The funny mathematics at ICM

The funny mathematics at ICM


    Have people really forgotten how they voted already

In their latest poll ICM had Labour with a 7% lead compared with the 3% that real voters gave the party in the election seven weeks earlier. Fine you might conclude – the ructions over the leadership are clearly causing problems. However from the detailed data from the survey, now out, it’s possible to draw a different conclusion.

  • More people said they would vote Tory now than told the interviewers that they voted for the party at the General Election
  • Fewer people said they would vote Labour than told the pollster had voted for the party on May 5th
  • Fewer people said they would vote Lib Dem than said they had gone with the party at the General Election.
  • On May 5th, as we all know, the vote split in Great Britain was LAB 36.2%: CON 33.2%: LD 22.7%. Yet from my mathematics it seems that ICM, was working on a split of LAB 38.1%: CON 31.1%: LD 22.25%. when doing the calculations for its latest survey.

    The challenge for any phone pollster is that it takes 5-6,000 unsolicited randomised phone calls to find 1,000 people who are ready to be be surveyed and those that do invariably are more pro-Labour than the population as a whole. This problem is not new – it’s just that those who respond to such calls tend to be this way.

    So to ensure that their samples are representative ICM and other pollsters do some complex mathematics – they weight the data in accordance with what happened at the election and then make an assumption that a proportion have forgotten what they did.

      And for reasons that I cannot fathom ICM seem to think that even after this short time there’s a big difference between the number of people saying they voted Labour and those that actually did do.

    The notion of interviewees overstating past vote intention for Labour might have been correct in the aftermath of the 1997 and 2001 landslides but I question whether it’s right in the current climate. It will be interesting to see how the other pollsters deal with this.

    Before May 5th I wrote several articles about the phone pollsters’ number crunching and placed a four figure sell spread bet on the Labour vote share which I said was inflated because the markets were believing the pollsters. At the time I said I was “putting my money where my mouth was” and was rewarded with big profits.

    Mike Smithson

    Comments are closed.