How a device to deal with shy Tories gives Brown’s party a 4% boost

If it’s any consolation today’s ICM poll in the Guardian could have been much much worse for Blair-Brown but for the pollster’s final adjustment that was introduced after the 1992 General Election to deal with the phenomenon of “shy Tories”.

For instead of the CON 39%: LAB 29%: LD 22% headline figures that have been reported the shares were CON 41% – LAB 27%. For what ICM do is to use the responses to the “how did you vote last time” question to apportion part of the group in the survey who said they would be voting but refused or said they did not know when asked to name the party.

In recent months this has only changed the headline figure by 1-2% but with today’s poll it had a much greater impact. As Julian Glover writes in the Guardian “For years, it was Tories who could be counted on to mumble “no thanks” when the pollsters came calling but still turn out on election day. Now Labour voters may be the ones inclined to keep their politics to themselves…Labour will have its fingers crossed that its supporters are still out there, somewhere, even if these days they don’t like to admit it.”

With three polls on consecutive days showing very different pictures it appears as though the pollsters are all over the place.

In fact most of the differences between this week’s polls can be explained by timing and different methodological approaches.

The Mori poll on Monday was a week old so the field-work took place before the Tory tax announcement. Also, as we have said many times before, the pollster does not use past vote weighting.

The field-work for the Communication Research poll yesterday took place, like today’s ICM survey, from October 20-22nd. CR, like Mori does not use past vote weighting but has a very fierce “squeeze” question to try to get a response from those who say they will vote but won’t name a party. They are asked how they would vote if it was a legal requirement to do so. I think that the very high CR “others” figure is partly explained by this question.

Both ICM and CR include the names of the three main parties in their question with the SNP and PC being included for Scottish and Welsh respondents. This usually helps the Lib Dems but CR had them at 14% with ICM recording a 22% figure. My guess is that when we see the final data a part of this discrepancy will be put down to ICM past vote weighting approach.

The next poll should be YouGov for the Telegraph which is expected to be out on Friday. Who knows what the Internet pollster will bring?

Betting News. There’s been a big move to Gordon Brown in the Labour Leadership markets overnight. He’s tightened from 0.42/1 to 0.38/1 – so a winning £100 bet would produce £4 less this morning than yesterday.

Mike Smithson