Why the YouGov polling on the London Mayoral election is wrong
It is a serious matter to call a poll wrong. For inherent statistical reasons even the best conducted polls will periodically produce rogue results and therefore the previous YouGov poll showing Boris Johnson 12% ahead in the race for London Mayor, out of line with other polling, could have been a statistical freak. However examination of the YouGov poll published today - which shows Boris Johnson 10% ahead - and the previous one shows them to be fundamentally statistically flawed for the following clear reasons.
Today's poll shows minority parties (that is all parties other than the three main ones) receiving only 5% of the vote and the YouGov poll published on 17 March showing them receiving 3%. This is totally implausible - in 2004 these parties got nearly 19% of the vote. This shows YouGov is using a sample systematically out of line with London's electorate and grossly under recording parts of the electorate.
YouGov's data does not accurately weight the proportion of black and Asian people in London. Ethnic minorities make up around a third of the population of London - far higher than the UK average. Yougov's Internet panels therefore do not accurately reflect London's population, most notably significantly under-counting ethnic minority Londoners and other social groups which more strongly support Ken Livingstone than Boris Johnson.
Today's YouGov polling on issues reveals results far out of line with normal telephone polling. Normal telephone polling carried out by IPSOS MORI only two weeks ago found Londoners supporting the £25 a day CO2 charge on gas guzzlers by more than two to one at 61% per cent to 27%, whereas YouGov reports that this lead is only 41% to 39%.
YouGov today claims to find that Ken Livingstone will get no more votes for Mayor than Labour will win in the London Assembly elections - both will get 37 per cent. In the 2004 election, Ken Livingstone polled 11% higher than the Labour Party in the Assembly elections and all other polling, plus the Labour Party's canvass returns, has shown Ken polling significantly ahead of Labour for the London Assembly.
Today's poll claims that Boris Johnson is 10% ahead among 18-34 year olds and 9% ahead among C2DEs - results that are totally at variance with normal telephone polling.
This YouGov internet polling is therefore wildly at variance with normal polling. The last YouGov poll could have been a rogue one but the results published today shows show what is involved is was not an individual rogue poll but systematic errors in methodology.
Conventional polling indicates YouGov has a record of significantly underestimating Ken Livingstone's vote in London. On the day before the last Mayoral election YouGov put put Ken Livingstone just two per cent ahead of the Conservative Steve Norris, leading to an Evening Standard headline of ‘It's Neck and Neck' even although Livingstone actually won by nearly 11 percent.
Ken Livingstone's campaign therefore states clearly that it does not accept this poll as accurate and, in light of its continuing errors, it does not regard YouGov polling on London as accurate.
The same comments do not apply to other pollsters.
What does show up in normal polling to an unusual degree is that what will decide this election is the turn-out - polls consistently show the higher the turnout the higher the support for Ken Livingstone.
If Boris Johnson were Mayor he would hit Londoners in the pocket with a £2 a week rise in bus fares and abolishing the policy that half of all new homes should be affordable for ordinary Londoners. He is an environmental vandal who opposed the Kyoto climate change treaty and promises to drop the £25 gas guzzler charge and his statements on ethnic minorities in London have been strongly objected to.
We believe the more they learn about Johnson's policies, the more Londoners are going to turn out to vote.