Is Caroline Lucas on course to be elected?
Much of the environmental blogosphere is getting het up about a new poll in the Brighton Argus, which claims to show that the Greens’ lead in Brighton Pavilion has been overhauled. According to the poll, Labour now lead, 16 points ahead of Tories, with the Greens in third on 19% – 16 points lower than they were in a December ’09 poll, which had put them in the lead. That’s a massive change for two months, and something that would really need explaining.
As Anthony Wells has argued on UK Polling Report, there are several reasons why we should be pretty wary about taking the new poll too seriously. The question is whether the differences between the two polls reflect a genuine change in attitudes, or are something to do with the methodology.
The issues that Anthony Wells identifies with the new poll are:
- Small-ish sample size among those that state a party (c. 200 in Brighton Pavilion);
- The new poll seems not to have adjusted for those who are unlikely to vote – this usually overestimates Labour’s support;
- The party-choice question in the new poll was phrased differently in the two polls – the new one did not name the Greens, when the December poll did name them;
- The poll wasn’t weighted to compensate for sampling bias in phone polling.
Of these issues, I think the first and second aren’t big problems in this case. It’s true that the sample size of those picking any party is fairly small, but it’s more than half of the equivalent sample in the December poll. The margin of error is around 7pts, so fairly large but not enough on its own to justify ignoring the poll completely.
While not factoring in likelihood to vote would normally be a problem, I think this time it doesn’t change much. Going back to the December data, we can look at the proportion of each party’s voters who say they are likely to vote (for argument’s sake, let’s take this as 6-10 on the scale, where 10 is certain to vote). The Tories have 82% of their supporters likely to vote, Labour have 80% and the Greens also have 80%. So cutting out those who are unlikely to vote would make little difference to the competitive position.
However, the other issues are probably more important. Normally, the phrasing of a question about voting intent is something like “The Conservatives, Labour, the Liberal Democrats and other parties would fight a new election in your area. If there were a general election tomorrow which party do you think you would vote for?” – which is usually a pretty good way to predict how people will actually vote.
The December poll specifically prompted on the Green party (not surprisingly, since it was commissioned by them), asking “Labour, the Conservatives, the Green Party, the Liberal Democrats and other parties will fight a new election in 2010 in your area. If there were a general election tomorrow which party do you think you would vote for?”, while the new poll didn’t mention the Greens (Caroline Lucas has said that the new poll didn’t prompt on any party; I can’t find the wording, but the appendix of the poll suggests that respondents were in fact prompted on the major parties). I would expect this to be a factor in bringing down the Greens’ numbers in the new poll.
That isn’t to say that asking the question without prompting on the Greens is wrong. The Greens are in a pretty unusual situation in Brighton Pavilion in being a nationally minor party with a credible chance of winning. It may be true both that prompting on the Greens overstates their numbers, and that not prompting on them understates their numbers. My guess is that because they’re quite high profile in Brighton, prompting on them gets an answer that’s closer to reality – but that’s just a guess.
Which leaves Wells’ point about weighting. As he says, it’s necessary to weight phone polls because people who are available and interested in answering phone polls are (for whatever reason) more likely to vote Labour. So if the poll methodology means that we’re getting a higher proportion of likely Labour voters, and we’re not weighting them down, it’s a pretty good reason to doubt the new poll.
But even if all of these issues could be argued away, there’s one reason above all why I think we should be suspicious about the results of the new poll. It looked at three constituencies, and showed the Greens on 19% in Brighton Pavilion (down 2pts on the notional 2005 result); 13% in Brighton Kemptown (up 6pts); and 10% in Hove (up 3pts).
The Green candidate in Brighton Pavilion is Caroline Lucas. The other two seats are being fought by local councillors, who presumably don’t have the kind of name recognition that the leader of the party has. It seems pretty hard to believe that introducing a high-profile candidate would actually reduce the Green Party’s share in the constituency compared with the last election.
So, whether it’s because of different wording, an absence of weighting, or some other reason, I don’t think the Greens in Brighton should be too worried about this new poll. Some have speculated that the poll was paid for by someone with an interest in creating a bad story for the Greens. My guess is that this isn’t conspiracy: just a poll that wasn’t weighted properly – and yet another case of journalists not understanding polling before writing an article about it.