Graduation, retention, time to degree, degrees conferred, earnings and debt for University of California undergraduates, plotted across entering cohorts — split by the populations the public data can distinguish. Pick a comparison, then add or remove groups. The purple tag on each chart states exactly what kind of quantity is plotted; the amber tag states what the x-axis means. Every point carries its definition and, where published, its cohort size.
Retention and graduation appear on this page in more than one series, published by different offices over different universes. They are not interchangeable, and the gaps between them are larger than most of the trends they are used to describe. Every chart therefore carries a universe badge, and a series line sits directly beneath it.
| Series | Publisher and universe | First-year retention | Four-year graduation |
|---|---|---|---|
| Systemwide | UC Information Center; all nine campuses pooled, by entering cohort | 91.7% (2021 entrants) | 74.0% (2021 entrants) |
| UC Berkeley | UCB Office of Planning & Analysis; the campus's own cohort definitions | 95.8% (2021 entrants) | 80.6% (2021 entrants) |
| UC San Diego | UCSD Institutional Research; the campus's own definitions, by school of major | campus definitions; not comparable with either series above | |
| By source school | UC Information Center; entrants from one high school or community college to the selected campus, or to all nine pooled | integer percents, cohorts of 10 or more | |
| Federal | IPEDS / College Scorecard; first-time full-time or Title IV students, indexed to the data year rather than the entering cohort | a narrower population; not comparable with the UC series | |
A figure quoted without its series is ambiguous. “74.0%” is systemwide four-year graduation for the 2021 entering cohort, not Berkeley's, which is 80.6%. “About 92%” is systemwide first-year retention, not Berkeley's, which is about 96%. Retention is not graduation: first-year retention counts students enrolled at the start of the second year and says nothing about completion.
Measures are grouped in the picker by what they mean, and every chart carries a purple tag: % of entering cohort — the share of everyone who started (the usual "graduation rate"); % of graduates — a split among those who finished, saying nothing about how many finished; average of graduates — a mean (time to degree, final GPA) among those who finished; count — raw degrees conferred, not a rate; composition — who is enrolled or receiving aid, an input rather than an outcome; dollars — median earnings or debt. Where a true cohort share is not published, the page derives a transparent estimate and labels it "(est.)" with its denominator stated under the chart, rather than silently switching meanings.
| Universe | Who is counted | Group definitions | Used in |
|---|---|---|---|
| UC data UC Information Center | All new UC undergraduates (freshman or transfer entrants), by entering cohort year. By-school tables count enrollees from that source school. | UC's ethnicity categories; Pell status in the entering year; first-generation = no parent with a four-year degree. | Systemwide groups, high schools, community colleges, GPA bands |
| Federal data IPEDS / College Scorecard | Completion (C150): first-time, full-time bachelor's-seeking fall cohorts, completing at the same campus within 150% of expected time (six years). Scorecard earnings/debt: students receiving federal Title IV aid only. IPEDS completions: degrees conferred, not rates. | Federal race/ethnicity (includes two-or-more); Pell in entry year; Scorecard first-generation = no parent with any postsecondary schooling (broader than UC's definition); family-income terciles at entry. | Campuses, majors, earnings & debt, first-generation (federal) |
| UC San Diego data UCSD Institutional Research | UCSD's own first-year and transfer entering cohorts (2013–2024), broken out by the school or division of the student's major. Cells under the dashboard's reporting threshold are suppressed at source. | UCSD's categories: ethnicity keeps Hawaiian/Pacific Islander and Unknown/decline separate; first-generation and Pell as flagged by UCSD. | UC San Diego deep dive |
| UC Berkeley data UCB Office of Planning & Analysis | Berkeley's own freshman and transfer entering cohorts (2010–2023 for graduation; retention through 2023), from the campus "Disaggregated Grad Rates" dashboard and Our Berkeley downloads. | Berkeley's categories: disaggregated Asian (Chinese, Filipino, Japanese, Korean, South Asian, Vietnamese…) and Latinx subgroups, gender incl. nonbinary, first-generation, and the campus's EOP-eligibility markers. Retention splits use coarser groups and entry college. | UC Berkeley deep dive |
Federal C150 rates are keyed here to the Scorecard data year in which they were published. Each value reflects a cohort that entered roughly six years earlier; comparisons across campuses or groups within a year are unaffected, but do not read the x-axis as entering-cohort year. The amber x-axis tag restates this.
Bands are terciles of the systemwide enrollee GPA distribution for each cohort and entry type (weighted-capped high-school GPA for freshman entrants, transfer GPA for transfer entrants) — verified: the published band boundaries are identical across campus selections. Boundaries move year to year and are printed under the chart; each point's tooltip shows its cohort's band range. The lowest band always reaches down to that cohort's minimum recorded GPA (for the 2021 test-blind cohort this minimum is 0.00), so the chart reports only the informative inner boundaries. Because the bands are systemwide, campuses draw very unevenly from them: transfer sheets, which publish band rates alongside counts, imply that Berkeley's 2015 transfer entrants split roughly 200 / 520 / 1,450 across the three bands while Merced's split 73 / 12 / 10.
True band rates. The page now draws these views from a summary-API extraction of the same dashboard, which exposes the underlying full-precision band rates and their denominators (numerator ÷ rate; the harvest's internal consistency check agrees to ~10−12) for every campus × group × band × cohort cell, including per-campus and subgroup × band rates that the crosstab export left unidentified. Denominators are the dashboard's own universe — students with a usable GPA (systemwide ≈96% of the published entering cohort at 2019). Later windows are published as increments sharing the band denominator, so cumulative rates are exact sums. Rates for cohorts whose 5th/6th- year (transfer: 3rd/4th-year) windows haven't closed appear as real zeros in the source; those cohorts are omitted from the affected measures rather than plotted.
Maturity. A cohort's 5th/6th-year completions only exist once those years have passed. Freshman timing shares and six-year estimates are therefore shown only for cohorts old enough for the full window (through 2019 in the current data); later cohorts are omitted rather than plotted as artificial spikes. Transfer windows (2/3/4 years) are closed for all published cohorts.
School-level cells are shown by the source only when the entering cohort has at least 10 students, as integer percents. GPA-band timing shares are hidden here when fewer than 25 graduates underlie them. Systemwide lines carry the published cohort sizes in tooltips; small groups (for example American Indian entrants) are noisy — read them with their Ns.
Entering cohorts 2020–2021 experienced pandemic disruption, and 2021 was the first test-blind cohort; 2020–21 high-school records also carry grade inflation. Charts shade these cohorts.
Individual students' grades, major switching, or reasons for leaving; anything about students below the published masking thresholds; and for Scorecard measures, anyone outside the federal-aid universe.
Tidy CSVs behind this page: systemwide rates · by high school · by community college · GPA band rates · GPA band counts · GPA cutpoints · UC San Diego deep dive · UC Berkeley deep dive · federal campus panel · majors · earnings & debt · first-generation curves. Build scripts live in build/.
Last updated 31 July 2026. Build dates for each data layer are given with the sources below; revision history is in docs/CHANGELOG.md.