Monday, 16 February 2015

The curious case of different predictions by leading polling analysts

There are some quite stark contrasts between leading opinion analysts 'predictions' of the results of the May 7th General Election, the question being how this comes about. YouGov, at least in the person of its leader Peter Kellner, reckons that 'Ed Miliband's prospects of becoming prime minister are fading' (Sunday Times February 16th, 'Marauding Scots threaten to keep Ed out of No 10'). Despite the focus  Kellner puts on the number of seats that might be won by the SNP, my attention is drawn to the thinking behind his projection of quite a sharp swing to the Conservatives away from Labour sometime by May 7th.

According to the article in the Sunday Times YouGov  puts Labour ahead of the Conservatives by 35 to 32 per cent of votes but predicts that by the time of the election this will have reversed so that the Tories lead Labour by 35 to 30. This,  he projects,  would give the Conservatives 293 seats as opposed to Labour on 270. If (as he suggests) the Liberal Democrats receive 30 seats, this arithmetic might suggest that Nick Clegg in a pro-Tory mood would just about keep David Cameron in No 10, perhaps with the tacit support of the DUP. I don't know. I suppose it depends on what Vince Cable has had for dinner and whether he can keep it down in this scenario.

But what is interesting is the disparity between YouGov and other pollsters projections, and also his notion that the Tory-Labour balance of votes will shift quite significantly this close to the election (in contrast to the relative stability in 2010).

Paul Whiteley, the much-respected and longstanding University of Essex elections analyst projects rather less change in the Labour-Tory balance of votes, and projects that Labour will get 291 seats compared to Tories 281. Surprisingly ( to many) he suggests that the Liberal Democrats will get 48 seats and that the SNP will make few gains. See Government result: Labour minority Government?

Meanwhile Chris Hanretty, the new boy on the election forecasting block, is putting the Conservatives and Labour close in terms of seats, but in the context of the SNP winning over 30 seats and the Liberal Democrats not quite 30. Hanretty's analysis is altered daily, or rather his software alters the findings, as the polls shift. Today (Feb 17th) Labour were half a dozen seats ahead of the Conservatives and the SNP were projected a dizzy 38 seats, Lib Dems a sad 26. I must say I think the Lib Dems will pick up more than 26 and I would be gobsmacked if the SNP get as many as 38 - I know they will make some gains, but given that the pollsters tend not to use the actual candidates names, given the likelihood of a differential turnout (converts to SNP may be less likely to vote) I suspect that on May 8th Jim Murphy will get some plaudits for minimising the scale of Labour losses (whether deserved or not).  Under Hanretty's current arithmetic a Labour minority government would most likely result. See Certainly in any scenario where the Tories plus Lib Dems plus DUP add up to less than 323, Cameron would be likely to be heading to the Palace to hand in his resignation after the weekend.

I would like to ask Peter Kellner, writing in the Sunday Times, how he projects the polls to shift rather more dramatically between Labour and Conservatives more than happened in the run-up to the 2010 election. YouGov's own analysis shows that the polls hardly changed between Labour and Conservatives in the last 2-3 months before the election. You can see this by going to the webpage - you have to scroll down to see the chart.

Maybe Kellner thinks the UKIP tide will melt away leaving a lot of Tory votes on the beach ready to be scooped up? Not that the weakening of the UKIP vote in the last few weeks has led to Conservative gains, and Nigel Farage may now emerge from his trench in South Thanet to sally forth around the country tilting at windmills, solar panels and the EU.

It should be pointed out that all of these pundits (to a lesser or greater Kellner extent) seem to be assuming that come election day Labour will have lost their lead, and that the Tories will get most votes. Maybe. Turnouts in Labour areas are usually lower than in more well-heeled Tory shires. Certainly if Labour did maintain a 1 or 2 point lead then Red Ed wouldn't take long to move into No 10. But the difference with Kellner is that he seems to be assuming a relatively radical shift in votes towards the Tories. This will keep Tory spirits high, and they live in hope that the same syndrome that appeared to affect Neil Kinnock in 1992 will fell 'Red Ed' (he looks rather pinkish to me, though) in May. But this shift did not happen in 2010 - the relative Tory-Labour positions remained remarkably static and typical of the final result. Why should there be a big shift to the Tories now ?

But Kellner is a skilful analyst, so I wouldn't want to rule his predictions out entirely. But I am intrigued. It's all good fun of course. Although Kellner might be keeping Tory spirits up by suggesting that they are still likely to end up in government, his predictions won't actually help David Cameron. Indeed they might even encourage people to try that bit harder to get out to vote Labour (or Green or even Liberal Democrat).

No comments:

Post a Comment