One of the most iconic components of election night in the United Kingdom is the 10pm exit poll. There will have been a heavy outpouring of regular polls, predictions and projections throughout the campaign, but the beauty of the exit poll is that instead of asking people what they intend to do it asks what they already did – all uncertainty thus being removed. The mood of political parties regarding their relative fortunes, the emotional state of all watching and the entire political narrative can be one way throughout the whole campaign and then change dramatically into something entirely different as soon as the bongs sound. The most prominent examples are 2015 – where it was widely predicted that the Conservatives would sink to around 290 seats and be level pegging with Labour, only to find that they’d actually gone up – and 2017 – where Theresa May had long expected to win by a landslide, but actually lost her majority.
Only a very small section of the people are surveyed – in the 2010 example Dimbleby said one hundred and thirty polling places, or one for every five parliamentary constituencies. The statisticians in charge of the polling companies are razor sharp in finding exactly the right places from which to check the political temperature. This makes it all the more remarkable that the polls’ predictions are so close to reality. Despite the protestations of the talking heads that “it’s too early to call” and Dimbleby’s own quip that “if it was dead accurate there’d be no need for anybody to go and vote”, generally the numbers shown are not far from the real ones.
To make the lead graphic, I skimmed through the coverage of the six UK general elections that have taken place in my lifetime, and compiled a spreadsheet of the seat totals projected for the two main parties as well as the actual numbers of seats won by those parties. The graph shows how much either party was under- or overestimated each time. The 2010 poll was nearly spot-on, with the Conservative figure exactly right and the Labour only three too low. The worst inaccuracy was the Conservative figure for 2015, and indeed this was the only occasion in this millennium where the overall result was mispredicted (it was a small Conservative majority rather than a hung parliament).
Exit polling explained – University of Warwick