Poll position (1 May 2008)
London elections 08: It’s been a hard-fought, close-run race in the mayoral election and a great deal is at stake – and that’s just for the pollsters
Polls on the London mayoral election have divided into two types – YouGov’s and everyone else’s. YouGov has consistently shown Boris Johnson with a sizeable lead, while the other polling companies have shown the race to be more or less neck-and-neck. As well as the actual election, there is something resembling an election within the opinion-polling industry as two contrasting methods of sampling public opinion fight it out.
A lot of well-established pollsters would love to see YouGov come a cropper; it is a brash, media-savvy outfit which has trumpeted a record of relatively accurate polling in public elections and even television show votes, and whose method (asking its panel for opinions via the internet) contrasts with the telephone or in-person polling as practised by most of the others. The anonymity of the net may produce less inhibited, or more considered, answers than person-to-person interviews, but getting a really representative sample requires quite a bit of weighting and tweaking. The same, to some extent, is true of person-to-person polling. Pollsters now need to get two things right – a good sample, and an accurate set of adjustments.
Polling is perhaps particularly difficult in London, with its enormous diversity, fast-changing population and Britain’s largest proportion of people eligible but unregistered to vote. Pollsters also have to screen for likely turnout, estimates of which range widely between the 37% reached last time to a scarcely credible 60%. Then, because the second preferences of people who voted other than for the top two will be reallocated, they need to get a representative view on how the Lib Dems, Greens and others will line up in the final stage of the count, which is difficult on a small sample.
Boris Johnson has to win pretty big for YouGov to have bragging rights about this election. Their mayoral campaign polls put him between six and 13 points ahead, and their last poll indicated a margin of 7 points. If the result is Johnson by more than six points, YouGov can chalk up the impressive feat of polling in an election in very difficult territory for pollsters. More conventional polling companies would need to ask themselves why they had the election so wrong, as the excuse of a late swing is probably not available, in contrast to the last big polling fiasco of 1992.
A Livingstone win, on the other hand, would vindicate Ipsos Mori and ICM, which have shown the late stage of the race to be extremely close. It would certainly not discredit YouGov, or internet polling, but it would at least require a bit of a rethink of its sampling and weighting techniques. However, a Johnson win by something like two to five percentage points would leave the clash of the pollsters unresolved.