YouGov's London polling

Various questions were asked last week about the age weighting on the YouGov poll which made it look as though there were far too many over 55s in the London sample. Putting on my YouGov hat, what actually went on under the bonnet is a bit more complicated than that, but suffice to say the weighting has been corrected. In an email to Mike Smithson Peter Kellner says he has re-run the voting intention questions for the previous YouGov Mayoral election polls with the new age weighting (and the ethnic minority weighting, which is being added to the next YouGov London poll - more on that later today) and found it made no difference to the final result.

This might at first sound counter-intuitive, ICM suggested a very strong contrast between voting intentions amongst white and non-white Londoners, and over 55s tend to be the most Tory group. The reason it doesn't make a big difference is that the poll is also weighted politically by party ID, which is an even stronger correlation. In other words, if you kept the sample the same, but with more ethnic minorities and less old people, you would get a sample with fewer Conservative identifiers and more Labour identifiers... except the sample actually has the same number of Conservative identifiers and Labour identifiers regardless, because to as weighted upon.