Why did the pollsters overestimate Labour’s vote?

While Labour won a landslide victory in last Thursday’s UK general election, the party won far fewer votes than predicted by pollsters.

Sir Keir Starmer’s party won 34.7 per cent of the votes in Great Britain, 4 points lower than the pre-election polling average of 38.9 per cent, according to the final polls released by 18 different pollsters.

One theory used to explain the miss is that there was a late swing away from Labour just before polling day.

Twenty-two per cent of voters made up their minds on the day, according to polling conducted after the election by Lord Michael Ashcroft. Voters who opted for the Greens were more likely than others to have decided late, with 31 per cent of them making up their minds on polling day.

You are seeing a snapshot of an interactive graphic. This is most likely due to being offline or JavaScript being disabled in your browser.

“Vote intention polls are potentially where the country was the day before the election, but there was a late-stage swing to the Greens and independents from Labour,” said Callum Hunter, data scientist at JL Partners. “There’s only so much we can do to counteract that, because we’re only taking a snapshot,” he added.

“Our modelled vote share was within 1 point for the Conservatives, but it was out for Labour and Green,” said Hunter. “If you take into account that on-the-day-surge, then we would have gotten it almost bang on.”

In general, the polls were more accurate for other parties, with the result for the Conservatives, Reform, Greens and Liberal Democrats all within 3 percentage points of the polling average.

You are seeing a snapshot of an interactive graphic. This is most likely due to being offline or JavaScript being disabled in your browser.

Another possibility is that the methods used to model which voters were actually going to vote on the day overestimated Labour’s support. Turnout fell to 59.9 per cent nationally, down from 67.3 per cent at the last election, but turnout fell more than the average in safe Labour areas.

In seats that Labour won at the last election, turnout fell 9.6 percentage points, compared with 6.5 percentage points in other seats.

If turnout fell evenly everywhere, calculations by the Financial Times indicate that Labour’s vote share would have been slightly higher, at 35.1 per cent.

“[Turnout] is an area where pollsters are really variable. Some of them do it on the stated likelihood, others use an empirical model relating measurement to action, while others preset what the voter population is going to look like based on historical data”, said Joel Williams of Verian.

You are seeing a snapshot of an interactive graphic. This is most likely due to being offline or JavaScript being disabled in your browser.

Another explanation is that the samples used were not representative of the wider electorate. Most pollsters use online volunteer panels, who are paid small sums of money to answer surveys, and then the sample is reweighted to reflect the national demographics.

Verian, the pollster that came closest to the final result, uses a different method called “random-probability sampling”, where they send out requests to a random list of addresses to build a panel of respondents, rather than letting people volunteer.

Verian’s final poll put Labour on 36 per cent, 1.3 points above their final result.

“It’s about finding people who aren’t particularly engaged in politics,” added Williams. “We did our absolute best to make sure our sample was good.”