Panelbase have got into the rather refreshing habit of releasing the datasets for their Sunday polls on the Sunday itself. With ICM it's not unusual to have to wait several days (although to be fair they were a bit quicker last time). First things first - here are what the Panelbase numbers look like if you only round them to one decimal place...
Should Scotland be an independent country?
Yes 48.3% (+2.1)
No 51.7% (-2.1)
Unless we count the Panelbase poll from last summer that used an unusual question sequence, that's the narrowest No lead shown by any pollster in the campaign so far. It even beats the famous Easter Sunday poll from ICM, which had Yes at 48.0% and No at 52.0%. It's also better for Yes than this week's Survation poll by a slightly wider margin than the rounded numbers would make you think - Survation had Yes at 46.6% and No at 53.4%.
And unlike Survation, there's nothing that leaps out at me in the Panelbase datasets that would raise any question marks over the swing to Yes - there hasn't been any radical upweighting of a particularly Yes-friendly group. As is the case in most polls, the No lead is slightly wider on the unweighted numbers, but that mainly happens because pollsters start with too many older and "ABC1" respondents, who have to be scaled back. In this case there were also slightly too many women. But there was no upweighting of respondents who recalled voting SNP in 2011, whereas Labour voters were actually upweighted quite significantly.
Evidence for the remarkable strength of the Yes campaign among men continues to build - in this poll Yes lead by 55.8% to 44.2% with male respondents, after undecideds are excluded. I understand the importance that is attached to closing the gender gap, but given the significance of social proof, I do wonder if there might still be scope for a substantial further swing to Yes among men. It certainly seems that in some male-dominated social circles (especially working-class ones) undecided voters are now more likely to encounter Yes voices than No voices, which must surely have an effect. In other words, I don't find it totally implausible that Yes might win with a gender gap that stays the same or widens further.
One interesting detail, which may just be coincidence, is that there were exactly the same number of respondents - 397 - who said 'Yes' to the referendum question, and who said they "very likely" to vote for independence if they thought there would be another Tory or Tory-led government after next year's UK general election. An additional 61 respondents said they were "quite likely" to vote for independence in those circumstances. So it almost looks as if the 397 respondents are the 'baseload' figure for Yes, with there being an additional potential source of support among anti-Tory voters who currently say they are undecided or voting No. Admittedly, though, some of the 397 drift away (albeit probably only to "quite likely to vote for independence") if they assume that Labour will win the general election.