One of the purported benefits of the Alternative Vote is that it would end the need to vote ‘tactically’ in elections, but this claim has been challenged by some commentators. Andy White of the Electoral Reform Society explains why, in practical terms, the claim stands up to scrutiny.
What is tactical voting?
A tactical vote is an insincere vote: one in which a voter’s ranking of candidates on the ballot paper does not reflect his or her true intentions, and which produces a more desirable result than if the voter had voted sincerely. For example, under our present system of ‘first past the post’ elections, it is (or was) common for Labour supporters in unwinnable seats to plump instead for the Liberal Democrat candidate, in order to prevent the Conservative from winning.
This strategy was illustrated at the 2010 general election in the seat of Eastleigh, where Chris Huhne beat his Conservative opponent by appealing to Labour supporters. You can see how this choice was presented to voters in the chart in the bottom right of this election leaflet.
The Gibbard-Satterthwaite theorem
The Gibbard-Satterthwaite theorem to which both Rafael Hortala-Vallve and Roger Mortimore referred in recent LSE blog posts, proves that all voting systems are susceptible to tactical voting, “in the sense that there are conditions under which a voter with full knowledge of how the other voters are to vote and of the rule being used would have an incentive to vote in a manner that does not reflect his preferences.”
Nobody disputes the theorem, but a key phrase in its construction is ‘full knowledge’. Tactical voting is easy under first past the post, because only vague knowledge is required to make a correct decision. If I am a UKIP supporter in a seat that has only ever been held by Conservative or Labour MPs, then I can confidently predict that a tactical vote for the Conservative candidate will have the desired effect.
Indeed, tactical voting is so simple and effective under first past the post that the phenomenon has its own dedicated websites and guides in tabloid newspapers. Dr Stephen Fisher of Oxford University estimates that nearly a tenth of all votes in British general elections are tactical (see this paper from 2001).
The proportion of seats susceptible to tactical voting (i.e. so-called ‘marginal’ constituencies, where more than one party stands a realistic chance of winning) is up to half of all seats in Britain, and these are the seats which decide our general elections.
Tactical voting with the Alternative Vote
Roger Mortimore has posted an excellent description of how, in theory, one might be able to vote tactically in an Alternative Vote election. I suggest readers follow his worked example before returning to my argument.
You can see in Mortimore’s example that detailed prior assumptions are required to vote tactically in an Alternative Vote election. In the example Mortimore gives, the swing needed for one candidate to overtake another (about 4 per cent) – and thus change the character of the contest – is within the margin of error of a typical opinion poll. Moreover, even if voters are able to coordinate an effective tactical strategy, the number of constituencies susceptible to it is greatly reduced (there are about fifty ‘three-way marginals’ in Britain).
I contend that the level of prior knowledge and coordination required to vote tactically in an Alternative Vote election is so great – and the risks so much clearer than the rewards – that there is no incentive to attempt it. Tactical voting under first past the post requires only a certainty that one’s favoured candidate cannot win. With the Alternative Vote, conversely, one must accurately project the vote shares and transferred preferences of three different candidates (taking into account the fact that one’s co-conspirators will also be voting insincerely). For these reasons, the Gibbard-Satterthwaite theorem does not contradict claims that the Alternative Vote would end tactical voting in British general elections.
I am a supporter of AV.
However, I am also a fan of honesty and fair information.
I can think of one situation where ‘tactical voting’ can operation under AV as currently proposed.
This is where someone does not put their favoured candidate first, but uses their first preference as a ‘protest vote’.
i.e. It is conceivable that in an area where migrants are considered a problem, supporters of any/all parties might vote (say) BNP first and then for their real candidate. They would almost certainly not want BNP elected, but want to indicate to their real candidate that they want immigration moved up the agenda. This could ‘backfire’ and the ‘protest’ candidate could inadvertently end up being elected.
This risk could be entirely removed by (following the adoption of AV) the implementation of ‘weighted votes’ in parliament. The technology for this is now trivial (unlike at the time of the Jenkins report) – if MP’s votes were weighted by the number of first preference votes their party received nationally, then votes would be obliged to put their first ‘real’ preference first.
Protest votes are not, strictly speaking, tactical.
A tactical vote is one which is both insincere AND improves the electoral chances of a favoured candidate (as I explained at the start of my article).
I agree – AV will remove the ability to vote tactically. It is theoretically impossible to be sure that a tactical vote will work and if it is tried it can have adverse affects – i.e. the tactical voting can loose an election which could be won without it. This means no-one will use it. This is totally different to the ridiculous FPTP system
This is my uneducated view of the situation – please correct me if I am mistaken.
