The local elections

I remember that a year ago my referral logs in the days running up to the local elections were full of people searching for local election opinion polls. They must have come away disappointed, because there were practically none. It looks as though they’ll be similarly disappointed this year. Local election polls are actually pretty rare because of the difficulty of identifying people in areas that do have elections and the problems of doing a bog standard poll giving people the choice of all the main parties when in thousands upon thousands of seats not all the main parties are standing.

The only real prediction based on evidence that we have is Ralling and Thrasher’s prediction, based upon analysis of recent local authority by-elections. Rallings and Thrasher predict local shares of the vote of CON 39%, LAB 24%, LDEM 29%, compared to 2003 when the seats were last fought this has the Conservatives up 4 points, Labour down 6 points and the Lib Dems up 2 points (these, it should be noted, are notional shares of the vote based on what they think would happen if there were local elections across the whole country, not actual shares of the vote. The notional shares of the vote on the BBC on the night of the election will be calculated on the same basis.) Rallings and Thrasher project that this would result in the Conservatives gaining around 330 seats, Labour losing around 500 and the Lib Dems gaining around 110.

Rallings and Thrasher’s local election predictions got some bad publicity last year after their initial prediction on Newsnight went horribly wrong – they predicted that the Conservatives would lose 95 seats, Labour lose 130 and the Lib Dems gain 190. In the event the Conservatives gained 317, Labour lost 320 and the Lib Dems stayed almost static. Rallings and Thrasher had produced revised predictions closer to the 2006 elections, but even they showed the Tories only gaining around 100 seats.

Despite last year’s mistake this doesn’t mean that the Rallings and Thrasher predictions should be ignored. In 2006, between the local by-elections that Rallings and Thrasher based their projections upon and the elections themselves Labour suffered the foreign prisoner release scandal and the revelation of John Prescott’s affair – a week that did a great deal of damage to their image. Perhaps if they hadn’t suffered those scandals the results would have been closer to Rallings and Thrasher’s prediction, we’ll never know. In both 2003 and 2004 Rallings and Thrasher’s predictions for the likely notional national shares of the vote in the local elections were actually very accurate so, assuming no scandals or other events break in the next couple of days, Rallings and Thrasher’s predictions are the best we have.

UPDATE: There is a useful list on 18 Doughty Street of which councils are counting on Thursday night and which are counting on Friday morning.

A new YouGov poll in Scotland carried out for the ESRC between the 17th and 23rd April has voting intentions in the constituency vote of CON 15%, LAB 30%, LDEM 12%, SNP 38%. In the regional vote support stands at CON 14%, LAB 27%, LDEM 12%, SNP 32% (and, presumably, others on a total of 15%). A seat projection based on these figures has the SNP with 47, Labour on 40, Conservatives on 18, Lib Dems on 18 with 5 Greens and one SSP.

The poll has a sample size of 1,800, so larger than previous surveys, but it was also carried out prior to the last YouGov Scottish poll for the Telegraph, so is less up to date.

A separate poll of Labour party members reported in the Sunday Times asked respondents about their voting intention in possible Labour leadership contests. In a contest between Brown, Michael Meacher, John McDonnell (presumably the poll was formulated before Michael Meacher claimed that one of the two left-wing candidates would stand down in favour of the other) and Charles Clarke, 80% of those expressing a preference said they would back Brown, with McDonnell on 9%, Meacher on 6% and Clarke on just 5%. The Sunday Times report suggests that a similar question showed Brown also crushing John Reid, but doesn’t give any figures. In the Deputy leadership race Hilary Benn remains in the lead amongst party members, with the support of 36% of those expressing a preference, followed by Alan Johnson on 19%, Peter Hain on 15%, Harriet Harman on 13%, Jon Cruddas on 10% and Hazel Blears on 9%.


YouGov April Poll

As well as the Scottish poll today also saw the publication of YouGov’s monthly GB poll. The topline figures with changes from their last poll are CON 37%(-2), LAB 32%(+1), LDEM 18%(+1). The changes in this poll alone are insignificant, but the drop in Conservative support does match ICM’s poll in the week, as does the Lib Dem increase (though it is obviously at a much lower level).

We’ll probably never find out for sure if there is a genuine weakening in Tory support, since unless there are any polls in the Sunday papers this is probably the final voting intention poll before the local elections, after which comes the deluge! Local, Scottish and Welsh elections results will themselves likely have an effect upon the polls as some parties are seen as “winners” as some as “losers”. More importantly the elections are widely expected to be quickly followed by Tony Blair’s resignation, which will change the whole landscape of British politics.

