Tuesday, June 3, 2014

Ipsos-Mori say that the Yes campaign have drawn level among male voters - but what does that mean?

The Ipsos-Mori representative who appeared on Scotland Tonight was clearly making an admirable effort to be even-handed in his analysis of his firm's new referendum poll, but I do still think he said a number of things that didn't really make sense.  He pointed to a bucking of the general Yes-friendly trend in the least affluent Scottish communities, and suggested this was directly attributable to Labour big-hitters like Gordon Brown beginning to speak out more frequently against independence.  In truth, the evidence that there has been any movement towards No in the poorest areas is extremely weak (the subsample sizes are just too small to draw such a firm conclusion), so to attempt to offer any sort of explanation for the phenomenon seems a classic case of "always seeing patterns in things that aren't there".  Even if an explanation had been required, though, I'm sure most of us could have come up with something rather more plausible than the quaint notion that Gordon Brown has a special hold over the working class.

More questionable still was the claim that, in spite of the big swing to Yes, the gender gap remains "as big as ever".  In fact, the gender gap is much bigger than before in this poll - and in a sense that's exactly why Yes have made progress.  They've seemingly made stunning gains among the male population, while only making up minimal ground among women.  Here is the gender breakdown after Don't Knows are excluded...

Should Scotland be an independent country?

Men :

Yes 50% (+9)
No 50% (-9)

Women : 

Yes 32% (+1)
No 68% (-1)

This is a timely reminder that the conventional wisdom stating that Yes can only win if the gender gap closes isn't necessarily true.  Yes can make net gains regardless of whether the gender gap gets wider still, stays static or narrows - what matters most is the general direction of traffic among both men and women.

All the same, the fact that a No-friendly pollster like Ipsos-Mori is showing a higher Yes vote among men than even some Yes-friendly firms tend to produce is bound to reignite interest in Stephen Noon's recent observation about the 'two-part' swing to the SNP that occurred in the run-up to the 2011 election, with male voters switching to the party in big numbers to begin with, thus generating a substantial gender gap of exactly the same type that we see now in referendum polling.  But the later converts were disproportionately women, meaning that by polling day the gender gap had been whittled down to an almost insignificant 3%.  So why would women move in the same direction as men, but just a little later?  There are two possible explanations - a) the fact that women are more risk-aware than men (which is not the same thing as being more risk-averse) means that they might weigh up the arguments for longer before committing to a change of vote, and b) women may be less likely to engage with an election/referendum campaign at all until polling day comes into view.

It's quite possible, therefore, that the current trends in the male vote may be a glimpse into the future of the female vote.  That's not inevitable - history doesn't have to repeat itself.  But the thought must at least have occurred to the No campaign, and a tied race among men will be causing them some alarm.

Having said that, I see no compelling evidence that the gender gap has actually increased recently - the pattern we're seeing in this individual poll is more likely to be caused by sampling variation.  That's not to say that the gains for Yes aren't real - the chances are that the movement to Yes among men has been overstated, and the movement to Yes among women has been understated, with the two errors balancing each other out in the overall figures.  We saw something similar recently in two back-to-back Panelbase polls which produced near-identical results, even though the gender gap was unusually small in one, and unusually large in the other.

*  *  *

A commenter on the last thread pointed out that the percentage of Labour voters that Ipsos-Mori say are planning to vote Yes is lower than other pollsters have been suggesting.  Now, that doesn't necessarily mean that Ipsos-Mori are getting it wrong - it may be that their telephone fieldwork is turning up a different type of Labour voter than is typically found in volunteer online panels.  But it's certainly worth making the point that if by any chance there is such a thing as Shy Yes Syndrome (ie. Yes voters who are too embarrassed to admit their intentions to pollsters), you'd expect it to be particularly prevalent among the traditional Labour support, and you'd also expect it to manifest itself more obviously in telephone polls than in relatively anonymous online polls.  So a suspiciously low Yes vote of 8% among Labour voters in this poll may well be significant - although crucially this refers to people who currently intend to vote Labour, rather than people who voted Labour in 2010 or 2011.

*  *  *

After the SNP vote was overstated by ICM and Survation for the European elections, John Curtice tentatively raised one or two small question marks over the methodological changes that pollsters have introduced recently. But Ipsos-Mori's findings are actually immune from those concerns.  Although the firm are probably foolish not to weight by recalled vote from 2011, the one good thing about it is that it means their methodology isn't as untested as that of some other pollsters, and therefore there's no reason at all to be sceptical when they say that the Yes vote has broken through the 40% barrier.

*  *  *

Probably the most misleading aspect of Ipsos-Mori's commentary is the tedious pretence that only their own polls exist/matter - ie. the claim that, even in spite of the big swing to Yes, there is still a relatively handsome lead for No, and Yes "need" to make up much more ground with time running out.  It would be more accurate to say that Yes "need" to make up a lot of ground to take the lead in an Ipsos-Mori poll.  As it happens, though, that isn't what the Yes campaign are trying to achieve.  They're trying to win the referendum, and it remains to be seen whether Ipsos-Mori's methodology is more or less accurate than others.  Indeed, after the March poll, I had completely written off the Yes campaign's chances of ever taking the lead in an Ipsos-Mori poll -

"One thing that we can certainly deduce from yesterday's Ipsos-Mori poll is that the Yes campaign are highly unlikely to go into polling day in September in the lead with that particular pollster. But the million dollar question is whether they actually need to do so."

I'm now more optimistic about the prospect of Ipsos-Mori showing a Yes lead at some point, but the latter point still stands - that may not be a prerequisite for referendum victory.  Given the current gap between Ipsos-Mori's figures and the average numbers across all pollsters, it's quite possible that No could have a 5-6 point lead with Ipsos-Mori on the eve of polling, but still find themselves behind on the average.

*  *  *

Judging from his comments on Scotland Tonight, Andy Maciver seems to be the latest victim of the No campaign's moronic propaganda.  He suggested that the Yes camp would be relieved by the Ipsos-Mori poll, because prior to the European elections "ICM and YouGov" had shown signs of movement back to No.  Er, excuse me...YouGov?  The pollster that in its more recent poll showed No slipping to its lowest lead ever?

But of course Maciver is referring to the alternate reality in which a Progressive Partnership poll showing a 9-point slump in the No lead is not a Progressive Partnership poll at all, but is instead a YouGov poll in disguise.  To reiterate the truth yet again (although clearly it's never going to sink in for some people), Progressive and YouGov are two entirely different companies with two entirely different weighting procedures, and BOTH showed a drop in the No lead in their most recent polls.


  1. Like a lot of people James, I have been struggling with the No vote seeming to score nearly 60% in the polls (when DK's are taken out) it just didn't seem to reflect the findings of those 'on the ground' I was concerned that there must be areas of Scotland where the No vote was popular by a landslide!
    However, I saw an interesting article from the Sun newspaper in which they showed a map of Scotland broken down into about 20 areas with people who backed the No vote written as a percentage on each area. If this graph is accurate the No campaign will be shitting themselves, because it had No's polling at around 46% on average with only 2 areas showing over 50% and then in low 50's. it looked to me that the area around Glasgow was at just over 30% for No, which would be a game changer indeed. Do you know anything about this graph and how they got the numbers?

  2. I should add that the Sun graph did not have any numbers for Yes or Don't Know.

  3. I think I saw that map on another website - from memory I think it might have been from ICM? If it was from just one poll it would have been almost meaningless, because the sample sizes for each region would be too small. If it was an average of several ICM polls it might be more useful.