Introducing PAPA – the PB All Pollsters’ Average

Introducing PAPA – the PB All Pollsters’ Average

PB All Pollsters’ Average Date CON % LAB % LD %
PAPA Current 24/11/10 37.6 38.4 13.4
PAPA October 2010 31/10/10 37.6 37.2 15.2
PAPA September 2010 30/09/10 37.8 37.0 14.8
PAPA August 2010 31/08/10 39.3 36.0 15.0
PAPA July 2010 31/07/10 40.0 36.7 15.0
PAPA June 2010 30/06/10 40.2 37.2 17.2

Labour edge upwards as the LDs dive

Here it is – the first PB All Pollsters’ Average (PAPA) as was trailed on a thread last week. Taking in a number of the points I have decided to make it dead simple. The figures at any stage will be based on an average of the latest poll of from each of the firms in the preceding month.

Unlike Anthony Wells’s average on UK Polling Report PAPA does not discriminate for or against any of the firms in any way.

Anthony has a time factor as well as an accuracy formula. At the same time he down-weights Ipsos-MORI because he does not approve of the firm’s methodology adding “if someone else wants to do an average that treats them the same go ahead.” That’s exactly what I’m doing.

With PAPA all firms are treated the same and all are included provided they have published a poll within the time period and satisfy the transparency requirements of the British Polling Council.

Only the last poll from each firm counts – unlike Anthony’s which can include several surveys from the same pollster. His current figures are based on seven polls – five of them from YouGov which, I suggest, rather defeats the objective.

The table shows retrospectively how PAPA has changed since the election with Labour now nudging ahead of the Tories for the first time and the Lib Dems at their lowest point.

Producing a polling average is very much a volte face for me which I acknowledge. I am introducing PAPA primarily because of what I see as defects in the Wells approach. If an average is to mean anything then all pollsters who have published a poll should be included.

Mike Smithson

Comments are closed.