Populus Euro-poll, and YouGov's European methodology

Today's Times has figures for European voting intention from the recent Populus poll - our first poll for the European elections from a pollster other than YouGov. The topline voting intention figures are CON 34%, LAB 25%, LDEM 20%, UKIP 6%, GRN 5%, BNP 2%.

On a uniform swing this would lead to the Conservatives winning 31 seats, Labour 20 seats, the Lib Dems 14 and UKIP only 3. The SNP and PC would both win 1, though a uniform swing calculation will underestimate their performance given that an increase in their vote would actually be concentrated entirely in Scotland and Wales.

These figures are very similar to those I reported for YouGov at the weekend. The BNP have significantly lower support - 2% compared to 4% - but that's to be expected given that Populus use live phone interviewers and YouGov use a more anonymous web-based survey, where there is less chance of a social acceptability bias.

However, on the subject of YouGov's voting intention Peter Kellner has a commentary on YouGov's European Parliament figures. Firstly Peter has provided figures for only those certain to vote, which given the likely turnout in the European elections he thinks will give a better prediction. These figures are CON 37%, LAB 22%, LDEM 19%, UKIP 7%, SNP/PC 5%, GRN 4%, BNP 4%. These would, on a uniform swing, give the Conservatives 32 seats, Labour 18, the Lib Dems 13 and UKIP 4 (on this calculation the SNP and PC would get one each, but since in reality their increase would be wholly in Scotland and Wales the SNP would get more). Unlike the unfiltered YouGov figures and Populus's figures, these show Labour declining compared to 2004, which does seem rather more feasible given the currently political situation.

Is Peter right to go for the "only those certain to vote" though? Well, in theory he certainly is - the European elections are a low turnout election and some sort of filter should really be applied. In practice it isn't quite so cut and dried, YouGov produced both sets of figures back in 2004, and filtering by likelihood to vote certainly gave far more accurate vote shares for Labour and the Conservatives. It was less good for UKIP and the Lib Dems though.

2004 YouGov European election polling All giving an intention - CON 24%, LAB 26%, LDEM 15%, UKIP 19%, GRN 6%, BNP 4% Certain to vote - CON 26%, LAB 24%, LDEM 13%, UKIP 21%, GRN 6%, BNP 3% Actual result - CON 27%, LAB 23%, LDEM 15%, UKIP 16%, GRN 6%, BNP 5%

Those figures bring us onto the other significant question - whether or not to prompt by minor parties. In Westminster voting intention questions all pollsters prompt by party name: would you vote Labour, Conservative, Lib Dem or SNP/PC? The reasons is that historically polls that did not prompt by party name used to underestimate the level of Liberal Democrat support - people forget they were an option. Pollsters do not, however, prompt by the name of smaller parties like UKIP or the Greens for similar emperical reasons, if they do, they end up overestimating support for them. It may not sound very fair, but it leads to the most accurate results.

However, with elections run on a PR system it's a trickier question. Parties like UKIP and the Greens have representation at regional or European level. No longer parties that can't win, or won't even have candidates in many constituencies, shouldn't they be prompted for? Again, the answer comes down to what works. In 2004 YouGov included minor parties in their prompt by party name, and it significantly over-estimated the true level of support for UKIP. In the Scottish elections in 2007 YouGov prompted by minor parties in the regional vote question and over-estimated Green support. In his commentary Peter also reveals that YouGov ran test surveys after the 2004 European elections to see what level of prompting gave the better recall result, and they too suggested it was better not to include prompts for minor parties in the question. As a result YouGov are not including minor parties in their main prompt for the European elections this time round.