The Ed Miliband campaign says they expect to win the Labour leadership on second preference votes. The Guardian reports the claim that their canvassing returns show supporters of Diane Abbott, Andy Burnham and Ed Balls prefer Ed to David Miliband by three to one.
That rather falls into the "they would say that, wouldn't they" category of electioneering, along with the David Miliband camp's insistence that they are increasingly confident of winning, and also think that David can take at least as many second preference votes as Ed.
But it is a good moment to assess the evidence for the widely discussed possibility that second preferences will decide the election. The answer is that this is a strong possibility that they will if the race is within five points on the first round, but that a candidate with a lead closer to double-figures is likely to hold onto it (unless second preferences really do break as strongly as 3:1 against them).
In one sense, second preferences will decide the result: second (and later) preferences will be counted if no candidate has 50% on the first round. That is pretty certain to be the case.
But for transfers to be decisive - and to change the winner - two other conditions need to be met. For candidate B (let us call him "Ed") is to win a large enough share of second preferences to close and overturn the first preferences lead of candidate A (let's say "David"), we would need to know two things:
What is the first preference share and lead?
(i) The race has to be close enough for the second preferences to make the difference. (A candidate ahead 49% to 31% is going to win comfortably - their rival would need more than 95% of other voters to prefer them).
How do the second preferences divide?
(ii) The first round leader must get fewer second preferences than their rival. If they can get an even split (or better), even the smallest lead will be held.
So, how close does would the race need to be on the first round for preference voting to change the winner?
Let's project some possible results, using an assumption that the leading two candidates share two-thirds of the vote on the first round - while three other candidates share one-third. (The Miliband versus non-Miliband split was 70:30 among party members and 60:40 among trade unionists in the YouGov poll at the end of July. Of course,
similar assumptions can be used to compare any run-off pairing, as long as you have a view about the first round result, and the share of transfers, as we don't have any run-off polling for other combinations).
To simplify, let's just look at who wins the party members' section. The winner will probably need to do this (though it ain't necessarily so, as Friday's Left Foot Forward post on split college decisions set out).
The YouGov poll showed that Abbott/Balls/Burnham (ABB) voters in late July divided 51% for Ed Miliband, 33% for David Miliband, 9% neither. (6% unsure at that stage). That would give a 55-35 split, with one in ten ballots not counting, and valid second/later preferences divide pretty much 3:2
If two Milibands did end up ahead, those ABB votes would split it 18% to Ed Miliband, 12% to David Miliband, with 3.33% of members' votes leaving the contest by not having a preference.
Let us make that one of four scenarios to how first and second preferences could interact if a third of the vote was up for grabs.
* If second preferences divide 50:50 between the leading candidates, the first round winner will win, as of course is also true if the leader wins the greater share.
* If second preferences divide 3:2 for Ed, David Miliband would need a 6%+ lead on first preferences. (This is shown by MiliD being ahead 38-32 among party members in the July YouGov poll, and ending up 50-50 in the members' section).
* If second preferences divide less strongly 55:45 for Ed, David Miliband would need a 3%+ lead on first preferences to stay ahead. Now a 35%-32% lead could be enough.
* If second preferences were to divide 3:1 for Ed - as his campaign suggests - then if one-third of voters did not vote for a Miliband first, he would pick up 22.5% of the vote on second preferences and David just 7.5%. (I am again writing off 3% of ballots). So David Miliband would now need a 15%+ lead on first preferences to retain a lead, for example 41% to 26% to avoid being overtaken.
My own view is that the YouGov poll remains the best evidence we have - so that the best rule of thumb is that a 5-6% lead on first preferences could well be necessary for David Miliband to stay ahead among party members.
Of course, the combined first round share of the top two could be higher or lower than 67%. If the leading candidates have more of the first round votes, second preferences matter less. If it is a lower share, they are more important. For example, if the Milibands already shared 75% of the first round, a 3:2 division would close the gap by 4.5 points (rather than 6), a 55:45 split by just 2 points (not 3), and a 3:1 split by 11 points (not 15%), again assuming 10% of the remaining votes don't transfer. (By contrast, if their joint first round share fell as low to 50%, even a 3:2 split of second preferences would eat up a 9% lead).
(A major reason transfers were so important in the 2007 deputy leadership was that the even six-way race meant the final two candidates only shared 37% of first round votes, with even the top three candidates having a combined share under 57%. Harman won on transfers in the final round - with a roughly 3:2 split (Cruddas' 30% splitting 17:13), in fact, Harman and Alan Johnson picked up almost exactly a 50:50 share of transfers overall, with Harman leading Johnson by 0.77% on the first round and 0.87% in the end, both overtaking Cruddas who began in front).
Will the party members vote prove decisive? It will do so if the other sections are evenly balanced.
A reasonable holding guesstimate of the Parliamentary section would be 55% to 45% for David Miliband, a lead of around 27 MPs or MEPs (149 to 122) if everybody's vote were still to count in the final round.
LabourList's running totals suggest declared first preference support for the Milibands is already 40% (David) to 29% (Ed) of this section, with their latest totals showing 109-79 (including MEPs), so an even division of the remaining 80+ Parliamentarians would see David ahead by 28-30 MP/MEPs (or 10-11%). A better estimate will be possible if large numbers of MP second preferences are placed on the record: these are the second preferences with the most electoral college weight! (That Tom Watson and Geoffrey Robinson have different second preferences shows that even MPs closest to Ed Balls will not act as a bloc; there are slightly more Andy Burnham than Diane Abbott votes, though the assumption that these MPs will transfer to David Miliband depends on unpacking political and north-west regional loyalties among Burnham nominees).
The likelihood of a David Miliband lead in the Parliamentary section makes a victory for Ed Miliband in the affiliates section necessary to put him in contention.
In July's poll, David Miliband did better in this section than among party members. But that was prior to any campaigning activity from the unions themselves. The David Miliband campaign will be hoping that the support of both The Mirror and Jon Cruddas (who was almost twice as popular among affiliated voters as among individual party members in 2007) helps to win votes in this section. But this is also where the Mandelson-Blair interventions are more likely to have done more harm than good to him campaign. And the second preference factor is likely to help Ed more than David in this section. Peter Kellner's health warning about differential turnout is important - 1994 turnout was 19% - with the frontrunner's higher name recognition likely to help most among those least likely to vote and so not following the contest.
I think Ed Miliband is likely to win this third section. If he could do so by a margin which offset any Parliamentary deficit, then the overall leadership would look very much a neck-and-neck race. That would also be one where the preference of party members' - whoever they choose - would (happily) definitely prevail.