Postgraduate Enrolment by Field of Study: Sizing Thesis Load Faculty by Faculty (2026)

The headline finding: the field ranking inverts between master’s and doctoral level. Business, administration and law is the largest master’s field in the EU-27 — 1,270,082 students in 2023, 23.0% of all ISCED 7 enrolment — and only the sixth largest doctoral field, at 64,470 and 8.9%. Natural sciences, mathematics and statistics runs the other way: 6.5% of master’s enrolment and 20.1% of doctoral enrolment. A graduate school that sizes its thesis provision from its taught-masters mix will be wrong about its doctoral load by a factor of three in both directions.

Every figure below is from Eurostat table educ_uoe_enrt03 (students enrolled in tertiary education by education level, programme orientation, sex and field of education), reference year 2023, dataset updated 11 August 2026, extracted through the free dissemination API on 31 August 2026. Fields follow ISCED-F 2013; levels follow ISCED 2011. Totals are all sexes, all programme orientations.

Master’s enrolment by field, EU-27, 2023

The ten broad fields sum to 5,439,061; the residual 78,053 sits in generic programmes and in enrolment not allocated to a field. For scale, the same table puts total EU-27 tertiary enrolment at levels 5 to 8 at 18,822,775, of which 11,169,561 is bachelor’s level.

Doctoral enrolment by field, EU-27, 2023

The ratio that actually sizes a support service

Neither table on its own tells a graduate school what to staff. The operational number is the ratio between them: how many master’s students a field carries for each doctoral candidate. It separates fields where the thesis workload is a large, short, seasonal cohort from fields where it is a small, long, continuous one — and those two shapes need different services.

The spread is 8.6-fold, from 2.5 to 21.5. An education faculty and a physics department sit at opposite ends of it, and an institutional service designed around the sector average of 7.6 is materially mis-sized for both. The taught-masters end of that range is the cohort that arrives all at once, which is the load pattern described in our piece on the masters dissertation season capacity cliff; the doctoral end is a continuous multi-year load that never produces a peak and never goes away.

Where the broad field codes will mislead you

Two of the ten codes are aggregates that no institution is actually organised around, and both hide a reversal.

Business, administration and law is two very different faculties. At ISCED 7 the EU-27 splits into 674,275 in business and administration against 463,326 in law — business is 1.46 times larger. At ISCED 8 it splits into 18,097 in business and administration against 34,152 in law. Law produces close to twice as many doctoral candidates as business and administration, from a smaller master’s base. Expressed as the ratio above: business and administration runs at 37.3 master’s students per doctoral candidate, law at 13.6. Reading the combined code as if it described your business school understates your law school’s doctoral load by roughly a factor of three.

Health and welfare is not one workload either. Within it, medicine runs at 665,864 master’s-level and 97,507 doctoral students (a ratio of 6.8), nursing and midwifery at 69,215 and 2,526 (27.4), dental studies at 84,895 and 6,183 (13.7), and welfare at 41,609 and 1,213 (34.3). A faculty of health sciences containing a large nursing school and a small medical school has a completely different thesis profile from one with the reverse composition, and the broad code cannot see the difference.

The same disaggregation is available for every broad field — humanities excluding languages runs at 3.0 while languages runs at 8.2, inside a single arts and humanities code. If your provision plan is departmental, pull the narrow codes.

The caveat that decides whether the number is usable

ISCED 7 is a level, not a description of what the student is writing. National systems place different qualification types at that level, and the effect is large enough to change a plan.

Two visible examples sit in the numbers above. EU-27 medicine reports 665,864 students at ISCED 7 — far more than any plausible population of medical candidates writing a master’s thesis, because long-cycle medical qualifications are counted at that level in several systems. And Germany reports 105,315 law students at ISCED 7 against 28,470 in business and administration, an ordering that appears in no other system in this comparison and that a graduate school reading the EU aggregate would never predict.

The practical rule: before you convert an ISCED 7 count into a thesis-supervision denominator, check which national qualifications your own system reports at that level. The figure is comparable as a level count and is not comparable as a count of dissertations. This is the same class of problem as the reporting delay described in how current your higher education data actually is: the number is correct and it is answering a different question from the one you asked.

National composition varies too much to borrow a benchmark

Doctoral field shares across six member states, 2023, same table:

The health and welfare row runs from 6.3% in France to 37.2% in the Netherlands. That six-fold spread is too wide to be a real difference in medical research capacity between two neighbouring systems, and it should be read primarily as a classification difference in where clinical training and clinical doctorates are counted. It is a good illustration of why a benchmark taken from a peer institution in another country is not evidence about your own establishment.

