The point of no return? Earth-shaking TELEPHONE poll from Ipsos-Mori gives the SNP an astonishing 28% lead with just weeks to go

It's hard to believe, but Ipsos-Mori have almost exactly replicated the result of their full-scale poll of three months ago, which so many people assumed to be an extreme outlier.

Scottish voting intentions for the May 2015 UK general election (Ipsos-Mori, 12th-19th January) :

SNP 52% (n/c)
Labour 24% (+1)
Conservatives 12% (+2)
Greens 4% (-2)
Liberal Democrats 4% (-2)
UKIP 1% (-1)

In spite of the fact that Ipsos-Mori are one of the UK's leading pollsters (arguably only ICM have a better pedigree), there's a tendency to assume that these numbers "cannot possibly be right", simply because they would produce such an unprecedented political earthquake if they were repeated in the general election.  People are therefore speculatively casting around for reasons why Ipsos-Mori "must" be wrong.  One theory doing the rounds is that their failure to weight by recalled referendum vote is the culprit.  Well, Ipsos-Mori don't weight by any sort of recalled vote - not Holyrood, not Westminster, not referendum.  So it cuts both ways, doesn't it?  Their referendum polls showing much bigger No leads than most other firms also had to be seen in that light, and yet their final poll showing Yes 47%, No 53% was pretty close to the mark.  Indeed, it may have been absolutely bang on the money, because we have some proof of a very late on-the-day swing back to No.  So there's no concrete evidence from recent Scottish polls that a failure to weight by recalled vote leads to inaccuracy.

A much, much sillier theory comes courtesy of (predictably enough) Mike Smithson, who seems to think the 'problem' is that Ipsos-Mori's call centre is based in Edinburgh, and that much of their fieldwork is therefore presumably carried out by people with Scottish accents.  I mean, where to begin?  It's NORMAL for telephone polling to be carried out by people with accents that are familiar to respondents.  It's when you diverge from that normal practice that there is cause for concern - for example, you wouldn't have had a French call centre conducting polls for the 2004 US presidential election, in case respondents were too embarrassed to tell a French interviewer that they were planning to vote for George Bush.  Scottish respondents are more likely to be honest about their voting intention when speaking to a Scottish caller, because it avoids any slight stigma that might otherwise be attached to an admission of voting SNP.  I cannot see the remotest basis for Smithson's belief that anti-independence respondents would be more honest with a non-Scottish interviewer - how does he square that with Ipsos-Mori being one of the most No-friendly firms during the referendum, for example?

This poll also takes a sledgehammer to Nick Sparrow's eccentric (and some would say downright cynical) insinuation that artificial momentum is being generated for the SNP as a result of misleading online polling.  Leaving aside a Survation poll that was conducted just after people voted on 18th/19th September, we've now had two telephone polls since the referendum, and both have shown much larger SNP leads than any online poll.