Methodology: How We Compute College Costs, Debt, and Earnings
Every number on this site traces to a named source file with a date on it. This page lists those files, shows the formulas we run on them, and states what we refuse to do. It is the reference behind every cost, debt, earnings and ROI figure we publish. When a page says "our math," this is the math.
The source files and their vintages
Six sources feed the site. Each published figure carries its source and vintage inline at the point of use — (Scorecard, June 2026), (FSA, Mar 31, 2026), (BLS, 2024). Licensing detail per source lives on the data-sources page.
Source | What we take | Vintage in use |
|---|---|---|
College Scorecard (US Dept of Education) | Median earnings 1 and 4 years after completion, median federal debt at graduation, average net price, completion rate, accreditor — institution and field-of-study files | June 10, 2026 release |
IPEDS (NCES) | Published tuition and fees, enrollment, net-price components | Collection year noted per page |
Federal Student Aid (US Dept of Education) | Loan portfolio: balances, borrower counts, debt-size bands, borrower location | March 31, 2026 (Q2 FY2026) |
Bureau of Labor Statistics | Education pays (CPS 2024 medians), OES wages (May 2024), Occupational Outlook Handbook | 2024 annual / May 2024 |
Federal loan rates (ED) | Fixed rates for loans first disbursed July 1, 2026 – June 30, 2027 | 6.52% undergrad, 8.07% grad, 9.07% PLUS — announced June 4, 2026 |
School tuition pages | Per-credit and per-term rates, required credits, mandatory fees | Fetched and dated per article |
How we compute the numbers
Medians, weighted by cohort size. A national earnings figure hides a pooling choice. When we pool the Scorecard field-of-study file, each program's median is weighted by its graduate count, so a 4,000-graduate program moves the field figure and a 40-graduate program barely does. Our 30-field ROI ranking pools 27,454 bachelor's programs covering 2,691,989 graduates this way (June 2026 file). Net-price averages are enrollment-weighted the same way: $13,788 per year at public four-year colleges and $26,597 at private nonprofits, computed across 816 and 1,221 schools respectively.
Which earnings number, against which baseline. We use program-level median pay 1 and 4 years after completion — newer and more specific than the "10 years after entry" institution metric, and a floor rather than a ceiling, since it misses mid-career growth. Scorecard earnings and debt cover federally aided graduates only, and small cohorts are privacy-suppressed; both caveats ride along wherever the figures appear. Our earnings baseline is strict on purpose: $48,360, the BLS 2024 median for a full-time high-school-diploma worker age 25+. The Scorecard file ships a softer $36,082 benchmark that includes younger workers; against that bar nearly everything looks good, which is why we do not use it.
The monthly-payment reframe. A balance is an abstraction; the payment is a budget line. Every debt balance and financeable total gets translated to the 10-year standard plan — 120 equal payments at the ED-published rate for the current award year. The $40,467 average federal balance becomes $460 a month at 6.52% (2026–27 undergrad rate). The column is a consistent yardstick, not a quote: existing loans keep the fixed rate they were issued at, so your real payment will differ.
Total cost, not per-credit stickers. A per-credit price is a multiplier, not a price. For any program we price: total cost = published rate × required credits + every recurring fee. Denominators get converted before comparison — Purdue Global bills quarter credits, and 1 semester credit equals 1.5 quarter credits by its own conversion, which turns a $371 sticker into a $556.50 semester-credit equivalent. Rates come from each school's own published pages, fetched and dated per article.
Payback years. Payback = total net cost ÷ annual earnings premium, where the premium is the field's median pay four years after completion minus the $48,360 baseline. We do not discount future dollars into net present value, and we hold premiums flat at the year-4 level, which understates fields with steep mid-career growth. Each payback figure is a floor, not a ceiling, and the assumptions are restated in every footer that uses the formula.
The "our math" mark. Any figure we computed rather than quoted is marked "our math" inline. Any quoted figure carries its source and vintage in parentheses where it appears, not in a bibliography you have to hunt through.
What we do not do
No prestige rankings. We rank by dollars: total cost, median earnings, debt, payback years. Reputation surveys, selectivity mystique and brand weight get zero columns. If a famous program's earnings do not support its cost, its row says so.
No pay-for-placement. No school can buy a row, a rank, or a kinder number. There is no paid-inclusion product, and we do not accept payment to edit a table.
No commercial influence on editorial. This site runs no program-search forms and takes no money from schools. If advertising or affiliate links are ever added, they will never change which schools appear, in what order, or with what figures. The separation rules are written out in the editorial policy.
No scraped competitor data. We take nothing from Niche, US News, CollegeSimply or other commercial ranking sites — their terms of service prohibit it, and their figures would import their methodology choices. Federal files and schools' own published pages are the record.
No cosmetic rounding, no filled gaps. $40,467 stays $40,467, never "about $40K." Where a source cell is missing — small cohorts are privacy-suppressed in Scorecard — the table shows an em-dash and a footnote, never an estimate dressed as data.
When the numbers refresh
We re-run the affected computations after each source release, not on a marketing calendar. College Scorecard: full re-computation of earnings, debt and net-price figures per ED release — the current one is June 10, 2026. FSA portfolio figures: quarterly, as the Data Center posts them. BLS: annually — CPS education-pays medians each year, OES each May release. Federal loan rates: each June, when ED announces the coming award year's fixed rates. School tuition pages: re-fetched at each affected page's refresh, with the fetch date printed in the footer. Every data page carries a "Data updated" stamp, and the stamp changes only when the underlying numbers were actually re-checked.
Corrections
A wrong published number should die fast. Report it through the contact page with the page URL and the figure in question. We check the claim against the source file. A confirmed error gets fixed in the table or text, and the page carries a dated correction note stating what changed. If the mistake was in our own computation, the note says "our error" in those words. Disagreement with a methodology choice — weighting, baselines, the no-discounting rule — is not a correction; if the criticism improves the method, we change the method and document the change on this page.
Related: Data sources & licensing · Editorial policy
