The Chamberlain Files | Homepage
Never mind the results – who won the battle of the polls?

Never mind the results – who won the battle of the polls?

🕔12.Jun 2017

The results are in and the remifications are still playing out. But Chris Game reflects one other set of winners and losers at General Election 2017 – the pollsters and pundits. 

It’s all part of the Election Night tradition. The BBC’s overnight results coverage, chaired by a venerable Dimbleby, surrounded by predominantly (with apologies to Kuenssberg and Maitlis) middle-aged male politicians and pundits, all pontificating about the exciting impact of the youth vote – with one or more elderly politicos disbelieving the exit poll’s seat predictions.

After the 2015 farce I really thought last bit of the tradition would have to be replaced. You’ll probably recall former Lib Dem leader Paddy Ashdown insisting repeatedly, that “if these exit polls [predicting his party being reduced to 10 MPs] are right, I’ll publicly eat my hat on your programme” – until the 10 turned out in fact to be 8.

Followed by the equally ridiculous Alastair Campbell, one-time Press Secretary to Tony Blair, promising to eat his kilt. And their both eventually having to do so – at least in the form of chocolate facsimiles.







Previously, in 2010, it had been Iain Dale, former Conservative candidate turned political commentator, and mentioned in my previous polling blog. He questioned the prediction that the Lib Dems would get only (that’s how long ago it was) 59 seats, and pledged to “run naked down Whitehall if that turns out to be true.” In one sense it didn’t – they managed only 57 – which I suppose was his typical politician’s excuse for breaking his promise.

It’s often said, though, that politicians hardly watch television unless they’re on it, so all this must have gone unnoticed by studio guest and still Defence Secretary, Sir Michael Fallon. For he, on learning this year’s exit poll prediction of a hung parliament (C 314 seats, L 266, LD 14, SNP 34), kept reminding us on Thursday night that “This is a projection, not a result. These exit polls have been wrong in the past”.

Ideally, one can anticipate two pleasing responses to whoever’s performing the swivel-eyed Ashdown/Fallon role. First, there’s increasingly likely nowadays to be a fellow politician around with a bit more sense, who understands that exit poll projections can’t be highhandedly dismissed as just another opinion poll, and can put them right.

This year it was Shadow Foreign Secretary, Emily Thornberry: “They’ve been right for the last 20 years or so, 30 years, so I think we’re on the verge of a great result.” And, if she hadn’t got carried away and risked the 30 years bit, then, as shown in the bar chart, she’d have been pretty well right.








If there’s no clued-up politician on hand, there’s ALWAYS the University of Strathclyde’s Professor John Curtice. As well as being the BBC’s polling supremo and thereby the UK’s most widely recognised political scientist, he’s President of the British Polling Council, and for the past several elections head of the team of academic analysts responsible for turning the pollsters’ exit poll data into seat predictions and a projected overall result.

Despite his public fame, Curtice isn’t (really) a show-off, and last Thursday night was clearly content to have Thornberry defend his poll’s record, so that he could smile down magisterially from his BBC studio balcony and just look, if I’m honest, smug – based on the knowledge that since 2001, the predictions he’s overseen have never missed a party’s actual seat count by more than 20, and of course they didn’t this year either (C – underestimated by 3; L +4, LD +2, SNP -1).

The quality of that record and the justification of the Curtice confidence is due to the methodology, which has evolved almost unrecognisably since the early, and very fallible, exit polls that were first used to open election night results programmes back in the 1970s. When, just to add to the confusion, the BBC and ITN would conduct separate polls, with sometimes, like October 1974, wildly varying seat predictions.

They certainly had their good years, but there was no way of knowing until after the results whether it was a good or bad year. That still seems to be Sir Michael Fallon’s approach – treating the now single exit poll, commissioned jointly by the BBC, ITV and Sky News, as if it were essentially the same, and as fallible, as an ordinary, even large-sample, election campaign poll.

It just isn’t. Most obviously, an exit poll isn’t starting from scratch. It’s not drawing a new sample each time to try to take a one-off snapshot of national opinion, then getting, as I described in my earlier blog, essentially self-selected respondents to divulge their future voting intentions and also the likelihood of their actually voting.

