Exit poll for the 2017 general election

The clock struck ten, and it became clear that Theresa May’s election gamble had misfired in spectacular fashion. And indeed, that Jeremy Corbyn—who was universally written off as leading his party to ruin—had shored his Labour party up, and maybe even made some advances.

The 10 O’Clock BBC headline is that the Conservatives have lost their majority, and that we are now in a hung parliament, the most dramatic turnaround from what was—when May first called it—an election designed to secure a landslide that all the self-proclaimed political sages, and I can’t pretend to be any exception, assumed was coming her way.

But—a big but—at 314 projected seats the Conservatives are not only well clear of Labour’s projected 266, but also within a dozen seats of the winning line of 326 which is needed for a majority of one. Expectations started very differently from last time, but the exit poll numbers are extremely similar to 2015, when the BBC projected the Tories would be on 316, and they came back instead with 330. Consequently, it becomes imperative to ask at this early stage of the night just how accurate the exit poll will prove.

“It would be a surprise if the final tally of any party were to move more than 20 up or down on the 10 O’Clock forecast”

It is only a sample of voters, and no sample ever matches the population as a whole. But recent history holds out scant hope to anyone staying up all night in hope of an entirely different result. Exit polls involve talking to people who’ve actually voted on their way out of the polling station, not internet panels which some people are much more likely to opt in to than others, so there are not the same biases. And they’ve performed well in recent elections, even as the reputation of conventional polling has crashed.

I can (just about) remember BBC suggesting at the start of its 1987 general election coverage that Margaret Thatcher would soon be back with a shrivelled majority of 20-something; in the event she got 102. But that was 30 years ago, and the expertise has improved a lot since then. Even in the next Tory victory of 1992, when the exit poll wrongly suggested a hung parliament, it gave the Conservatives a definite edge on vote share, unlike most of the surveys of the campaign. Through the Tony Blair years it was on the money, and then especially in 2010 when it correctly picked up that Cleggmania had disappeared as rapidly as it arrived, and furthermore that the Tories would fail to win. Then in 2015, it was the exit poll that woke the nation up to the reality that instead of the expected deadlock election, the Conservatives were streets ahead of Labour. As in 1992, however, it rather understated the shock, predicting the Tories would fall ten seats short of an outright majority, instead of securing the outright win of 12 seats they achieved in the end.

Who knows how far tonight’s seat numbers will shift, but on the strength of recent experience it would be a surprise if the final tally of any party were to move more than 20 up or down on the 10 O’Clock forecast. A move of 20 up for the Tories would take them to 334, and a majority of more or less what they started out with. May gambled on doing better, and will soon be gone even in the event that that happens. If Labour slips 20 down, by contrast, it is still no catastrophe for them. Corbyn will still have held his own, and is secure in post. If the final numbers were to move 20 in the other direction, there would be a political earthquake and the Corbynistas would start to sound credible when they protested that with a bit more loyalty from Labour MPs they could have romped home.

So a lot still hangs on a long night of results. And there is still an outside chance that this exit poll will prove to be more wrong than any recent predecessor. But—barring an unprecedented disaster on the part of the boffins—tomorrow’s headline seem clear. May’s party may have emerged as the largest, but she is—without doubt—the great loser from calling an election she always vowed she would not.

5106525535cbc670011d858.80805783