SurveyMonkey subsample finds 54% of Scottish public want independence

I'm not sure if we're ever going to see datasets for last night's GB-wide SurveyMonkey poll for The Sun, but one person who clearly has seen more detailed results is Hugo Gye, a political correspondent for Sun Online.  He revealed an intriguing titbit last night...

"And one fairly surprising finding: in this survey, 54% of Scottish voters say they would support independence..."

Now, of course, that result is merely taken from the Scottish subsample of a GB-wide poll, but it's an unusually large subsample. The GB sample was around 6000, so presumably around 500 were in Scotland - roughly the sort of figure required to make a poll credible. The snag is that the subsample probably wasn't correctly weighted, but it's an interesting straw in the wind all the same. It arguably reinforces the message from recent full-scale Scottish polls, which is that support for independence is either holding steady or perhaps creeping up a little.

* * *

A number of people have written to me in recent days, asking about 'constituency polling' that YouGov appear to be conducting in several Scottish seats. It's actually not at all what it seems - people responding to standard GB-wide polls are just automatically being asked a supplementary question about how they will vote in their own constituency. I'm not an obsessive follower of the articles on YouGov's website, so I'm not sure if there's been a public explanation yet, but I can think of a few possibilities -

1) It might be for internal testing purposes, ie. to see after the election is over whether the headline or constituency question was more accurate. (Famously, and to the surprise of many, the constituency question proved the less accurate of the two in Ashcroft's polling for the 2015 election, especially in Lib Dem-held seats.)

2) It might be to help reassign respondents who initially say they will vote UKIP or Green in constituencies where those parties aren't standing.

3) It might be an attempt to build up constituency samples over the course of several weeks that are large enough to make a stab at predicting the election seat-by-seat.