Monday, 9 September 2013

Fantasy Polling

I was ignoring the Ashcroft 'poll' on the basis that it was so schonky that it might have been invented as a tale by CS Lewis during his morning ablutions but some eejits took the thing at face value and never bothered looking at it properly.  Some of this may be due to a lack of staff in newsrooms these days, if we're to be charitable.  Some people have already explained some of the problems with it, including Severin Carrell at the Grauniad, Gary Dunion at Bright Green and even John Curtice at What Scotland Thinks.

The published 'results' actually came from three different polls.  Here are some things that are wrong with the large poll:

1.  There is no indication of who did the poll, it may have been a call centre in the Turks and Caicos islands and some random people doing doorstep interviews.  There's no indication that those conducting the polls have any idea what they're doing, whether they're members of the Polling Council, or what rigor they apply to any analysis they do.  The results are presented in a similar style to Populus or ComRes but those companies would publish the results if they'd carried out the research.

2.  The age demographics ('weighted' figures) in the large telephone are out, 18-24 years olds over-represented by 8.6%; those in the 25-34 bracket under-represented by 10.4%; in 35-44 under by 18.2%; in 45-54 over by 3.3%; in 55-64 over by 11.5%; and those who have passed the 65 year mark over-represented by 6.5%.  Given that we see in poll after poll that there are differences in independence and party support in teh different age groups, this is important.

3.  I had a quick look and can't find Scottish figures for socio-economic groups (feel free to look harder than me) but these are the 'weighted' figures given for them - AB; 2910: C1; 3100: C2; 2139: DE; 1677.  Households are classified on the SEG of the chief income earner and include people retired from those professions and are; A- Higher managerial, administrative, professional e.g. Chief executive, senior civil servant, surgeon; B - Intermediate managerial, administrative, professional e.g. bank manager, teacher; C1- Supervisory, clerical, junior managerial e.g. shop floor supervisor, bank clerk, sales person; C2 - Skilled manual workers e.g. electrician, carpenter; D- Semi-skilled and unskilled manual workers e.g. assembly line worker, refuse collector, messenger; E - Casual labourers, pensioners, unemployed e.g. pensioners without private pensions and anyone living on basic benefits.

The SEG ratio used by Yougov for Scotland in the DevoPlus poll recently used A, B, and C1 together and C2, D and E together for percentages of 47% in ABC1 and 53% in C2DE.  The large Ashcroft poll used a split of 61.2% ABC1 and 38.8% C2DE. 

Panelbase, in the SNP-commissioned poll, had AB at 20.7% while Ashcroft had that group at 29.6% - a whopping 9% bigger chunk of the population.  Panelbase for C1 was 27.3% while Ashcroft was 31.5%; C2 Panelbase 15%, Ashcroft 21.8%; DE Panelbase 37%, Ashcroft 17.1%.

Ashcroft has massively overcounted the most affluent members of Scottish society and massively undercounted the least affluent.  Given that we've seen a difference in independence and party support in the different SEGs, this is also important.

4.  There are references to notes in the results (letters beside the numbers) but the notes are missing.  Notes often give important information about the results and certainly put them in context - why would they be missing?  Here's a possible explanation.  As John Curtice noted, Ashcroft took two polls after the large one which showed independence in a better place than in the large one, but he used the figures from the earliest in this strange release.  That looks likes an attempt to mislead or misdirect.  Are the notes missing because they were at the end of the tables and the other questions and answers didn't suit the agenda?

Ashcroft isn't a political party, it's not as if some of the questions he asks are for campaign planning and need to be kept confidential to avoid giving opponents an advantage.  Unless, of course, being a Tory chap, he's giving that party some advantages in terms of polling information, in which case, I hope the donation is declared.  Perhaps Better Together is getting the advantage of the secret questions - although that particular organisation has promised not to take donations from abroad and I understand that the noble lord is a tax exile.  In any case, we don't know what the notes were and so we don't know whether they would have changed our impression of the results.

The other polls
I thought about doing the same for the other polls but, to be honest, I don't see the point.  I did do the SEGs for them, though, and they're below.  Large is the Ashcroft abomination examined above, VI is the Holyrood voting intention and FUP is the Follow-Up-Poll - the other two Ashcroft 'polls' re-released today to provide a headline on the cheap.  The Panelbase and YouGov polls are there for comparison - you'll note how close these two are on SEG.



Ashcroft polls
Panelbase

FUP
VI
Large
SNP
AB
32.3
28.3
29.6
20.7
C1
30.3
33.2
31.5
27.3
C2
20.1
12.6
21.8
15.0
DE
17.3
25.9
17.1
37.0






Ashcroft polls
YouGov

FUP
VI
Large
DevoPlus
ABC1
62.6
61.5
61.2
47.0
C2DE
37.4
38.5
38.8
53


 

1 comment:

david cashley said...

Strikes me that if the unionists want to manipulate the polls they really should be doing it in line with their "better together" strategy of scaring the bejesus out of anyone daft enough to listen. A poll that shows Yes support languishing means those dithering or feeling like using a wee protest vote will feel quite safe opting for a cheeky wee Yes. If their "strategy", and I use the term loosely, was consistent then they should in fact be claiming that the yes vote is streaks ahead and we all need to panic before the excavators start burrowing a crevice Eastward from the Salway. Now, where's my hard hat?