‘Prediction is very difficult’, said the Danish physicist Niels Bohr, ‘especially about the future’. But a recent academic conference at the University of Manchester tried to do exactly that with the forthcoming election. The conference was featured on the ever-excellent politicalbetting.com website, and what followed was a shower of abuse from readers of the site, many of whom didn’t like what they were reading. Much of this was ad hominem, much downright abusive, and much of the worst (or best, depending on how you look at it) has since been removed from the site. One of the allegations was that the academics were engaged in an exercise in ‘groupthink’.
I’m personally quite sceptical about a lot of election forecasting. With the exception of David Sanders’ work on the economy and voting – which famously got the 1992 election right, when the polls and most commentators got it wrong -- I’m not convinced that it adds much value, beyond what you could get anyway from glancing at the opinion polls, and applying a uniform swing. Too often a massive amount of work goes into developing models that deliver little more than a series of SOTBOs (that is, Statements Of The Blindingly Obvious). And for much of the conference, I felt like a character in a Bateman cartoon: the man who was sceptical about election forecasting.
However, the one charge of which my colleagues are absolutely free is that of group think. The different papers presented at the conference came from a variety of different institutions, all working separately, using completely different methods and approaches, and drawing on different data. Some used measures of party support, others measures of prime ministerial/leaders approval. One paper forecasted using local council by-elections, another looked at the public’s expectations of the election. They then weighted and filtered these data in different ways.
And yet here’s the thing: despite working independently, almost every single paper forecast a hung parliament, one in which no party had an overall majority. They differed over which party would hold most seats (most papers predicted the Conservatives), but they almost all predicted than no party would have an overall majority. That’s not group think. That’s the sort of consensus that should make even those of us who are sceptical sit up and take notice. Of course, they could all be wrong, and we’ll know that in a month and a bit, but if so, they will be independently wrong.
And here’s another thing. As politicalbetting.com recognised a few days later, the betting markets – which are also sometimes used as a form of prediction, for those who believe in the wisdom of crowds – are now heading in the same direction.
Professor Philip Cowley