Will Straw has had a go at modelling the first round of the Labour leadership contest for Left Foot Forward. The model is not entirely convincing, as Straw acknowledges. It is probably as decent an attempt as anybody has yet made to address the gaps in what we don't know about the Labour leadership contest. Yet it also highlights how large are the gaps in the public information, making any such exercise extremely provisional at this stage.
Left Foot Forward suggests that the race looks neck-and-neck between two Milibands.
So how can you win if you are not a Miliband? An Electoral College, using the Alternative Vote, means that to win Balls, Burnham or Abbott must finish ahead of and knock out at least one Miliband at some point in the count, before going on to defeat the final candidate (quite possibly another Milband) with half of the electoral college in the final round.
So, first, catch your Miliband, as Mrs Beeton might have put it.
Political logic might suggest that Ed Balls or Diane Abbott would aim to finish ahead of Ed Miliband, before seeking to defeat David Miliband in the final round. Andy Burnham might think he had better prospects of finishing ahead of David Miliband, and hoping to win transfers from him to defeat the field.
Second, try to knock out your Miliband straight away. I doubt that any candidate could win the college without finishing in the first two on the opening round. That ain't necessarily so. Harriet Harman was never lower than second in the 2007 deputy leadership, though sequentially behind two different candidates, in polling behind Jon Cruddas on the first round before trailing Alan Johnson from rounds two to four, and leading at the end.
It is theoretically possible to win the electoral college from third. In practice that probably depends on something close to a near-tie among three candidates on the firsst count. One would not just need to gain disproportionately from transfers as the first two candidates are eliminated to move up into second, but also to remain within range of overtaking the leader to win the college on the final round. (Since votes tend to transfer towards the centre, this hypothetical scenario may also depend on a candidate from one edge of the party being in the lead, as in 2007 when the leftest candidate was ahead of first-past-the-post, being eliminated in third under AV).
Third, once you have knocked out one Miliband, you have to hope that votes do not transfer in significant numbers from one Miliband to another. Alternatively, you might rely on one of your rivals to simultaneously defeat the other Miliband so as to avoid this potential problem of a fraternal cartel.
Straw's LFF model shows that it should be possible to get quite a strong grip on the Parliamentary section of the Electoral College: the smallest section in which the majority of voters are on the record about their first preferences at least.
If the Milibands do have two-thirds of this section between them on the first ballot on first preferences, this shows how challenging it will be for other candidates to get into the final run-off. If Abbott, Balls or Burnham can not switch PLP votes for the first ballot, they would have to perform very well in the member and affiliates section to catch their Miliband from the start.
If there were a Miliband v Miliband run-off, then the key question in this section is how the transfers from other candidates will affect the PLP totals in the final round.
After that, all is mere speculation at this stage. There is pretty much no usable public information on the voting intentions of party members, until there is a poll of party members themselves (as opposed to Labour-supporting voters, where at this stage the results are still going to pick up name recognition, ). In the absence of anything better, LFF uses a LabourList self-selecting internet poll of members. But I doubt any weight can be placed on it at all. Though, if anybody could (perhaps through the use of a truth serum) get the honest canvassing guesstimates of each campaign team and average them out, one might have a useful straw in the wind.
The CLP nominations are an interesting indicator - though they carry no weight at all in the electoral college themselves. It is very difficult to then translate them back into individual votes, since the nomination have a strong first-past-the-post (albeit under AV) bias towards the leading candidates.
The Labour party website gives the following results from the first 185 CLP meetings (28% of CLPs). I am not aware of any reason why that cross-section which has already nominated would have any particular demographic or political bias.
David Miliband - 86
Ed Miliband - 58
Andy Burnham - 20
Diane Abbott - 15
Ed Balls - 6
The CLP nominations support the narrative of a fairly close Miliband v Miliband contest. Given that these too are determined by transferable voting (albeit in a "caucus" rather than ballot format), David Miliband's current lead might be one reason to cast doubt on the prevailing assumption about how strongly second preferences would prove decisive for Ed Miliband.
If almost every CLP does hold a nomination meeting, then in excess of 10,000 party members (over 5% of party members in the ballot) would take part and potentially influene and be influenced by these meetings, though in most cases only members of the General Committee can vote. (Bassetlaw is deciding both its nomination and, perhaps more significantly, its MP John Mann's vote with a primary ballot open to all members and supporters, but is an unusual case).
If an average of 15 party members take part in each one (or, alternatively, if the average was 25 across two-thirds of CLPs). Some members report nomination meetings are about twice as well attended as a routine meeting, though any aggregate is a guess. The 50 members who took part in Islington South last Friday would probably be well above average. The group of participants is, by definition, distinctive from the broader membership.
The affiliates section is the hardest section of all to predict, not least because it is anybody's guess what turnout will be.
It will surely be higher than the 8% who voted in this section in the 2007 deputy leadership contest. I have heard suggestions that this was 30% in the 1994 leadership contest, but don't have a solid source for this. (That was a different universe in terms of political communication, which might suggest it could be possible to mobilise a higher turnout, though this has so far been a lower profile contest in the national media).
If there were approximately 4 million ballots, that would be the difference between 320,000 votes and 1.2 million. Whichever it is, the total vote in this section will have the same weight as that of the party membership. The voters would have considerably different characteristics if the electorate was triple what it was in 2007, and one impact would probably be to considerably dilute the impact of union endorsement on turnout compared to 2007.
Again, what is missing is a good data source to put into the LFF model on affiliate voting, with the LabourList poll again being used as a proxy alongside nomination effects based on the 2007 contest.