The Scottish referendum tomorrow is a great set piece test of the accuracy of opinion polling. We have snapshots from a wide range of pollsters, both online and telephone all suggesting a very close finish, with No ahead only a few points at 52 per cent, versus 48 per cent for the Yes. But what if the polls are wrong? There are plenty of potential challenges.
There are still a group of around 10 per cent of people who are undecided or could change their minds. Will they split roughly 50/50 as with everyone else or not? Will they all decide to support Alex Salmond at the last minute? At Ipsos MORI our approach to dealing with this is to ask which side of the argument they are leaning towards, Yes or No. We include all the “leaners” with the side they are leaning to, as we have routinely done with every final voting poll we have ever published. At the moment they split Yes/NO bang in line with everybody else and move the overall result by less than 0.1 per cent. That leaves us with “don’t knows” who say they are certain to vote, but still can’t decide at all – around 4 per cent of all our respondents. However, even if they split by say 2 to 1 either way, they only move the headline figures by 1 point – and, at the moment anyway, even if they voted 100 per cent Yes they would still not pull the Yes vote over 50 per cent. So undecided voters are not that big a challenge to pollsters.
For me the real challenge is whether one side’s supporters are more likely to actually vote on the day. Is there a possible differential turnout with “silent” No voters more likely to make it down to the polling station or post box than the apparently vociferous Yes voters? Or vice versa? Precedent suggests “Yes” could be more susceptible to this. The rationale for this view is that Yes supporters tend to be younger and more working class – both groups who are less likely in normal elections to actually vote than older more middle class people (who tend to support “No”). However the “once in a lifetime” nature of this event, the clearly effective grass roots Yes campaign and voter registration drives mean that this may well not happen. We will see. In 2010 pollsters tended to overstate the Lib Dem share of the vote for this reason – LibDem supporters from Nick Clegg’s surge turned out to be “softer” than Labour and Conservative voters who were both accurately predicted – even though LibDem supporters told pollsters they were dead certain to vote, on the day some of them didn’t!
Finally there is the idea of “silent Nos” – that there is a spiral of silence making some intimidated “No” voters less likely to agree to take part in surveys at all, or to say they are undecided or refusing to say how they will vote and biasing the sample. The challenge for us is spotting them in the polling data and how to treat them. If “shy Nos” really don’t want to take part in even internet surveys, or completely private phone calls, then even with samples that are demographically matched to Scotland’s population, we will be understating the size of a No vote. We will see. We are polling until the last possible point today in our final poll to see if we can detect movement one way or the other!
Ben Page is chief executive of Ipsos MORI