I was a touch bemused by the supposedly "relieved" reaction of some of Jim Murphy's media cheerleaders to the news that the Ashcroft constituency poll of East Renfrewshire put their hero just 1% ahead of his SNP challenger Kirsten Oswald. The reality is that, even if you take that result at face value, it's a "statistical tie" - the standard 3% margin of error means that it's impossible to know whether Murphy or Oswald is in the lead. But it's even worse than that for Murphy, because the poll's methodology was almost bending over backwards to produce a result that would flatter him...
1) Because weighting by 2010 vote recall was foolishly used, the SNP vote was massively downweighted, and the Labour vote was significantly upweighted. For the headline results, the 120 people who recalled voting SNP in 2010 were scaled down to count as just 67. A proportion of that adjustment may have been warranted if SNP supporters are simply keener to speak to telephone pollsters, but it seems likely that for the most part, people who are mixed up between how they voted in 2010 and 2011 were wrongly downweighted.
2) The results of the second voting intention question (asking respondents to think about the candidates in their own constituency) were used for the headline figures, rather than the results of the standard voting intention question, which favoured the SNP more. There's nothing wrong with asking voters to think about local circumstances (in fact it's a very good idea), but that should really be part of the wording of a single voting intention question, so that people don't feel that they "ought" to give a different response at the second time of asking.
3) The Greens received 3% of the vote in the poll, and yet someone who attended the Greens' spring conference emailed me today to say that they'd heard the party won't be standing in East Renfrewshire. If true, it seems plausible that those votes will break more for the SNP than for Labour. Admittedly, it doesn't always work that way when lower preference votes are transferred in local council elections, but it's murderously hard to see the average Green supporter being a fan of Jim Murphy.
4) Even with the help of all these factors which boosted Murphy's position in the poll, he was STILL fractionally behind the SNP. The only reason he was eventually reported as being slightly ahead was the "spiral of silence adjustment" which assigns a large portion of undecided voters to the party they voted for last time around (and thus almost always aids the incumbent). That procedure is claimed to improve the accuracy of the results, but to the best of my knowledge it hasn't been tested in extraordinary circumstances like these.
So there are good reasons for supposing that Murphy may actually be trailing in East Renfrewshire. OK, he still has time to turn things around, but if he fails to do so his brief leadership of Scottish Labour will come to a very abrupt and ignominious end. It may be theoretically possible that the rule insisting that the leader must be a sitting parliamentarian could be finessed, but it would be politically unthinkable. So who would replace him, and become the FOURTH Scottish Labour leader since the 2011 election?
It's safe to assume that it won't be another Westminster MP, because there'll be so few of them left. Neil Findlay would almost certainly put himself forward again, in which case Kezia Dugdale would be under huge pressure to stand as the "stop Findlay" candidate. She's the current darling of the unionist press, so the leadership would be hers for the taking if she wants it - but does she? If not, Jenny Marra would be a possible Plan B, but she seemed just as reluctant as Dugdale to sup from the poisoned chalice last time around.
* * *
UPDATE : SNP draw level with the Liberal Democrats in new Britain-wide Ashcroft poll -
Conservatives 34% (n/c)
Labour 30% (-1)
UKIP 15% (+1)
Greens 8% (+1)
SNP 5% (+1)
Liberal Democrats 5% (-2)
Plaid Cymru 1% (n/c)
The Scottish subsample figures are : SNP 54%, Labour 21%, Conservatives 16%, Liberal Democrats 5%, UKIP 3%, Greens 1%.