The almanis Brexit market has moved sharply towards “Remain” over the past day-or-so, now assigning it around an 80% likelihood. That’s even punchier than Betfair whose odds are showing a 72% implied probability. But why are forecasters showing such confidence in a “Remain” victory in the face of fairly equivocal polling evidence. Here are two very good reasons.
The common theme though is that a referendum is not like a normal election and we shouldn’t read polls in the normal fashion. In elections, people tend to know how they’ll vote and how they voted last time. Polling is therefore a fairly simple matter of putting together a representative sample and asking the right questions – they got the first bit wrong in last year’s general election though. In this referendum, far more people will remain undecided and have no comparable voting history. That makes it all a bit more tricky and we need to dig a little deeper to get at the truth.
Exhibit 1: Pay closer attention to the phone polls, not the online polls
Both the “Leave” and “Remain” camps have had polling leads reported over the past couple of months (summarised here). Those polls are neatly split: those conducted by phone have tended to show big “Remain” leads while those carried out online have suggested “Leave” has its nose out in front. A paper published by Matt Singh of Number Cruncher Politics and James Kanagasooriam of Populous suggests that it is the phone polls we should be listening to.
Phone pollsters, it seems, are far more able to tempt undecided voters off the fence, something they are able to do because they are actually engaged in a conversation with them. These undecided voters are just as likely to vote as those with clearer views so their opinions matter just as much but they’re only captured by phone pollsters. It seems that, when pushed, they have a strong tendency to fall into the Remain camp.
Exhibit 2: Pay closer attention to people’s vote expectations, not their voting intentions
While phone pollsters can delve a little deeper into the uncertain minds of voters than their online cousins, there are other ways to entice a richer assessment. Most people have a pretty good idea of how they’re going to vote, and how a good number of their friends and colleagues will vote too. If you ask someone which way they expect the referendum to turn out, you effectively throw your net across the whole group and derive a broader analysis.
But so much for theory. David Rothschild and Justin Wolfers of the Brookings Institute published a paper back in 2012 which included some actual analysis. This showed, fairly conclusively, that you’re better off predicting outcomes of narrow races using voter expectations than voter intentions. They looked at US state presidential elections between 1958-2008 and found that, where expectations and intentions predicted different outcomes, the expectations were correct in 78% of cases.
As luck would have it, ORB International have been asking for both intentions and expectations and the results are below. The difference is stark.
Note also that Ladbrokes, the UK bookie, has reported 90% of the money staked on the referendum in the past month has been on “Remain”. That’s another powerful indication of expectations.