Would Tory leads be as big if older methodologies were still used?
One factor that has completely changed the backcloth against which UK political life operates has been the almost total overhaul of the polling industry since the 2001 election to deal with what was the systemic problem of Labour over-statement.
It is my contention that if the 2001 line-up of pollsters and polling methods were still in place then the current Labour poll deficits would be on a much smaller scale and there would not be quite the same crisis facing Brown.
For the whole morale of a party and the internal confidence in the leadership are governed to quite a large degree by the polling numbers. Yes by elections and local council results also provide pointers but it’s the regular grind of the monthly polling that really sets the scene. And if those numbers are almost always over-stating Labour, as was happening, then that can have a huge impact.
After 2001 we saw the Daily Telegraph moving to then new internet pollster, YouGov, which surveys only those on its own polling panel on which it has a mass of data which allows it to weight by party identifiers. Then the Times started their relationship with another newcomer, Populus, which followed the ICM lead from the mid-1990s in dealing with sample bias by weighting according to how respondents said they voted at the previous election.
Even in the current parliament, since 2005, the revolution has continued. In March 2007 the Independent’s pollster, ComRes, introduced past vote weighting which had a dramatic impact its figures. Then in the past few weeks the two most long-standing UK pollsters, Ipsos-MORI and ICM have made refinements that have the effect of improving the Conservative position against Labour.
At the heart of the problem was that whichever interview methodology you used Labour supporters were almost always over-represented in samples. Quite why this was not always been clear but as part of their post London Mayoral election polling review MORI found that there was a disproportionate number public sector workers in samples. New weightings have been introduced to deal with this which will be to the detriment of the Labour position.
It’s when the pollsters get a result wrong, like at the 1992 general election or this year’s London Mayoral race, that polling firms review their approaches.
Even ICM, the pioneer of modern telephone polling, has made changes following Boris’s victory so that more account is taken of what those interviewed said they did at the last election.
In my betting I’ve always followed “the golden rule” – that the most accurate survey is the one with Labour in the least favourable position. Maybe the polling revolution will have changed that and that it is the Conservatives who are being over-stated? We shall have to wait and see but I don’t think we have quite got to that stage yet.