Since the end of July, YouGov finds a 2% swing to Ed Miliband among party members, and 13% among unions. The poll supports this blog's suggestion that MP (31-40 on first preferences) and union sections (36-29 on first preferences) could well balance out, happily making the party members votes decisive.
Having declared the campaign almost over - a small exaggeration - it should follow that this late poll is more likely to reveal what has happened than to affect the last week or so of voting. That's mostly the case. But are there areas where it could still make a difference?
Turnout and GOTV
That it remains a knife-edge election will motivate both campaigns - and perhaps more importantly, their volunteers - to take up their Get Out the Vote operations to the wire. New phonebank sessions are being planned this morning.
A close election is likely to see membership turnout up on the 69% of party members who voted in 1994. It seems likely to be 75%+ and perhaps higher still. Among new members - those joining since May make up 20% of the party membership - it could well be highest.
That it is so close can be seen to reinforce both major campaign's core arguments about why the choice matters. At this very late stage, most members will want both campaigns to continue to ensure the race is not excessively personalised on either side. (The poll also suggests that some 'negative' interventions - the Mandelson/Blair critiques - may well have been counterproductive with voters for the outcome they wanted).
Remaining GOTV efforts could possibly have more impact in this section than among party members - simply because there is a much greater swathe of the electorate which will not have voted. A 13% swing to Ed Miliband between YouGov polls - compared to 2% swing among party members - in part reflects that the first poll significantly overstated where David Miliband support was likely to end up, through pre-campaign name recognition, but it does suggest the campaign has made an important difference in this section.
There is a potential vulnerability for the David Miliband campaign in its strongest section, where YouGov has a 56-44 margin. This is the most volatile section at this stage, simply because each individual vote has a much greater weight.
Where MPs think one or other candidate is better for the party's chances, they will vote for them. But perceptions of the likely winner can certainly influence what professional politicians think of a close race. (This is dramatised by a contested claim in the Sunday Times report: "One source claimed backbenchers were being warned that refusing to support the former foreign secretary could prove a “career-ending moment”). So the question, highlighted by PoliticalBetting this morning, is "Could this poll cause MPs/MEPs to switch brothers?". If the poll had been 10 days ago, this could have been significant, but any effect now is more limited. Firstly, the timing: most MPs have probably voted - though it may be rational for them to hang on, particularly if not sure about a potentially decisive second preference. Secondly, the poll remains close, while any late public switch would look opportunistic, again suggesting this is more likely to effect undeclared MP second preferences (where votes have not been cast).
This highlights that an important missing part of the overall electoral college jigsaw is how the 80 or so MPs voting for Abbott, Balls and Burnham cast second preferences (and how far they have already done so). YouGov assumes an even split to leave David Miliband 12% ahead in this section - but there has been no substantial report or survey on this in the media or online. Ed Miliband may well have a couple of undeclared first preferences: he is thought likely to have Gordon Brown's vote, and may have a fair shot at Harriet Harman's. Every second preference up for grabs will count, and any remaining big catch for either side - what about Yvette Cooper? - could be seen as an important indicator.
Perhaps the "college balance" risk now shifts: a month ago, it was for Ed Miliband to demonstrate whether he could tie the two sections with a strong affiliates' lead, so taking it all to the members. Is it now more for David Miliband to show that he can hold on to an offsetting Parliamentary lead, again meaning everything hangs on the close membership vote?
The poll will have less impact on remaining voting than on the way the final fortnight of the campaign is reported and discussed. That it is so close - and the potential of a surprise result - will probably increase the amount of coverage, and shift its focus.
The media/Westminster village assumption had been hardening around the idea that David was winning: a somewhat self-reinforcing cycle influenced by MiliD doing better with newspaper endorsements, being ahead in the betting and having had an effective communications strategy in timing the release of MP and other party endorsements. There has been very little reporting of the ground campaigns among party members or unions to test media assumptions.
Reports suggesting this is a "shock poll" partly suggest a lack of close media scrutiny of the election or the previous poll, where members came out 50-50. The evidence before this poll was that the race was neck-and-neck, despite media assumptions to the contrary.
Now, media assumptions and narratives may well swing too far the other way. The poll is not the result - being very close and depending on unpredictable patterns of transfers. It is probably still most sensible to now regard the Miliband brothers as co-favourites in a neck-and-neck race.
This may well prove the last poll of the campaign - not least because any follow-up may well provide a similarly inconclusive result. And so it does look very much as though everybody will have to wait for the envelope to be opened and the official declaration, a scenario canvassed here recently.
So, when will we find out the result? In Manchester on September 25th, of course. Yet I expect it may well be in Peter Kellner's YouGov powers to pretty much declare the result 7-10 days out. (A full projection would require some political intelligence on MP second preferences too) ...
So the spoilsports may have it. Whatever your candidate preference, the demands of political drama may depend on hoping that any post-voting poll remains on a margin of error knife-edge. So it remains possible that - as in 1981 and 2007 - the Labour membership could gather in Manchester still not sure what the result will be on the night.