Is Rentoul’s polling commentry mis-leading?
Could over-stating Labour be a thing of the past?
John Rentoul in his latest column for the IoS makes some observations about the polls which I believe are out of date and could give a misleading view of what will happen at the next election.
He writes: “…Of course, the quirks of the system mean that the Conservative share of the vote has to be eight percentage points larger than Labourâ€™s to win a Commons majority. But I suspect that the opinion polls still overstate Labourâ€™s support. In the last four election campaigns, Labour has been overestimated by an average of five points. If that is still true, the Tories need be only three points ahead in the opinion polls to win. At the moment the average Tory lead is 15 points..”
Looking at what happened in 1992, 1997, 2001 and 2005 and drawing the conclusions he does ignores the polling revolution that we have seen which I believe will go a long way to dealing with the problem of Labour over-stating.
In 1992 when John Major was returned after securing a vote margin of 7% was not the industry’s finest hour. In fact the common term to describe it was as a “debacle”. Although there was the odd survey suggesting that the Tories might be ahead the overall picture was terrible.
But none of the pollsters operating then are still in business carrying out surveys in the same way. ICM became the pioneer with past vote weighting and “won” the 1997 polling race by quite a margin.
The following election, 2001, saw ICM coming out top again and there was a first survey by YouGov which had its own mechanism for dealing with Labour over-statement.
In the 2001- 2005 period the Times switched from MORI to the fledgling Populus and Gallup was replaced at the Telegraph with YouGov. MORI continued but changed to reporting its headline figures in terms of those certain to vote only. NOP used past vote weighting, got the final numbers precisely right, and was rewarded by being sacked by the Independent.
ComRes, now polling for both the Indy and the Indy on Sunday, switched to past vote weighting in March 2007 though it does it in a way that’s different from ICM and Populus. The other big change since the last election was the abandonment by MORI of face to face polling and the introduction of public sector worker weighting.
So all five pollsters that carry out monthly surveys now have special measures to deal with the issue and my guess is that they’ll all do pretty well next time. Certainly the idea of an average overstatement of 5%, as suggested by Rentoul, is going to be wrong.