Now people get all huffy, "You can't take any notice of sub samples" "They don't make any sense on their own"
But I say they do eventually make up the big picture, if a supermarket announces that they had a million pounds worth of sales, then it would break it down to how much the butchery, bakery, clothes etc etc by department by department, it wouldn't say it doesn't matter here is the headline rate.
So I take the same theorem to sub samples, if I tweet a stat from a sub sample I put the caveat in but in itself it is a small piece of information which made up the whole number.
now yesterday YOUGOV released a poll LINK to DATA TABLES
You have two sections one which includes all respondents, so intention to vote for a party, or I won't vote, or I don't know who to vote for. What the pollster then does is strip out the don't knows and won't votes to give the headline figure. They do also weight samples to reflect the numbers of peoples from each area, age group, men/women to make it a fair reflection.
But!!!!!!
Do these figures, figure out? Here is a couple of examples. Now if I have made a school boy error, come and correct me, but this just looks plain wrong.
example 1
If there were a General Election tomorrow who would you vote for?
Scottish sub sample
161 respondents weighted to 158 so they had just about the right mix of Scottish respondents.
including the don't knows showing in percentages (in brackets how many respondents that would be aprox +1/-1)
SNP 41% (65)
CONSERVATIVES 16% (25)
LABOUR 14% (23)
LIBDEMS 6% (9)
UKIP 3% (5)
OTHERS 1% (1)
WON'T VOTE 5% (8)
DON'T KNOW 13% (22)
Now an issue I have is that the percentages are rounded, so that is why they total 99% I have used the weighted numbers and got mine to total 158 so mine is as close a guesstimate, but should be no more than 1 or 2 respondents out either way.
So for the headline rate you strip out the 30 Won't votes / Don't Knows leaving 128 using the numbers in brackets so the respondents who expressed an opinion to vote for that party you then work out the percentage. Seems sound enough maths to you? here goes
SNP 51%
CON 20% (actually 19.53 but used rounding)
LAB 18%
LD 7%
UKIP 4%
OTHERS 1%
and yes because of rounding comes to 101 but here is the rub on yougov when they have removed the WV/DK
SNP 50% Yep near enough spot on
CON 25% (Whoa even my rounding up didn't find 5%)
LAB 13% actually went down from the 14% inc DK/WV
LD 5% also went down when it should have gone up slightly
UKIP 3% well the numbers are low so nominal change expected
GREENS 4% now in the DK/WV these were actually under others at 1% so jumped to 4%
Now YOUGOV do put in a line of how they also add likelihood to vote but really! They think those who say they will vote Labour decrease when those who don't know or won't vote are taken out. I Don't think so.
Now pollsters have been damned in the past by not being able to find the quiet Conservative voter, but have pollsters now gone the other way and over compensated and might explain the high margins for the Conservatives?
Example 2
Lib Dem voters from 2015, a rare breed. But yougov found 112 of them and they thought that was too many so trimmed it down to a weighted number of a 103.
Now I am going to tackle how I think the Lib Dems are doing far better than the polls suggest in a future blog post. But lets just tackle this
in the DK/WV sample you got the numbers
A Lib Dem voter from 2015 says they will now vote
LD 49%
LAB 15%
UKIP 9%
CON 8%
OTHERS 2%
Won't Vote 1%
Don't Know 17%
won't bother with the brackets as the sample is weighted to 103 you might as well take it 1 for 1
take out the 18 won't vote Don't Knows leaves you with 85 take 1 off for the tolerance of 103 for a 100% you have 84
So 49 Lib Dems staying Lib Dem would be 58.333% Yougov when excluding the don't knows and won't votes have it at 51%
But here is the queerest number
You have 8 Lib Dems from 2015 who have now decided they want or intend to vote Conservative of the 84 left in the sample you can near as damn it say that's 10% correct. True. or just under if you are picky.
Anyways what percentage of Lib Dems from 2015 who now say they will vote Conservative do you think that YOUGOV have as a figure.
28%
Did feel like highlighting the number, making it bolder. Yep when YOUGOV excluded the Don't knows got rid of the mighty 1% of former lib dem voters who say they won't vote. They found nearly 3 out of 10 said they would vote Conservative. Yet before they could only find less than 1 in 10.
Not discounting the 9 Lib Dem council by election wins mostly from the Conservatives.
A link to those results you can find here >>> Council election results May 2016 - April 2017
So folks what do you think? As I don't know..................
