Should PB have its own polling average?

Should PB have its own polling average?

Poll/publication Date CON % LAB % LD %
YouGov/Sun 17/11/10 40 40 11
Ipsos-MORI/Reuters 14/11/10 36 39 14
ComRes/Independent 30/10/10 35 37 16
Angus Reid/ 28/10/10 35 37 15
Populus/Times 24/10/10 37 38 15
ICM/Guardian 24/10/10 39 36 16

As regulars will know I have always been strongly against polling averages. However I am coming to a view that we do need something which provides an indicator of the broad trend by taking the last survey from each of the firms that regularly carry out voting intention polls.

Anthony Wells, the senior YouGov executive who runs UK Polling Report, established such an average in the run-up to the 2010 general election and continues to publish it. Unfortunately the way it is set up means that the final figures are totally dominated by YouGov.

Only three pollsters feature in the latest average, see below, but five out of the seven polls used are YouGov. Factor in Anthony’s time weighting, older polls count for less, and 97% of the current published average is based on the online firm.

If an average is to have any meaning then it should be an amalgam of all the polls with only their last surveys being included.

So I am considering establishing what I’m provisionally calling the “The PB Polling Indicator” which will include the last polls from each of the firms that have reported in the previous month.

There will be a weighting so that the most recent count for more. Also the polls that did best at the general election will count form more than those further down the list. So ICM and Populus will have the top weightings here with Angus and BPIX the bottom ones.

These are only ideas at the moment and I would welcome input from the PB community.

This is what the latest UKPR average is based on:-

Pollster/Client End Date C L LD Lead Weighting
YouGov/Sun 2010-11-16 37 42 10 Lab +5 0.93
YouGov/Sun 2010-11-15 40 42 10 Lab +2 0.67
YouGov/Sunday Times 2010-11-12 39 41 10 Lab +2 0.39
YouGov/Sun 2010-11-11 40 40 10 Tied 0.19
YouGov/Sun 2010-11-10 42 37 11 Con +5 0.03
ComRes/Independent 2010-10-30 35 37 16 Lab +2 0.14
Angus Reid/ 2010-10-28 35 37 15 Lab +2 0.03

Mike Smithson

Comments are closed.