SNP still on course to defeat Jim Murphy and David Mundell in electrifying Ashcroft constituency polls

If you were hoping that further constituency polls from Ashcroft would cover interesting seats that haven't yet been polled (such as Glenrothes or Orkney & Shetland), it looks like you may be disappointed - instead he's decided to look at two key seats for a third time.

East Renfrewshire :

SNP 39% (+30)
Labour 36% (-15)
Conservatives 20% (-10)
Liberal Democrats 3% (-6)

(This would be an SNP gain from Labour.  Jim Murphy of Labour would lose his seat.)

Dumfriesshire, Clydesdale & Tweeddale :

SNP 42% (+31)
Conservatives 31% (-7)
Labour 17% (-12)
Liberal Democrats 5% (-15)

(This would be an SNP gain from the Conservatives.  David Mundell of the Conservatives would lose his seat.)

As before, the percentage changes listed above are from the 2010 result, rather than from the previous poll in each constituency.

Although Labour have gained ground on the SNP since the most recent poll in East Renfrewshire, it's important to stress that they've done this almost entirely by squeezing the Tory vote.  The SNP vote is only down 1%, which is a trivial change that can easily be explained by the margin of error.  Labour have failed to reverse the dramatic gains made by the SNP last time (they jumped from 33% to 40%).  If we take the headline numbers at face value - and for reasons I'll come to in a moment, it may be wrong to do so - this seat is clearly in "too close to call" territory, but the fact that the SNP still seemingly have their noses in front at such a late stage in the campaign is extraordinary enough.  The fieldwork was carried out over the last week and concluded only yesterday, so it's bang up to date.

In Dumfriesshire, the SNP only made modest gains in the second poll last month - they crept up from 34% to 36%, although that was sufficient to move them into a slender lead over the Tories.  Now they've made a big jump to 42%, and that can't be explained entirely by direct gains from the Tories, who have only slipped back 3%.  A troubling thought must surely be occurring to the unionist parties here - it may well be that it's the SNP who are the beneficiaries of tactical voting in this seat, and that Labour supporters are moving across specifically to maximise the chances of a Tory defeat.  OK, Dumfriesshire is an unusual constituency in that it's Tory-held, but it's scarcely implausible that there may also be many Labour supporters in other parts of the country who would be willing to vote SNP to turf Lib Dem incumbents out.

As always with Ashcroft polls, we can't exclude the possibility that the SNP vote is being significantly underestimated due to two methodological quirks.  Firstly, the headline numbers are weighted by 2010 vote recall, which is known to be unreliable due to people who switched to the SNP in 2011 getting the two elections mixed up.  In Dumfriesshire, respondents who claim to have voted SNP in 2010 have been downweighted from 114 to 73, and in East Renfrewshire they've been downweighted from 127 to 68.  Secondly, there is the spiral of silence adjustment, which artificially reallocates a portion of Don't Knows to the party they voted for in 2010.  In normal circumstances, that procedure improves accuracy, but these are not normal circumstances.

So it's not totally inconceivable that the SNP have a somewhat more comfortable lead in East Renfrewshire, and are already practically out of sight in Dumfriesshire.