But why are the pollsters being less transparent?
In what we think is Martin Baxter’s first prediction for the next General Election a big increase in Labour seats is projected from the calculation which involves applying the average swing in the latest polls to what happened on May 5th and applying them on a uniform national basis in each seat.
The Baxter calculator is one of the great tools for those who try to forecast and bet on elections but it has to be handled with care. In the run-up to May 5th it was predicting Labour majorities of well in excess of 100 and is very vulnerable to polling errors which continue to overstate Labour.
In his commentary on what happened in May Martin writes “Electoral Calculus predicted the winner correctly, but overestimated the majority. This was caused both by opinion poll error, regional swing and local factors.”
One problem at the moment is that the pollsters are being far less transparent about their methodology than they were in the run-up to the General Election.
Under the British Polling Council rules there is a requirement for the detailed data from each published survey to be made available in two working days but in some cases this is not happening. Two weeks ago today the July Guardian ICM poll was published. The main data set has yet to be published. I emailed the firm a week ago and have had no response.
Pollsters that use randomised unsolicited phone calls usually have to use complex maths to make sure that what they have found is representative. One assumption that’s made that leads to Labour over-stating is that more people are likely to “remember” voting Labour than actually did so. To judge each survey we need to look at the detailed figures to examine the scale of this.
Come on ICM – get your full data published