How we turn the UK's polls into a forecast for the House of Commons (650 seats, 326 for a majority): the 27 polls of the last 30 days, corrected for how the UK's final polls have missed before. Today's average: Labour 28.4, Reform UK 21.9, Conservative 21.7, Lib Dem 9.9.
Today's polls, read our way
England, from the 5 pollsters whose raw tables we have loaded for the last 28 days, next to the headline average of all pollsters.
| Labour | Conservative | Reform UK | Lib Dem | Green | Restore Britain | Lab lead over Con | |
|---|---|---|---|---|---|---|---|
| Headline average, all pollsters | 29.7 | 23.1 | 22.2 | 10.3 | 9.5 | 3.2 | +6.6 |
| Our adjusted average | 28.6 | 24.3 | 21.5 | 10.5 | 9.8 | 3.4 | +4.3 |
This is our headline forecast: see the seats →
Poll by poll (latest 15)
Each figure is Labour's lead over the Conservatives in England, in points. Raw: from the pollster's raw tables. Published: the pollster's own headline figures. Adjusted: after our corrections. ● marks the polls in today's average above; the others are older polls by the same pollsters.
| Poll | Raw | Published | Adjusted |
|---|---|---|---|
| Opinium ●20 Nov 2026 | +8.6 | +8.6 | +5.6 |
| Find Out Now ●19 Nov 2026 | +7.5 | +8.7 | +4.5 |
| YouGov ●16 Nov 2026 | +4.0 | +3.0 | +1.0 |
| More in Common ●15 Nov 2026 | +6.5 | +3.9 | +3.5 |
| Find Out Now12 Nov 2026 | +6.5 | +7.1 | +3.5 |
| Survation ●12 Nov 2026 | +10.0 | +8.6 | +7.0 |
| More in Common9 Nov 2026 | +5.7 | +2.7 | +2.7 |
| YouGov9 Nov 2026 | +8.1 | +5.1 | +5.1 |
| Opinium6 Nov 2026 | +7.7 | +7.7 | +4.7 |
| Find Out Now4 Nov 2026 | +6.9 | +7.1 | +4.0 |
| YouGov2 Nov 2026 | +5.5 | +2.0 | +2.5 |
| More in Common1 Nov 2026 | +7.3 | +6.4 | +4.3 |
| Opinium30 Oct 2026 | +10.4 | +10.4 | +7.4 |
| Find Out Now29 Oct 2026 | +7.3 | +6.6 | +4.3 |
| More in Common26 Oct 2026 | +5.7 | +2.9 | +2.7 |
Our adjusted average
The steps
- Start from what people said: the pollster's weighted table of everyone asked, before their turnout and don't-know steps. Don't knows and won't-votes are set aside.
- England figure: the pollster's England column, or its English regions added up by their weighted bases.
- Take off half the average past miss of raw answers, party by party (right).
- Average the latest poll from each pollster that publishes its raw tables, equal weights, last 28 days.
Half, not all: taking off the whole average miss did better on average but worse in 2017, and a correction has to help every time, not just on average.
The correction today
| Party | Average raw miss | We change raw answers by |
|---|---|---|
| Conservative | -4.0 | +2.0 |
| Labour | +2.0 | -1.0 |
| Lib Dem | +0.1 | -0.1 |
| Reform UK | +1.3 | -0.6 |
| Green | +0.5 | -0.3 |
Other parties (Restore Britain, the SNP, Plaid) have no raw-answer history, so they are not changed, only rescaled.
Each call comes from the same two records: how the published final polls missed in England at each election since 2015, and how the raw answers behind them missed. We take half the raw answers' average miss off today's raw answers. The full amount did better on average but overshot in 2017, and we want a correction that helps every time, not just on average.
