Sunday Polls

Three polls this Sunday, from ICM, NOP and MORI.

An ICM poll for the Sunday Telegraph finds 52% of Scots in favour of independence. This is a similar figure to that found in the first of ICM’s new regular polls for the Scotsman, but contrasts with the figure YouGov found in their poll for the Telegraph in the week, when only 32% of people supported independence with 50% supporting the status quo. I suspect the reason is due to the wording of the question – YouGov asked a detailed question asking respondents to chose between retaining the present Scottish Parliament, and making Scotland an independent state outside the UK but within the European Union. ICM ask a far more straightforward question on whether Scotland should become an independent country.

ICM also asked whether respondents in England would like to see an independent Scotland – 59% would, suggesting that people in England are actually keener for Scotland to leave the UK than Scots themselves are. 48% of English respondents also said they wanted independence from the UK for England. 68% of people in England said they would like an English Parliament with similar powers to the Scottish Parliament, and 62% said Scots MPs should not “be able to vote on English laws when English MPs cannot vote on Scottish laws”.

Asked about the number of Scottish MPs in the cabinet however there was relatively little concern. 76% of English respondents said it didn’t matter, only 21% thought it did. ICM also asked about sporting loyalties. When Scotland were playing a foreign team, 70% of English respondents said they would support Scotland, with 14% supporting their opponents. When the same question was asked about supporting the England team in Scotland, 48% of Scottish respondents said they would support England with only 34% saying they would support England’s opponents. This echoes a poll before the last World Cup which found that Scottish people would support the England team…which no one seemed to believe!

There are voting intention figures in the ICM poll – topline figures are CON 37%, LAB 31%, LDEM 23%, but the article in the Sunday Telegraph suggests they are figures for just England, which is unusual. As it happens they are very similar to the GB voting intention figures from ICM’s poll for the Guardian published on Wednesday, with Labour down slightly and the Lib Dems up slightly in comparison.

Secondly, an NOP poll for ITV’s Sunday Edition asked a series of questions asking people to compare Gordon Brown with David Cameron. Cameron had substantial leads on having fresh ideas (39% to 10%) and on being in touch with modern Britain (33% to 17%). Brown had a very slight lead on protecting teh UK from terrorists (22% to 21%). As ever, the picture these sort of comparison polls creates really depends on the things the people commissioning the poll chose to ask about – we know where the two mens’ strengths lie in the terms of their public perceptions – Brown is seen as competent, reliable, strong, experienced and so on. Cameron is seen as likeable, fresh, modern and in touch. If you ask about the former sort of attributes Brown comes out looking stronger, if you ask about the latter Cameron does. What was interesting was the “Best Prime Minister” question, where Cameron recorded his largest lead over Brown so far (though fo course, this is the first time the question has been asked by NOP) – 29% for Cameron compared to 19% for Brown and 5% for Menzies Campbell.

According to Tim Montgomerie on Conservative Home, the poll also shows that “24% told GfK NOP that they would prefer Charles Kennedy to return as leader and slightly more (7%) preferred LibDem Home Affairs spokesman Nick Clegg to the embattled Sir Menzies (6%)”, though I’m not sure what sort of question or structure this was in response to.

Finally, MORI’s monthly poll is in the Observer. The topline voting intention figures with changes from MORI’s last poll are CON 35% (nc), LAB 33% (-4), LDEM 20% (+2). While figures are not given in the Observer, the poll also apparantly shows a sharp fall in David Cameron’s approval rating.

The SNP have today released the voting intention figures from a new YouGov poll that shows the SNP ahead in both votes. The topline figures for the constituency vote, with changes from the YouGov poll the SNP commissioned back in April are CON 14% (nc), LAB 29%(-1), LDEM 16%(-4), SNP 36%(+10), Others 6%.

On the region vote the topline figures, with changes from the April poll (though if I recall correctly the SNP didn’t release the figures at the time), are CON 15%(+2), LAB 26%(+2), LDEM 15%(-4), SNP 30%(+7), GRN 8%(-3), SSP 3%(-4), Solidarity 1%(+1).

The SNP project that these figures would translate into seats as SNP 49 (+23), LAB 36 (-14), Con 16 (-2), Lib Dems 17 (nc), GRN 6 (-1), SSP 1 (-5). I’m not quite sure how they have done this, using uniform swings at a regional and constituency level I get far fewer SNP seats – only 39 to Labour’s 38 – but Scottish politics is not my forte, so I may have made an error in the distribition of the top-up seats somewhere, or the SNP may have used regional swings or a turnout adjustment. If the SNP projections are correct then Labour and the Lib Dems would not be able to command a majority in the Scottish Parliament, even alongside the Greens.

