How do exit polls work and how accurate are they?

David Dimbleby calls exit polls 'the worst invention ever' - but you might not guess why.

The release of the exit poll is one of the defining moments of every general election.

Broadcast when polling stations close at 10pm, it offers a considerably more reliable prediction of the final result of the vote than the surveys released in the weeks before.

Peculiarities about the UK electoral system make it extremely difficult to predict the outcome.

But the British exit poll is widely regarded as one of the best of its kind in the world, and helped turn its co-creator, Professor John Curtice, into a leading expert on voting behaviour.

Here’s how it’s put together, and how accurate its results have been in the past.

How do exit polls work?

The model used today was created in 2005 by Prof Curtice and statistician David Firth.

All exit polls work in fundamentally the same way: voters are asked which candidate they went for as they leave their polling station.

But with limited time and resources, organisers can’t ask every single person in every single station – far from it.

Pollsters from Ipsos talk to around 20,000 people across 130 polling stations – a drop in the ocean compared to the millions who turn out to the 40,000 stations across the UK.

Voters are selected at random throughout the day and asked to ‘vote’ again on a mock ballot paper which is placed into the exit poll. (Around 80% agree, in case you were curious.)

At regular intervals throughout the day, interviewers talk send their results so far back to Ipsos’ HQ.

But the work doesn’t stop there. Layers of statistical wizardry are then used to turn the small sample of results into something that accurately represents the entire country.

This is crucial because voting behaviour varies between different types of polling stations and parts of the country.

Voters can also refuse to vote, vote by post, or vote in person but refuse the survey, potentially skewing the result.

Experts say one of the main ways this is accounted for is the careful study of how voting behaviour changes from election to election.

For this reason, the polls are pretty much always carried out in the same places.

How accurate are they?

Experts say it isn’t technically possible to describe the accuracy of the exit poll, which is usually measured as a ‘margin of error’, because it isn’t based entirely on random sampling.

But the projection is usually extremely close – usually off by only a few seats.

For example, in 2010, it correctly predicted the exact number of seats won by the Tories.

Indeed, since 2005 it has called the majority exactly right three times.

In two cases, it was less than ten seats off from the final result.

The exit poll’s largest miss was in 2015, when the projected majority fell short of the actual number of seats by 30.

In fact, it’s so accurate that David Dimbleby – arguably the face of election night – has called it the ‘bane of the broadcaster’s life’.

He told The Guardian earlier this month the exit poll is ‘the worst invention ever’ because it takes the fun out of his job.

‘It’s like a thriller and you’re given the answer before we’ve even started on page one,’ he added.

Get in touch with our news team by emailing us at webnews@metro.co.uk.

For more stories like this, check our news page.