Shock poll points to Scotland splitting - but is it accurate?

Only those who spent the weekend on Mars will be unaware that, with ten days to go before the independence referendum, a "shock poll" from YouGov has, for the first time, given Scotland's Yes campaign a two-point lead over the No campaign: 51 per cent to 49 per cent.

If the polling is accurate, and the momentum really is behind the pro-independence camp, it looks as though Britain as we know it will break up, with Scotland going its own way.

There's a catch, however: what the headlines haven't told us is that a second poll, from Panelbase, was published over the weekend which said quite the opposite. As Mike Smithson of Political Betting comments: "Clearly one of them can’t be right."

Subscribe to The Week

Escape your echo chamber. Get the facts behind the news, plus analysis from multiple perspectives.

SUBSCRIBE & SAVE

Sign up for The Week's Free Newsletters

From our morning news briefing to a weekly Good News Newsletter, get the best of The Week delivered directly to your inbox.

From our morning news briefing to a weekly Good News Newsletter, get the best of The Week delivered directly to your inbox.

Sign up

Only one thing is certain right now – and that's that there will be a bad-tempered scrap between the polling companies in the aftermath of the 18 September referendum.

What made the YouGov poll stand out – bringing talk of a "constitutional meltdown" from commentators, a fall in the pound, and a promise from George Osborne of a timetable for further devolution if the Scots vote to stay in the Union – was that the same company has for months been showing double-digit leads for the No camp.

And what made the Panelbase poll just as interesting is that they are the only pollsters to have ever come up with a Yes lead in the past - and yet their latest survey shows the No camp ahead: 52 per cent to 48 per cent.

First news of the YouGov poll came from Rupert Murdoch, owner of the Sunday Times, who commissioned it. He tweeted that the result was "a huge black eye for [the] political establishment, especially Cameron and Miliband".

It could hardly have been more awkward for David Cameron: news of the poll came as he was arriving at Balmoral for his annual visit to the Queen's holiday home. The Sunday Times claims the Queen is "horrified" at the idea of the Union splitting up. (Alex Salmond wants to keep the Queen as head of state, but not everyone in the SNP agrees.)

Peter Kellner, president of YouGov, says of the two-point Yes lead: "The fact that the contest is too close to call is itself remarkable, as Better Together [otherwise known as the No campaign] seemed to have victory in the bag. Month after month, they held a steady lead, averaging No 58 per cent, Yes 42 per cent.

"In the past four weeks support for the Union has drained away at an astonishing rate."

Only Conservative voters have resisted Alex Salmond’s advances, says Kellner: 93 per cent of them still plan to vote No. But all other sections of Scottish society are "on the move":

• Among Labour voters, the Yes contingent is up from 18 per cent four weeks ago, to 35 per cent now;
• Among voters aged under 40, there's been a jump from 39 per cent saying Yes to 60 per cent;
• Among women, the Yes vote is up from 33 per cent to 47 per cent.

All of which looks pretty conclusive – until you look at the findings from Panelbase. It finds no movement since its mid-August poll: it still shows a four per cent lead for the No campaign.

Over at Political Betting, Mike Smithson, a veteran poll-watcher, has delved into the archives to show why there's reason to be sceptical about YouGov's findings.

In April 2010, after the first of the TV debates ahead of the last general election, when Nick Clegg did so well against David Cameron and Gordon Brown, YouGov gave the Liberal Democrats a stunning 34 per cent share and a three per cent lead over the Conservatives.

In the event, of course, they did nothing like as well, coming third behind the Tories and Labour with 23 per cent of the vote.

Referendum polling isn't easy, of course: there's no previous vote with which to make a comparison. With general election polling, you ask respondents how they voted last time and how they plan to vote this time. That's not an option with a one-off referendum.

The last referendum held in the UK – readers will be forgiven if they've forgotten all about it – was in May 2011 for the AV voting system. YouGov did not have a good day: there was a gap of more than 15 per cent between their final polling and the actual result.

Then there's the "weighting" question. Pollsters have to "weight" their samples to make them more representative of the electorate: to give the most simplistic example, if the sample shows 60 per cent men and 40 per cent women, and yet the electorate is 50-50 in terms of gender, then you weight it accordingly.

Exactly how YouGov does its weighting will be the subject of considerable scrutiny should the poll prove to be way off when we learn the referendum result on 18 September.

After all, the unweighted figures showed that of the 1,084 people sampled between 2 and 5 September, 475 people said they would vote Yes to independence, while 538 said they would vote No. With weighting, YouGov has increased the Yes vote to 514 and reduced the No vote to 489.

There are further opinion polls from Scotland due this week – from TNS and Ispos/MORI – which hopefully will give us an idea of whether the YouGov poll was an "outlier".

In the meantime, of course, YouGov has got the headlines, not Panelbase. And the question is - whether YouGov has got it right or wrong - will its findings have helped add to the momentum behind the Yes campaign?