Election polling post-mortem

A brief election post-mortem before I get some rest - hopefully we will have an actual London result by the time I finish writing! It is almost exactly a year since the polling error at the last general election. Yesterday's elections were the first real test of the polls since then (there was accurate polling for the Labour leadership election, but polling party members really is a completely different exercise).

Scotland

Taking Scotland first, all the polls obviously had the SNP winning, but that was hardly a challenge. Perhaps the bigger challenge was second place. In the event Labour narrowly held onto second place in the constituency vote but were pushed into third in the regional vote - the polls conducted in the last few days of the campaign did get this right, but all the Scottish polls did underestimate the level of Conservative support, and apart from YouGov's final poll there was an overestimate of SNP support in the regional vote (though many of the polls finished some time before the election - the TNS face-to-face poll in particular - so it may be that SNP regional support dropped in the final week.)

Wales

YouGov was the only company to poll in Wales, and thei final poll held up very well, with Labour, Conservative, Lib Dem and Plaid all well within the margin of error. The only fault was an overstatement of UKIP support.

London

As I write, the mayoral results STILL haven't been announced, and given how late they were in 2012 I'm not waiting up to write about them. Based on the live count of the first 90% of ballots the polls seem to be roughly in line with the expected result, and projections of the second round score suggest the polls are going to be close to it. You'll apparently find out around midnight so you can compare to the polls below... but I intend to be asleep.

All in all, the performance of the polls was far more credible than last year, though

it looks like there may still have been some issues with the Tories in Scotland (and to be fair, most of the polling companies have been very explicit in saying they are still addressing their issues and developing their methods - the problems of last year are not going to be addressed overnight).

On a personal note - I'm most relieved the broad narrative was right. After the general election there were plenty of people saying how they knew the Tories would win, their instincts told them they would, how could those silly pollsters not spot it? Well, many of us silly pollsters thought the Tories would end ahead of Labour too: questions on leadership and the economy favoured them, we expected the polls to move towards the Tories... but the data just kept on showing the parties neck-and-neck, and ultimately a pollster's job is to measure the answers the public give us, not report what we think they should say. We trusted the data, but it turned out to be wrong.

This time round it was the other way round. I never quite believed that the Conservatives could come second in Scotland. Yes, Scottish Labour was a mess, but Scotland would surely never vote for the hated Tories. My instincts said it wouldn't happen in the end. A few months ago when YouGov were the only company showing Labour and the Tories neck and neck in Scotland I worried whether we'd get egg on our faces, but the data said it was happening, and I had to have confidence in the methodology corrections we'd made and in what the data was telling me... and this time, the data was telling the right story and the Tories really did come in second. Phew!