Where CS graduates go
144,101 US computing dissertations, 2000–2025, linked person by person to what their authors did next. The career data does not know what anyone studied; the dissertations do.
Ask a computing PhD student where their cohort ends up and you get anecdote. Ask a department and you get the placements it chose to publish. Ask a national survey and you get a snapshot taken at graduation with nothing attached about what the person actually worked on.
This is the join that was missing: this person wrote this dissertation, on this topic, then went here, and was still there five years later. 144,101 US computing dissertations from 2000 onward, 100,644 doctoral and 43,457 master’s, matched one-to-one against a career panel covering tens of millions of US graduate degree-holders.
The headline
Of doctoral graduates from the 2015–2022 cohorts, one year after the degree: 50% are in industry, 23% in academia, 5% at research institutes and national labs, and a further 8% at a named company the panel cannot assign to a sector. The remaining 11% have no position recorded anywhere. That last group is the share of a US-trained computing PhD cohort that the American labor market never sees: people who left the country, or who never kept a public professional record. In a field where a large majority of doctorates go to international students, it is one of the more consequential numbers on this page.
Why the career data could not answer this on its own
Resume-derived labor panels have covered US graduate degree-holders for years. What they do not know is what anyone studied. In the panel used here, field of study takes 18 distinct values across 175 million education records, and Computer Science is not one of them. The nearest category is “Information Technology”; most computing degrees land in the catch-all “Engineering”, and 62% of records leave the field blank. So the career side knows where someone went and roughly what they earn, and cannot tell you they were a computer scientist. The dissertation supplies exactly that, in the author’s own words.
What is in here
Eight sections, each one a different cut of the same linkage.
- Part 01 – the national destination split, and how the academic share erodes over the decade after the degree rather than at graduation.
- Part 02 – what they worked on. Deep learning went from 1.3% of dissertations in 2012 to 50.1% in 2024, and most of that is outside machine learning.
- Part 03 – placement by degree-granting institution, a several-fold range.
- Part 04 – how two datasets that share no identifier were joined, and how the dissertations were labelled.
- Part 05 and Part 08 – one institution measured the same way as everyone else, then its last decade rebuilt in full.
- Part 06 – the 2021–2025 cohorts: what each subject got hired to do and what it paid.
- Part 07 – the 2023 hiring freeze, which cost one cohort 11 points of industry placement in a single year.