Exit polls make Conservatives largest party in hung parliament

PM unveils Mahatma Gandhi statue in London

The exit polls commissioned by BBC, ITV and Sky News have the Conservatives on 316 seats, by far the largest party, while Labour are predicted to win 239 seats.

These findings are in stark contract to pre-election polls. And based on these figures the current coalition government would have the numbers to continue in office.

The exit poll was based on interviews with 22,000 people in 140 polling stations.

Although the exit poll was spot-on in both 2010 and 2005 elections, a BBC exit poll in 1992 falsely forecast a hung parliament when, in the event, the Conservative party secured a majority.

The challenge for Thursday night’s exit poll is that if the result is as close as polls suggest, even an “error” of 10 seats – which statistically would be minor – could be the difference between winning or losing the election.

Because of this it is important to remember that it is an exit poll – not actual results. Forecasting is already a formidable task, and even more so in an election that is so tight.

The Guardian’s final pre-election projection of the polls had the Conservatives and Labour tied on 273 seats each.

Based on those figures, David Cameron was expected to no longer have the numbers to command the confidence of the House of Commons. This is because the sum of those parties inclined to support a Conservative-led government – the Lib Dems, the DUP and Ukip – was expected to fall at least 10 seats short of the required 323 seats.

Instead, if the results confirm the exit poll this figure would tally up to 336 seats.

Meanwhile, in the final Guardian analysis, those parties that would vote a Conservative government down – Labour plus the SNP, SDLP, Plaid Cymru and Greens – added up to a clear majority. But if the exit poll pans out that bloc would win only 306 seats.

The Guardian’s projection was not alone in drawing this conclusion from the polls: while there were slight differences between the final distribution of the seats – with support for the Tories ranging from 271 to 285, compared with 262 to 273 for Labour – no forecast was predicting that Cameron would have the numbers to continue as prime minister.

Beyond the two main parties, the exit poll has the SNP winning 58 seats, the Lib Dems on 10, Ukip two.

Since last year’s independence referendum, the SNP has commanded a 20 point lead over Labour in poll after poll. So when it came to Nicola Sturgeon’s party, the question was only ever about how many of Scotland’s 59 seats they would eventually gain. The Guardian’s projection put the party on 52 seats, a gain of more than 46 seats since 2010.

By holding on to 27 constituencies, in the Guardian projection, the Lib Dems were expected to cling on to about half of its seats, thanks to a strong ground campaign and constituency polling, which showed a very different picture to the party’s more modest nationwide figures.

But if the results confirm the exit poll’s 10 seats, Nick Clegg’s party will have fallen well short of expectations.

Meanwhile, Ukip had seen its polling gradually dip to 12.5% as the campaign dragged on. Despite proclamations last year that the party was competitive in dozens of constituencies, forecasts never put Farage’s party on top in more than a handful of seats.

The exit poll seems to have confirmed expectations. However, uncertainty around how Ukip supporters would vote in key Tory-Labour marginals was one of the unknown factors identified as being critical to whether Cameron’s and Miliband’s party would win in tens of constituencies where the outcome could be decided by a relatively small number of voters voting tactically.

The Green vote, which appeared to have been squeezed based on the evidence of the final polls, was expected to translate into only one seat.

Pre-election polls had Labour and the Conservatives parties in a deadlock. Most polls pointed to a late swing to Labour as more undecided voters broke for Miliband’s party than for Cameron’s in the final days of the campaign.

The Guardian’s own poll with ICM had Miliband’s party 1 point ahead following weeks of consistent Conservative leads. The same movement was evident in other phone polls – Ipsos Mori put the Tories 1 point ahead from a 5 point lead a week earlier. The same trend was also recorded in the final ComRes poll.

Meanwhile, internet polls had the two parties virtually tied. In the Guardian’s average of polls, the Tories were marginally ahead on 34% compared with Labour on 33.5%.

The main difference between these regular polls and an exit poll is that the former asks “who will you vote for?” while the latter asks “who did you vote for?”. Exit polls are based on thousands of interviews carried out outside polling stations.

The starting point to consider when conducting an exit poll is the choice of places to poll, as only 100 or so people are interviewed in each. The challenge is selecting constituencies and polling stations that together are somewhat representative of the country as a whole. This is difficult because, alongside demographic factors, results vary between elections (a district representative of the result at the last election may not be at the next one).

Once the responses are collected, the second task is to model the data into a projection by taking into account various factors such as demographics, past voting and exit poll patterns, differential swing and turnout between different locations, and the likely outcome all this pooled information from the polled constituencies implies for constituencies elsewhere.

It is important to note the key word here is “likely”: an exit poll provides an estimate result. It assigns a probability to outcomes across the country based on the data collected.

Once assigned, the probabilities of all constituencies are added up to generate a forecast of the total number of seats for each party.

Powered by Guardian.co.ukThis article was written by Alberto Nardelli, for The Guardian on Thursday 7th May 2015 22.36 Europe/London

guardian.co.uk © Guardian News and Media Limited 2010