As I said yesterday, I'm now pursuing a slightly different tack. We could proceed on the basis that support had moved evenly between the parties in all of the different constituencies by uniform national swing. That is certainly a handy tool and in the absence of any evidence that it's wrong, it should always be the fallback position.
But it is apparent that in some cases at least support has not moved uniformly. And there we should try to do better - to identify where it has moved differently and how that might affect different constituencies. It will help us decide whether the bookies have got their odds right.
The single most dramatic development of the last few years has been the rise of UKIP. In 2010 UKIP polled 3.1% of votes and came third in only four constituencies. It is now routinely polling in double figures and in some polls well into the teens. While as I noted on Saturday UKIP is odds against to win any seats and its chances of winning significant numbers of seats are slim, if it gets voters in the numbers currently being polled, its voters could swing the outcome of many constituencies. So where have UKIP's voters come from? And will they stick around for the general election?
UKIP's voters have been little-studied, and the first major academic study is set out in "The Revolt On The Right":
From analysis of opinion polls, the authors conclude: “Contrary to those who argue that Ukip’s voters are middle class Tories, we actually find that its base is more working class than any of the main parties”.
Blue-collar Ukip voters outnumber their white-collar counterparts by 42 to 30 per cent. In contrast, 44 per cent of Conservative supporters are middle class; while for Labour, the middle classes have a narrow 36-35 per cent lead. 57 per cent of Ukip supporters are over the age of 54, while just over one in 10 are under 35. 55 per cent left school at 16 or earlier, while only 24 per cent went to university.
However, Ukip suffers from “a persistent failure to appeal” to three key groups of voters – women (put off by the “chauvinistic and anti-feminist views” of Ukip members and politicians); young people (who find the party “almost comically out of touch” with their own worldview) and the ethnic minorities (because of its “strident and often emotive language about immigration.”)
If you want to go into more detail, I can strongly recommend the excellent election data site:
We know something of UKIP voters' political background, thanks to YouGov:
And it's clear that the Conservatives are hit hardest by this. Fully 45% of UKIP supporters were 2010 Conservative voters - a far greater percentage than any other single segment of UKIP support. That's over 5% of the electorate as a whole. So in constituencies where UKIP is likely to be disproportionately strong, the Conservatives are likely to be disproportionately affected. But it shouldn't be forgotten that in seats where UKIP is likely to perform poorly, the Conservatives can expect to do better than otherwise expected.
It seems that UKIP's rise has favoured Labour. It has taken a disproportionate number of Conservative voters, resulting in Labour being relatively better off by about 4% than it otherwise would be. Where we have no particular evidence to the contrary, we should assume for now that relative swing will be uniform.
This uniform swing is of course already factored into the polls, so we do not need to make an additional adjustment. What we need to do is determine which seats may see a disproportionate swing to UKIP and adjust our betting strategy accordingly by tending to favour Labour and write down the Tory chances. Oh, and to identify which seats may be resistant to UKIP's charms, where we should do the reverse.
It seems from local election results that UKIP's new support is geographically concentrated. This image from after the last local elections tells the story (click on the map to see it in detail):
It seems a safe bet that the deeper the colour, the more impact UKIP is going to have on the outcome, even if it doesn't win. Note also that large areas have proved resistant to the purple tide so far (though this year will be the first test of their newfound popularity in metropolitan areas). These include seats with UKIP at relatively short prices to win, like North Devon and South Suffolk. And, obviously, UKIP is basically an English affair. I suppose EIP would be harder to pronounce.
Let's look at this map a while. Even bearing in mind that the metropolitan areas are not on the same footing, UKIP still look as if they are a rural south eastern party. UKIP seem to be particularly strong in East Anglia, and particularly the fenlands. But even here they have no county council seats in Norwich, Ipswich, Cambridge or Peterborough. There are other areas of strength - notably Kent and Sussex, but with the exception of south Essex, UKIP councillors are mainly elected in rural areas.
Now obviously council seats are awarded on a winner-takes-all basis, and healthy vote shares don't always register in such tables. We know that northern English urban areas have potential UKIP voters, because UKIP gathered respectable vote shares in the Rotherham, Middlesbrough, South Shields and Wythenshawe & Sale East by-elections (the contrast with Manchester Central and Cardiff South & Penarth is marked). But in the absence of enough data to judge reliably, we should not make any allowance either way as to how UKIP will fare in these areas in the general election. As and when we get more data, we can come back to that. The coming local elections in May are going to be a very important data point on this critical question.
If UKIP's support is disproportionately in rural south east England, it's going to be disproportionately in Conservative safe seats. That's a sideshow from the main Labour/Conservative fight (and arguably reflects Labour's inability to speak to those areas of the country).
Leaving aside other regional variations, we should tend to look more kindly on Conservative prospects in more northern marginals where UKIP have so far failed to impress and less kindly on Conservative prospects in marginals on the eastern coast and in the Thames estuary. All other things being equal, this consideration would make you think that Weaver Vale would be a more likely Tory hold than Waveney, and Rossendale & Darwen a more likely Tory hold than Cleethorpes. To give an example, if UKIP's vote is 24% in Waveney and 8% in Weaver Vale, that would disproportionately benefit Labour relative to the Conservatives in Waveney to the tune of 5% if UKIP's voters' previous allegiances are comprised in the ratios identified by YouGov in both seats (and conversely give the Tories a better chance in Weaver Vale).
Of course all other things are not equal. There are other geographical considerations at play. And it is far from clear how much of its current support UKIP will hold onto for the next election. So next we need to look at where UKIP's current support might go to.