History in the making
Blimey. The news is full of headlines like Scottish independence: ‘Yes’ vote takes shock lead in poll for first time.
Not what I expected!
I have a post half-written about the issue itself (I come down on the No side, but have very mixed feelings), but I wrote it very much in the expectation that it’d be a fairly clear No on Sept 18th. I need to update my thinking here on what’s likely to happen. Here’s what I’d drafted a week or so ago:What do I think will happen? I don’t have any special insight. I’ve been talking to friends, family and colleagues who have a vote and I haven’t found a profoundly different picture to what I’ve picked up from the media. So I think the vote is most likely to be a No, but not by a huge margin. A Yes is definitely not impossible, though. We’ve no good basis for grounding the polls against what-actually-happens (since the last poll on the topic was some time ago), and we know that it’s entirely possible for there to be a fairly large disconnect without that.
That last bit very much still applies. You can see this in the spread of poll results - the polling firms are all adjusting the raw data in different ways, because they don’t have a clear, validated model to adjust towards. There was the famous ‘Shy Tory’ effect in the 1992 General Election - after the result, it seemed in retrospect that people had been reluctant to tell pollsters they were planning to vote Conservative. There may be a similar one going on here. My guess is that a ‘Shy No Voter’ effect is much more likely than a ‘Shy Yes Voter’, though. Why didn’t I expect this latest shift? I’d seen lots of polls that showed a clear and stable lead for No. If we go back to late 2013, these results are typical, and had been like that since before the poll started: Yes/No/Don’t Know 34/57/10 - Ipsos Mori 33/52/15 - YouGov 38/47/15 - Panelbase Around then I vividly remember talking to a group of colleagues from elsewhere in Europe on the topic. The ones with connections to Catalonia and Flanders were particularly interested. I felt a bit cornered as the only British person present, but I eventually opined that while a Yes was possible, and had significant support, it really didn't look very likely since No was polling in the 50s and Yes in the 30s, with no sign of movement. Very little happened in the polls from the start of the campaign until about the end of this July/early August, when the gap started to close: 35/55/11 - YouGov, 40/54/7 - Ipsos Mori 41/48/11 - Panelbase Then we had the second TV debate on August 25th, which most people thought was a clear win for Alex Salmond, and the polls after showed some movement, and now (early Sep) we have: 44/48/8 - Panelbase 47/45/8 - YouGov Cross-checking the BBC Poll Tracker against the Wikipedia one, and this other independent (?) one, I spotted that the BBC sorts them as No/Yes/Don’t Know, while Wikipedia sorts them as Yes/No/Don’t Know. Why ever would you not do it with Yes first? It makes it harder to understand the results. I can see that in election polls you might reasonably decide to list the parties in order of their showing in the first poll, and for consistency in your poll tracker you keep them in that order even if one party overtakes another. (This is the polite explanation for why UK poll reports routinely list Conservatives first.) But for a referendum Yes/No question, Yes is such an obvious and natural thing to list first. This is a minor point, and I’m not holding this up as serious evidence of No bias. But I have felt that the BBC’s coverage does seem to have something of No bias. I lean towards wanting a No result myself, so it must look like painful and egregious No bias to Yes voters. Oh, and mucking around in the poll data reminds me that it’s technically not the first poll in the campaign to show a lead for Yes. There was a Panelbase poll for the SNP back in late August 2013 that had 44/43/13 on the referendum question. It sticks out massively from other polls at the time that found a 10-point or more lead for No. If you glance at their report you see why. They’ve done a beautiful version of the Yes, Minister trick. The first question is ‘Do you agree or disagree with the following statement: “Scotland could be a successful, independent country”?’ Then they ask ‘Who do you trust to take the best decisions for Scotland: the Scottish Government or the Westminster Government?’. And only then do they ask voting intention in the referendum. They could’ve got an even clearer lead for Yes if they’d also asked “The Tory party draws almost all its political power from outside Scotland. Are you happy for Tory policies, such as the Poll Tax, the Bedroom Tax, tuition fees, and creeping privatisation of the NHS, to be applied in Scotland?”. (This is also why it's significant that it's YouGov that's showing a Yes win. Panelbase you could perhaps dismiss more easily as an outlier, since they've leaned Yes compared to other pollsters all the way through.) Another point from my old notes: I’m interested in how both sides are framing the debate. The old saw says that there are at base only two political slogans: More Of The Same, and Time For A Change. The Yes campaign is very cleverly going for both: it’s clearly a change, but it’s also framing independence as a way of seeing off the threat from Westminster to the things Scots hold dear and want to preserve, like the NHS and social security. So far as I can make out, the No campaign has struggled to make the small-c-conservative argument, which I think is a strategic failure, because although very few Scots are Conservative, many of them are conservative.
