There have been a number of encouraging polls for the pro-independence campaign this year, but there were two in particular that stood out - the ICM poll in January showing the No lead fall from 17 points to 7, and the Survation poll last month showing the No lead drop from 20 points to 9. The ICM poll was followed up by a poll from the same company showing a reversion to the mean (although a look at the actual unrounded figures later revealed that Yes had retained most of its gains - in the January poll the No lead had fallen from 17 points to 6.4, and in February it had only recovered to 11.3). It must be a matter of great concern to Blair McDougall and the anti-independence campaign that the same thing has failed to happen with Survation - the new poll today suggests that the No lead hasn't recovered at all from its slump to nine points, which dramatically reduces the chances that last month's figures were artificially high for Yes due to normal sampling variation.
In the referendum, voters will be asked, “Should Scotland be an independent country". If this referendum were held today, do you think you would vote "Yes” or "No"?
Yes 39% (+1)
No 48% (+1)
The headline figure for Yes is the highest that any pollster has shown thus far in referendum year. In terms of the trend, this poll is strikingly similar to the most recent YouGov poll, which put both Yes and No up by one point, taking Yes to its highest level of support of the campaign to date. That's a pattern that will typically favour Yes more, because a static headline lead as the number of undecideds decline is generally an indication that the underlying No lead is slipping back. On this particular occasion, the movement hasn't been sufficient to tip the balance on the rounded figures, with the No lead remaining at 10 points when Don't Knows are excluded (although on the unrounded figures the Yes vote has increased from 44.8% to 45.2%, meaning that the No lead has in reality declined by just under 1%).
Yes 45% (n/c)
No 55% (n/c)
These numbers are particularly timely in that they help to put last week's outlying Ipsos-Mori poll in its proper perspective. STV's spin on that poll, suggesting that it proved George Osborne's bullying tactics over the currency had "worked", has been shown up as the utter garbage that observers of the polling landscape always knew that it was. The true picture has been one of significant gains for the Yes campaign between September of last year and January of this year (with the No lead on this blog's Poll of Polls slipping from approximately 20% to 14%), followed by consolidation in the period to March. Osborne's intervention has failed to make even the slightest dent in that consolidation, and as John Curtice points out today, that remains the case even though voters have now had plenty of time to digest the significance of the developments on the currency, and are no longer simply lashing out emotionally against Westminster arrogance.
Indeed, Ipsos-Mori themselves have broadly corroborated the trend - they showed the No lead slumping in the wake of the publication of the White Paper by some 5%, of which the unrounded figures of last week's poll suggest that barely over 1% has been clawed back (something which is highly likely to be due to sampling variation). That good news story for Yes was masked by the fact that Ipsos-Mori's methodology continues to generate a much, much higher lead for No than other pollsters.
Which brings me neatly to another crucial point about how we should be interpreting the online polls we've seen since January. While the trend is obviously important, what we should also be looking out for in each new poll are clues that will help us resolve the mystery of the true state of play. It's all very well observing that Yes have either closed the gap or failed to close the gap, but any such observation self-evidently has a very different meaning depending on how big or small we think the gap was in the first place. We'd obviously be considerably more optimistic about the Yes campaign's chances of overturning a ten-point deficit over the course of the official campaign period than we would be about their chances of overcoming the 25-point deficit that Ipsos-Mori are uniquely suggesting. And that's the thing - the polls from ICM and Survation this year have essentially tipped the balance, and made it look much more likely that the true gap is closer to ten points. Until just a few short weeks ago, Panelbase were very much the outliers in showing a modest gap, and it was easy to sneeringly dismiss their figures (as so many unionist commentators did) due to the fact that they were less experienced than No-friendly pollsters such as YouGov. But now two other online pollsters have joined them in showing near enough identical numbers - and one of those just happens to be the UK's so-called "gold standard" pollster. ICM's last two polls have shown No leads of 7% and 12% (or closer to 6% and 11% if you look at the unrounded data), while Survation's last two polls have both shown a No lead of 9%. Panelbase's figures now look positively mainstream, while Ipsos-Mori have increasingly stuck out like a sore thumb on the extreme end of the spectrum.
YouGov remain the odd ones out among the online pollsters in that they continue to show a somewhat bigger No lead, but even they have drifted away from Ipsos-Mori and are now in a more middling position. The fact that they have failed to fully converge with their online cousins ICM, Survation and Panelbase may be at least partly due to their eccentric weighting procedure. Although they finally gave in to sanity back in September and started to weight by recalled Holyrood vote rather than by Westminster party identification, they now split voters who recall voting SNP in 2011 into two distinct categories - "SNP" and "SNP, but Labour in 2010". That may sound fair enough, given that it's reasonable to assume that there may be significant differences between those two groups. But the snag is that the "SNP, but Labour in 2010" group clearly isn't intended to include (as you'd be forgiven for expecting) all respondents who voted Labour in 2010 but switched to the SNP in 2011. We know that because the SNP are weighted to have an advantage over Labour even without that group. So on what basis are some Labour-to-SNP switchers being assigned to the regular "SNP" group, while others are assigned more logically to the "SNP, but Labour in 2010" category? To the best of my knowledge, YouGov have failed to provide any explanation, and the fact that they always seem to significantly upweight the responses of the arbitrarily-defined "SNP, but Labour in 2010" group is bound to raise a few suspicions that something fishy is going on.
