Poll - 39.7% Yes, 41.4% No, 18.9% DK

There was a poll…

A rare poll, a rattling poll…

A poll down indyref-ee-oh

So, see that TNS-BMRB poll that gave such surprising results?  There were lots of comments about how it wasn’t weighted to the 2011 election results given that the question was asked.  One chap, a certain Professor John Curtice, said:

“Still, here perhaps we should enter another small word of caution. Unlike either YouGov or Panelbase, TNS do not weight their results so that their respondents’ reports of how they voted at the last election (the 2010 UK election in the case of YouGov, the 2011 Holyrood contest in Panelbase’s case) more or less match the actual outcome of that election. However TNS did actually ask their sample how they voted in 2011 – and found that rather fewer said they voted SNP than indicated they backed Labour. Such a divergence from what actually happened in 2011 is bound to raise questions about whether TNS’s sample adequately reflects the nation’s political balance.”

So I thought I’d do a wee calculation to weight it the way that Panelbase weights it and the result is Yes 39.7%, No 41.4%, Don’t Know 18.9%.

Before anyone feels the blue touch-paper burning their bum, this isn’t the way to weight a poll properly; it’s just a calculation for my own amusement.  Also, there are some parts of the calculation that I’d criticise if anyone else did this calculation (feel free to tell someone).  Here’s how I did the calculation with instructions for you to follow if you’re exceedingly bored –

1.      Open the tables for the poll - http://www.tns-bmrb.co.uk/assets-uploaded/documents/som-data-tables-august-2013_1378220251.pdf

2.      Go to Table 1 on Page 3

3.      Copy the weighted numbers against the parties in the end box (Vote in 2011) – ignore “Can’t remember” and “Did not vote” because those who can’t remember probably didn’t vote and people who didn’t vote in 2011 are least likely to vote in 2014 and the two categories together add up to 42% against the 2011 turnout of 50% (if I remember rightly).  That, and it made the calculation all watery (I don’t have the raw data, just the tables).

4.      Note that there is no vote percentage there for Green, SSP, Solidarity or any party not in the big three plus the Lib Dems – they must all be included in “Other”.

5.      Calculate what percentage those numbers translate to in terms of party share (SNP comes out at 36.5%, for example).

6.      Compare that to the actual 2011 result to get a ratio.

7.      Use the ration to change the numbers to what they would be had it been weighted to the 2011 result.

8.      Bob is no longer your Auntie but is now showing the contest neck and neck – hoopla!

9.      There are plenty caveats but you can work them out for yourself.

Here’s a wee table (just copied from my spreadsheet) -

	poll	%	Yes	No	DK	2011 const	Yes	No	DK
SNP	215	36.50	119	42	54	45.39	148	52	67
Lab	231	39.22	32	149	49	31.69	26	120	40
Con	58	9.85	3	48	7	13.91	11	68	10
LD	68	11.54	8	48	13	7.93	82	33	9
Oth	17	2.89	2	11	4	1.5	1	6	2
	589						268	279	128
							39.71026974	41.37094892	18.91878134

If you want even more fun, use the unweighted figures and you get even closer – 40.1% Yes, 41.1% No, 18.8% Don’t Know.  Here’s the table:

	poll	%	Yes	No	DK	2011 const	Yes	No	DK
SNP	230	39.05	119	42	54	45.39	138	49	63
Lab	249	42.28	32	149	49	31.69	24	112	37
Con	63	10.70	3	48	7	13.91	10	62	9
LD	69	11.71	8	48	13	7.93	81	32	9
Oth	18	3.06	2	11	4	1.5	1	5	2
	629						254	261	119
							40.07452711	41.11113968	18.81433321

Next week I’m making a submarine out of a sponge and a bar of soap.