The Unionist right wing press seem to be getting themselves over-excited about an #indyref2 poll from the Sun. They claim it shows a big drop in support for independence. It doesn’t. Here’s a 🧵 as to why.
Firstly, the sample is smaller than the normal Scotland wide polling sample of about 1,000. This means that the margin of error will be higher (4.4 points compared with 3), and the actual % figures quoted will be less reliable.
https://twitter.com/whatscotsthink/status/1571775446517223424
Also as @WhatScotsThink note, there is no previous #indyref2 poll from this company to compare with. This is important because each pollster has its own approach to weighting & determining likelihood to vote. This means comparing polls from different pollsters is problematic.
Some pollsters don’t consider likelihood to vote in their figures, some use only those 8 to 10 certain, others use 6 to 10, others use a weighting system based on the likelihood scale.
Setting all this aside, most pollsters weight their samples based on the #indyref 2014 vote. In my view this is flawed, because of the demographic changes over the past 8 years. Among those who voted in 2014, the current split would be closer to 47/53.
So, weighting based on 45/55 inflates the No vote sample. And the voting intention figures in polls is very closely linked to the level of #indyref vote share in the weighted sample. The exception is ipsos who weight only on demographics. Properly weighted polls put Yes ahead.
But even just taking the findings at face value, the question asked about ‘if there was a referendum tomorrow’. That is in the middle of the mourning period and a couple of days before the Queen’s funeral. The referendum is scheduled for next October.
Even those sympathetic to independence would surely think this to be a little insensitive. But despite this, only 48% stated they would vote No. The same proportion as in the Panelbase poll last month. The only difference is a shift from Yes to Don’t Know.
The Don’t know proportion, at 10%, is on the high side. I don’t know how Deltapoll deal with likelihood to vote. But either they are taking a fairly liberal approach, or the timing of the question has inflated the DK figure.
So, in summary:
High margin of error
No comparable figures
Unknown weighting approach
Unknown approach to turnout
Poor framing of question
No increase in No vote
If it shows anything it’s that people don’t want a referendum in the days before the Queen’s funeral. #indyref2
High margin of error
No comparable figures
Unknown weighting approach
Unknown approach to turnout
Poor framing of question
No increase in No vote
If it shows anything it’s that people don’t want a referendum in the days before the Queen’s funeral. #indyref2
https://twitter.com/heraldscotland/status/1571451435853676545
• • •
Missing some Tweet in this thread? You can try to
force a refresh
