Calm before the storm?

I'm mildly taken aback by the lack of polls this evening, partly because this referendum is rather a big deal and we're less than three weeks away from it, and partly because we know for a fact that there have been YouGov and Panelbase polls in the field over the last week. From the way the Panelbase questions were described to me, that poll bore all the hallmarks of a Yes Scotland internal, and therefore the fact that it hasn't been passed to the Sunday Herald for publication may indicate that the Yes vote hasn't reached another all-time high (ie. 49% or higher). However, we certainly shouldn't jump to the conclusion that it's a bad poll, because even if it had merely equalled the previous record of 48% (which in truth would be a fantastic result after Panelbase's recent methodological change), it would probably have been withheld due to the danger of the No campaign saying : "Yes Scotland's own poll shows that they have failed to gain support since the debate, blah, blah, blah".

Equally surprising is the silence from YouGov, because we know they have been conducting at least two independence polls over recent days. One of them used question wording that is typical of not-for-publication Better Together internal polls, although the epic length led some people to wonder if it might be for an academic study. But the other poll had "for a newspaper" written all over it, and I fully expected to see it tonight. Perhaps it'll still turn up at some point over the next couple of days.

While we have this little lull, I thought I'd amuse myself by conducting an experiment. As you probably know, Panelbase and YouGov recently took the very sensible step of changing their methodology to correct for the fact that they have a disproportionately high number of English-born people in their samples. ICM, Survation and TNS-BMRB have thus far failed to follow suit, but there's something of a mystery (in my mind at least) about Ipsos-Mori's position. They do always ask for their respondents' country of birth, and they give that question the grand classification of "DEM 6", which implies that they weight by it - and yet their weighted numbers don't seem to be in line with the census results. So just out of interest, I did a rough reweighting of Ipsos-Mori's most recent referendum poll, using YouGov's census-derived target figures for country of birth. It had a slightly bigger effect than I expected - it increased the Yes vote (after Don't Knows are excluded) by a full 1%, and thus decreased the overall No lead by a full 2%. That may not be a dramatic transformation, but it's certainly not to be sniffed at.

One other thing that intrigues me is that Ipsos-Mori's small sample of respondents who were born outside the UK and Ireland is relatively favourable for Yes, and produces a No lead that is smaller than exists even among Scottish-born respondents (after the certainty to vote filter is applied). If memory serves me right, that finding is not untypical of previous Ipsos-Mori polls, and yet it's in complete contrast to both Panelbase and YouGov, both of whom show huge No leads among people born outside the UK. Of course, what makes Ipsos-Mori unusual in this campaign is that they're not reliant on volunteer online panels, and given that you could easily imagine that there might be special problems in recruiting non-native-English speakers to those panels, you'd be forgiven for wondering whether the people that Panelbase and YouGov do have are really representative. So that's yet another potential area of uncertainty.

* * *

UPDATE : A couple of people have mentioned on Twitter that they've been interviewed by Ipsos-Mori over the weekend - one of them tried to find out who the client was, but to no avail. Please God let it just be a coincidence that STV have another big debate coming up on Tuesday night. Are they never going to bloody learn that the purpose of a debate is to help people decide how to vote based on the arguments, not to tell them at the outset how they're already voting? (And through the distorting lens of one of the most No-friendly polling firms at that.)