Faced with opinion polls that continue unrelentingly to put the No side ahead, albeit by differing margins, the Yes side appears to be putting its faith in the so-called ‘missing million’. The term is used somewhat loosely; sometimes it refers to those who are eligible to vote but are not on the electoral register and sometimes to those on the register who do not usually vote. However, central to the idea is that, if as anticipated the turnout in the referendum is very high, many people will vote in the referendum who do not usually do so. In particular, if turnout is 75% (which some would consider a conservative estimate) that would mean a million or so more votes would be cast than the two million that were counted in the 2011 Scottish Parliament election (when just half of those eligible to vote did so).
The Yes side believe that those who do not regularly vote will prove to be fertile ground for the independence cause, not least because many of them live in some of the more deprived parts of Scotland where Yes support tends to be higher. They then go on to ask how good the polls are at capturing this section of Scotland’s population. It is suggested that the ‘missing million’ may well not be accessible by phone, may not be particularly inclined to sign up to take part in internet surveys, and would certainly seem particularly unlikely to agree to participate in a survey about politics.
So how good are the polls at securing responses from those who are not regular voters but may well vote on the 18th September? And is there any reason to believe that support from irregular voters could deliver victory to the Yes side on 18 September?
As it happens all of the polls, apart from Ipsos MORI, ask their respondents what they did in the 2011 Holyrood election. The figures are not always reported as clearly as we would like; sometimes we cannot separate those who voted for a small party on the constituency ballot from those who did not vote at all (but fortunately only 1% voted for a small party on that ballot anyway), while some polls (not unreasonably) allow people to say they cannot remember what they did in 2011 (while of course anyone aged under 21 was not eligible to vote three years ago). But in the following table we show for the most recent poll of each of the five pollsters for whom the information is available, the proportion of people who said they did not vote in 2011, could not remember whether or how they voted, or were not eligible to do so. In each case we first of all show the relevant proportion for the unweighted data (and thus the actual number of people interviewed who fell into that category) and then the same proportion after the poll was weighted to make it more demographically and politically representative.
Table 1 Reported Levels of Non-Voting etc in the 2011 Scottish Parliament Election
* includes those who said Don’t Know, Can’t Remember, Not Eligible etc.
** includes those who voted for a small party as well as all those who say they did not cast a vote
As the table shows, the polls vary considerably in the proportion recorded as not having voted in 2011. In TNS BMRB’s most recent poll, as many as four in ten (prior to weighting) said they did not vote or could not remember what they did in 2011. TNS BMRB’s polls are conducted face to face as part of an omnibus survey that covers a wide range of topics, not just the referendum, and that may well help ensure that they are more successful at securing the participation of those with little interest in politics. In the other polls, all conducted over the internet and often only including questions about the referendum, the figure is more modest, at between a fifth and a third.
It would seem that the Yes side have a point, as even TNS BMRB’s poll contained fewer people who did not vote in 2011 than was the case according to the official result. However, when we look at the proportion of non-voters after the data have been weighted, in two cases, TNS BMRB (again) and ICM, the proportion is then quite close to the official 50% abstention rate. The reason for this is simple – these are the only two companies to weight their data so that the proportion who said they did not vote or cannot remember what they did more or less matches the official abstention rate. In the case of the other companies, however, their weighting procedures make relatively little difference to the proportion who say they did not vote.
Still, given this contrast, if the ‘missing million’ are more inclined to vote Yes we might anticipate that the polls conducted by TNS BMRB and ICM would have been reporting a relatively high level of support for Yes. But while ICM are one of the pollsters that tend to produce a relatively high Yes vote (on average they have put the Yes vote at 45% – after Don’t Knows are excluded – in the polls they have conducted since April), TNS BMRB has been one of those that has tended to paint a less optimistic picture for Yes (42%).
However, we can also look directly at how those who claim they did not vote in 2011 say they will vote in the referendum. In Table 2 we show the proportion of this group in each poll who said that they will vote Yes (after Don’t Knows are excluded, and as one might anticipate the proportion of Don’t Knows is rather higher than it is amongst voters as a whole) and compare it with the proportion of all voters in the same sample who said that they intend to vote Yes.
Table 2 Percentage Yes support amongst those who abstained in 2011 and amongst all voters
N.B. In each case those who say they don’t know how they will vote in the referendum are excluded.
In four of the five polls the level of support for Yes was in fact lower than in the sample as a whole – and markedly so. On this evidence, rather than serving to boost Yes support, the ‘missing million’ are apparently keener on remaining in the UK than are voters as a whole. Only in one case, that is YouGov’s most recent poll, is there any indication that those who did not vote in 2011 might be a little more inclined to vote Yes. In short, in so far as some polls at least may have too few of the ‘missing million’ in their sample, the effect may be to lead them to over- rather than under-estimate Yes support.
There is, of course, a possible caveat. Perhaps those people who abstained in 2011 who are successfully contacted by the pollsters are not representative of all those who are potential members of the ‘missing million’. Perhaps there is a certain kind of non-voter that none of the polls succeed in contacting who is more inclined to vote Yes. But given what most polls are telling us about the inclinations of those 2011 non-voters whom they can contact, the behaviour of any such group is going to have be very distinctive indeed if the votes of the missing million are to prove decisive for Yes on polling day.