How Many Postgraduates, and How Many Staff to Supervise Them? (2026)

The headline finding: on the two figures a business case usually reaches for, Germany looks generously staffed and France looks stretched — 2.8 postgraduates per academic staff member against 8.1. Switch to the doctoral population alone and the ordering reverses: France has more staff per doctoral candidate than Germany, 1.9 against 2.4. The same two tables, the same year, opposite conclusions. Which means the denominator is doing the work, and any single “student–staff ratio” in your paper is a claim about definitions rather than about capacity.

Both numbers come from Eurostat’s UOE collection, which is free, keyless, and queryable in one HTTP request. The figures below were retrieved on 19 August 2026.

The two source tables

Students: educ_uoe_enrt01, “Students enrolled in tertiary education by education level, programme orientation, sex, type of institution and intensity of participation”. Dataset last updated 11 August 2026; reference year used here 2023; unit number; both sexes; all sectors; all intensities.

Staff: educ_uoe_perp01, “Classroom teachers and academic staff by education level, programme orientation, sex and age groups”. Dataset last updated 5 August 2026; reference year 2023; unit number; both sexes; all ages; ISCED level 5–8 combined, which is the finest cut this series publishes.

That last detail is not a footnote. There is no ISCED 7 or ISCED 8 staff figure to be had, so every ratio below divides a postgraduate numerator by a denominator that also teaches short-cycle and bachelor’s students. Everyone who publishes a postgraduate staffing ratio from this source is doing the same thing; most do not say so.

Enrolment: master’s and doctoral, 2023

The France–Germany doctoral gap is the first thing worth noticing: France enrols almost as many master’s students as Germany and barely a third as many doctoral candidates. Spain, on less than half of France’s master’s population, enrols 40 per cent more doctoral candidates. Doctoral scale is not a function of system size.

At EU-27 level the direction of travel is up on both levels but faster at the top: ISCED 8 enrolment rose from 660,592 in 2018 to 721,447 in 2023, a 9.2 per cent increase, against 7.0 per cent for ISCED 7 over the same period.

Academic staff, ISCED 5–8

Germany reports 3.6 times as many tertiary academic staff as France while enrolling 1.24 times as many postgraduates. No plausible account of teaching practice explains a gap that size. What explains it is who each system counts: Germany’s tertiary staff count sweeps in a large body of research and teaching personnel, including doctoral candidates employed on academic contracts, that France’s reporting treats differently. The number is correct in both cases and comparable in neither without adjustment.

The two ratios, and why they disagree

Both columns below are derived by us from the two tables above, not published by Eurostat.

The left column spreads over a factor of 3.1. The right column spreads over a factor of 1.4, and it puts France at the resourced end rather than the stretched one. A ratio that flips its ranking when you change the numerator is not measuring capacity; it is measuring the boundary between master’s and doctoral provision in each system.

The usable conclusion for an institutional paper is narrow and worth stating plainly: use these figures for your own country’s trend and for order of magnitude, and do not use them to argue that another country resources supervision better than you do. The same denominator discipline that applies to case counts applies here, and it is set out at length in why your integrity violation statistics measure enforcement rather than misconduct.

One cross-check worth running: stock against flow

Enrolment is a stock and completions are a flow, and dividing one by the other gives a crude but genuinely informative indicator no published table carries.

Germany’s federal statistical office recorded 28,171 doctoral examinations passed in 2024 (Destatis, revised 5 February 2026; 2023: 26,570; 2022: 27,692). Against the 205,302 doctoral candidates enrolled in 2023, that implies a stock-to-flow ratio of about 7.3 — the average number of years a candidate would spend registered if the population were in steady state and every registration eventually ended in an examination.

Neither condition holds, so read 7.3 as an upper bound rather than an average duration: the stock includes registrations that will end in withdrawal and registrations that are dormant. It is nonetheless the right order of magnitude to bring to a discussion about how long your own candidates are on the books, and it is a number a graduate school can reproduce for its own institution in an afternoon. What it cannot be turned into is a completion rate, for the reasons in where graduate completion and attrition data actually comes from.

The output side of the same population — how many theses per year actually reach an open index — is counted from a different free API in how many doctoral theses are produced each year.

