BMG

[robert1]

My standard observation.

Given that the final poll by BMG before the election over-estimated Labour by 5% and under-estimated the Conservatives by 2%, have they adjusted their polling approach or are they as inaccurate as ever?

[The Bishop]

Why the assumption that the particular circumstances that obtained at the last GE will remain the case for all time - any more so than for the 2015, 2017 or 2019 elections?

Its this sort of constant change that makes polls so difficult, and given growing voter volatility and unpredictability that is likely to be the case for the foreseeable future. Maybe we need to accept that to a degree, rather than expecting perfection every time. Polling can still have value despite this.

[redtony]

2 POLLS one giving the Tory's a 1 point lead another giving Labour a 6 point lead with the field work on the same days one must be an outer but which one

[robert1]

It is precisely because circumstances vary that I made my comments. Unless they make changes the polling companies are working on the 'assumption that particular circumstances....will remain the case for all time'. They do make changes of weightings etc and often clearly identify them.

According to Prof Will Jennings of Southampton University the polls this year were the most inaccurate they have ever been and according to Martin Boon, of Deltapoll, they have shown a persistent under-estimation of right wing parties and over-estimation of their left wing counterparts.

Without making adjustments polls will be starting from a false base.

Polls are an exceptionally useful tool particularly when looking at ebbs and flows of opinion directionally. I just want polls to be as accurate as they possibly can be and therefore start from an accurate base. 

[John Chanin]

Nov 3, 2024 8:03:10 GMT robert1 said:

It is precisely because circumstances vary that I made my comments. Unless they make changes the polling companies are working on the 'assumption that particular circumstances....will remain the case for all time'. They do make changes of weightings etc and often clearly identify them.

According to Prof Will Jennings of Southampton University the polls this year were the most inaccurate they have ever been and according to Martin Boon, of Deltapoll, they have shown a persistent under-estimation of right wing parties and over-estimation of their left wing counterparts.

Without making adjustments polls will be starting from a false base.

Polls are an exceptionally useful tool particularly when looking at ebbs and flows of opinion directionally. I just want polls to be as accurate as they possibly can be and therefore start from an accurate base. 

My own view is that directionality on a consistent basis forms the most useful use of polls.  Two other points I have made before. Unlike Robert I don't want the polls to be an exercise in Mystic Meg prediction of what an election outcome might be.  I want them to report public opinion as it is polled, and then commentators can use it if they wish to make predictions.  And following on from this I want much more prominence given to the number and percentage of don't knows (including won't says).  By all means publish data (if you have it) on what the don't knows did at the last election, but don't try and use this to guess how they might behave when faced with a ballot paper.

[london(ex)tory]

x.com/towler/status/1862975020072366355?s=46

[Pete Whitehead]

Dec 1, 2024 8:33:47 GMT london(ex)tory said:

[The Bishop]

Which is actually an improvement for Labour on their previous poll.