An exit poll is working from the known result, constituency by constituency, of the previous election, and trying to estimate the extent of opinion change since then, by interviewing a sample of perhaps 1 in every 10 or 15 voters who’ve actually voted at about 100 polling stations, often the very same ones that were used in that last election.

That’s not remotely to say it’s easy, but the exit poll’s tried and tested methodology does contain several factors that greatly increase the statistical confidence one can have in its findings, compared to those of a standard opinion poll.

The more so in this election, when those standard polls, having all in their final polls significantly understated the Conservatives’ 7% lead over Labour, have been making different adjustments to their sampling and weighting methodologies.

I suggested that, as a consequence, we could reasonably expect at least one or two of the final polls this time to get pretty close to the actual result, even if as a set they were all over the place. It wasn’t a terribly daring prediction, but it was right.






To be fair, most of the nine major polling companies got considerably closer this time to the Conservatives’ final 42.4%, and seven were within 2%. Only three, though, were within 2% of Labour’s 40%, and on average they were 4% adrift.

But, as might have been expected, one was virtually spot-on: Survation, a British market research company who carry out political polls for, among others, Sky News – so congrats to them.

As noted above, pollsters themselves are generally more than happy nowadays to leave seat projections to the academics, and from their spread of percentage lead projections you can understand why.

They’re a Richard Hammond-scale car crash just waiting to happen, as Curtice and his academic colleagues discovered themselves in 2015, when their only fractionally wrong seat projections pointed them to a hung parliament, rather than a narrow, but decisive, Conservative majority.

Which is what makes the YouGov ELECTION CENTRE projection at the foot of the table so remarkable. I described in my previous blog how, in the last week or so of the campaign, YouGov were carrying out two separate and methodologically completely different polling exercises.

In addition to their regular polls for The Times, like the one in the main table above, they also produced – and, more remarkably, published on a different part of their website – an aggregated model, based on some 7,000 interviews each day, showing updated daily voting intentions and estimated results for every one of the 650 UK parliamentary constituencies.

I described it merely as rash, and many within the polling industry thought they were barmy, risking such a massive hostage to fortune. So too did The Times, who, while publishing YouGov’ s regular polls, kept pretty quiet about the daily Election Centre stats.

As someone who had nothing whatever to lose, apart from the time involved, I reproduced the first set of estimates for a wide selection of West Midlands constituencies – made, it should be emphasised, before the final week of the campaign and the harder evidence of Labour’s generally improving poll ratings was avilable.










I’ve done so again, this time with details showing how impressively accurate those estimates proved to be, even down to Ian Austin’s 22-vote scramble in Dudley North. The final column of actual percentage majorities for what are mainly Labour seats highlights, incidentally, just how remarkable Labour’s performance was in Birmingham especially – the party’s 62.5% vote share nearly 7% higher than in 1997, when Tony Blair’s New Labour achieved an overall Commons majority of 177.

I concluded that last blog by saying that, however accurate or inaccurate the model’s detailed estimates proved to be, “I really hope this YouGov initiative doesn’t end up getting completely trashed. It deserves better than that.” Of the many things to worry about after this election, I really don’t think this will be one.

Similar Articles

Mayor 2020: on the right track?

Mayor 2020: on the right track? 0

The race to win the West Midlands mayoralty may be the election contest to watch

Just when you thought the election was over…

Just when you thought the election was over…

As Members of Parliament began to be sworn in yesterday, the long campaign for Mayor

The Day of Truth

The Day of Truth

So, here it is. The polls will open in a few minutes. Contain your excitement

PM: gave unlawful advice; frustrated Parliament

PM: gave unlawful advice; frustrated Parliament

"Scenes." As young people would say, writes Kevin Johnson. "Unlawful." "Unequivocal." "Historical." These words are not,

WMCA: Nothing to see here…move along

WMCA: Nothing to see here…move along

As the Prime Minister prepared to address leaders ‘up North’ gathering for the Convention of

About Author

Chamberlain Files Weekly

Don't miss a thing! Sign up for our free weekly summary of the Chamberlain Files from RJF Public Affairs.
* = required field

powered by MailChimp!

Our latest tweets

Published by

Published by


Our community