Data sources

What it captures. The Integrated Postsecondary Education Data System is the federal census of every Title IV-eligible college and university. It covers institutional metadata (control, locale, mission), admissions (applicants, admits, yield, test scores), enrollment by demographics, completions by CIP code, graduation and retention rates, transfer-in and transfer-out counts, tuition and fees, financial aid, and faculty.

How we use it. IPEDS is the spine of the site. It populates all 5,777 college profiles, every state and city location page, the transfer-pathway feeder/destination counts, and the completion-by-major figures on Worth-It and major pages.

Strengths. Universal coverage, mandatory reporting, stable schema across decades, and direct lineage to other federal datasets via UNITID. When a number must be defensible against a college's own internal records, IPEDS is the version that wins.

Limitations. Lags by one to two years (current vintage on Worth-It pages: 2024). Transfer-in and transfer-out figures are reported only by some institutions and don't track students across schools — for that, see the gap discussion under methodology limitations.

What it captures. The Department of Education's flagship outcomes dataset, derived from linked IRS records rather than alumni surveys. It includes median earnings at 1, 6, 8, and 10 years after enrollment; earnings percentiles (25th and 75th); share earning above $25k; median debt at graduation (completers vs non-completers); estimated monthly loan payments; loan-repayment rates at 3, 5, and 7 years; and net price by family-income bracket ($0–30k, $30–48k, $48–75k, $75–110k, $110k+). Also includes a field-of-study earnings file at the 4-digit CIP × credential level — the same institution-by-major granularity that powers our Worth-It analyses.

How we use it. Earnings, debt, and net-price headlines on college profile pages. Earnings inputs to all 54 Worth-It major analyses (institution-level and field-level). Cost figures on Compare Paths (2-year vs 4-year). Primary input to the Best Value rankings.

Strengths. Tax-record provenance — these are actual earnings, not survey responses. Granular family-income net-price brackets let us show what a school costs for a specific family, not the fiction of the published sticker price.

Limitations. Only captures graduates with US tax records, so excludes those who leave the country, work outside the formal economy, or are self-employed without 1099s. Field-level cells are suppressed at small enrollment counts (privacy protection); when a metric is missing on a Worth-It page, this is almost always why.

What it captures. Post-Secondary Employment Outcomes is a Census/LEHD product that joins state-reported graduation records to quarterly-wage employer records. The result is earnings percentiles (p25, p50, p75) at 1, 5, and 10 years after graduation, broken down by institution × degree level × CIP × cohort. PSEO also reports flow data: which industries and which states employ a school's graduates.

How we use it. The 10-year median is the headline earnings figure on Worth-It pages — chosen over Scorecard's institution average for its cleaner CIP × institution joining. The percentile spread (p25 vs p75) is what lets us flag fields where the median hides high variance, and the 1→5→10-year trajectory is what the trajectory charts visualize.

Strengths. Percentiles, not just medians, expose risk — two majors with the same $50k median but different p25 figures have very different downside profiles. Multi-year horizons capture trajectory: some fields start low and grow steeply, others start higher and plateau.

Limitations. Coverage is uneven — it depends on each state choosing to participate, and the experimental status means some institutions or majors are not yet represented. State-of-employment data is shaped by where graduates can find work (high-cost metros), which can distort the purchasing-power picture.

What it captures. The landmark Chetty et al. mobility research. For each institution, it measures the probability that a student from the bottom 20% of the family-income distribution reaches the top 20% as an adult (the headline mobility rate), along with the school's access (share of students from the bottom quintile) and success (share of bottom-quintile students who reach the top quintile). The fundamental equation is mobility = access × success — a school can be high-mobility either by admitting many low-income students or by being unusually effective at propelling them upward, or both.

How we use it. Mobility is the headline metric of the Best Colleges for Economic Mobility ranking and is surfaced on college profile pages where available. It is the only source on the site that answers "does this school actually change lives?" rather than "do its students earn well?"

Strengths. Tax-record-linked, multi-decade, and uniquely designed to separate selection (who got in) from treatment (what the school did for them). Some mid-tier public universities have higher mobility rates than Ivies because they admit more low-income students — a finding only visible through this lens.

Limitations. Static cohort. The dataset tracks students born 1980–1991, who entered college roughly 1998–2009. Mobility ranks describe a durable institutional track record, not a forecast of what today's freshmen will experience. We disclose this on every page using OI data.

What it captures. The Occupational Employment and Wage Statistics survey reports annual mean and median wages and employment counts for ~830 SOC-coded occupations, at the national, state, and metropolitan-area level. Reference period: each May.

How we use it. OEWS gives us current-year wage ground-truth per occupation — the figure cited on major pages when we say "the median wage for registered nurses is $X." It also drives the wage charts on career profiles.

Strengths. Establishment-survey methodology — wages come from employer payroll records, not worker recall. Metro-area granularity exposes regional wage patterns that national figures hide.

Limitations. Captures wage-and-salary employment only, so excludes self-employed workers (relevant for trades and creative fields). Occupation taxonomy is SOC, which sometimes splits or aggregates roles in ways that don't match industry usage.

What it captures. The 10-year occupational outlook — projected employment growth rate, projected annual openings (replacements + new jobs), and the typical entry-level education for each occupation. Joined to majors via the BLS-published CIP→SOC crosswalk.

How we use it. Labor-market signal on Worth-It pages and major pages — "is this field actually growing, plateauing, or shrinking?" When a major maps cleanly to one occupation (nursing → registered nurses), the numbers are exact. When it maps to several adjacent roles (biology → medical scientists, biological technicians, environmental scientists, …), the headline figure is a weighted blend across the mapped occupations, and the page flags this with a "blended" tag and a low tiering confidence indicator.

Strengths. Integrated with OEWS via SOC, so growth and wage numbers are directly comparable. Biennial refresh keeps the outlook close to current macroeconomic conditions.

Limitations. Projection, not measurement — the further from the base year, the wider the uncertainty. CIP→SOC mapping is many-to-many for most majors, which is why we expose the per-occupation breakout alongside the blended headline.