1.1·Where the evidence comes from

Where does this evidence come from?

We collected articles and reports published between January 2023 and June 2026 from education media, government, parliament, unions, civil society, research organisations and others. That is 9,367 documents from 63 sources across the three countries. It reflects the structure of the debate, but it is not the full picture. This page shows who publishes the documents behind our analysis, and how that changes over time.

How we assembled the corpus

What counts as in scope. Public education-policy material published in England, Scotland or the Republic of Ireland since January 2023. Ten kinds of publisher are represented: education media, government, parliament, unions, civil society, think tanks, research organisations, professional bodies, advocacy organisations and regulators.

How a source gets in. A source is included only where its material is public and its terms permit responsible collection. Collection is rate-limited and runs under the study’s research-ethics process. We keep each source’s type and drop its name, so no outlet can be identified from anything published here.

How often it runs. The scrape runs weekly, because feeds drop older items if you wait, and the site itself is rebuilt monthly. The collection is not frozen: it grows, and the figures on these pages are those of the most recent rebuild.

What happens to a document. Documents are reduced to one row each, with duplicates removed and page furniture stripped out. Place and organisation names are removed before modelling, so the models group on the language of the subject rather than on proper nouns. Each document is then split into passages, because one document usually covers several subjects.

What the counts mean. We collected 9,426 documents from 63 sources. 9,367 of them carry a substantive category and are the basis of every figure in the analysis; the remaining 59 had no substantive topic and are set aside. The counts on this page are of everything collected, so they are the larger number.

Not every source spans the whole period. 22 of the 63 sources do not run the full window, covering 1,651 documents. A source that starts late or stops early changes what the record holds in a given year without anything changing in the debate itself, so any reading of movement over time carries that with it.

What this does not cover. Anything published behind a paywall, anything not published openly at all, and anything a source stopped publishing before we began. What an organisation chooses to put online is not the same as what it does or thinks, and the record reflects the first.

In simple terms: we collected what these organisations published openly, kept the kind of organisation and dropped the name.

Who published the documents we collected

How to read this (expand)

Each bar is a share of that country’s own collected documents, so the bars within a country sum to 100%. The three collections are very different sizes, and that is itself a finding: England’s is roughly six times Scotland’s. Source counts are the number of separate publishers of that kind, not the number of documents. These counts cover every document collected, including the small number later set aside because no substantive category applied, so they do not sum to the 9,367 used in the analysis. Choosing a single year switches the chart to seven grouped kinds of publisher rather than ten, because that is the level the year-by-year figures are held at: think tanks and research organisations are counted together as research, and advocacy, professional bodies and regulators fall under other.

Period
England Scotland Rep. of Ireland Bars are the share of each country’s own collected documents.

England

Education media
54.0%  2 sources
Government
13.1%  3 sources
Union
9.5%  3 sources
Think tank
6.1%  4 sources
Professional body
5.6%  3 sources
Research organisation
5.5%  3 sources
Civil society
2.8%  3 sources
Parliament
2.1%  3 sources
Advocacy
0.9%  2 sources
Regulator
0.6%  2 sources

Scotland

Government
36.6%  2 sources
Civil society
19.8%  3 sources
Union
13.3%  2 sources
Parliament
9.4%  2 sources
Think tank
8.8%  2 sources
Regulator
4.6%  1 source
Professional body
3.2%  2 sources
Education media
2.8%  2 sources
Research organisation
1.4%  1 source
Advocacy
no source of this kind collected

Ireland

Government
29.8%  2 sources
Parliament
22.1%  2 sources
Union
17.1%  3 sources
Civil society
7.9%  2 sources
Education media
7.6%  3 sources
Research organisation
6.8%  2 sources
Advocacy
4.6%  1 source
Professional body
2.5%  2 sources
Regulator
1.7%  1 source
Think tank
no source of this kind collected
0%20%40%60%

Who is missing?

What to take from this

Our dataset does not capture everything that was said. It reflects who is publishing and how often, and this impacts the choices we make and what we find.

What this shows

Who published the documents we collected in each country, how many sources of each kind we found, and how much of the collection they produced.

How to read it

A larger share means more documents, not more influence. A few prolific sources can outweigh many quieter ones, so source count and document share tell different stories.

What this cannot tell us

It does not show who is influential, who is right or who represents whom. It shows only what we could collect, shaped by what organisations publish publicly and how often.

Why this matters

This mix shapes everything that follows. Education media produced more than half of England’s collection; government produced more than a third of Scotland’s. Every finding on this site rests on that foundation.