The Conservatives England: polls 23.1, our bet 24.3 (+1.2)
We think the polls understate the Conservatives by 1.2 in England. The final polls understated them at 3 of 4 elections since 2015 (by 2.6 on average), and what people told pollsters before any adjustment understated them at all 4.
| Miss (poll − result) | 2015 | 2017 | 2019 | 2024 |
|---|---|---|---|---|
| Published final polls | −5.7 | +0.6 | −1.6 | −3.5 |
| Raw answers | −6.6 | −2.1 | −3.3 | −4.0 |
Labour England: polls 29.7, our bet 28.6 (−1.1)
We think the polls overstate Labour by 1.1 in England. The final polls overstated them at 3 of 4 elections since 2015 (by 0.7 on average), and what people told pollsters before any adjustment overstated them at 3 of 4.
| Miss (poll − result) | 2015 | 2017 | 2019 | 2024 |
|---|---|---|---|---|
| Published final polls | +1.5 | −3.9 | +0.9 | +4.3 |
| Raw answers | +1.6 | −1.5 | +1.8 | +6.1 |
Reform UK England: polls 22.2, our bet 21.5 (−0.7)
We think the polls overstate Reform UK by 0.7 in England. The final polls overstated them at all 4 elections since 2015 (by 1.4 on average), and what people told pollsters before any adjustment overstated them at all 4.
| Miss (poll − result) | 2015 | 2017 | 2019 | 2024 |
|---|---|---|---|---|
| Published final polls | +0.8 | +2.1 | +0.8 | +2.0 |
| Raw answers | +0.9 | +2.1 | +1.1 | +1.0 |
The Greens England: polls 9.5, our bet 9.8 (+0.3)
We think the polls understate the Greens by 0.3 in England. The final polls understated them at 1 of 4 elections since 2015, and what people told pollsters before any adjustment understated them at 1 of 4.
| Miss (poll − result) | 2015 | 2017 | 2019 | 2024 |
|---|---|---|---|---|
| Published final polls | +0.9 | +0.6 | +0.1 | −0.6 |
| Raw answers | +1.1 | +1.0 | +0.4 | −0.4 |
The Lib Dems England: polls 10.3, our bet 10.5 (+0.2)
The Lib Dems: close to the polls. Past misses point both ways (the final polls overstated them at 2 of 4 elections and understated them at 2), so we make no real call.
| Miss (poll − result) | 2015 | 2017 | 2019 | 2024 |
|---|---|---|---|---|
| Published final polls | +1.8 | +0.7 | −0.6 | −0.9 |
| Raw answers | +2.0 | +0.7 | −0.5 | −1.7 |
Red: the polls overstated the party. Blue: they understated it. Published figures are every pollster's final poll; raw answers are the final polls whose full tables survive (18 polls). Today's raw answers come from 5 pollsters: Opinium, Find Out Now, YouGov, More in Common and Survation.
How we get from the pollster average to our bet
Every number in England, in points, for today. The gap between the pollsters' average and our bet has three parts, and we show each one.
| Step | Conservative | Labour | Reform UK | Lib Dem | Green |
|---|---|---|---|---|---|
| 1. Average of every pollster's published figure (5 pollsters, latest poll from each in the last 28 days) | 23.1 | 29.7 | 22.2 | 10.3 | 9.5 |
| 2. The same, for only the 5 pollsters whose full tables we hold | 23.1 | 29.7 | 22.2 | 10.3 | 9.5 |
| 3. What those people actually told the pollsters, before the pollsters' turnout and don't-know adjustments | 22.3 | 29.6 | 22.2 | 10.5 | 10.1 |
| 4. Our correction: half the amount raw answers have missed by, averaged over 2015, 2017, 2019 and 2024 | +2.0 | −1.0 | −0.6 | −0.1 | −0.3 |
| 5. Our bet (step 3 corrected, then scaled back to 100%) | 24.3 | 28.6 | 21.5 | 10.5 | 9.8 |
| Our bet minus the pollster average (5 − 1), and where it comes from | |||||
| Which pollsters we can read (2 − 1) | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Undoing the pollsters' own adjustments (3 − 2) | −0.8 | −0.1 | −0.1 | +0.2 | +0.6 |
| Our correction, after scaling back to 100% (5 − 3) | +2.0 | −1.0 | −0.6 | −0.1 | −0.3 |
| Total difference from the pollster average | +1.2 | −1.1 | −0.7 | +0.2 | +0.3 |
Step 1 is the figure every other forecast uses. Steps 2 and 3 are the same pollsters' latest polls, so the first split shows how much of our difference is simply which pollsters publish full tables. The Scotland and Wales figures are left as polled. Each poll's own page shows these steps for that poll.
Why we measure the polls in England
Every poll you see in the news is a Great Britain figure. We work with England's figure from the same poll, taken from its published tables. That is what lets us predict the next miss.
