We ran a separate topic model for each country to preserve national differences. Together, the models identified 138 topics: 75 in England, 30 in Scotland and 33 in the Republic of Ireland. Explore which topics emerged, how much of each collection they account for and how that changes over time.
TRACE analysed 9,367 publicly available documents dated from 2 January 2023 to 25 June 2026: 6,512 from England, 1,033 from Scotland and 1,822 from the Republic of Ireland. The documents came from 63 named sources, including governments, parliaments, inspectorates, researchers, unions, campaigners and media organisations.
Documents were divided into passages. A separate topic model was fitted for each country so that country-specific patterns could be examined without the much larger England collection determining one shared topic structure. Each model represented passages numerically and grouped passages that used similar language.
The models identified the groups, but they did not determine what those groups meant. One researcher reviewed the principal terms and representative material associated with each topic and assigned its label. Those labels have not been independently reviewed by an education-policy expert.
The models produced 75 topics for England, 30 for Scotland and 33 for the Republic of Ireland. These totals reflect differences in collection size, model settings and the material available to each model. They must not be interpreted as evidence that one country’s education debate was more diverse than another’s.
Minimum cluster sizes were set separately for the three models, 120 for England, 60 for Scotland and 70 for the Republic of Ireland, to produce usable topic structures from collections of different sizes. These are modelling parameters, not substantive thresholds for what counts as an educational issue.
In simple terms: the model grouped similar language; a person interpreted and named the groups.
The model groups passages that use similar language. A researcher examines each group and assigns a label describing what it appears to represent. A fourth step is not shown: each document’s passage results are then rolled up to a single principal category, which is the unit every figure on this site reports.
Each topic is a cluster identified by the model, and each document is assigned once to its strongest match. Bars show a topic’s size relative to the largest topic in that same country: a longer bar means more frequent, not more important. Percentages are each topic’s share of its own country’s documents and sum to 100% within that country, so they are not comparable between countries. What can be compared is where a topic sits in the order within each country: a topic can rank third in one record and twelfth in another. The shares also carry differences in source coverage and publisher mix, and a smaller collection gives a less precise share: Scotland’s is a sixth the size of England’s. Trend lines show change from 2023 to mid-2026, while arrows show changes in rank. A small topic was uncommon in these sources; that does not mean it was unimportant. A researcher assigned the topic names. Open a topic to see how many publishers contributed to it and whether one of them wrote more of it than they usually write: a topic can rank highly because many people discussed it, or because one organisation published a great deal about it, and those are different things. Select a topic to read a short description.
Choose topics
The picture changes depending on when you look. A shift in the topics may reflect a change in the documents collected, not the debate itself.
The topics identified in each country, how much of the collection they account for, and how that changes from year to year.
A larger topic contains more documents. Interpret its size alongside the leading publishers, the country’s source mix and the fact that 2026 runs only to June.
Size is not importance. An issue may appear absent because it was not collected, was grouped elsewhere or was given another name.
This is a view of each country through the documents we collected. It is a starting point for questions, not a complete map of its education debate.