I've been having a look through the newly-released Survation and TNS-BMRB datasets, and here are a few bullet-points -
* Survation have joined Panelbase in showing that Yes have an outright lead among male voters. TNS-BMRB still show a narrow No lead, but even so, we now have two out of seven BPC pollsters suggesting that one gender is already backing independence. TNS-BMRB actually show that the No vote is almost identical among men and women, suggesting that the real difference between the genders is that there are fewer women in the Yes column, and more women in the undecided column. Although Survation suggest that the No vote is a few points higher among female voters, there is again the suggestion that women are far more likely to be undecided at this stage.
* Remember how the one caveat that was raised about the sensational ICM poll a few weeks ago concerned the radical upweighting of the responses of young voters, who happened to be the most pro-independence part of the overall sample? Well, almost exactly the opposite has happened in the Survation poll, with the very small sample of 16-24 year old voters proving to be the least favourable to independence other than over-65s, and with that sample being massively upweighted from 59 actual people to 138 'virtual' people. So if sampling problems have in any way distorted the results of this poll, it's more likely to have led to a slight understatement of the Yes vote than of the No vote - in other words it's conceivable that Yes should be a touch higher than 45%.
* Although Survation have broadly come into line with Panelbase and ICM in respect of the weighting procedure, they haven't done so in respect of the question, which still asks how people would vote if the referendum was held "today". It's widely theorised that this wording has the effect of artificially increasing the No vote, although admittedly there is no hard evidence of that.
* Survation show a significant increase in the SNP lead over Labour in Holyrood voting intentions, although the apparent shift in opinion may be largely illusory due to the introduction of a new methodology.
Scottish Parliament constituency ballot :
SNP 44% (+6)
Labour 31% (-5)
Conservatives 13% (+1)
Liberal Democrats 6% (-3)
Scottish Parliament regional list ballot :
Liberal Democrats 8%
So a gap has opened up between ICM, Panelbase and Survation in one camp who are all showing a substantial SNP lead, and YouGov and Ipsos-Mori in the other camp who are suggesting that the race is very tight. The question of who is getting it right is vitally important, because the first camp also happens to be showing the most Yes-friendly referendum figures at the moment. One doubt that will be lingering in some people's minds is this - is it realistic to think that the SNP have such a commanding nationwide lead when there was a significant pro-Labour swing in the Cowdenbeath by-election last month? Answer : Yes, it's perfectly plausible, because by-elections often exist in their own little bubbles (Glenrothes 2008 being one classic example). But it's certainly a reasonable doubt to raise, and we may have to wait until the European elections to get a stronger clue.