New Welsh polls

This week there have been two new polls for the Welsh Assembly elections. The first was by Beaufort Research for the Western Mail and conducted between April 16th-23rd. We don’t know much about how Beaufort’s polls are weighted or adjusted other than that the poll is based only on those certain to vote. The full figures are:

Constituency vote: CON 19%, LAB 36%, LDEM 13%, PC 26%
Regional vote: CON 20%, LAB 35%, LDEM 12%, PC 26%

This is Beaufort’s first media poll of the campaign, though Plaid Cymru did release figures from a private poll conducted by Beaufort earlier in the campaign which showed figures of CON 14%, LAB 37%, LDEM 14%, PC 30%. It wasn’t clear whether they were regional or constituency figures, but either way Plaid are now lower than in Beaufort’s previous poll and the Conservatives higher… or, of course, there is a difference in how the polls were weighted or filtered and they aren’t comparable.

The second poll is by NOP for ITV Wales, conducted between April 19th and 23rd. It is not a brand new poll, but a call back to around half the sample who participated in their poll earlier in the campaign. Their voting intention figures are below, with changes from the last poll, and show a big advance for Plaid at the expense of the Conservatives.

Constituency vote: CON 19%(-4), LAB 32%(-4), LDEM 15%(nc), PC 26%(+6)
Regional vote: CON 18%(-6), LAB 34%(-1), LDEM 15%(nc), PC 24%(+4)

The levels of support for the Conservatives and Plaid Cymru are pretty similar in both polls, with Labour doing slightly worse in NOP’s figures and the Lib Dems slightly better. I am somewhat dubious about the NOP poll though. As a callback poll it should be very stable indeed – the basic truth is that people really don’t change their voting intentions that much, most of the change in voting intention polls is just sample error – yet in a matter of three weeks we have party support altering by 6 points. It seems strange.

Two new YouGov polls today, one covering Scotland, the other their monthly GB poll for the Telegprah. The Scottish voting intention poll, with changes from their last Scottish poll, has constituency figures of CON 13%(-1), LAB 30%(nc), LDEM 15%(nc), SNP 39%(+2). Regional support stands at CON 13%(nc), LAB 27%(-1), LDEM 11%(-2), SNP 31%(-4), Green 9%(+5), SSP 3%, SCUP 3%, Solidarity 2%, others 2%.

The latest figures show the SNP lead in the constituency section growing to 9 points, but falling to 4 points in the regional section, largely it seems to the benefit of the Greens. However, this change is probably partially to do with the way the question was asked. Normally YouGov ask voting intention questions by giving respondents a list of the main political parties or “other” to chose from, people who select “other” are then given a list of minor parties to chose from. This is the way that earlier YouGov polls in Scotland were conducted. Today’s poll however gave people a single list of all the Scottish political parties with seats in the Parliament to chose from.

This change is likely to increase the number of people saying they will support the minor parties. Including the name of the three main parties in the question tends to increase the level of Lib Dem support by a couple of percent as people were reminded that the Lib Dems were an option, so it is likely to have the same effect with smaller parties.

Theoretically it sounds fairer to include all the party names in the prompt, even minor and fringe parties, since when people come to vote all the party names will be on the ballot paper (or at least, they will in PR elections with regional lists. In general elections then in the majority of seats people won’t have a Green or BNP candidate on their ballot paper.) Unfortunately, what is “fairer” isn’t necessarily the same as what is more accurate. Back in 2004 YouGov’s poll prior to the European elections prompted by all the party names, and it resulted in a poll that predicted too high a level of support for UKIP. In that case, including all the minor parties in the prompt made the results less accurate, even though the wording seemed “fairer”.

It is impossible to say until after the event which approach will give the more accurate results. In this case, as Peter Kellner explains here, YouGov have chosen to go with a question that includes all the minor parties in the initial question because the results seemed more plausible – at the last election to the Scottish Parliament “other” parties received 22% of the vote. The last YouGov poll using the old version of the question wording showed the other parties together getting only 11% of the vote, half what they got in 2003, the new question shows the other parties together on 19%, still down on their 2003 support but not catastrophically so. We’ll have to wait until next Friday to see which question wording really does produce the more accurate figures.