- One contest, not three. England, Scotland and Wales have different parties. The SNP and Plaid Cymru only stand in Scotland and Wales, so a GB share mixes three separate races. In 2015 Labour's vote fell from 42% to 24% in Scotland but rose in England. The GB figure hid both moves. In England every voter chooses between the same parties, so a poll's error there is a like-for-like miss.
- It is where the seats are. 543 of the 650 seats are in England. We forecast them from England's vote shares, by swing from the 2024 England result. So the error that moves our seat forecast is the England error, and that is the one we measure and correct.
- It is how we correct each poll. For each pollster we take the England figure from its raw answers, then take off part of the average amount raw England figures have missed by before. The past miss and the poll it corrects are measured the same way, so the correction applies like for like. A GB miss would carry Scottish swings (the 2015 collapse, for example) into English seats.
- It costs little precision. About six in seven people in a GB poll live in England, so the England figure rests on nearly the whole sample. Scotland and Wales get their own polls and their own record of polling error (Scotland & Wales).
The catch: England figures and raw tables are only published consistently from 2015, so this record covers four elections. For the longer history we use Britain-wide final polls back to 1945 (Long view). They show which way the polls tend to miss, but they can't feed the correction.
If 2024 is anything to go by, Labour is overstated by much more
At the last election Labour did far worse than the polls said. In England the final polls had Labour 4.3 points too high and the raw answers behind them 6.1 too high. Labour's lead over the Conservatives was overstated by 7.8 points.
Our bet takes only 1.0 off Labour's raw answers, because it averages four elections and then halves the result, and in 2017 the polls understated Labour by 3.9. If today's polls are exactly as wrong as in 2024, Labour would win 276 seats rather than our 319, with the Conservatives on 200 and Reform UK on 40.
So should we lean on the most recent election? We tested it the way a real forecast works: each election predicted only from the ones before it. Average miss per party in England (Labour's own miss in brackets):
| Election | Pollsters | Our method half, all earlier elections | Half the previous election's miss | All of the previous election's miss |
|---|---|---|---|---|
| 2017 from 2015 | 1.58 (−3.9) | 1.17 (−2.3) | 1.17 (−2.3) | 2.03 (−3.1) |
| 2019 from 2015 and 2017 | 0.79 (+1.0) | 0.90 (+1.8) | 1.15 (+2.6) | 1.44 (+3.3) |
| 2024 from 2015, 2017 and 2019 | 2.19 (+4.8) | 2.18 (+5.7) | 2.00 (+5.2) | 1.43 (+4.2) |
| Average | 1.52 | 1.42 | 1.44 | 1.64 |
These tests use the final polls whose full tables survive, so the pollsters' Labour miss here differs a little from the figure for all final polls above.
On Labour alone, our method has been further out than the pollsters in 2019 and 2024. The pollsters' own turnout and don't-know adjustments lowered Labour at 4 of 4 elections, and that was the right call at 3 of them (not 2017). We set those adjustments aside and take off 1.0 instead. Their adjustments took off more than that in 2017 and 2024. Across all parties our method is ahead; Labour is the party we watch most closely, in the same direction as the 2024 warning.
Best on this test: our method. Assuming the polls repeat the last election's miss has not paid off: misses swing from one election to the next (Labour was understated in 2017 after being overstated in 2015), so we keep averaging. We show the 2024 case because it is the most recent evidence, and the one to watch.
What the raw answers show
For each general election since 2015 we went back to the final polls' own tables and took the figures before any turnout or don't-know adjustment: everyone asked, weighted to the population, don't knows set aside. This is how far those raw answers were from the result in England, where every party stands.
| Election | Polls | Conservative | Labour | Lib Dem | Reform UK | Green | Lab lead over Con |
|---|---|---|---|---|---|---|---|
| 2015 | 2 | -6.6 | +1.6 | +2.0 | +0.9 | +1.1 | +8.2 |
| 2017 | 4 | -2.1 | -1.5 | +0.7 | +2.1 | +1.0 | +0.6 |
| 2019 | 6 | -3.3 | +1.8 | -0.5 | +1.1 | +0.4 | +5.1 |
| 2024 | 6 | -4.0 | +6.1 | -1.7 | +1.0 | -0.4 | +10.1 |
| Average | -4.0 | +2.0 | +0.1 | +1.3 | +0.5 | +6.0 |
Points, raw answers minus the England result. Red: raw answers overstated the party; blue: understated. The Reform column is UKIP in 2015 and 2017 and the Brexit Party in 2019.
