A semi-regular question I get asked here is why Yougov, with a panel of 250,000 people, can’t do polls with much larger sample sizes so we can see what is happening in a particular type of seat, region, county, city and so on. A normal voting intention poll does tell us how the parties are doing compared to last month, or last year or the last election in terms of overall popularity, but they really are quite limited in terms of telling us what would happen at an election. To do that, we need to know what is happening in different parts of the country, different “battlegrounds”, seats of different marginality (or in a perfect world, individual seats). With the normal sample size of around 2000 we can’t do that, though it doesn’t stop people trying: I’ve given up telling people that the cross-break for respondents in Scotland in a standard poll is so small as to be utterly meaningless.
The reason huge polls allowing us to look at small groups of seats hasn’t happened before is the cost. For YouGov every person they ask costs an extra 50p, for companies conducting phone polls, every interview takes an extra 20 minutes or so of phone charges and interviewer’s wages. Few newspapers would cough up that amount of money for a poll that would be fascinating for activists, anoraks and apparatchiks, but wouldn’t be much better at selling papers than an ordinary poll. Enough beating around the bush, PoliticsHome has gone for it: a poll of 34,000 odd people, in marginal seats, with the fieldwork carried out by YouGov. To declare my interest – the survey design and analysis was almost entirely done by me.
How we did it was a bit different from usual. Firstly, we didn’t lump all marginal seats into one big pile and assume that they all behaved together. Instead we split them into different sub-groups of seats that we thought had shared attributes – so all those seaside towns on the Conservative target list were taken as a single group, all the commuter seats around London were grouped together, the group of marginals in West Yorkshire was another, those in the East Midlands along the M1, the Metropolitan West Midlands were taken together and so on. We also took out different battleground seats, so Lib Dem held seats in the South West were treated seperately, so were Lib Dem seats elsewhere, and Lib Dem vs Labour seats. Naturally, Scottish and Welsh marginals were also treated seperately. This shows some clear trends, with the Conservatives doing noticably better or worse in different parts of the country.
Secondly was the way we asked the question. Past marginal polls in seats contested between Labour and the Conservatives haven’t tended to be that bad, but seats where the Lib Dems are in contention have normally produced bizarre results showing the Lib Dems facing annihilation, even when they haven’t been doing that badly nationwide. My theory is that this is because when people are asked how they would vote in “an election tomorrow” they give their national party preference, which isn’t necessarily the same as how they would vote in their own seat. For example, a Labour supporter living in Cornwall might say Labour if asked how he would vote in a poll, but actually vote tactically for the Liberal Democrats in his own seat to keep the Tories out. A Conservative supporter in Lewes, for example, might say Conservative in a poll on how he would vote, but actually vote Lib Dem in his own seat because he admired Norman Baker.
The Liberal Democrats benefit the most from tactical voting and from “incumbency vote”, so disproportionately suffer in polls like this. Hence in this poll we used different questions: first we asked how people would vote “in an election tomorrow”, then we asked them whether they voted for their first preference or tactically, then we asked them to consider their own seat, the situation there and the parties and candidates that were likely to stand. This didn’t make a vast difference in most seats (though there was evidence of a personal vote for Labour incumbents too, albeit a smaller one), but it made a massive difference in Liberal Democrat seats – it really did make the difference between annihilation for the party, and relatively limited losses.
Anyway, enough teasing you – the report is published and available for download from the PoliticsHome website here. I’ve written a lot of in depth analysis of the data for PoliticsHome’s report which can be downloaded on their website, so rather than regurgitate it here I’d urge you to go and download the report from there and read it yourself. As well as voting intention there is a huge amount of stuff on messages, what would make floating voters switch, what drives peoples votes and so on – plus three articles by me on what I think the strategic lessons for each party should be.
To summarise the voting intention bit, it projects a Conservative majority of 146 seats – the full top line figures are the Conservatives on 398 seats, Labour on 160 and the Lib Dems on 44. The important part of the study though is to show where the swing is larger or smaller, and how well the Liberal Democrats are doing against Labour and the Conservatives.
Firstly, the common perception is that the Conservatives are doing badly in the North, so-so in the Midlands and best in the South. The actual picture is somewhat different. The Conservatives worst regions are Scotland (no surprise at all there – the anti-Labour vote has gone almost entirely to the SNP), parts of the North and, perhaps surprisingly to some, London. In the North the Conservatives are doing comparatively badly in West Yorkshire, which tallies with the poor local election results they’ve had there, and Cumbria. However, in the rest of the North-West they are doing very well, as they are in the North-East – though there are few seats there where a Conservative swing will translate into seats gained. The comparatively poor Conservative performance in London actually tallies with the mayoral elections, which saw Boris Johnson winning the race, but not on the sort of huge swing that national polls would have suggested (and alternative interpretation, of course, could be that Boris in power has put people off the Tories!).
Where the Conservatives are doing best is not the South, the London commuter belt is actually pretty average for them. Their strongest showing is in the Midlands, producing some surprising projections such as Geoff Hoon’s Ashfield falling and the Conservatives winning a seat in Coventry.
Perhaps the most interesting part of the poll are the Lib Dem findings. Standard national polls throughout this Parliament have been very poor for the Liberal Democrats (the last couple of days have been an exception, but we won’t know for weeks whether that is anything more than a temporary conference boost), and their supporters have been consoling themselves with claims that Lib Dems seats don’t necessarily reflect the polls and how the personal vote of their MPs somehow shields them from drops in support, Chris Huhne wrote an article making exactly that sort of argument on LibDemVoice a week or two ago. Under normal circumstances though all of this is guesswork: we really don’t have the evidence to judge. This poll gives us some.
The topline results of the poll show a swing of about 5 points from the Lib Dems to the Conservatives, enough for them to lose 19 or so seats, but not the sort of slaughter that uniform swing calculators normally project. The bad news is that the gains from Labour that the Lib Dems hope will counter-balance any losses to the Conservatives have almost entirely failed to materialise. In seats that are marginal between Labour and the Liberal Democrats the anti-Labour vote is going to the Conservatives (which in effect, means Labour hold most of these seats). The consolation for the Liberal Democrats is that relatively few people in these seats realise they are in a Lab/LD marginal – if the party can successfully position themselves as THE party to beat Labour in those seats, something they have great experience in doing, they could do far better in those seats and start gaining Labour seats to balance losses to the Tories.
Looking beneath the surface in those Lib Dems seats though produces some fascinating results – without the “locally prompted” question I mentioned earlier the Lib Dems really would be facing annihilation, but asking how people will vote in their own local constituency produces incredibly different results in Lib Dem seats, with much, much higher Lib Dem support. That’s partially tactical voting, but it’s also likely to be down to Lib Dem MPs’ personal votes which, as Chris Huhne suggested, probably really are much bigger than those of many MPs from other parties. One of the other things we asked in the survey was for people to rate their MPs on the attributes they considered important and, on almost every count, people with Lib Dem MPs rating their own MP far more highly than constituents of Labour and Conservative MPs rated theirs. Lib Dem supporters often claim their MPs work harder or perform better or are more popular or similar stuff, and most other people dismiss it as partisan guff – but in the perception of voters, it’s true.
Anyway, the full report is here to read.
(Before someone else spots it, a mea culpa. Obviously if 33% of people in Twickenham thought Zac Goldsmith was their Tory candidate they would have been horribly mistaken, it was 33% of people in Richmond Park and they were quite correct. Unfortunately Andrew Rawnsley spotted my rather shameful error after the report went to print. Ooops)