When the exit poll dropped at 10pm the numbers seemed unbelievable.
A few hours later, it looks as if the exit poll may have even underestimated the number of seats the Tories will win and overestimated Labour’s share.
When all ballots will have been counted the Conservatives will find themselves on the brink of an outright majority - an outcome based on pre-election polling was deemed near impossible.
Cameron on track to remain prime minister after electoral triumph.
Although we still haven’t received projections of a vote share, it is already clear that the polls and their pollsters have had a bad election.
In the end the debate between online and phone polls, and different methodologies, proved irrelevant.
Although during the course of the campaign the latter had shown several Tory leads, the final crop of polls were roughly anticipating a tie.
The challenge with trying to understand what went wrong isn’t simple. The same methods (often used by the same companies), in different countries, are most of the time accurate. Indeed they were just five years ago, at the last general election in 2010.
But at times, like on Thursday, or at the recent Israeli general election, polls get it wrong. At this stage it’s impossible to know why. It could be simply that people lied to the pollsters, that they were shy or that they genuinely had a change of heart on polling day.
Or there could be more complicated underlying challenges within the polling industry, due for example to the fact that a diminishing number of people use landlines or that internet polls are ultimately based on a self-selected sample.
In Scotland the polls seemed to have had a better night (or maybe Scots were less shy with pollsters). Expectations that the SNP would win nearly all of Scotland's 59 seats have materialised. The paradox is that most pollsters thought the pre-election figures north of the border were less likely than the expected results in England.
In the end, the outcome in Scotland was roughly in line with the polls. The result in England wasn’t - same companies, same methods, but very different results.
The Guardian
