Mark's Musings

A miscellany of thoughts and opinions from an unimportant small town politician and bit-part web developer

Tactical Voting and AV


There is a claim often made by supports of AV that it’s less susceptible to tactical voting that FPTP. This is simply false; not only is AV equally susceptible to tactical voting but the results of tactical voting can be strongly counter-intuitive and lead to outcomes that are contrary to the wishes of the voters. The result of a closely-contested AV ballot can be determined by the order in which candidates are eliminated, meaning that it can be advantageous tactically to vote for a candidate you dislike in order to manipulate the elimination sequence.

The best way to demonstrate this is by means of an example. Imagine a hypothetical constituency where there are three leading candidates: Conservative, Labour and LibDem. For the sake of argument, the Conservative candidate has a healthy, but not unbeatable, lead and the other two are neck and neck. Here’s our hypothetical set of first-preference votes:

Co: 15,000 votes
La: 9,000 votes
LD: 8,000 votes

Under FPTP, the Conservative candidate would win. If we assume that LibDem voters would split their second preferences equally between the Conservative and Labour candidates, then we’d get this result in AV:

Co: 15,000 + 4,000 = 19,000
La: 9,000 + 4,000 = 13,000

So, the Conservative still wins. But now let’s reverse the order of the last two candidates:

Co: 15,000 votes
LD: 9,000 votes
La: 8,000 votes

Now, we can assume that almost all (for the sake of argument we’ll assume that it’s all) Labour voters will put the LibDems second. So now we have this result:

Co: 15,000 + 0 = 15,000
LD: 9,000 + 8,000 = 18,000

This time, the LibDem candidate wins. By improving their position from third to second, they pick up the second preferences of the third-placed Labour candidate to overtake the Tories in the second count. So, it is clear that the result can be affected by who gets eliminated first.

But what about tactical voting? If you’re a Conservative supporter, what do you do? Well, what you want least to happen is for Labour to come third and be eliminated first. How do you avoid that? Vote tactically – for Labour! If 2,000 Conservative voters give their first preference to Labour, then we get this on first preferences:

Co: 13,000 votes
La: 10,000 votes
LD: 9,000 votes

Now, the LibDem is eliminated, and the second preferences split equally again, giving this:

Co: 13,000 + 4,500 = 17,500
La: 10,000 + 4,500 = 14,500

And now the Conservative candidate wins again. Voting tactically for an opponent secures the election of their favoured candidate.

It gets even more complicated if you’re a Labour voter. Under this scenario, Labour has no way of winning, as they can’t win on first preferences and won’t pick up enough second preferences to win if the LibDems go out first. But, as a Labour voter, you really don’t want a Conservative victory – if you can’t have a Labour winner, then you at least want the LibDem to prevail. The solution here is for Labour voters to switch to the LibDem candidate as an “anti-Tory” vote, just as they would under FPTP. But the difference is that, under FPTP, 6,001 Labour voters would need to switch to ensure a LibDem victory. Under AV, only 501 need to do so – as that’s enough to put Labour last and ensure that the LibDem candidate wins on second preferences.

As for the LibDem supporters, there’s nothing they can do to manipulate a result which guarantees them a win. All they can do is hope that they beat Labour on first preferences and pick up enough second preferences to overtake the Tories. Their best hope of achieving that is to persuade Labour voters to vote tactically, while persuading Conservative voters not to.

Incidentally, before anyone asks, the presence of fringe candidates on the ballot won’t make any difference to this analysis provided they don’t pick up any more than a hundred or so votes each. If there are fringe candidates getting thousands of votes then that does change things – it doesn’t mean that tactical voting won’t work, but it may require a change of tactics! Just to illustrate, let’s add a Green candidate with 2,000 votes to the second example where the LibDem came second overall and won on second preferences:

Co: 15,000 votes
LD: 9,000 votes
La: 8,000 votes
Gr: 2,000 votes

Let’s be a bit more sophisticated here and assume that Green voters will mostly, but not entirely, put Labour as their second preferences:

Co: 15,000 + 0 = 15,000
LD: 9,000 + 400 = 9,400
La: 8,000 + 1,600 = 9,600

No candidate has reached the 50% mark yet, so there’s going to be a third round this time. But look – the Labour candidate is now second, so the LibDem candidate is the next to go out. Making the same assumption that we did before about the split of LibDem second preferences, we now get this:

Co: 15,000 + 0 + 4,500 = 19,500
La: 8,000 + 1,600 + 4,500 = 14,100

And this time the Conservative wins. Before, when there were only three candidates, the LibDem won. By adding a Green candidate to the mix, the Conservative comes out on top.

Obviously, these are made-up examples (although using at least moderately plausible figures). In the real world, the transfer of votes to second preferences will be a lot more unpredictable. But the underlying principle remains the same. So if anyone tells you that AV isn’t susceptible to tactical voting, or that there’s only ever one candidate with the potential to win, then you can point them here to show that they’re wrong!