The No campaign has been a bit rubbish. As
andrewducker entertainingly puts it, “it seems that the three unionist parties are now falling over themselves to offer proposals for what Scotland would get if it voted No in time for the ballot. No, wait. They're _apparently_ offering to set up a convention which will get a bunch of people together to discuss proposals to possibly hand over some powers in the future.” It’s worth noting, however, that the scheme Andrew is rightly dissing is mentioned only ina report from the Observer, which has no named source. Such reports - in the Observer, no named source, shockingly misconceived plan to be announced shortly - have, in my experience, most often turned out to be materially wrong in their predictions. Charitably, this is because the sources have used the Observer as a way of testing the reaction to their proposal, and when it is been negative, they think better of it. Less charitably, the Observer or their unnamed source is just making it up. Picking up another of Andrew’s points, refusing to include a ‘Devo Max’ option on the poll is starting to look like a spectacular mistake by the unionist parties, and the Westminster Government in particular. I was pretty confident that such an option - once fleshed out a little more - would win handily, and still am. Excluding it - when it was polling so well - was always going to look like the Westminster Government trying to keep power from the Scots. It now very much also looks like an attempt to raise the stakes that may backfire badly. Or it may not, just. I heard from
ciphergoth that bookies’ odds on NoYes had jumped to 31% (from about 20-25% from memory). One poll does not a result make. But it is true to say that the Yes campaign has a lot of momentum, and the No campaign does not have much time. Edit: Betfair had Yes on 33% this morning. And it turns out that rubbishy Observer story was only partially rubbish - George Osborne is now on record saying "we'll totally be announcing plans to give Scotland more powers, real soon now". Which is really quite rubbish. Why only now, with only days to go before the vote? Does he think wavering Scots are stupid? This is, I suspect, really not going to help the No vote. This entry crossposted to http://doug.dreamwidth.org/276603.html, where there are
comment(s) not shown here.
That last bit very much still applies. You can see this in the spread of poll results - the polling firms are all adjusting the raw data in different ways, because they don’t have a clear, validated model to adjust towards. There was the famous ‘Shy Tory’ effect in the 1992 General Election - after the result, it seemed in retrospect that people had been reluctant to tell pollsters they were planning to vote Conservative. There may be a similar one going on here. My guess is that a ‘Shy No Voter’ effect is much more likely than a ‘Shy Yes Voter’, though. Why didn’t I expect this latest shift? I’d seen lots of polls that showed a clear and stable lead for No. If we go back to late 2013, these results are typical, and had been like that since before the poll started: Yes/No/Don’t Know 34/57/10 - Ipsos Mori 33/52/15 - YouGov 38/47/15 - Panelbase Around then I vividly remember talking to a group of colleagues from elsewhere in Europe on the topic. The ones with connections to Catalonia and Flanders were particularly interested. I felt a bit cornered as the only British person present, but I eventually opined that while a Yes was possible, and had significant support, it really didn't look very likely since No was polling in the 50s and Yes in the 30s, with no sign of movement. Very little happened in the polls from the start of the campaign until about the end of this July/early August, when the gap started to close: 35/55/11 - YouGov, 40/54/7 - Ipsos Mori 41/48/11 - Panelbase Then we had the second TV debate on August 25th, which most people thought was a clear win for Alex Salmond, and the polls after showed some movement, and now (early Sep) we have: 44/48/8 - Panelbase 47/45/8 - YouGov Cross-checking the BBC Poll Tracker against the Wikipedia one, and this other independent (?) one, I spotted that the BBC sorts them as No/Yes/Don’t Know, while Wikipedia sorts them as Yes/No/Don’t Know. Why ever would you not do it with Yes first? It makes it harder to understand the results. I can see that in election polls you might reasonably decide to list the parties in order of their showing in the first poll, and for consistency in your poll tracker you keep them in that order even if one party overtakes another. (This is the polite explanation for why UK poll reports routinely list Conservatives