UK Decides is an independent UK political data and reference site. This document explains how polling figures, polling averages, election results and constituency information are selected, calculated and presented. It should be read alongside our Sources and Editorial standards pages.
1. Scope and source hierarchy
We distinguish between observed data (a published poll or declared election result), derived statistics (such as an arithmetic polling average) and forecasts or modelled estimates. They are not interchangeable.
For polls, the preferred source is the polling organisation’s original publication or data tables. For election results, we rely on official declarations and authoritative election datasets. Secondary reporting can provide context, but is not a substitute for an available original source. Links to the principal organisations are collected on the Sources page.
2. Polling database: records and verification
The polling database covers published UK Westminster voting-intention surveys. Entries identify the polling organisation, fieldwork dates, reported party percentages and a source link; sample sizes are shown where verified and available.
Fieldwork dates and publication dates are different. Fieldwork describes when respondents were surveyed; a release may appear days later. For comparisons over time, fieldwork is the more informative date. Where a reported percentage is absent, it should not be inferred from the other parties’ figures.
Pollsters may differ in sampling, turnout modelling, weighting, treatment of undecided respondents, party prompting and geographical coverage. UK Decides records their published headline Westminster vote shares; it does not retrospectively reweight the underlying respondents or make different pollsters’ methods identical. See Why pollsters disagree and How to read polling tables.
3. Current Poll of Polls: calculation specification
The current Poll of Polls is a rolling summary of national Westminster voting-intention polling, not an election forecast. The published calculation uses the following rules:
- Window: qualifying polls fall within the latest 30-day inclusion window.
- Selection: use the most recent qualifying poll from each polling organisation within that window.
- Frequency: each pollster contributes at most one poll, irrespective of how often it publishes.
- Aggregation: calculate an unweighted arithmetic mean of the selected pollsters’ reported vote shares for each party.
- No additional modelling: the headline average does not apply sample-size weights, pollster house-effect adjustments or a constituency seat projection.
In notation, for party p and its n qualifying reported values, the published average is:
Average(p) = [sum of the n reported percentages for p] / n
The number of contributing polls can differ by party. For example, Restore Britain is averaged only across selected polls that publish a separate Restore figure; an unreported value is not treated as 0%. The constituent polls and their source links appear beneath the current average so readers can inspect the inputs.
The 30-day window balances recency with coverage from pollsters that publish less frequently. Giving each polling organisation one contribution prevents a high-frequency publisher from dominating the average merely through the number of releases. This does not eliminate house effects or methodological differences.
4. Monthly averages and historical revisions
The monthly Poll of Polls table uses a different selection window from the current rolling average: the latest qualifying poll from each pollster in each calendar month. Those selected figures are averaged across pollsters for the month. The accompanying count shows how many pollsters contributed.
Where a party is not separately reported in any qualifying selected poll, the table may display a dash rather than manufacture a value. A reported zero and a missing result are distinct observations. Historical trends should be read with these gaps in mind, rather than assuming the gaps represent zero support or a measured change.
UK Decides is continuing to verify and add earlier polls, including records going back to January 2026. A monthly value may therefore change when another eligible historical poll is added or a record is corrected. The chart and table should be treated as a reproducible summary of the database as currently recorded, not as immutable historic publications.
5. Interpreting the numbers
Polls are estimates of public opinion during their fieldwork periods, not counts of votes cast. Sampling uncertainty is only one source of error; non-response, turnout assumptions, question design and systematic differences between pollsters also matter.
The gap between the highest and lowest current pollster figures, where displayed, is a range of published measurements, not a statistical confidence interval or a ranking of pollster quality. The Poll of Polls does not estimate how many Westminster seats a party will win.
For further background, see Margins of error, How weighting works and How polling averages work.
6. Election results and constituency boundaries
Election results are separate from polling records. Where available, constituency vote totals, party shares, majorities and turnout are based on declared results and authoritative parliamentary or election datasets. Turnout and vote share use different denominators and should not be confused.
Constituency boundaries change. Historical results are attributed to the constituencies and election years to which they apply; they should not be presented as though they were measured on today’s boundaries. Where a notional result or other boundary-adjusted comparison is used, it must be identified as an estimate rather than an officially declared constituency result.
Statistical and demographic fields may have different publication years and geographic bases. Readers should check the reference period before comparing census information, population figures and election outcomes. See our constituency directory and guide to comparing election results over time.
7. Corrections, provenance and editorial control
Each data page should provide enough context to establish what was measured, who published it and when. If an original source corrects its tables, we may amend the corresponding record and recalculate derived figures. Changes to an underlying poll can affect both the current average and monthly history.
UK Decides does not select or exclude polls because their results favour a particular party. Inclusion is governed by source verification and the published calculation rules. We do not endorse parties, candidates or electoral outcomes; our editorial standards explain the separation of data, explanation and opinion.
To report an incorrect poll, source, election figure or constituency detail, please contact info@ukdecides.com, ideally with the relevant page and an original source. Substantive corrections are made to the affected record and, where necessary, its derived displays.
Methodology reviewed: 14 October 2026. Calculation details describe the published UK Decides Poll of Polls at the time of review.