What the pollsters' own adjustments did
Pollsters know their raw samples lean, and their turnout and don't-know steps try to fix it. Here is the Labour lead over the Conservatives in the same polls, before and after those steps, against the result.
| Election | Raw answers: lead error | Published headline: lead error | What the adjustments did |
|---|---|---|---|
| 2015 | +8.2 | +7.1 | Helped, but not enough |
| 2017 | +0.6 | -4.5 | Made it worse |
| 2019 | +5.1 | +2.7 | Helped, but not enough |
| 2024 | +10.1 | +7.8 | Helped, but not enough |
Lead error = poll's Labour lead over the Conservatives minus the real one, in points. In 2017 the turnout models (for example Ipsos counting only those certain to vote) moved polls towards the Conservatives by far too much.
Did it work?
We tested it the strict way: for each election, the correction was worked out only from the other three, as if we did not know the result, and compared with what the pollsters published.
| Election | Our average miss per party | Pollsters' average miss | Our Lab–Con lead error | Pollsters' lead error | |
|---|---|---|---|---|---|
| 2015 | 1.79 | 2.13 | +5.5 | +7.1 | Better |
| 2017 | 1.29 | 1.58 | -3.3 | -4.5 | Better |
| 2019 | 0.64 | 0.79 | +1.9 | +2.7 | Better |
| 2024 | 2.18 | 2.19 | +7.8 | +7.8 | Same |
| Average | 1.47 | 1.67 | 4.6 | 5.5 |
Average miss = mean of the absolute errors for the Conservatives, Labour, Lib Dems, Reform family and Greens, in England. Lead errors in the average row are absolute.
Everything we tried
| How much of the past miss we take off | Average miss per party | Average Lab–Con lead error | No worse than pollsters | |
|---|---|---|---|---|
| 0% | 1.99 | 6.0 | 1 of 4 | |
| 25% | 1.72 | 5.2 | 2 of 4 | |
| 50% | 1.47 | 4.6 | 4 of 4 | Used |
| 75% | 1.34 | 4.1 | 4 of 4 | |
| 100% | 1.46 | 4.2 | 3 of 4 |
0% is the raw answers with no correction, which did less well than the pollsters. We also tested giving don't knows back to the party they voted for last time (25% to 100% of them). It cut the Labour–Conservative lead error a little but raised the party-by-party error and no longer beat the pollsters every time, so it is not used.
How the method did when tested
Each election predicted using only the other elections' evidence, so the test never sees the result it is predicting. Average miss per party in England, in points (lower is better).
| Election | Pollsters | Pollsters minus past miss | Our method | Lab lead miss theirs · ours |
|---|---|---|---|---|
| 2015 | 2.13 | 1.75 | 1.79 | +7.1 · +5.5 |
| 2017 | 1.58 | 2.54 | 1.29 | −4.5 · −3.3 |
| 2019 | 0.79 | 0.62 | 0.64 | +2.7 · +1.9 |
| 2024 | 2.19 | 1.82 | 2.18 | +7.8 · +7.8 |
| Average | 1.67 | 1.68 | 1.47 | 5.5 · 4.6 |
Taking the average past miss off the published figures ("pollsters minus past miss") is the obvious fix. It is not reliable: it made 2017 worse (2.54 against 1.58), so on average it gains nothing. Starting from the raw answers is what made the difference. The full method, step by step →
What the next elections will tell us
- Four elections is not many. The Conservative pattern held every time, but the size varied from about 2 to 7 points. Our ranges on the forecast still use the full spread of past misses.
- Not every pollster publishes its raw answers. Today's reading uses 5 pollsters; more will be added as we build readers for other pollsters' tables (Deltapoll, BMG, Freshwater and others have none yet). Where a pollster has no raw tables, it is not in our projection at all.
- "Raw" is not identical between pollsters. Some already set aside people certain not to vote, and almost all weight to how people say they voted last time. Each poll's page shows exactly which table we used.
- England only. The correction is measured in England, so Scotland and Wales are left as polled. Their own records show no steady lean to correct: see Scotland and Wales.
- The pattern could break. The Conservatives were in government at all four elections. If who answers polls changes, the correction will be wrong. That is why it is shown next to the plain polls, not instead of them.
Figures come from each pollster's own published tables. Every poll page links to the file and table used.
Figures as of . Published by Polling Forecast.








