Panoramic Panelbase poll produces steady picture on independence

Panelbase have become the fourth pollster to conduct an independence poll to mark the run-up to the anniversary of the referendum - will ICM and Survation join the party over the coming days?  Today's poll removes any lingering doubt as to the explanation for TNS and Ipsos-Mori both showing a clear majority in favour of independence - it appears that was caused by different methodology, rather than by a change in public opinion over the summer.

Should Scotland be an independent country? (Panelbase)

Yes 47% (n/c)
No 53% (n/c)

I'm slightly troubled by John Curtice's reaction to this poll and yesterday's YouGov poll.  He quite fairly makes the point that they undermine the idea of a very recent swing to Yes, but you'd think an equally important point to make is that we're now left with no idea whether Yes or No are ahead right now, or indeed whether Yes or No have been ahead for much of the last year.  The only conclusion that is consistent with all of the polls we've seen is that the state of play has been relatively steady following a Yes surge (of uncertain scale) in the immediate aftermath of the referendum, but beyond that we have a clear split between Ipsos-Mori and TNS on the one hand, who are showing a Yes lead, and YouGov, Panelbase, Survation and ICM on the other hand, who are (mostly but not always) showing a slender No lead.

Instead, the subtext of what Curtice goes on to say seems to be that more credence should be given to the No lead shown by Panelbase and YouGov, because both firms weight by recalled referendum vote - something that TNS and Ipsos-Mori don't do.  This overlooks (or at least downplays the significance of) three vitally important points -

1) The failure to weight by recalled referendum vote is not the only factor that distinguishes TNS and Ipsos-Mori from the other firms, or even the most important one.  They are also the only pollsters that use a 'real world' data collection method, meaning they actually go out and find a fresh sample of respondents for every poll - in contrast to the other firms who are reliant on volunteer online polling panels.  That difference in itself could easily explain the divergence we are seeing - it could be that there is now a real world majority for independence that the online firms are simply unable to pick up.

2) TNS do weight by past vote recall, but they use the 2015 general election as their reference point, rather than the referendum.  It's not at all clear that switching to referendum weighting would be any better at correcting for the presence of slightly too many 'eager Nats' in the sample.

3) The Yes lead in the Ipsos-Mori poll was so big that it stretches credibility to suggest that weighting by recalled referendum vote would have magically produced an outright No lead in line with the online firms.  Not impossible, admittedly, but unlikely.

*  *  *

I was planning to update this blog's independence Poll of Polls last night for the first time since the referendum, because we now have more than enough recent polls to make it a meaningful exercise.  But I suddenly realised that there's a discrepancy in the reported figures from the YouGov poll.  The What Scotland Thinks website says that with Don't Knows included, the numbers are Yes 43%, No 49% - which, even taking account of rounding issues, is completely irreconcilable with the headline figures of Yes 48%, No 52% with Don't Knows excluded.  Hopefully the mystery will be solved when the datasets emerge.