Ashcroft week poll

[Pete Whitehead]

Oh sure - I wasn't alleging bias in this case

[marksenior]

The problem with the 4 Lab/LD Ashcroft polls is that the sample people polled voted ( or say they voted ) overwhelmingly Labour in 2010 . Indeed even after weighting adjustments Labour outpolled the Lib Dems in the sample in 2 of the seats . Some of this may be down to false recall although as per Kellner's recent go at Survation some pollsters have more false recall than others or alternatively their samples are simply not representative .
It may or may not be significant that in Lord A's Con /LD polls there was hardly any false recall in 2010 LD voters . Having just checked , nor was their much false recall found in the Con/Lab marginal polls either .

[greenchristian]

Jul 2, 2014 8:18:31 GMT Tony Otim said:

Jul 2, 2014 7:58:11 GMT Pete Whitehead said:

Channel 4 news saw fit to make a song and dance about this poll yesterday with particular reference to Bradford East. Apparently the suggestion that Labour might gain Bradford East indicated that Miliband was on course for a majority. In fact the LD result there, in terms of their retention of vote share is surprisingly respectable

Is it that surprising? Their vote there has held up much much better at local level than in Manchester, Norwich or Brent, so it's pretty much in line with what I'd expect.

For constituency polling those figures look surprisingly credible, although I suspect the Green figure in Withington and UKIP in Norwich might be slightly on the high side.

From a Green perspective those look OK - Pavilion too close to call, running second in Norwich South.

Brighton is OK, but not good. Norwich is decent. I notice that he prompted for UKIP but not us in the opening question, which probably depresses the Green figures a little bit in both constituencies and explains the high UKIP figure in Norwich.

[thirdchill]

Would have been better to include some marginals other than the ones that are the most likely to fall to labour. Birmingham Yardley or Redcar would have been good to include. Cambridge would have interesting as well.

[Deleted]

Weekly Ashcroft polls really are a waste of time at the moment

Mike Smithson @msmithsonpb
This week's Ashcroft phone poll has
CON 27
LAB 34
LD 11
UKIP 15

[Deleted]

looks more sensible this week

Patrick Wintour @patrickwintour
Ashcroft National Poll: Con 32%, Lab 36%, Lib Dem 7%, UKIP 14%, Green 6%

[iain]

It looks similarly erratic.
These polls should clearly not be taken seriously.

[The Bishop]

Maybe not, but is the supposed "gold standard" of polling asking us to believe that UKIP support has dropped 7% in one month really much better?

[iain]

I am sceptical, but the guardian article said they dropped by a similar amount last year (6 IIRC)

[The Bishop]

It was probably wrong back then, too.

[iain]

Though on the Yougov polls, UKIP have also dropped off

[East Anglian Lefty]

Some drop-off is to be expected. Halving support isn't.

[The Bishop]

For the record, the latest random collection of numbers figures are Lab 35 Con 27 UKIP 17 LibDem 7 Green 7.

[Deleted]

Polling experts like Anthony Wells says the polls are fine, but no.. Vote UK has spoken!

[woollyliberal]

The difference between polls in the series is so great that either the margins of error are huge, Ashcroft is very unlucky with outliers or the methodology is a bit iffy. You can't really tell which it is from the published data.

Whatever it is, the numbers swing so wildly, that they have to be taken with quite a lot of salt. It spoils the benefit of a weekly poll because you can't trust an individual poll in the sequence. By the time you've averaged a few week's worth, you may as well look at a monthly poll.

[The Bishop]

Worth remembering here that Ashcroft *starts* with a sample of about 1k - once those are filtered to just the likely/certain voters (not sure which formula he uses exactly) the number is much less; the most recent poll which showed a Tory lead ended up with barely 400 people. When the numbers are that low, the MoE becomes higher and bigger swings are thus likely.

To a lesser degree, some other pollsters (notably the ICM and MORI phone surveys) have the same problem.

[Philip Davies]

Jul 23, 2014 12:37:39 GMT The Bishop said:

Worth remembering here that Ashcroft *starts* with a sample of about 1k - once those are filtered to just the likely/certain voters (not sure which formula he uses exactly) the number is much less; the most recent poll which showed a Tory lead ended up with barely 400 people. When the numbers are that low, the MoE becomes higher and bigger swings are thus likely.

To a lesser degree, some other pollsters (notably the ICM and MORI phone surveys) have the same problem.

YouGov, Opinium, Populus, Angus Reid and ComRes have ~2,000 as their usual sample size while Lord Ashcroft, ICM, Ipsos-MORI, Survation and TNS-BRMB have ~1,000 as theirs. That doesn't mean that a sample size of 2k should be given twice the weight of one of 1k. The differential is proportional to the square root of the sample sizes.

[woollyliberal]

Today's Ashcroft poll is in line with other pollsters.

Con 32
Lab 34
LD 9
UKIP 14
Green 6

[The Bishop]

Jul 28, 2014 15:11:15 GMT woollyliberal said:

Today's Ashcroft poll is in line with other pollsters

Make the most of it

[john07]

Jul 29, 2014 7:38:20 GMT ArmchairCritic said:

Jul 22, 2014 15:41:14 GMT @rlemkin said:

Polling experts like Anthony Wells says the polls are fine, but no.. Vote UK has spoken!

I think Anthony, who used to be a member of the old site, means that Ashcroft isn't doing anything absurdly wrong. But the fact that his polls are so at odds with other pollsters, not to mention his own earlier polls, means that it is not unreasonable to question them. Logic dictates that they cannot all be accurate.

If you include the margin of error, say a 95% confidence interval then they can all be right.