How to pull these yourself

The dissemination API takes no key and no registration. The pattern is a base path plus the dataset code plus dimension filters:

https://ec.europa.eu/eurostat/api/dissemination/statistics/1.0/data/educ_uoe_enrt01?format=JSON&lang=EN&geo=DE&isced11=ED7&isced11=ED8&sex=T&unit=NR&time=2023&sector=TOT_SEC&worktime=TOTAL

Four practical notes. Dimension codes are dataset-specific and an invalid one returns a 400 whose message names the offending dimension, so the error response is the fastest way to learn the schema. Repeat a parameter to request several values. The staff table takes isced11=ED5-8 and age=TOTAL, and the aggregate geo code is EU27_2020. And the response carries an updated timestamp, which is the field that belongs in your footnote alongside the reference year.

Three caveats that belong in the footnote

• Reference year, not release year. These tables were refreshed in August 2026 and describe 2023. That is the ordinary shape of official statistics and it is covered in how current your higher education data actually is. Record reference year, release date and extraction date.
• Headcount, not workload. The staff series is a count of people, not of full-time equivalents or of supervisory capacity. A department with many fractional appointments looks better staffed than it is.
• UK and US are not in here. Neither is in the EU-27 aggregate; use HESA and IPEDS respectively, and do not splice their definitions into a Eurostat table without reconciling them first.

Used with those three lines attached, the pair of tables does the job most business cases actually need: a defensible order of magnitude for the population, a defensible order of magnitude for the people supporting it, and an explicit statement of what the comparison cannot carry. The operational consequence — that supervisory attention is the scarce resource in all five systems — is the subject of what your supervisors are actually spending review time on.

If you would like these figures reproduced for your own jurisdiction and set against your institutional registry, request an institutional evaluation and we will build the extract with you.

Frequently asked questions

How many academic staff are there in EU higher education?

1,533,943 in 2023 across the EU-27, up from 1,486,989 in 2022, counting ISCED levels 5 to 8 combined (Eurostat educ_uoe_perp01, dataset updated 5 August 2026).

How many postgraduate students are there?

At EU-27 level, 5,517,114 at ISCED 7 and 721,447 at ISCED 8 in 2023 (Eurostat educ_uoe_enrt01, updated 11 August 2026). Country figures for 2023 are in the first table above.

Can I get a staff figure for doctoral education alone?

No. The series publishes tertiary staff at ISCED 5–8 combined, so any postgraduate ratio built from it carries short-cycle and bachelor’s teaching in the denominator. State that when you publish the ratio.

Why do Germany and France differ so much on staff numbers?

Because their national reporting counts different populations as tertiary academic staff. Germany reports 483,305 against France’s 133,307 while enrolling only 1.24 times as many postgraduates, which is a definitional gap rather than a capacity gap.

Which ratio should I use in a business case?

Whichever you use, publish both the numerator and the denominator definitions beside it, and use it for your own country’s trend rather than for cross-country ranking. The two ratios in this article disagree about which country is better resourced.

Is the API really free?

Yes. The Eurostat dissemination API needs no key and no registration, returns JSON, and reports its own last-updated timestamp in the response.

How do I find the right dimension codes?

Send a request with a wrong one. The 400 response names the dimension it did not recognise, which is faster than reading the metadata pages for most datasets.

What is the stock-to-flow figure for Germany?

About 7.3, dividing 205,302 doctoral candidates enrolled in 2023 by 28,171 doctoral examinations passed in 2024. Treat it as an upper bound on average registered duration, not as an average.

Can I use that as a completion rate?

No. It contains no information about how many registrations end without an award, which is a separate and much harder measurement problem.

How current are these figures?

Reference year 2023, released in 2026. That is normal for the UOE collection and is the reason your footnote needs a reference year, a release date and an extraction date rather than just a source name.

What about part-time and fractional staff?

The unit used here is a headcount. Full-time-equivalent variants exist within the collection but are not comparably populated across all countries, so mixing them into one table quietly changes what the column means.

What is the single most common error with these tables?

Quoting a ratio without its denominator definition. It is the error that produces the France–Germany result above, and it survives peer review inside institutions because nobody asks what was counted.