Rogue Kindle Survey is not a Political poll

The Argus play an important part in Brighton & Hove’s community and political life. I have it delivered every morning.

So it’s particularly disappointing that they’ve published the results of a survey in the guise of a proper political poll. What do I mean by that? Well respected political pollsters like ICM, MORI etc use agreed procedures set by the British Polling Council. They weight results using measures to make the result more representative and a better predictor of election results. They also put a huge amount of thought into the questions to improve the likelihood that the results are accurate predictors of electoral behaviour.

Of course polls make mistakes and there are elections where pollsters collectively get it wrong (e.g. Major’s surprise 1992 victory over Kinnock). But the results The Argus quotes are clearly way out of line. Greens scored 22% in Brighton Pavilion for the 2005 General Election, so a 12% Green vote-share is completely incongruous with Greens’ 31.4% in the 2009 Euros, 41.6% in the 2007 Regency by-election and 35% in the Dec 2009 ICM poll (see all the graphs). Furthermore, while I’m not keen to promote Tory chances, it’s absurd to suggest Labour are 10% ahead of the Tories in Brighton Pavilion in the face of a clear national Conservative poll lead of 6-10% and all recent local elections having Tories in a firm second place. This is a rogue survey, it doesn’t deserve to be given the status of poll.

For these figures The Argus cites a survey by Kindle Research, who look to be a small technology research consultancy, not political pollsters. From the information I’ve seen Kindle did ensure demographics were representative of the constituencies – they asked 336 people in each of the three Brighton & Hove parliamentary constituencies. This is a small sample size compared with 533 for the Dec 2009 ICM poll. They also didn’t prompt for ANY political parties. Political polls habitually prompt, because that’s what a ballot paper does when electors come to vote. Kindle also failed to filter out non-voters, which is a basic first step in political polling. I also understand that the political questions were tagged onto the end of a set of completely unrelated questions on a different topic. Kindle admit to having a 10% margin of error (compared to ICM’s 4.3%) but I think it’s even greater than that given not only the small sample size but the type of questioning without prompts or mention of the constituency.

In Brighton Pavilion Greens are fielding Caroline Lucas our high-profile leader, an MEP for 10 years and we have more than doubled our number of councillors in the city since 2005. The results The Argus cites just aren’t credible and do them a disservice.

The full Kindle Research results for Brighton Pavilion:

Labour 26%

Conservative 16%

Green 12%

LibDems 5%

UKIP 1%

Other 1%

Would not vote 11%

Undecided 19%

Refused 7%