It's most likely to be something to do with our old friend False 2010 Recall. My guess is that even YouGov now accept that a large number of people who say they voted SNP in 2010 did not actually do so, and are taking that into account by effectively weighting to an imaginary 2010 result in which the SNP did somewhat better. The problem is that they're probably not making a big enough adjustment, as evidenced by the fact that they still can't find enough Labour-to-SNP switchers in their raw figures, and have to persistently upweight the ones that they do have (who may well not be representative in terms of referendum voting intention - again, we don't know for sure because YouGov don't provide those figures).
To return to today's Survation poll, there are also voting intention figures for Holyrood and Westminster. The good news is that the Holyrood constituency figures are broadly in line with Ipsos-Mori, which suggests that Survation's better figures for Yes can't be explained away by them having a "Nat-heavy sample". As I've suggested before, one of the few potential explanations I can think of is that Ipsos-Mori are for some reason interviewing different types of SNP and Labour voters, who are perhaps more conservative with a small 'c'.
Holyrood constituency ballot :
SNP 45% (+1)
Labour 34% (+3)
Conservatives 13% (n/c)
Liberal Democrats 5% (-1)
Holyrood regional list ballot :
SNP 40% (-1)
Labour 28% (n/c)
Conservatives 11% (-2)
Greens 8% (+5)
Liberal Democrats 7% (-1)
UKIP 5% (+2)
Westminster general election :
SNP 38% (n/c)
Labour 36% (+3)
Conservatives 15% (-2)
Liberal Democrats 5% (-1)
It's interesting to ponder whether Labour should be concerned about being behind even on Westminster voting intention. In many ways, these figures mirror what was going on at this stage in the electoral cycle (or possibly slightly earlier) last time round, and of course Labour were able to recover in time for the general election with the assistance of the rigged TV leaders' debates which wholly excluded the SNP. However, much will depend on the outcome of the referendum - if it's a Yes I would expect voters to rally behind the SNP in unprecedented numbers to strengthen Scotland's negotiating hand.
Survation have changed their methodology on the Holyrood regional list question to make it much easier for respondents to express a preference for one of the smaller parties, so the changes there may be largely illusory.
* * *
SCOT GOES POP POLL OF POLLS
For obvious reasons there's not a huge amount of change in this update of the Poll of Polls, although the No lead has fallen a smidgeon when Don't Knows are excluded. On the headline numbers, Yes has nudged back to within 0.2% of its high watermark in the campaign to date.
MEAN AVERAGE (not excluding Don't Knows) :
Yes 34.7% (+0.1)
No 49.0% (+0.1)
MEAN AVERAGE (excluding Don't Knows) :
Yes 41.5% (+0.1)
No 58.5% (-0.1)
MEDIAN AVERAGE (excluding Don't Knows) :
Yes 42.0% (n/c)
No 58.0% (n/c)
(The Poll of Polls is based on a rolling average of the most recent poll from each of the pollsters that have been active in the referendum campaign, and that adhere to British Polling Council rules. At present, there are seven - Angus Reid, YouGov, TNS-BMRB, Survation, Panelbase, Ipsos-Mori and ICM. Whenever a new poll is published, it replaces the last poll from the same company in the sample. Changes in the Poll of Polls are generally glacial in nature due to the fact that only a small portion of the sample is updated each time.)
As hinted at earlier, online polling is producing distinctive (and on the whole more Yes-friendly) results in this campaign. That could conceivably be because online pollsters use volunteer panels who may not be entirely representative of the general population, although weighting procedures are in place to deal with that issue. The explanation is much more likely to be that respondents feel less inhibited when faced with a computer screen, thus overcoming the problem of 'shy Yes syndrome' - and if so, it's possible that online polling is producing the most accurate results. So I thought it might be interesting to have a look at the average for online pollsters only. To make it fairer, I'll exclude Angus Reid (who haven't reported since last summer), and I'll include YouGov in spite of concerns over their No-friendly methodology.
MEAN AVERAGE OF ONLINE POLLSTERS (not excluding Don't Knows) :
MEAN AVERAGE OF ONLINE POLLSTERS (excluding Don't Knows) :
MEDIAN AVERAGE OF ONLINE POLLSTERS (excluding Don't Knows) :
Not a gargantuan difference, but it certainly does look healthier for Yes on that measure.
* * *
UPDATE : I've altered an earlier part of this post to reflect the fact that Survation are actually showing that the No lead has declined by 0.8% on the unrounded figures, after Don't Knows are excluded. I was looking at the wrong table earlier - the correct one is filtered by likelihood to vote.
Yes 45.2% (+0.4)
No 54.8% (-0.4)
On the headline numbers including Don't Knows, the No lead has fallen from 8.9% to 8.3%.
Yes 39.3% (+1.6)
No 47.6% (+1.0)
* * *
UPDATE II : For once, the prize for daftest comment on this poll goes not to Blair McDougall, but to Michael Marra of the 'Five Million Questions' project -
"The hill for the Yes campaign to climb remains considerable. Everyone who is currently undecided has to make the leap to independence. Winning from here for Alex Salmond would be an astonishing achievement."
Which would be a truly fabulous point if it wasn't for the inconvenient fact that we have plentiful polling evidence to suggest that a substantial minority of people who currently say they will vote either Yes or No are open to changing their minds before polling day. Indeed the clue is in the question that Survation ask - "do you think you would vote Yes or No?" The idea that the No campaign somehow have 47.6% of the vote already "in the bank", and that the hopes of Yes rest totally on winning over almost all of the Don't Knows, is ludicrous beyond words.
If the Yes campaign win, they'll do it by a combination of winning over the majority (but not necessarily an overwhelming majority) of Don't Knows, and peeling off soft No voters. I would have thought that point was self-evident, but apparently it does need to be pointed out to some. Not quite such a mountain to climb when you look at it that way, is it?