first.) But for a referendum Yes/No question, Yes is such an obvious and natural thing to list first. This is a minor point, and I’m not holding this up as serious evidence of No bias. But I have felt that the BBC’s coverage does seem to have something of No bias. I lean towards wanting a No result myself, so it must look like painful and egregious No bias to Yes voters. Oh, and mucking around in the poll data reminds me that it’s technically not the first poll in the campaign to show a lead for Yes. There was a Panelbase poll for the SNP back in late August 2013 that had 44/43/13 on the referendum question. It sticks out massively from other polls at the time that found a 10-point or more lead for No. If you glance at their report you see why. They’ve done a beautiful version of the Yes, Minister trick. The first question is ‘Do you agree or disagree with the following statement: “Scotland could be a successful, independent country”?’ Then they ask ‘Who do you trust to take the best decisions for Scotland: the Scottish Government or the Westminster Government?’. And only then do they ask voting intention in the referendum. They could’ve got an even clearer lead for Yes if they’d also asked “The Tory party draws almost all its political power from outside Scotland. Are you happy for Tory policies, such as the Poll Tax, the Bedroom Tax, tuition fees, and creeping privatisation of the NHS, to be applied in Scotland?”. (This is also why it's significant that it's YouGov that's showing a Yes win. Panelbase you could perhaps dismiss more easily as an outlier, since they've leaned Yes compared to other pollsters all the way through.) Another point from my old notes: I’m interested in how both sides are framing the debate. The old saw says that there are at base only two political slogans: More Of The Same, and Time For A Change. The Yes campaign is very cleverly going for both: it’s clearly a change, but it’s also framing independence as a way of seeing off the threat from Westminster to the things Scots hold dear and want to preserve, like the NHS and social security. So far as I can make out, the No campaign has struggled to make the small-c-conservative argument, which I think is a strategic failure, because although very few Scots are Conservative, many of them are conservative.
The No campaign has been a bit rubbish. As
andrewducker entertainingly puts it, “it seems that the three unionist parties are now falling over themselves to offer proposals for what Scotland would get if it voted No in time for the ballot. No, wait. They're _apparently_ offering to set up a convention which will get a bunch of people together to discuss proposals to possibly hand over some powers in the future.” It’s worth noting, however, that the scheme Andrew is rightly dissing is mentioned only ina report from the Observer, which has no named source. Such reports - in the Observer, no named source, shockingly misconceived plan to be announced shortly - have, in my experience, most often turned out to be materially wrong in their predictions. Charitably, this is because the sources have used the Observer as a way of testing the reaction to their proposal, and when it is been negative, they think better of it. Less charitably, the Observer or their unnamed source is just making it up. Picking up another of Andrew’s points, refusing to include a ‘Devo Max’ option on the poll is starting to look like a spectacular mistake by the unionist parties, and the Westminster Government in particular. I was pretty confident that such an option - once fleshed out a little more - would win handily, and still am. Excluding it - when it was polling so well - was always going to look like the Westminster Government trying to keep power from the Scots. It now very much also looks like an attempt to raise the stakes that may backfire badly. Or it may not, just. I heard from
ciphergoth that bookies’ odds on NoYes had jumped to 31% (from about 20-25% from memory). One poll does not a result make. But it is true to say that the Yes campaign has a lot of momentum, and the No campaign does not have much time. Edit: Betfair had Yes on 33% this morning. And it turns out that rubbishy Observer story was only partially rubbish - George Osborne is now on record saying "we'll totally be announcing plans to give Scotland more powers, real soon now". Which is really quite rubbish. Why only now, with only days to go before the vote? Does he think wavering Scots are stupid? This is, I suspect, really not going to help the No vote. This entry crossposted to http://doug.dreamwidth.org/276603.html, where there are
comment(s) not shown here.