Quote:
Originally Posted by Chris
Indeed, however it's also worth nothing that ICM and Survation aren't very good at getting definitive answers out of people. For some reason they come back with relatively high levels of 'don't know' compared to YouGov. I'm not sure why this is, although I suspect there is something in the methodology that makes people more likely to give a firm answer with YouGov.
*If* there is any significance to the variation in 'don't know' between polls, it would tend to indicate that people who say they are undecided, when they do actually make a choice between Yes and No, are more likely to go for No.
|
Hopefully. It's odd to have polls so different from each other but aligning in different places. You have three converging around 40% and three converging around 45%. It will be different methodologies as they're too consistent for it to be anything else. Panelbase aligns, albeit a point or two higher, with ICM and Survation but for previously discussed reasons they're not massively trusted.
I wonder if it's sampling. How do you sample for a one-off event? I.E Do you go with the Holywood elections with might mean the sample is made up of a higher proportion of SNP voters than actually exists in the electorate or do you use some other criteria? I know I looked at Mori and they asked the question 'How would you vote in the next election?' and sampled the Independence Question on those findings (two different polls). Do you use the Scottish Election or the Westminster one? The first would oversample SNP the latter would undersample them! A combination? AGHHH.
It could be they're all messed up and Independence doesn't correlate well with political affiliation which would mean all the polls are suspect and we have no idea what is actually happening. Which is worrying. It would be interesting to see the Internal polling from the campaigns as they'll have a better idea of the questions to ask.