The pollsters predicted a Labour landslide – but still got the election wildly wrong

While all of the major polling companies picked the winning party, outlandish predictions of a 20-point lead failed to materialise, writes Andrew Grice – and may even have influenced voters

Some Labour and Tory politicians were so alarmed by this year’s poll-driven election, they want to ban the publication of polls in the run-up to polling day (AP)

Polling companies are keeping their heads down after getting last month’s general election wrong. Political polls are their shop window and, unsurprisingly, they don’t want to advertise their mistakes to their more lucrative commercial clients.

When you ask pollsters why they messed up in July, they say they forecast a Labour landslide and that’s what happened. But to suggest they got it right is disingenuous.

Does it matter? Yes. The pollsters forecast a 20-point Labour lead, and the result was a 10-point winning margin. It was their worst collective performance since their annus horribilis of 1992, when the Conservatives’ “surprise victory” would not have been a surprise if the polls had been accurate.

The reaction this time is much more muted because the race was less close than 1992, and so pollsters did not name the wrong winner. They have got away with that one. That shouldn’t spare their blushes, and they should launch an inquiry into what went wrong, as they did after another bad election in 2015.

There’s no sign of one this time. British Polling Council members will compare notes next month, but issued a remarkably complacent statement saying: “The polls told the story of the election campaign, including the prospect of a Labour landslide and the rise of Reform UK as a serious electoral force.”

This year’s mistakes have attracted little media attention. Perhaps newspapers don’t want to remind readers of their coverage. Yet Labour’s supposed 20-point lead shaped the entire campaign. It was self-fulfilling: Keir Starmer was regarded by the media as the PM-in-waiting, and – surprise, surprise – polls showed he was seen as the best PM, while Rishi Sunak was cast as a loser.

If the polls had shown Labour 10 points ahead, might the Tories have avoided some of the internal feuding over their poor campaign? Might fewer Tory supporters have drifted off to Reform, allowing Sunak to close the gap with Labour? Might Labour have then wobbled, made mistakes and created precious momentum for the Tories? At the next election, the media should assume the lead of the party ahead in the polls is half what the pollsters say it is.

A few pollsters are owning up. Martin Boon, a veteran pollster who founded Deltapoll, told a Westminster briefing this week: “I don’t think my industry deserves another free pass. We have to get it right. Otherwise, there might not be a polling industry for much longer.”

There were some special factors this time – a low turnout, a record level of anti-Tory tactical voting, a last-minute switch away from Labour, a revolt against Labour’s stance on Gaza (pollsters admit to a poor performance on ethnic and religious minorities).

But their fundamental mistake is one that has occurred time and again since 1970 – polling firms exaggerate the Labour vote and underestimate Tory support, partly because their samples include too many left of centre voters. In seven of the last nine UK elections, the final polls of the campaign have overestimated Labour’s position to the Tories’ detriment. Boon says this is “bias”, not “error”.

True, it’s a worldwide problem for polling firms. They have every interest in fixing it and have tried tweaking their methodology, but July’s result shows they have failed.

Pollsters have a point when they say the problem is not their surveys but their prominence in the media. As someone who has commissioned and reported on political polls since 1988, I was once addicted to their easy headlines but now think they shouldn’t dominate election coverage. There are many more polls and polling companies than ever. But elections should be more than just a horse race; policies deserve more media scrutiny and the latest poll less.

Yes, the Tories were always likely to lose after 14 years in power and the worst campaign I have ever witnessed. There was a “swings and roundabout” polling effect which harmed both main parties. The assumption Sunak was heading for a catastrophic defeat weakened his campaign, but Labour strategists believe it persuaded some voters they didn’t need to vote for Labour, who suffered a last-minute drift to the Greens and independents.

Another possible factor is “shy Tories” reluctant to admit their support for the party. This has been charted over the years by Robert Hayward, the Tory peer and polling expert. He believes Starmer would still have become PM if the polls had been accurate but that “we would have had a very different election campaign”. True.

Some Labour and Tory politicians are so alarmed by this year’s poll-driven narrative they want to ban the publication of polls in the run-up to election day, as several European countries and Canada do. Despite the noise, I doubt a ban will happen. Private polls would be commissioned and leaked to the media or published abroad. A more realistic solution is for both the pollsters and the media to learn from their respective mistakes.