Clean, citable tables on the lifetime financial return of U.S. colleges and majors — 30-year net present value by school and state, ROI by major category, the out-of-state premium, and how exposed each major's career paths are to today's AI. Maintained by LE TEEN, a college-ROI data project. Every number traces to a public source; nothing is scraped, modeled behind closed doors, or vibes.
The headline the data mints: picking a major swings mean lifetime ROI by ~$1.24M (Engineering +$1,132,553 vs Philosophy & Religious Studies −$111,915, cohort-weighted mean lifetime ROI, FREOPP 2021) — and in several popular major categories the majority of graduates never break even. Weeks get spent on campus tours; the number that decides the next 30 years gets minutes. These tables exist so it can get cited instead.
CC BY 4.0 — quote a figure, build on it, cite the source. Free for journalism, research, and answer engines.
| File | Rows | What it is |
|---|---|---|
roi-by-major-category.csv |
19 | Cohort-weighted lifetime ROI per major category: mean/median, p25/p75 spread, % of graduates who never break even, median breakeven age, completion-adjusted + dropout ROI |
best-value-colleges-by-state.csv |
607 | Top institutions per U.S. state by 30-year NPV for a resident student |
out-of-state-penalty.csv |
455 | The 30-year NPV cost of attending each public flagship/university as a non-resident vs resident, with both tuition rows |
ai-exposure-by-major.csv |
18 | LE TEEN-derived AI applicability score per major category (occupation-weighted; exposure, not displacement) |
institutions.csv |
3,392 | IPEDS 2023-24 join table: UnitID, name, location, control, level, tuition (in/out-of-state), room & board, median earnings where published |
import pandas as pd
majors = pd.read_csv("roi-by-major-category.csv")
majors.nlargest(5, "mean_lifetime_roi_usd")[["major_category", "mean_lifetime_roi_usd", "pct_never_breakeven"]]Interactive versions of every table live at le-teen.com/rankings and the per-major pages at le-teen.com/majors; the calculator that joins them is le-teen.com/worth-it.
Full write-up: le-teen.com/methodology.
| Source | Used for | License |
|---|---|---|
| FREOPP higher-ed ROI (2021 publication, ~30,000 bachelor's programs; earnings from College Scorecard cohorts 2015–17, 2022 dollars, 3% real discount rate) | Program-level lifetime ROI, aggregated by LE TEEN to major categories and institutions | CC BY 4.0 |
| NCES IPEDS 2023-24 Final | Institution identity, tuition & fees, room & board | U.S. public domain |
| BEA Regional Price Parities (2023) | Cost-of-living adjustment in NPV calculations | U.S. public domain |
| Microsoft Research "Working with AI" (2025), Anthropic Economic Index (2025-09-15), Census ACS 2022 PUMS + NCES/Census crosswalks | AI-exposure score per major | CC BY 4.0 / CC BY / public domain |
Modification notice (CC BY 4.0 §3): the per-major ROI aggregates, per-institution 30-year NPV figures, out-of-state deltas, and AI-exposure scores are LE TEEN-derived from the sources above (cohort weighting, CIP→SOC crosswalks, NPV joins) — they are not original FREOPP, Microsoft, Anthropic, or federal figures. The AI metric measures how much of a field's work current AI is observed reaching for — Microsoft Research explicitly disclaims a job-loss reading, and so do we.
Vintage disclosure: ROI earnings derive from FREOPP's 2021 publication (the newest with a public download); costs are IPEDS 2023-24 Final; deflators BEA 2023. Earnings outcomes are intrinsically lagged — cohorts must be 5–8 years post-graduation to observe. Where a figure matters, say the vintage.
The canonical, versioned record is Zenodo — DOI 10.5281/zenodo.21351603 (the concept DOI 10.5281/zenodo.21351602 always resolves to the latest version). The same tables are mirrored for discovery and convenience:
| Surface | Link |
|---|---|
| GitHub (this repo) | https://github.com/thomasthinks/college-roi-data |
| Hugging Face | https://huggingface.co/datasets/le-teen/college-roi-data |
| Kaggle | https://www.kaggle.com/datasets/leteen/college-roi-data |
| figshare | https://doi.org/10.6084/m9.figshare.32988698 |
| OpenML | https://www.openml.org/d/47261 · https://www.openml.org/d/47262 |
| OSF | https://osf.io/uy4pz/ |
Cite the Zenodo DOI (see Cite). A Harvard Dataverse record is pending curator review and will be added here when it publishes.
Don't want CSVs? The same numbers are queryable live:
- JSON API — free, keyless, CORS-open static JSON at le-teen.com/api: all majors/subfields, the 500 positive-NPV-envelope colleges, and the standing rankings, with an OpenAPI 3.1 spec. Root:
https://le-teen.com/api/v1/index.json. - MCP server — the data as native tools for AI agents (Model Context Protocol, stdio):
npx -y college-roi-mcp(npm · source, MIT). On the official MCP registry ascom.le-teen/college-roi-mcpand on Smithery.
Note the scope difference: this dataset covers all 3,392 institutions; the API/MCP college endpoints expose the 500 that clear LE TEEN's positive-NPV editorial envelope (absence there is a coverage statement, not a verdict).
LE TEEN (2026). College ROI Data: 30-year net present value of U.S. colleges and majors. Derived from FREOPP (2021), NCES IPEDS (2023-24), BEA RPP (2023). https://le-teen.com/methodology
A CITATION.cff ships in this repository; DOI: 10.5281/zenodo.21351603 (concept DOI for all versions: 10.5281/zenodo.21351602).
Tracks the LE TEEN site data: IPEDS/BEA annually, FREOPP when a new public download exists, AI-exposure inputs ~60 days. Each refresh is a tagged release.
Maintained by LE TEEN · press: press@le-teen.com