353 seats? The mega-poll's Labour majority comes from a split right, not tactical voting

The mega-poll in the Daily Mail on 16 October gave Labour 353 seats and a majority of 56. Our projection, built from the average of every poll, has Labour short of a majority on 316. Those are very different stories about the same electorate, so we took the mega-poll apart to see where the 37 seats come from.

The answer is not where the coverage pointed. The release made much of tactical voting: a third of Labour voters, it said, would back the Conservatives to keep Reform UK out. But when we run the poll's own vote shares through our seat model, with no tactical voting at all, Labour gets 358 seats. The majority is in the vote shares, not the seat model.

What the poll found

The poll was an MRP by Electoral Calculus and Find Out Now for the public affairs firm PLMR, with 6,699 respondents interviewed between 25 September and 2 October. MRP turns a large national sample into an estimate for every constituency by modelling how different kinds of voter behave and applying that to the people who live in each seat.

It put Labour on 28 per cent, Reform UK on 21, the Conservatives on 18, the Liberal Democrats on 11, the Greens on 8 and Restore Britain on 8. Its seats: Labour 353, the Conservatives 90, Reform UK 81, the Liberal Democrats 70 and the SNP 28.

Labour's share is not the difference

Labour's 28 per cent is almost exactly where the other polls have it. Our average of every recent poll puts the party on 27.4. If the mega-poll had found a Labour surge, the majority would be easy to explain. It did not.

What it found is a weaker Conservative Party and a much stronger Restore Britain. The Conservatives are on 18 per cent in the mega-poll and 21.5 in our average. Restore Britain is on 8 per cent, against about 3 in the average. Find Out Now's own regular polls since mid-September have put the Conservatives about one and a half points lower than other firms have, so part of the gap is the firm, and part may be the size and timing of this sample.

That split on the right is worth dozens of seats to Labour. In a seat where Labour wins on 30 per cent against a divided right, every point that moves from the Conservatives to Restore Britain makes the Labour lead safer. We moved three and a half points from Restore Britain back to the Conservatives, to bring both near the average, and changed nothing else. Labour fell from 358 seats to 330, and the Conservatives rose from 97 to 145.

Tactical voting points the other way

If the mega-poll's model were assuming heavy tactical voting against Reform, it would give Reform fewer seats than a simple swing model on the same vote. It gives Reform more. On the mega-poll's shares our uniform swing gives Reform UK 46 seats; the MRP gives it 81.

When we add the switching the release describes, a third of other parties' voters moving to whoever is best placed to beat Reform, our model cuts Reform to 10 seats and lifts Labour to 377. Whatever the MRP assumes about tactical voting, it is not what produces Labour's majority. You can try the same settings yourself in the tactical voting section of our prediction page.

Within our range, at the top of it

None of this makes the mega-poll wrong. Our own 80 per cent range for Labour runs from 232 to 373 seats, and 353 is inside it. A majority on 28 per cent of the vote is possible under first past the post. Labour won 411 seats on under 34 per cent of the UK vote in 2024 for the same reason: its opponents split.

What the mega-poll shows is how much rests on the right staying divided. On today's average the Conservatives are level with Reform UK on about 22 per cent and Restore Britain is a minor party. On the mega-poll's figures the two are much closer to sharing the right-wing vote, and that is what turns 28 per cent into 353 seats.

The record of mega-polls

MRPs have a good record on the shape of an election and a weaker one on seat totals. In 2017 YouGov's MRP called a hung parliament when most forecasts had a Conservative majority. In 2024 every final MRP gave Labour too many seats, from 425 to 516 against the 411 it won, because the polls underneath overstated Labour by about four points and the seat count magnified the error. A big sample shrinks random error, not the lean every poll from one firm shares. At about 11 respondents per constituency, this one measures no seat directly.

We keep stand-alone MRP releases out of our polling average, for that reason and because one very large sample would outweigh every other poll for weeks. We do compare every published one with our own projection. Our new page, MRPs and mega-polls, plots all 29 published since the last election against our projection on the same day, and our research paper MRPs and mega-polls sets out the full record and how we treat them.

The question to ask of this one is not whether tactical voting will hand Labour a majority. It is whether the right is really split that evenly. Most other polls say it is not, yet.