How the constituency pages work
Every number on PollCheck's 632 constituency pages is computed from a named source - election results, official statistics or PollCheck's projection model. This page explains each data layer once, so the pages themselves can stay focused on the data.
The current projection
The "current projection" on every page is PollCheck's demographic swingometer applied to the rolling seven-poll average of GB-wide polls. Each constituency has per-party sensitivity multipliers derived from demographic regressions, so national vote-share movement translates into different local movement seat by seat. It is not a true MRP: vote-share changes are applied through those multipliers rather than estimated from raw respondent data.
The projection answers one question - how would this seat vote if a general election were held today, at the current poll average. It is not a forecast of the next general election, and on seats with a pending by-election it is explicitly not a by-election forecast: by-elections have different turnout, candidates and tactical dynamics.
The projection updates automatically as new polls are added. The same model drives the whole-country swingometer, so seat-level and national figures always agree.
General election results
2024 results come from the UK Parliament Election Results portal and House of Commons Library briefing CBP-10009. The 2019 row in the history table is Parliament's notional recalculation of the 2019 vote onto the 2024 boundaries, which makes it directly comparable to 2024. Results for 2010, 2015 and 2017 are from House of Commons Library results files, on the boundaries in force at the time.
Boundaries and predecessor seats
All pages use the constituencies as drawn by the 2023 boundary review (first used in July 2024). Where a seat was created or substantially redrawn by the review, pre-2024 rows in the history table show the predecessor seat that covered most of the same area, identified by geographic overlay. Those rows are flagged "predecessor" in the table and their vote shares and majorities are not directly comparable to post-2024 figures. Turnout changes are likewise only shown between elections fought on the same basis.
Ward-level GE2024 estimates
The "GE2024 winner" column in the ward table, and the ward-winners map layer, use Britain Elects / New Statesman modelled estimates of how each council ward voted at the 2024 general election (article by Ben Walker, underlying data). Ballots are not counted by ward at general elections, so these are estimates modelled from constituency totals, with an average margin of error of about 4 points per ward. Treat single-ward margins under ~8 points as toss-ups. The estimates do not cover every seat; where they are missing the ward layer is omitted.
The council layer
Local council elections are a separate ballot from Westminster general elections - different candidates, much lower turnout, and strong local factors. The council section shows them as context, not as an input to the Westminster projection.
What the "Most recent council ward results" table shows, in order of preference:
- May 2026 results where the council held elections on 7 May 2026, sourced from Democracy Club via PollCheck's locals dataset.
- The most recent earlier result (2014-2024) where the council was not in the 2026 cycle, from DCLEAPIL v1.0 (Jason Leman, drawing on Andrew Teale's LEAP dataset and Democracy Club). English councils rotate their election cycles, so "most recent" can be several years old.
- Sitting councillors from OpenCouncilData where no contested ballot is on record for the ward - a "who holds the seat today" view rather than an election result, so no vote shares are shown.
The "Shift since GE2024" column is only computed for May 2026 results, because earlier council votes pre-date the general election. Where the same party won the ward at both elections it shows that party's share change; where the ward flipped it shows the conventional two-party swing (the average of the loser's fall and the winner's rise).
Some seats also show a county council table: counties are a separate upper tier of local government, and several counties voted on 7 May 2026 even where their districts did not. County vote shares are computed from declared per-candidate ballots.
Councils rotate ward boundaries too. Where a council redrew its wards for 2026, results are matched to the old wards where possible and flagged where a ward was absorbed into a larger one.
Demographics
Demographic tiles use the ONS Census 2021 for England and Wales, aggregated to constituency level with the ONS LSOA-to-constituency best-fit lookup, and Scotland's 2022 Census for Scottish seats. The 2016 EU referendum figure is a constituency-level estimate (Hanretty 2017 method) - the referendum was not counted by constituency. Median age is computed from Census age-band tables.
The constituency map
The map has two layers: the ward-winners layer described above, and a demographic layer that splits the seat into census neighbourhoods - around 1,500 residents each in England and Wales (LSOAs), around 800 in Scotland (Data Zones). Boundary geometry comes from the ONS Open Geography Portal.
Similar constituencies
The "constituencies most like this one" cards are the five seats with the closest demographic profiles, matched on census variables. Each card lists the features where both seats deviate from the England mean in the same direction. The comparison is demographic only - the politics shown on the cards is context, not part of the match.
MP details
MP names, parties, start dates, government and opposition roles, and photos come from the UK Parliament Members API.
Something look wrong?
Every figure should trace back to one of the sources above. If you spot a number that does not, please let us know - data corrections are applied across all 632 pages at once.