European elections - final predictions

Having asked around most of the pollsters, I'm not expecting any more European election polls before the big day. So, with all the polling done, let's bring together the final polls of the campaign and see what they tell us about the likely result.

ComRes/Green (31/05/09) CON 24, LAB 22, LDEM 14, UKIP 17, GRN 15, BNP 2

YouGov/Telegraph (29/05/09) CON 27, LAB 17, LDEM 15, UKIP 16, GRN 9, BNP 7

ICM/Sunday Telegraph (28/05/09) CON 29, LAB 17, LDEM 20, UKIP 10, GRN 11, BNP 5

Populus/Times (28/05/09) CON 30, LAB 16, LDEM 12, UKIP 19, GRN 10, BNP 5

The Conservative and Labour levels of support seem comparatively clear. ICM, YouGov and Populus all show them in similar positions, the Conservatives in the high twenties (or on 30% in Populus's case), Labour at around 16% or 17%. The exception to this is ComRes, who have Labour much higher and the Conservatives significantly lower - I suspect this is at least partially a result of their poll not being weighting by past vote (why I don't know, since as far as I can tell it was conducted as part of the same survey as their Indy poll, which was weighted by past vote).

Less clear is how much support the Lib Dems and the minor parties will get. Support for the Liberal Democrats varies wildly - ICM have them all the way up in second place on 20%, which would be a good advance on their 2004 performance. In contrast Populus have them down at only 12%. YouGov have them somewhere inbetween on 15%. The contrast confused me at first - Populus and ICM normally use methods that are very similar indeed, so why the big contrast? Having looked in more detail, it probably boils down to an issue I've mentioned several times when talking about European polls - whether minor parties are included in the prompt.

Populus asked people whether they would vote for "Conservative, Labour, Liberal Democrat, SNP [Scotland only], Plaid Cymru [Wales only], UKIP, BNP, Green or another party?" - and then gave people who said another party a further list of the other micro-parties on the ballot paper.

In contrast, ICM said to people "Some people we have spoken to have said they will vote for one of the main national parties. By which I mean Conservative Labour, Liberal Democrat, while others have told us they will vote for another party. Will you vote for a main national party or for another party?" People who said they'd vote for a main party were then given a list of the 3 (4 in Scotland and Wales) main parties, people who said they'd vote for another party were given a prompt that listed the Greens, BNP, UKIP, Jury Team and Christian Party.

The effect of this seems to be to push people who aren't voting Labour or Conservative towards the Lib Dems in the ICM poll, while including it in the main prompt appears to have boosted UKIP in the Populus poll.

Which one is right? Well, as I've said earlier, YouGov have runs tests in parallel, and learnt from past mistakes, and it seems that for online questionaires you get more accurate results by not prompting by minor parties. However, it doesn't follow that the same applies to polls done by phone where the interview affect is also thrown into the mix. It's likely that having an interviewer depresses the vote for "others" slightly, since people might feel slightly embarrassed about saying they support what other people might see as a strange fringe party.

That brings us onto the subject of the smaller parties. Firstly we should deal with UKIP, who in European elections at least are up there with the other main parties in terms of support. Populus indeed shows them in second place. Again, the difference is down to the prompting, but my guess is that they'll be closer to the levels of support YouGov and Populus show than that from ICM.

Prompted or not, the Greens have a similar level of support in ICM, YouGov and Populus's polls - all have them within 1 point of 10%. The poll they themselves commissioned has that strange 15% support for the Greens, but without seeing what question was asked it's impossible to know it is the result of unusual prompting.

That leaves us with the BNP. Here the big issue is interviewer bias, people are almost certainly more unwilling to admit to supporting a party widely seen as extremist and racist when they are talking to a human interviewer. My expectation therefore is that YouGov are likely to get closer to the true level of BNP support. Whether that is enough for them to win a seat is a different matter. On a uniform swing YouGov's final figures would give them two seats, but it really depends how well they do regionally. They could get seats even on a lower overall percentage if their advance is concentrated in the North West.

So, a day before polling begins we can expect the Conservatives to top the European election poll with around 28, 29%, twelve or so points ahead of Labour. The Lib Dems and UKIP are more difficult to predict - it depends which of the pollsters' questions are best at getting to the truth, but personally I expect both of them to be close to Labour, meaning we could see any of those parties come anywhere from 2nd to 4th. The Greens should increase their vote to around 10%, unless ComRes have picked up something others have missed. Finally I'd expect YouGov, who don't have to worry about interviewer effect, to be the best predictor of BNP support, meaning they'll get around 7%, probably enough for a seat.

Or, of course, everyone could be wrong. It's also worth remembering that all these polls were conducted at least 4 days before polling day. A lot of postal votes will already have been cast by then, but still, that period of time is enough for a further shift in support.