The polls
Over at Political Betting (Tories for London Mayor...) they're making a fuss about the latest ICM poll.Headlined, 'How ICM made a 17% drop in Labour support into just 1%' the point they make is this.
341 people told ICM that they had voted Labour last time. Yet only 291 said that they were planning to vote for the party on May 5th – a decline of about a sixth...YET the headline figure that was published was that Labour would get 41% - just one percent down on what the party got four years ago'
All this is based on ICM’s view that people do not tell the truth when asked how they voted four years ago. To deal with this they have a complex equation which weights towards Labour and away from the Lib Dems and the Tories.Are ICM right or are ICM wrong? We’ll know on May 6th.
This gives the impression that ICM are somehow favouring Labour. They're not though. In fact the reverse is true. In that Mirror survey, which ICM tried to conduct randomly, they found 291 people who said they planned to vote Labour and only 201 who said they planned to vote Conservative, a lead of 13%. They decided that this was unrepresentative, because when they asked people how they voted in 2001 it favoured Labour too much. So when they published their figures they said there were only 272 Labour voters and 214 Conservative ones, a lead of only 9%. ][Weighting for likelihood to vote brought this down further, but that's not important right now].
Political Betting are being misleading. The poll does not show a 17% fall in Labour support from the last election, as no-one believes its estimate of voting shares in last election. This is because of the phenomena of 'false recall', and it's a real phenomena.
You can argue it both ways. But you can't - as commentators on Political Betting have been -- argue that if ICM showed the 'real numbers' it would show a smaller Labour lead. It would show a larger one.
posted by Matthew @ 4/20/2005 07:49:00 PM
Comments: 0
