# Cohort data: sources, units, and limits

Cohort compiles public data about US colleges and California high schools. This
document tells you how to read the numbers the tools return.

## Units

- Rates are fractions from 0 to 1 (0.12 means 12%). This includes acceptance
  rates, graduation rates, demographic shares, aid shares, admit rates by major,
  cross-admit shares, and UC destination shares.
- Money is US dollars per academic year unless the field says otherwise.
  Median earnings are annual.
- SAT and ACT fields are the 25th and 75th percentiles of enrolled students.
- Enrollment is total headcount.
- Data years vary by table; each row that has a cycle or cohort field carries it.

## Sources

| Data | Source | Notes |
|---|---|---|
| Institution characteristics, admissions, enrollment, demographics | IPEDS (National Center for Education Statistics) | Public domain. Most recent published year. |
| Costs, net price by income, debt, graduation rates, earnings 10 years after entry | College Scorecard (US Department of Education) | Public domain. |
| Program-level earnings and debt | College Scorecard field-of-study files, with Census Bureau PSEO earnings as a fallback (earningsSource says which) | Public domain. |
| Program taxonomy | CIP 2020 (NCES) | Public domain. |
| Admission plans, deadlines, early-round statistics, admission factors, recommendation-letter rules, test policies | Compiled from each college's Common Data Set and admissions pages | Verify deadlines on the college's site before relying on them. |
| Admit rates by major | Published by the college (mostly University of California and California State University campuses) | Absence means not published. |
| Essay prompts | Quoted from each college's application materials | Copyright belongs to the college. Attribute and link. |
| UC source-school admissions (per high school, per campus) | University of California Information Center | Counts under three students are suppressed and returned as null with suppressed=true. |
| High school college-going outcomes | California Department of Education | Most recent graduating cohort. |
| UC admit destinations | University of California Information Center | Where each campus's admits enrolled, top 25. |
| Cross-admit pairs | Parchment | Fraction of students admitted to both who chose A, with a confidence interval and sample size. |
| Class destination lists | Collected by Cohort from school-published lists | Self-reported, may be incomplete. |

## What Cohort does not know

- Rankings. Cohort has no ranking and will not invent one.
- Current-year deadlines beyond what is stored. Deadlines change; confirm on the college's site.
- Anything about an individual applicant. Nothing here predicts one student's chances.
- Financial aid offers. Net price by income band is an average for that band.

## Known data issues

- Graduation rates: the fields `graduationRate4yr` and `graduationRate6yr` are mislabeled in the current data release. `graduationRate4yr` holds the College Scorecard completion rate within 150% of normal time for four-year institutions, which is the figure usually quoted as the six-year graduation rate. `graduationRate6yr` holds the same measure for two-year institutions and is null for four-year colleges. Treat `graduationRate4yr` as the six-year rate for bachelor's-granting colleges until the data pipeline is corrected.
- CIP codes in program and major tables are stored without dots and without leading zeros (for example 1107 for 11.07). Two-digit families 01 to 09 cannot be matched by prefix.
- Cross-admit and UC destination shares are stored as percentages and are converted to fractions before they are returned.
- Program enrollment share (`enrollmentShare`) is not populated in the current release; program lists are ordered by median earnings instead.
- `isOpenAdmission` is false both when a college is not open-admission and when the fact is unknown (about 2% of colleges).
- College-going outcomes are matched by school name; a few names are shared by schools in different cities.

## Reading the numbers responsibly

- An acceptance rate is one signal. Pair it with test ranges, GPA range, and
  the admit rate for the intended major where available.
- A null is missing data, not zero.
- When comparing colleges, hold the year constant: all headline fields come
  from the same IPEDS and Scorecard release.

Cohort is free and built by a high school student. https://collegecohort.com