UK election swingometer

It is an interactive UK seat projection model. You set the GB vote share for each party and it works out who wins each of the 632 constituencies it models. Behind it is a survey model. It learns how people vote by age, sex, education, ethnicity, how they voted in 2024 and where they live, from the British Election Study. It then applies that to the people who live in each seat, from the Census, and to each seat's 2024 result, and matches the totals to the polling average. When you move a slider, each seat moves by an amount the model has worked out for that seat, so a national shift lands differently in different places.

In autumn 2026 PollCheck replaced its seat model. The old model took each seat's 2024 result and moved it with the polls, using multipliers based on the seat's demographics. The new one estimates each seat's vote from what British Election Study respondents say, weighted to the seat's population, and adds a few judgement calls. The seat history on PollCheck was recalculated with the new model, so past projections will not match what was shown at the time.

A uniform national swing calculator applies the same percentage-point change to every constituency. This model distributes changes demographically instead. A rise in Reform support concentrates in Leave-voting, non-graduate seats; a rise in Green support concentrates in graduate, younger, urban seats. So the same national poll can produce a very different seat outcome here than under uniform swing.

The headline seat counts are the model's central projection. The ranges come from running the model many times with random polling error and seat-level error added. Seat classifications (Safe, Likely, Lean, Toss-up) show how often each party wins the seat across those runs.

It captures broad patterns - how age, education, ethnicity, past vote and the Brexit vote push different areas in different directions. It handles Scotland and Wales with their own polling, and picks up signals from recent local and devolved elections. It cannot predict surprises: a popular independent candidate, an MP's personal vote, a local campaign surge, or a pattern too local for the survey sample to see. Use it to test scenarios.

The model projects a vote share for every party in every constituency, then runs those through first-past-the-post: whichever party has the highest projected share in a seat wins it. Add up the winners across all 632 seats and you get the seat totals.

Yes. The Poll scenario panel lets you set the GB vote share for each party. When you raise one party the others adjust proportionally to keep the total near 100%, and you can lock any party to pin its value while the rest move. The pill row above the sliders loads a specific recent poll instead, or the default 7-poll average, which averages the seven most recent polls to smooth out individual poll noise. Polls are conducted for Great Britain, or sometimes just England and Wales, so they do not always include the SNP or Plaid Cymru; for missing parties the tool falls back to the baseline value, which is why a selected poll may not sum to exactly 100%. Restore Britain is the exception: a poll that does not ask about it sets it to 0, because its supporters are then counted in the parties that poll does report. The model normalises internally.

The current model has no turnout data, so the turnout sliders are hidden. The headline projection already includes a small amount of tactical voting. The "If people vote tactically" section further down starts from the headline and adds more: voters for smaller parties can switch to one of the two front-runners in their seat. The slider runs from the headline, where nothing more switches, to everyone who says they would switch doing so. You can also choose which parties' voters switch, and how Conservative voters split between Reform and Labour where those two are the top two.

Scotland and Wales are in the same survey model as England, with regional effects. The SNP and Plaid Cymru are set to their nation's polling, and the SNP and Plaid Cymru sliders only affect their own nations. Where each party is strong within Scotland and Wales also takes a signal from the May 2026 Holyrood and Senedd constituency results. In Wales, the overall level blends what the GB polls imply with recent Welsh Westminster polls, and the Plaid and Labour seat pattern is half this model and half PollCheck's previous model. Both Welsh steps are judgement calls rather than estimates.

Northern Ireland's party system is completely separate, and neither GB polls nor the British Election Study cover its parties, so the model does not project its 18 seats. They are shown as a static placeholder held at the GE2024 result, included only so the full 650-seat House and the 326 majority line are represented. They do not respond to the sliders, but you can reassign them by hand in the Northern Ireland panel.

The UK uses first-past-the-post. The Green vote is spread relatively thinly across many seats rather than concentrated in a few, so the Greens come 3rd, 4th or 5th in most seats and only win where their support is highly concentrated, as in Bristol Central in 2024. This is a feature of first-past-the-post, not a model error. Under proportional representation 12% of the vote would translate to roughly 12% of seats, about 76. Move the Green slider to see how many seats the model gives them at any national share.

Winner colours each constituency by its projected winner, the election-night view. Margin shows how comfortable the projected win is, with deeper colour for a wider margin. Party heatmaps show the vote-share gradient for one party, and the vs 2024 toggle turns that into change from the 2024 result, blue for up and red for down. The Census layer menu shades seats by a demographic variable, a reference layer that helps explain why the model predicts what it does.

Survey data: British Election Study Internet Panel (Fieldhouse, E., J. Green, G. Evans, J. Mellon, C. Prosser, J. Bailey, R. de Geus, H. Schmitt and C. van der Eijk), Waves 1-31. Population: Census 2021 (ONS) for England and Wales and Scotland's Census 2022 (NRS), both under the Open Government Licence. The 2024 General Election result is the starting point for every seat. Polling is a 7-poll average of recent national polls, plus Scottish and Welsh polls.

Yes. The "What would it take?" solver lets you pick a party and a target - winning the most seats or winning a majority - and it shows the vote share that would be needed. The answer is always measured from the current polling average, so it does not change between tries.