How to vote for Labour’s leader: It’s more complex than it looks

By Brian Barder / @brianlb

This is about making your vote in the Labour party leadership election this week do what you want it to do, and how not to make it self-defeating. It’s not about who I think you should vote for. So in the examples I give, I have given the five candidates pseudonyms: A, B, C, D and E. For the purposes of this article, I am taking the view that the only two candidates for the Labour leadership who possess the basic qualities required in a party and national leader are “C” and “B”; and that of those two, “C” seems to me to have the edge over “B” on policy and values, and moreover to look reasonably ‘papabile’. However, of the five candidates, only “A’s” views on almost all major issues chime with mine – on the UK nuclear deterrent, Iraq and Afghanistan, other foreign wars, prisons, terrorism, civil rights, taxation, the economy, and more.

Unfortunately, I don’t think that A has the personality or skills to be an effective party leader or a credible candidate for election as prime minister (sorry, A). Moreover, I don’t think A can win enough first or second preferences in the parliamentary or union sections of the electoral college to stand a realistic chance of winning the Labour Party leadership election itself, despite being likely to do well in the members section – although it’s never safe to base one’s votes on assumptions about how everyone else will vote . And I’m appalled by the thought of “D” becoming leader: indeed, were “D” to win this election, I would leave the party. So I don’t want to use my preferences in any way that could possibly help to bring that about. Those, then, are my views of the principal candidates. How should I translate them into voting preferences?

I was originally going to vote in the Labour leadership elections for 1. A, as having the right principles and policies but very little chance of winning; 2. C, as possessing the best leadership qualities and a good chance of winning; 3. B, also with excellent leadership qualities but marginally less attractive on policies than C; 4. E, about whom I have no opinion; and 5. D, who I think would be a disaster as leader. But I have been persuaded by an expert’s analysis of the voting system for the leadership election that this order of preferences would be risky, and possibly even self-defeating. I now plan to vote 1. C, 2. B, 3 E, 4. A, 5. D. Here’s why.

I had intended to give A my first preference, even though I don’t think A can win, on the basis that if A gets a respectable number of votes, that will oblige the new leader to take A and A’s supporters’ views seriously, thus moving the party as a whole in the right, or rather left, direction. It seemed reasonable to give my first preference to A, in the confidence that at some stage in the rounds of counting and reallocating losers’ second preferences, A will come bottom of the poll and so be eliminated, and that my second preference, will thereupon be reallocated to C. Mission accomplished. But is it? Now read on….

My original plan seemed to be confirmed by the advice in a letter in the Guardian of 27 August which advocated exactly what I was proposing to do (the names changed once again — this is about how to vote effectively, not who to vote for, remember?):

“Seumas Milne is absolutely right that those who want to return the Labour Party to its correct place within the political spectrum should ensure C beats B. However C is young and untested, and his leadership will not just be determined by his platform; it will be shaped by context. The first context will be the dynamic within the party following the result. A radical confident leadership from C will most likely emerge if the starting point is a strong A vote that transfers to him. It is therefore imperative that A supporters hold firm in their first preferences, determined, as Milne describes, “to see a voice for the left in the country’s main party of reform”. Vote A 1, C 2.
Basildon, Essex”

However, I had a nagging suspicion that the writer of that letter – and I – might be missing something here. So I sought the advice of a Labour supporter who understands the electoral system better than I do (no, not my wife, although no doubt she does, too). Here is what she said in reply:

“Let’s start with the simple case, in which your only interest is who gets elected (that is, ignore for a moment your desire to use your vote for the additional purpose of “sending a signal” as well).

Then it is all simple. One of the principal merits of AV (some would say the only merit) is that it is simple for the voter to know what to do, even if it is not simple to explain how the system works. The voter should simply number the candidates according to her ranking of them. There is no way to vote tactically. Even if a voter knew exactly how everyone else was going to vote (which of course she doesn’t) there would not be a reason to do anything other than number the candidates in order.

To answer specific questions that have been raised:

– In an election with five candidates, putting preferences by four of them, and leaving the fifth blank is identical to ranking the last candidate fifth. (Your last preference votes won’t ever be counted because it is not possible for four candidates to be eliminated and their second or lower preferences re-allocated in a five-candidate election.)

– If you want “anyone but D”, your best strategy is not to put a number next to D. Putting a number next to a candidate can’t harm them. If it is the lowest possible number (5 in a five-candidate election) it won’t help them either (see above). If it is any number other than last, it might help them. If you don’t want them, don’t vote for them.

– Your votes have to start with 1 and go down as far as you have preferences. In some elections a minority of voters put a “1″ next to the candidate they want and “5″ next to someone they detest, without putting the numbers 2, 3 and 4 in between. Some electoral officers will count this first preference; most will just declare the paper spoiled.

Now we make it more complicated, by acknowledging that some voters want to use the election not only to choose a winner but also for the secondary purpose of sending a signal. In the normal case, the signal a voter wants to send is aligned with her preferences in the election (that is, if a voter wants to send the signal that she likes what D has to say, she is also likely to think that D would make the best leader). In that case, we are back to the simple case – the voter should simply number the candidates according to her preferences.