However, as some people have noted, there is also a YouGov poll in the Scottish edition of the Daily Telegraph which has contrasting figures. The Telegraph figures are CON 15%, LAB 32%, LDEM 15%, SNP 32%, GRN 4%, SSP 1% in the constituency section and CON 17%, LAB 29%, LDEM 15%, SNP 28%, GRN 8%, SSP 2% in the regional vote. On these figures Labour would remain the largest party.

So why the difference? Strictly speaking the differences could just be sample error, but I think the differences are more likely to be a result of very subtle differences in the way the polls were conducted.

As I’ve said before, all reputable pollsters will put voting intention questions at the start of a poll to make sure other questions don’t skew the answers. Neither of these polls are an exception – there were no leading questions or suchlike in the SNP poll, it was perfectly legitimate. However, what happens when you’ve got more than one voting intention question? Unlike the SNP poll the Telegraph poll included a Westminster voting intention question, and that was asked before the Scottish Parliament question. Naturally Labour had a healthy lead in the Westminister voting intentions (LAB 36% to the SNP’s 25%, CON 18%, LDEM 16%) and I suspect people were more likely to say they were going to vote Labour in the Scottish Parliament if they’d just said they would in the Westminster Parliament.

The second difference was in the options given for minor parties. The Telegraph poll used YouGov’s normal formula, giving people the choice of the main 4 parties and “other”, with people who chose other then being given a further list of minor parties. The SNP poll was slightly different – the constituency poll gave the 4 main parties and other (though with no follow up question), but the regional question listed the Greens, SSP and Solidarity along with the main 4 parties. This appears to have boosted the minor parties’ support – 15% in the SNP poll compared to 12% in the Telegraph one (which one is the better approach is a difficult question. On the face of it it is fairer to list the smaller parties along with the big four but in practice it can result in too high a score for the small parties. YouGov’s over-rating of UKIP in the last European elections, for example, was because they were listed in the main question.)

A third difference was that the SNP poll asked a separate question up front of how likely people were to vote, while the Telegraph one just gave “would not vote” as an option along with the rest. This had the potential to make a big difference, but in practice it doesn’t seem to have. The proportions of non-voters in the two polls are pretty much the same.

So which figures are “correct”? Both are fair polls and neither are outlandish enought to look like an obvious rogue. The Westminister voting intention question in the Telegraph poll might well have skewed later answers slightly, as respondents like to give consistent answers. That said, the 36% figure for SNP support in the constituency section of the SNP poll does seem quite high, and it could be an outlier. Alas, in recent years we haven’t had solid series of Scottish polling data to be able to tell when a poll seems out of line and when it doesn’t. Hopefully as we move towards next years’ elections we will get more regular polls.


ICM November Poll

The headline voting intention figures for ICM’s monthly poll for the Guardian, with changes from last month’s poll, are CON 37% (-2), LAB 32% (+3), LDEM 22% (nc). Last month’s ICM poll showed a huge 10 point Conservative lead, with a Labour score that seemed anomalously low, this appears to be rather a return to the norm.

Asked how they would vote were Gordon Brown Labour leader, the figures changed to CON 40%, LAB 32%, LDEM 20%. So far all hypothetical polls of this type have shown Labour trailing, and in the vast majority of cases they have performed worse than their current standings if Gordon Brown’s name is mentioned in the question. There is normally a consistent pattern of an increase in the Conservative vote at the expense of the Liberal Democrats, probably the result of some voters disillusioned with Tony Blair returning to Labour from the Liberal Democrats, while at the same time some voters switching from Labour to the Conservatives. It is important to note firstly that it could also be a result of mentioning David Cameron in the question – Populus tested this and found that there was a similar, but smaller, shift in support if you mentioned Cameron and Blair’s names in the question. Secondly, it is a purely hypothetical question – people aren’t very good at predicting how they react to political events in the future and have no way of knowing what Brown will actually do once he becomes Labour leader.

On the subject of the Labour leadership, there was a MORI poll for the Sun yesterday comparinng Brown and John Reid. Overall 36% of people thought Brown would make the better Prime Minister, compared to 22% for Reid. Reid’s support was alrgely amongst non-Labour supporters, amongst Labour supporters Brown was overwhelmingly preferred by 59% to 15% for Reid.

