Poll differences

[right]

Are the 2017 General Election polling differences purely down to turnout models?

[carlton43]

All communities have their own form of charlatanism that suits the mood of the period. The I Ching, Cartomancy, the Crystal Ball, Divination from the entrails of birds or the Reading of Tea Leaves. Each offer a tissue to grab at when the fevered mind wants reassurance and comfort or afflictions to rail at.

Polls would work well in a settled homogenous nation with fairly static views and two main political parties. When a nation has morphed into a multicultural soup with a large addition of oddballs, aliens and febrile fashion-following airheads flitting from factoid to rumour to tweet to gormless nostrum.............They merely serve to record the state of play in fevered imaginations at a given hour but barely representing what the idiots thought even the following day.

In short they are slightly less use than Fortune Telling column of the tabloids.

[John Chanin]

Jun 4, 2017 7:38:25 GMT right said:

Are the 2017 General Election polling differences purely down to turnout models?

No. There are still some phone surveys around, and there are still variations in what is weighted and how.  However the way turnout is being dealt with is the major difference between companies, and why there is a much larger spread than would normally be expected.   

As I have said elsewhere I think those companies downweighting turnout of young people to 2015 levels are onto a loser, but that doesn't necessarily mean that other companies have it right. 

For those of us with an interest in polling per se, rather than just its relevance to the election, this is fascinating.

[marksenior]

Jun 4, 2017 7:38:25 GMT right said:

Are the 2017 General Election polling differences purely down to turnout models?

The simple answer is mostly .

If you look at all the latest polls and the data tables ( ORB does not seem to be available yet ) there is a clear pattern .

Ipsos Mori being a telephone poll and having their own unique weighting methods are in group of 1 and cannot be assessed with the other pollsters .

The other pollsters all have similar data pointing to a Conservative lead of around 2 to 3 % . The latest Yougov poll is a slight outlier with a virtual dead heat between Con and Lab in its raw data .

The pollsters then get to work and do their own in house weighting and other adjustments . These lead to widely differing final headline VI figures with Comres and ICM at one extreme and Yougov Survation and Opinium at the other .

Clearly all the pollsters cannot be correct in their methodology , they can all be wrong , of course . I don't think we will know until 10 pm on Thursday which are right .

Finally all the pollster are still weighting down the Lib Dems by around 1 to 1.5 % . It may be that they are all wrong in doing this , it may be that 2015 was an exception and they should be adding 1 to 2 % as in previous elections . In the greater scheme of things this does not alter the main questions which is what is the true gap between Con and Lab .

[tonygreaves]

The French polls for the Presidential election were all remarkably stable and accurate (both rounds) and appear to be so for the Assembly elections next week. Why the difference compared with our lot?

[Sibboleth]

If you look at things in terms of swing rather than lead then the differences between the pollsters seem less drastic...

Survation +3.0
YouGov +1.5
Norstat +1.5
Ipsos/MORI +1.0
Opinium +0.5
SurveyMonkey +0.5
Panelbase -0.5
ORB -1.0
Kantar -1.5
ICM -2.0
ComRes -2.5

[Deleted]

So the overall guess is basically no swing. Oh well...