h1

The post general election polling weightings

August 13th, 2010
Party ID/PV YouGov ID ICM past vote Populus PV ComRes PV
CON 28.5 25 26 24
LAB 32.5 21 21 20
LD 12 16 19 16
OTH 3 7 7 7
NONE/DK/Rfsd 24 31 28 31

How the four firms try to get “balanced samples”

I’ve put the above table together as a reference point for whenever the contentious subject of political weighting comes up.

YouGov, ICM, Populus and ComRes all seek to ensure that their samples are politically as well as demographically balanced. For this the three phone firms ask how interviewees voted at the general election and then each applies a bespoke formula to deal with the issue of false recall.

The online firm, YouGov uses a party ID weighting system which is harder to comprehend and has often been contentious. Certainly since the general election it has, as can be seen by the numbers, favoured the big two at the expense mostly supporters of SNP/PC/UKIP/GRN/BMP for whom it allocates a weighting of just 3%.

As a result YouGov is generally showing a big two aggregate bigger than other pollsters firms and an “others” aggregate that is much smaller.

One factor for the phone pollsters since the election has been that a greater proportion of of those interviewed “remember” voting Lib Dem than actually did so at the election. Accordingly some scaling back has to take place and this will affect future false recall calculations.

The other pollster that carries out regular surveys, Ipsos-MORI, does not apply a political weighting.

Mike Smithson