Poll after poll has been showing leads for the Leave campaign lately and we’ve seen a massive shift in momentum towards Brexit. I’m going to cut a lonely figure here and explain why we are still much more likely to see a Remain vote on Thursday. There are a few reasons for my confidence:
Polls don’t exist in a vacuum
When you work in the prediction game, as I do, one of the things that keeps you up at night is the thought of self-denying prophesies. An example would be the millennium bug. We all believed it would happen and, although some scare stories were undoubtedly overblown, that belief forced us to address the problem. As it turned out, the year 2000 was ushered in with no IT-related disasters.
Now look at what happened in the Scottish independence referendum of 2014. We’re now just under a week away from the EU referendum polling day and, at this time during that Scottish referendum, the “No” campaign (the equivalent of our “Leave”) had just gained an unexpected lead in some polls.
But polls don’t exist in a vacuum. People and politicians see these polls and respond accordingly.
On Tuesday 16th September 2014, two days before the big day and not, I would venture, entirely unrelated to this poll surge, the three main UK party leaders signed a “vow” to the Scottish people promising them full scale devolution if they voted to remain within the United Kingdom. I would fully expect a similar offer to be extended to UK voters this coming Tuesday 21st June 2016.
Also, any “Remain” supporter who thought their vote didn’t matter because the result was already in the bag will have reassessed in light of recent events. There are reasons why politicians like to play down their expectaions: among other things, they don’t want their voters lulled into a false sense of security.
Polls underperform predictions in referendums
I mentioned in an earlier blog that research from the US had shown measures of voter expectations significantly outperforming measures of their own voting intentions in predicting the outcome of US presidential primary races. In these primary races, as in referendums, pollsters suffer from a particular weakness: a lack of voting history with which to weight their survey samples.
Naturally, pollsters can’t ask everyone what their voting intentions are but they go to heroic lengths to ensure that the small sample of voters who they survey are representative of the electorate as a whole. In normal elections, their main weapon in this regard is the voting history of their survey respondents. In US primaries and referendums, where the choice on the table is nearly always unique (rather than the same old Democrat vs. Republican or Labour vs. Tories) there is no such information available.
In these cases, it turns out that when you ask people what their expectations of the result are, rather than their voting intention, they are able to deliver a fairly balanced view of the mood among their family, friends and colleagues. The “wisdom of crowds” effect can then do its work and it normally proves a better guide.
The polling firm ORB has been asking voters both for their voting intention and their expectations since March and the stark results are reproduced in the chart below. The people of the United Kingdom still overwhelmingly expect to wake up on Friday within the European Union.
ORB expand a little on their reason for including the expectations question here. They make particular reference to May’s Irish referendum on gay marriage. While the voting intention polls all under-estimated the “no” vote by around 7%, the public at large were able to predict the outcome accurately.
A poll is not a prediction
A friend who runs a polling firm never tires of pointing out to me that a poll is not a prediction. It is a snapshot of what people are thinking at any one time. Things change. And in referendums, time and again people swing back towards the status quo on voting day. Expect a surge back towards “Remain” in the only poll that really matters on 23rd June.