One further figure worth knowing before anyone quotes a national comparison: Germany reports 10 doctoral students in nursing and midwifery in 2023, against 442 in Spain. That is a structural feature of how nursing education is organised and classified in Germany rather than a measure of research activity, and it is exactly the kind of number that becomes a misleading slide when it is lifted out of its metadata.

How to pull your own version of this table

The Eurostat dissemination API is keyless, needs no registration and returns JSON. The pattern is a single request per level and country:

https://ec.europa.eu/eurostat/api/dissemination/statistics/1.0/data/educ_uoe_enrt03
  ?format=JSON&lang=EN&freq=A&unit=NR&sex=T
  &isced11=ED7&geo=DE&time=2023&iscedf13=F04&iscedf13=F09

Four working notes, each of which cost time to discover:

• Repeat a parameter to request several values of the same dimension, as with the two iscedf13 values above.
• An unfiltered request returns HTTP 413 with an asynchronous-processing message rather than data. Always filter at least the level, the country and the year.
• An invalid dimension name returns a 400 that names the offending dimension. That is the fastest way to discover the schema of any Eurostat table.
• The response carries its own updated timestamp. Put it in the footnote of whatever you produce, because the same query will return revised values later.

The narrow ISCED-F codes work in exactly the same call: F041 business and administration, F042 law, F0912 medicine, F0913 nursing and midwifery, F061 information and communication technologies. The dimension carries 183 codes in total, which is enough resolution for almost any departmental question.

What this changes in a provision plan

Three decisions move once the field cut is in front of a committee rather than the institutional total.

1. Where a seasonal service is needed and where a continuous one is. Fields above roughly 15 master’s students per doctoral candidate carry their thesis load as an annual cohort with one deadline. Fields below about 5 carry it as a permanent, individually supervised population. The same budget buys a different service in each case.
2. Which faculty a pilot should run in. A pilot in a field at one end of the ratio does not generalise to the other end, which is worth stating in the pilot design rather than discovering in the evaluation — see our sequence for measuring whether a writing support intervention worked.
3. What the supervision denominator is per faculty. The staff side of that calculation, and the reason two Eurostat tables disagree about which country is under-resourced, is set out in how many postgraduates and how many staff to supervise them. The output side — how many completed theses a system actually produces — is queryable separately, as described in the free field-by-field count of doctoral theses.

If you would like to work through the field cut for your own institution and what it implies for provisioning, request an institutional evaluation and we will build the table against your own registry extract.

Frequently asked questions

Which field has the most master’s students in the EU?

Business, administration and law, with 1,270,082 students at ISCED 7 in 2023 — 23.0% of all EU-27 master’s-level enrolment. Health and welfare is second at 1,113,773.

Which field has the most doctoral students in the EU?

Health and welfare, with 151,311 students at ISCED 8 in 2023, or 21.0% of the EU-27 total. Natural sciences, mathematics and statistics is close behind at 144,750.

Why does the ranking change between master’s and doctoral level?

Because fields differ enormously in how many taught-masters students they carry per doctoral candidate. The EU-27 average is 7.6, but education runs at 21.5 and natural sciences at 2.5. Professional master’s markets and doctoral research capacity are not the same population.

Is the ISCED 7 figure a count of master’s theses?

No. It is a count of students enrolled at level 7, which in several national systems includes long-cycle professional qualifications. Check which of your own awards are reported at that level before treating it as a dissertation denominator.

Can I separate law from business in this table?

Yes. The narrow ISCED-F codes F041 and F042 split the combined code, and the split is worth pulling: at doctoral level law is nearly twice the size of business and administration in the EU-27, which the combined code hides entirely.

What is the most recent reference year available?

2023, in a dataset updated on 11 August 2026. Education statistics reach the database roughly two academic years after the year they describe, so the current cohort is never in the table.

Does Eurostat charge for this data?

No. The dissemination API is free, requires no registration and no key, and returns JSON. A single filtered request per level and country is enough to reproduce every table on this page.

Why is the health share so different between France and the Netherlands?

The reported doctoral share of health and welfare is 6.3% in France and 37.2% in the Netherlands. A six-fold gap between neighbouring systems is not a research-capacity finding; it reflects where clinical training and clinical doctorates are classified. Treat cross-country field shares as indicative only.

How granular can the field breakdown go?

The field dimension carries 183 codes, from the ten broad fields down to narrow codes such as nursing and midwifery, dental studies, electronics and automation, and building and civil engineering.

Which figure should go in a business case?

Your own registry extract, cut by faculty, with the Eurostat table used only to show whether your composition is unusual. A local denominator survives scrutiny; a sector average invites the question of whether it applies to you.

How often should this be refreshed?

Annually, and re-pull rather than reuse: the API response carries an updated timestamp and earlier years are revised. Record the timestamp alongside the figures in any paper that goes to committee.