It is more complicated for a voter who wants to send a signal that is not aligned with her actual preferences for leader. Say a voter prefers C to be leader, but wants to send the signal that she prefers A’s policies and views to C’s. If the voter is trying to pursue these two objectives simultaneously, that reintroduces the possibility of tactical voting. The optimum strategy for the voter in this special case depends on (a) the relative weight the voter attaches to these objectives; and (b) what the voter thinks other voters will do.

In the actual case at hand, in which the voter reasonably expects that A has almost no chance of winning, she might put A first and her actual preference for leader (in this case, C) second. This would achieve the secondary objective (sending a signal) but there are two ways in which it might backfire on the first objective (choose C to be leader):

*first, there is a possibility (probably small in this case) that A might actually win, which is not what the voter intends;

*second, if enough people who want C as leader nevertheless put someone else first for the purpose of sending a signal, then there is a chance that C could be eliminated early on. Suppose E goes out first, and most of his second preferences go to A. Then C might conceivably still come below A in the next round, and he’d go out before A does. A would go out next, and then it would be a straight race between B and D. So in this story, the voter who has put A ahead of C to send a signal would have inadvertently made it more likely that D gets elected leader, even though she has correctly anticipated that A actually has no chance of success — and has tried to do nothing that might help D to win by putting them bottom of her preferences.

Whether the voter regards this as a risk worth taking depends on (a) the relative weight she attaches to the two objectives of electing the right leader and sending the signal; and (b) her view of the probability that C might be eliminated ahead of A. (Note that this is not the same question as whether A might get more first preferences than C).

I have to say, my guess is that it is probably quite rare for a voter to prefer one candidate but want to send a (misleading) signal implying wrongly that she prefers another. So in general, the aphorism that there is no tactical voting in AV is correct.

So the conclusions are:

*In the case where your objectives are limited to choosing a leader, you should number them in order of preference as far as you have preferences, and then stop. If you specifically don’t want a particular person, don’t put a number next to them — or, if you do, make sure it’s the last one (i.e. 5, using all four of the other numbers as well).

*In the case where you want to send a signal that is different from your preferences for leader, it is more complicated. You should only vote for a “signal” candidate ahead of the candidate you really prefer if you are pretty confident that the signal candidate will be not only eliminated, but also eliminated ahead of your true preference.”

I am persuaded by this. If the argument in the Guardian letter, no doubt also being advanced by others in Labour groups and forums up and down the country, influences enough voters to do what the letter recommends, which of course is also what I had been planning to do, it could conceivably cause C to be eliminated before A, who would then either be eliminated in turn – or else even go on to win, if (for example) we have all guessed wrongly how the voting in the parliamentary and trade union sections is likely to go. The more first preferences go to C, the less the risk that C might be eliminated before a candidate who has received a significant number of purely gesture preferences. If you think that of the five candidates C would make the best leader of the party, best leader of the opposition and potentially the best Prime Minister, you should give C your first preference, and resist the temptation to use it to make a political statement in favour of a candidate whose views you like but who you know lacks the personality and other attributes to lead the party, the opposition or the country. If – out of loyalty, sentiment or bloody-mindedness – you insist on giving your first preference to, say, A or E, you can limit the damage, or at any rate the risk, by being careful to give your second preference to C. Take care to number the rest of your remaining votes also in order of your assessment of their leadership qualities: even your fourth preference may be counted and thus affect the outcome, if your first three are all eliminated. Only your fifth preference will not in any circumstances be redistributed or counted: reserve that for the candidate whom you think a seriously unsuitable choice for leader. Either number that candidate 5, or don’t give that one a number at all – provided that you have numbered all the rest 1 to 4. My expert, you’ll have noticed, suggests numbering only those you are genuinely prepared to see elected, in order of your preferences, and then stopping; my inclination is to number all five, to be on the safe side.

It comes down to this. Whichever your preferred candidate and order of preference, a first or even second preference vote for a candidate who you know lacks the qualities required of a future leader but for whom you want to make a gesture of support, is not only a waste of your opportunity to influence the election’s outcome in favour of the candidate whom you really want to win: it may actually damage the latter’s chances. With First Past the Post there’s scope for tactical voting: e.g., if you’re a Labour supporter who used to think that the Lib Dems were the next best thing and the Tories the worst, and you vote in a constituency where Labour always comes a poor third, it’s sensible, or used to be, to vote Lib Dem. Now that the Lib Dems have (temporarily, we hope) become a branch of the Young Conservatives, tactical voting under FPTP becomes a little more complicated, at least until the Lib Dems repent. With AV, tactical voting is effectively impossible: in the present case, you should always give your first preference vote to the candidate who you think will be the best leader, however good or bad you rate his or her realistic chance of winning, and your second to the second best choice, and so on down to no. 5.

Next year we shall all need to consider, in the light of all these ifs and buts, whether we really prefer this Alternative Vote system for electing our MPs to the existing system of First Past the Post – the choice which will confront us when we vote in the referendum promised as blood money for the LibDems by the ConDem coalition in its founding document. But that’s for another day and another post.

Meanwhile some of us will have the opportunity to vote this week for a new leader of the Labour Party. I conclude from the analysis above that those who share my view of the best (and worst) achievable outcome should vote 1. C, 2. B, 3. E, 4. A, 5. D. Sending other signals is an unaffordable luxury.

Any questions? (No, I’m not allowed to reveal the real names. You can make whatever substitutions you like.)