifies only if it supplies the same national measure or an officially comparable restated series. These are the named primary sources for Public confirmation (A2); media accounts alone do not qualify. Apply Resolution date (A3), the Resolution window, and the Observation period convention.