Over the weekend figures from a new TNS/System 3 poll of voting intentions in the Scottish Parliament were released. The Sunday Herald’s report rather strangely describes it as being “leaked”, but perhaps that’s just to make it more exciting :). In fact it was conducted on System Three’s own behalf, and the fieldwork dates were the 26th-31st of October.

The topline voting intention figures, with changes from System Three’s last poll conducted at the tail end of August (it was published in September, and the numbers tally with the numbers the Sunday Herald ascribes to their September snapshot, so we don’t seem to be missing a set), are as follows. On the constituency vote: CON 12%(nc), LAB 38%(+2), LDEM 14%(-3), SNP 30%(+2), GRN 3%(nc), SSP 3%(-1), and on the regional vote: CON 9%(-2), LAB 30%(+2), LDEM 17%(-2), SNP 33%(+6), GRN 6%(-2), SSP 4%(-2). The poll suggests that in both votes support is beginning to focus upon Labour and the SNP at the expense of smaller parties.

In recent years Scottish opinion polls have been few and far between, so we don’t really have enough data talk about differences in party support between pollsters. Looking at recent Scottish opinion polls though, on the constituency vote there does seem to be a noticable contrast between the levels of Labour support recorded by System Three, which in their last 4 polls have consistently shown Labour at 36% or above, and the figures produced by other pollsters, who have all put Labour around about a 30% level of support. In the regional vote (the more important one in terms of overall seats) the pattern is less clear.

For the record, System Three use a methodology very similar to MORI’s. Quota sampling is used to construct samples, interviews are conducted face-to-face in people’s homes and samples are weighted using gender, age, social class and working status. Like MORI, there is no weighting by past vote (though the initial sampling points are chosen to be politically representative – i.e. there are the correct proportion of sampling points in Labour held constituencies, Lib Dem held constituencies and so on).

At the moment System Three are not doing any filtering or weighting by people’s likelihood to vote (though they will do this in pre-election polls). If likelihood to vote in Scotland follows a similar pattern as in Great Britain as a whole – i.e. if Labour voters are the least likely to vote – then this would go part of the way towards explaining the difference in figures.

The Times’s report of this month’s Populus poll highlighted what appeared to be a striking 7 point difference in the Conservative lead between how men would vote and how women would vote, a gap that grew to 12 points when Populus asked how people would vote with Brown as leader. There various bits of commentary on how David Cameron is winning the women’s vote. But is he?

I warned at the time that the big twelve point difference on the hypothetical question with Brown as leader might well be based on very small sample sizes – indeed it was. In fact, if you look at the gender breaks in Populus’s polls there is no obvious pattern, it bounces up and down from month to month. Back in April the Conservative lead amognst women was also 6 points higher than amongst men, but come July the Tory lead was 8 points higher amongst men.

ICM’s gender splits don’t show a steady pattern either – in their October poll male and female voting intention was almost identical, but in their two previous polls the Conservative lead was far, far larger amongst men than women. In their last poll for the Sunday Telegraph the Conservatives had a 15 point lead amongst men and a 2 point lead amongst women. In their September poll for the Guardian though, the Conservative lead was 10 points higher amongst women.

Therefore it looks as though the apparantly gender gap in the polls wasn’t significant at all, it’s just the result of the small sample sizes you get on crossbreaks throwing up erratic figures. The exception, however, is YouGov. YouGov have larger sample sizes to start with, and don’t filter or weight by likelihood to vote, leaving them with a larger sample and – in theory at least – less volatile cross breaks. Looking at the Conservative lead amongst men and amongst women on YouGov’s polls since the last election there is a noticeable pattern.

On YouGov’s crossbreaks the Conservatives have recorded a higher lead amongst women voters than amongst men in all but one poll since September 2005 (the exception had identical leads amongst men and women). On average the Conservative lead amongst women has been six points larger than amongst men. In comparison, prior to September 2005 YouGov recorded leads amongst women that were almost identical to those amongst men (if anything, the Conservatives did marginally better amongst men). It does tie up exactly with David Cameron’s emergence as the frontrunner in the Conservative leadership campaign, but certainly judging by YouGov’s figures the Tories have opened up an advantage amongst women since last Autumn.


The bottom line, therefore, is that the gender gap in one single Populus poll probably doesn’t signify much at all – but looking at the longer term picture women voters do seem to have swung more heavily to the Conservatives than male voters have.