Were there enough 2005 Lib Dems in the MORI sample?
My apologies for the third post in succession on the Ipsos-MORI poll but it produced numbers that were so startling that it deserves further scrutiny.
Yesterday it was all doom and gloom for the Lib Dems over the MORI 12% yet a close study of the data raises the question of whether were there enough 2005 Lib Dem supporters in the sample?
It will be recalled that on May 5th 2005 the party led then by Charles Kennedy secured 22.7% of the vote. Labour by comparison chalked up 36.2%. So for every two Lib Dem voters last time there were just over three Labour voters. You would have thought then when respondents were asked how they voted last time that MORI would have found broadly similar proportions. They didn’t.
When the field-work was done over the weekend they found (see the panel above) about seven 2005 Labour voters for every two Lib Dems which, surely raises questions over whether the sample was politically representative.
This would not have mattered too much if this had been an ICM or Populus poll. A weighting adjustment based on the past vote responses would have been attached so that the intentions of those who voted Lib Dem last time would have been scaled up and those who voted Labour would have been scaled back.
It’s hard to make calculations because of demographic weightings but my back of an envelope sums suggest that with the same data the other telephone pollsters would have put Nick Clegg’s party at 16-17% almost all of the difference coming from Labour.
As I argued here yesterday getting the Lib Dem share right will be critical for pollsters as we head into the general election.
This touches a raw nerve within the polling industry because Ipsos-MORI is the only remaining national pollster that does not weight by past vote or party identifier. As a result UK Polling Report has taken special measures to marginalise MORI in the new polling average that it has just introduced.
Statement of disclosure. I have been a member of the Lib Dems since its foundation and have been a parliamentary candidate and a councillor. Would I have been raising this issue if MORI had found a much higher proportion of the sample saying they voted for the party last time than actually did so? I like to think that I would.