by guest contributor Edmund Griffiths

The last ‘supporting nominations’ from constituency Labour Parties are now in. Not every CLP has chosen to nominate a leadership candidate; but, among the 390 (61%) that have, the breakdown is:

Corbyn 39.0%
Burnham 28.5%
Cooper 27.9%
Kendall  4.6%

This is almost the last solid evidence about the state of the race that we shall get before the results are actually announced. There will presumably be opinion polls; but opinion polling faces some very grave methodological questions following its failure to predict the results of the general election and the Greek referendum (a matter to which I hope to return elsewhere), and this is anyway a difficult election to poll: not everybody who will sign up as a party member or registered supporter has yet done so, and in fact the electorate will only definitively come into existence two days before the ballot papers are sent out. We would be unwise, therefore, to put much faith in poll returns.

How reliable are the CLP nominations, though, as a guide to how Labour Party members will vote? And, if they are reliable, what do they tell us? Which candidate do they suggest is likeliest to win?

This election is being fought under a new set of rules: but, even so, the best way to interpret the CLP nominations is on the basis of a comparison with earlier contests. The previous rules were in fact only put into effect twice: in 1994 and in 2010. (The election of 2007 was uncontested; elections before the introduction of ‘one member, one vote’ in 1993 do not provide relevant comparative data.) Of the two, 2010 is more helpful to us than 1994: (1) because it was more tightly contested, (2) because it happened more recently, and (3) because the number of CLPs giving nominations was larger—396, pretty much the same as this year’s figure, compared to only 109 in 1994. (Perhaps in 1994 fewer CLPs fully appreciated the role of their nominations under ‘one member, one vote’.) Nonetheless, it is instructive to look at both sets of results, comparing each candidate’s share of CLP nominations with their share of Labour Party members’ first-preference votes in the actual election. (These votes were then combined with those of MPs, MEPs, trades unions, and socialist societies, to form an ‘electoral college’ under rules that are no longer in force.)

In 1994, there were three candidates: Margaret Beckett, Tony Blair, and John Prescott.

Candidate CLPs Votes
Blair 53.2% 58.2%
Prescott 35.8% 24.4%
Beckett 11.0% 17.4%

In 2010, meanwhile, there were five: Diane Abbott, Ed Balls, Andy Burnham, David Miliband, and Ed Miliband.

Candidate CLPs Votes
D. Miliband 41.5% 44.1%
E. Miliband 37.9% 29.9%
Burnham 11.3%   8.5%
Abbott 5.0%   7.3%
Balls 4.3% 10.1%

The first thing we see from examining these numbers is that the CLP nominations did provide at least a rough guide, on both occasions, as to how Labour Party members would end up voting. The largest discrepancy anywhere is the 11.4-point gap between John Prescott’s CLPs and votes; and the mean discrepancy is 5.7 points (worse than the margin of error opinion polls claim—but better than they sometimes achieve). It is therefore moderately likely, although far from certain, that the share of first-preference votes cast for each candidate will be in the following ranges:

Corbyn 33.3% – 44.7%
Burnham 22.8% – 34.2%
Cooper 22.2% – 33.6%
Kendall   0.0% – 10.3%

And it is more likely, although still not certain, that they will be within the wider ranges:

Corbyn 27.6% – 50.4%
Burnham 17.1% – 39.9%
Cooper 16.5% – 39.3%
Kendall   0.0% – 16.0%

We can therefore take it as established that Liz Kendall is not competitive. Even on the most optimistic reading of her chances, the best she can look forward to is an unexpectedly strong last place. Any of the other three, however, can in principle still win from this position. But can we deduce anything further from the CLP numbers? In particular, is there any reasonable basis for deciding which candidates’ support is likely to be understated by these figures—and which overstated?

If we look back at the 1994 and 2010 CLP nominations, we notice that both sets modestly understate the support that the front runner would eventually receive (by five points in 1994 and 2.6 points in 2010), and significantly exaggerate the support for the candidate placed second (by 11.4 points in 1994 and eight points in 2010). They also tend to understate the support for candidates in third and subsequent places; although Burnham bucked this trend in 2010 by doing rather worse in first preferences than he had in CLP nominations. Is there a plausible mechanism, other than chance fluctuations in a small sample, that would lead to this pattern emerging? I think there is. In both 1994 and 2010, the front runner (Tony Blair, David Miliband) was seen as offering a politically distinct ‘Blairite’ message, while the second- and third-placed candidates (Prescott/Beckett, Ed Miliband/Ed Balls) were more ‘centrist’. And—crucially—CLP nominations are not decided on first preferences alone. It seems very likely that supporters of Beckett or of Balls transferred their backing to Prescott or to Ed Miliband, at CLP nomination meetings, once their favourite candidate had been eliminated—thereby depriving Blair or David Miliband of nominations from some CLPs where the front runner had been ahead in the first round. In the first-preference votes cast during the actual election, however, these supporters reappeared: Beckett and Balls did substantially better than the CLP nominations had suggested, and Prescott and Ed Miliband did worse.

This time, once again, there is a front runner who stands for a clearly marked change (albeit to the left rather than the right), and there are other candidates who are comparatively ‘centrist’ and comparatively interchangeable. The pattern differs somewhat from 1994 and 2010, in that it is not clear which candidate is in second place: neither Burnham nor Cooper has succeeded in becoming the obvious ‘no change’ candidate. It is therefore likely that both of them have benefited somewhat from transfers (whether Burnham-Cooper, Cooper-Burnham, Kendall-Cooper, or Kendall-Burnham). So the CLP nominations probably overstate support for Burnham and Cooper, and understate it for Corbyn and Kendall; although the error is likely to be quite moderate. As a ‘best guess’, we can say the CLP nominations point to a first-preference vote something like:

Corbyn 42%
Burnham 24%
Cooper 24%
Kendall  10%

If those were indeed the results in the first round, who would actually win? How would the second and third preferences be allocated? The best evidence (although it is not very good evidence) as to second preferences comes from a survey of readers of LabourList If we accept those figures and make some additional (reasonable but rather Corbyn-pessimistic) assumptions, we can create a model that will allow us to predict the final result based on an estimate of first preferences. And if we then feed the 42-24-24-10 ‘best guess’ into the model, it comes up with a final result that looks like:

Corbyn 50.1%
Cooper 49.9%

It would clearly be absurd to offer that as a prediction. When we have been casually throwing around errors of five to ten percent, we cannot very well claim to call the election down to a victory margin of maybe 500 votes. And it would be quite possible to juggle the figures and come up with plausible assumptions that led to a narrow Cooper win or a narrow Burnham win. What is much harder is to find a halfway plausible scenario under which the final round is other than very close. As clearly as we can make it out, based on the CLP nominations, this leadership election is a dead heat. Corbyn is almost certain to win the first round; but the actual result is on a knife edge.

Or it would be, if it were only Labour Party members voting (as it is at CLP nomination meetings). Given that party members seem to be divided basically fifty-fifty, in the final runoff, between Corbyn and either Cooper or Burnham, the votes of registered supporters and affiliate members will become decisive. If large numbers of people sign up to vote, and Corbyn wins among them by a big margin, then his chance of winning overall is very good; if only a few sign up, or if those who do sign up are no more likely to back Corbyn than party members are, then the election is looking much, much too close to call.

Edmund Griffiths is the author of ‘Towards a Science of Belief Systems’ (Palgrave Macmillan, 2014).