ICM

[Deleted]

ICM's methodology means had those respondants been with Yougov it would have been in the 20s. Thats firmly in the winning seats territory.

[Sibboleth]

Can't play those games.

[Deleted]

The 'jaw-dropping' line sounds like another example of pollsters' increasing tendency to pre-spin their own polls. As Anthony Wells pointed out on Twitter, when you're only doing one a month it's a bit harder to admit that you've just published a rogue.

[The Bishop]

Have just seen Anthony Wells explain that 4% BNP figure on UK Polling Report. Genuinely gobsmacking

[Deleted]

It is and you have to ask how reliable does it make the poll as the rest of the 3% has to be given to someone ??

[erlend]

Remember %s do not automatically add up to 100 due to rounding.

[Philip Davies]

May 14, 2013 13:12:49 GMT The Bishop said:

Have just seen Anthony Wells explain that 4% BNP figure on UK Polling Report. Genuinely gobsmacking

That exemplifies the flaw of estimating a whole population based on an n of 1 from a particular subgroup.

[Deleted]

New post: ICM/Guardian - CON 29, LAB 36, LD 12, UKIP 12

note the large drop in UKIP support as Anthony Wells says prob more methodological

[The Bishop]

So, "others" are at over 10% in this poll - up on last month. They aren't giving 4% to the BNP again, are they??

[Deleted]

Jun 11, 2013 13:02:02 GMT The Bishop said:

So, "others" are at over 10% in this poll - up on last month. They aren't giving 4% to the BNP again, are they??

They have sampled that old racist Welsh lady again. Last time the BNP were on 4 she was weighted to 28 people.

[The Bishop]

Jun 11, 2013 15:45:34 GMT joe said:

Jun 11, 2013 13:02:02 GMT The Bishop said:

So, "others" are at over 10% in this poll - up on last month. They aren't giving 4% to the BNP again, are they??

They have sampled that old racist Welsh lady again. Last time the BNP were on 4 she was weighted to 28 people.

It has now emerged that this latest ICM effort is giving 7% to the SNP - that is, 7% *across GB*, equivalent to about a 60% share in Scotland. Ho hum

[erlend]

Actually Scotland is less than 10% of the UK. So 7% must be more like 80% in Scotland.

[Deleted]

Jun 16, 2013 22:38:44 GMT erlend said:

Actually Scotland is less than 10% of the UK. So 7% must be more like 80% in Scotland.

Don't let the SNP press office know.. they love kooky sub-samples!

[Tony Otim]

Jun 16, 2013 22:38:44 GMT erlend said:

Actually Scotland is less than 10% of the UK. So 7% must be more like 80% in Scotland.

Isn't that assuming that the whole 7% actually comes from Scotland. I wouldn't be surprised to find that part of it didn't...

[erlend]

That's an example of bad polling. Do they give a prompted list of parties? If so they shouldn't prompt SNP/PC in England

[Arthur Figgis]

*insert obvious joke about Corby*

[marksenior]

All the SNP voters were in Scotland . The Scottish subsample was SNP 55% Con 18% Lab 16% LD 10% UKIP Zero . The sample size after removing Don't Knows etc was just 44 voters weighted up to 55 . Once again an important message to ignore subsamples in opinion polls .

[The Bishop]

The problem comes when such freakish subsamples - as with the largely mythical 4% for the BNP in last month's poll - affect the overall result.

[marksenior]

Jun 17, 2013 12:38:02 GMT The Bishop said:

The problem comes when such freakish subsamples - as with the largely mythical 4% for the BNP in last month's poll - affect the overall result.

This applies to every single opinion poll , if the sample had had 60% Labour voters in this subsample it would have still affected the overall result but have gone unremarked .

[The Bishop]

True, but the argument has always been that strange subsamples in the English regions tend to even out in a poll as a whole.

Because of the differing political setup in Scotland and (to a lesser degree) Wales, any freaky "result" has more of a distorting effect on the whole.