Monday, 28 October 2013

Modelling the next election: vast uncertainties remain

There's been a bit of a stir over recent days concerning a new statistical model of polling and popularity running up to the next election. Stephen Fisher, Lecturer in Politics at Oxford University, has published a new working paper using past polling trends to look at where we might get to in May 2015.

The results are pretty stunning, to be honest. Forget all that stuff about how impossible it is for the Conservatives to gain an overall majority because of the inbuilt bias of the electoral system, their non-existence north of Birmingham, the relative unpopularity of their sitting MPs, Labour's poll lead, and any number of other hurdles that have been thought to stand between David Cameron (above) and an overall majority.

No, Dr Fisher uses past data on governments' catch-ups over the last months of every modern Parliament, and the polls' record of overstating Labour support, to show that the Conservative Party is in fact pretty likely to win an overall majority. You can read the whole paper here, for free - it's only 19 pages long - and it repays a good long ponder.

Now there's a lot that's potentially going wrong with this analysis. One could equally point (more qualitatively) to historical parallels pointing the other way - showing that Labour aren't doing too badly on past form, and that it's very, very rare for incumbent governments to increase their vote shares. More narrowly, what one might call macro-approaches - that focus on the headline numbers - find it difficult to grapple with the internal dynamics of electoral switching. It's perfectly possible that Liberal Democrat switchers to Labour will stay put all the way up to election day, boosting the party's score from the low 30s predicted in this model to the mid-30s. No-one really knows. We're in uncharted waters. Sure, the numbers on single-party governments are strong, and other countries more used to coalitions back them up: the biggest governing party will indeed surge back into contention. But what if Labour's numbers don't sag in return? There seem to have been precious few switchers from Conservative to Labour, so those two processes are not mirror images of one another. It's a more parohical question that explains why Nick Clegg is trying to tack leftwards to get some of his 'lost' voters back. No-one knows whether he will succeed.

But this approach has the virtue of clarity. It has the advantage of showing us its workings - including the massive margins for error due to the assumed inaccuracies of past polls, to name but one uncertain element. It's an academic work that Dr Fisher is happy to take comments on, as you can see from the tenative nature of his draft and his careful and measured response to a tidal wave of comments. This blog has always been very, very sceptical that Labour will get as far as people have thought since the 'omnishambles Budget' of 2012, and there's been lots of recent evidence to back up that point. That party's progress in local elections, and in the polls, is anaemic, leaving the way open to a strong Conservative recovery - especially during a rapid economic upturn.

But saying that a Conservative majority is so likely - more likely than not - is a big, big punt. I'd like to say that you can tune in during May 2015 to see if regression analysis really does turn out to be a good guide to the future, but of course so much will have changed by then - so many events will have intervened - that it's not much of a test of whether the elements modelled here are the critical ones.

But it's a start - and it's better than a hunch, in anyone's language.

No comments:

Post a Comment