The polls got it wrong again, overestimating Keir Starmer

Few things in British politics seem familiar this morning. So it’s a welcome feature of this election that the pollsters have seemingly, once again, got it wrong. They overestimated the Lib Dems in 2010. They failed to spot a Tory majority in 2015. They largely didn’t foresee Brexit or Theresa May losing her majority. Now add to that list the apparent overestimation of Labour’s lead.

Or, more technically, vastly overestimating Keir Starmer’s lead over the Tories. At the time of writing, it looks like Labour have only got 35 per cent of the vote, to deliver 410 or so seats. Hardly shabby, thanks to the vagaries of first-past-the-post. But that’s only an 11-point lead. The BBC’s poll tracker gave Labour an 18-point lead yesterday morning. By that, Starmer has under-performed expectations.

More importantly, Starmer has also undercut the various MRP predictions made for him. Survation – usually the most reliable – gave Labour 470 in their eve of poll survey, with the Tories reduced to only 68. Last night was bad. But it wasn’t that bad. Survation aren’t alone. YouGov’s final MRP gave Labour 39 per cent of the vote.

Some pollsters were more successful than others in guessing parts of the political picture: the Lib Dem surge, the Tory shellacking, the exact manifestation of Reform’s vote. But even with the uncertainty, Starmer’s win today seems much less impressive than the pollsters had suggested. Blair 1997 in seats on Blair 2005 in votes is beatable in 2029. Predicting him an uber-landslide didn’t pass the sniff test.

Why did they get it wrong? Are we a nation of shy Tories? Did too many Labour supporters opt to stay at home? Is there a systematic failure to talk to the right people? Or is the real truth that a lot of this polling is a load of hogwash?

They hedge and tweak and fiddle, like a surgeon nipping and tucking some ageing starlet out of recognition. But if the original data is faulty, they miss it. Polling day really is the only poll that matters, and the only way we have of checking the marking of the pollsters against results. They keep getting egg on their faces. But still we get duped.