The Ipsos MORI poll that was originally expected to appear last night has now been published. It will doubtless raise the spirits of the No camp, but whether it should is less clear.
The Yes vote is put at 31% while No are reckoned to be on 59%. That represents no less than a 28 point lead for the No side. It is a much more favourable result for the No side than the average of all the recently published polls, in which the No lead stands at just 16 points.
However, the result is exactly the same as in the last Ipsos MORI poll in May. That too put Yes on 31% and No on 59%. So the poll does not provide any evidence of a swing to No. Rather it is in tune with the broad message of all the other recent polls that there has been little if any change in the balance of public opinion.
What of course Ipsos MORI’s findings do underline is the existence of systematic differences between the polls. Ipsos MORI’s polls have tended to produce larger leads for the No side and this poll is simply true to this past form. Rather than telling us anything new about the standing of the two sides, the poll looks more likely to intensify the debate about the different methods used by the pollsters.
In the last week three pollsters, Panelbase, ICM and YouGov have all produced results that were not dissimilar to each other. Once the Don’t Knows have been excluded they have all put the Yes vote at between 38% and 44%. But now in this Ipsos MORI poll the equivalent figure is just 32%.
There are two big differences between Ipsos MORI’s methods on the one hand and those of those three other pollsters on the other. First Ipsos MORI’s polls are conducted by ringing people up at random over the phone. The other three posters rely on responses from those who have previously indicated a willingness to reply to questions over the internet. Advocates of the former approach argue that it is more likely to secure a representative sample. Advocates of the latter argue that in practice it can be used to create a sample that is politically representative.
Second, in their most recent polls at least all three internet pollsters have weighted their samples so that the way that people say they voted in 2011 more or less reflects the actual outcome of that election. Ipsos MORI eschew that approach because they fear that people do not remember well enough what they did two years ago. But it does mean that we have no indication as to whether their sample looks like a group of people who are politically representative of Scotland as a whole or not. All that we can tell is that in fact the effect of the weighting that Ipsos MORI have done on their latest poll was to increase the estimated Yes vote by two points, and decrease the No vote by three.
The debates about the merits of internet versus other forms of polling and about whether polls should or should not be weighted by how people say they voted in the past has been going on in the polling industry for some years. If these differences persist during the forthcoming year, the Scottish referendum could prove a vital test of which arguments are correct – though of course if they do persist, the polls will continue to sow more confusion than clarity!
But if there are doubts about exactly how big the No lead is, there are certainly other warning signs in the Ipsos MORI poll that the Yes side cannot simply dismiss. First, although there is little change in the proportion who say they do not know how they will vote, the proportion saying they might change their minds about which way they will vote has increased from 19% in February and 23% in May to as much as 30% now. As ICM reported at the weekend, such voters are rather more numerous amongst existing Yes voters (34%) than No ones (28%). In short the Yes vote looks the more fragile – though in truth it appears that neither side ‘s campaign has been effective in firming up the support of its potential voters.
Second, while the poll confirms that those with a strong sense of Scottish identity are most likely to be Yes supporters, even amongst those who say they are Scottish and not British less than two-thirds (63%) say they will vote Yes. Meanwhile Yes support falls away as soon as someone acknowledges being British at all. Thus even amongst those who lay claim to a British identity but also say that their Scottish identity is the stronger, the level of support for the Yes side (39%) is lower than that for No (48%). The Yes side has not maximised its support amongst its ‘natural’ constituency let alone managed to reach out to those for whom being British is part of who they think they are.