There are two theoretical stoppers which show that it is not theoretically possible to guarantee a result from tactical voting – there is also a theoretical issue with blocks pratical attempts which I’ll describe later.
Firstly – The Gibbard–Satterthwaite theorem states that one of three situations must prevail, the ability to tactically vote is just one of them. The other two are that one person will have the overriding decision (i.e. the count will go down to the last vote counted) and/or some candidate(s) can not win. Until the election is held and all votes counted, which of the three situations has held is not known. Therefore there is no way of knowing for sure if tactical voting will even work theoretically.
Secondly – Also even if someone had a crystal ball to see which of the three outcomes were to materialise and the outcome indicates tactical voting could work, the party investigating tactical voting would need a very power crystal ball to work out exactly how each person is going to vote. This of course is not possible and therefore tactical voting won’t work. (ok so an argument could be that perhaps practically only a significant majority of cast votes needs to de divined but the cut off point is not known so every vote allocation is really needed to predict a result with a guaranteed result. This of course is not possible until after the election.
Now say a party is brave and tries it. the best method would be to do a poll of the population on the day (or very close to) the day of the election as they will need to estimate accurately how everyone is going to vote. They sample a statically significant random distribution in the voting population and then they would be able to estimate that tactical voting would work for them. They would then need a way to inform all of their supporters to vote in a certain way. Noting that some of them would have already voted by post which the party would have to take into account, some of the voters won’t bother turning out, some will vote their own way anyway because the message wont get to them or they won’t do what they are told for whatever reason. Also they would need to know the details of who is going to vote for them in the first place (actual people not just statistical guesses because they will need inform them of what to do). Also all as the other parties have got to do is to get some of their members to say they are voting for someone different and that would wreck the initial stats and could lead the tactical voting party from a marginal win to a marginal loss for their efforts. There is no way people will know for sure what is going on well enough to risk it
In short there are two many variables that are unknown and not accurately knowable to allow any claim that tactical voting in AV could work. Anyone who suggests it (if I am only half right above) either doesn’t understand the complexity involved or are trying to cover up for the dreadfully unfair FPTP which wastes millions of votes ever election.
The big difference in FPTP is that people have nothing to loose by tactical voting. If someone supports conservative in a labour stronghold with Lib-Dems second then the ‘temptation’ to tactically vote is very strong. The conservatives have 3 choices – (1) they can waste their time by voting for their own party (it may make them feel better to have shown support but for all the good it did them they may as well have stayed at home). (2) They decide that to stay at home because they realise that its a waste of time going out. They think that the world is a better place if it wasn’t run by labour and vote for lib-dems. This is real and happens a lot and its all because we don’t have a preferential system. In AV the conservative supporters will be able to vote for their first choice first but then vote for their second choice second – brillient. Why on earth do people oppose this great way of electing our MPs.
If anyone understands this better than me (may not be that hard to find someone) please let me know if I have cocked up somewhere
“For these reasons, the Gibbard-Satterthwaite theorem does not contradict claims that the Alternative Vote would end tactical voting in British general elections.”
Sorry to be picky, but I think the logical conclusion to your argument is that “Alternative Vote would _substantially reduce_ tactical voting in British general elections.”, not “end” as you state.
It is foolish to think that, just because G-S is defined from the perspective of a perfectly-knowledgeable voter, that it implies that perfect-knowledge is required for a voter to be tactical.
In fact, the fear caused by imperfect knowledge is a stronger motivating force than perfect knowledge: G-S is more like a lower bound on tactical-susceptibility, rather than an upper bound.
I’ll quote my own comment from Mortimore’s article:
“I disagree with some people’s assessment that ‘it’s so confusing, so everyone will just vote honestly.’
“No, what they’ll actually tend to do is vote for what they consider the least-risky choice; whichever of the two leading parties they like most (or, if you prefer, dislike least).Which, by the way, is exactly the same tactical decision that voters are encouraged to make under FPTP: to pick the “lesser of two evils”, driven mostly out of fear. Humans are, by nature, risk-averse.
“The actually change from using AV rather than FPTP would be very minimal because of precisely this similarity; the biggest difference would be the increased cost for counting the ballots.”
That the L-Ds have come so far is an astounding denouncement of Labor and the Conservatives; Duverger’s law would suggest that this situation cannot hold for long however. One of the three parties will fall or be subsumed, sooner rather than later. And AV would not prevent that.
That wasn’t the conclusion I drew, though, Dale. I know that tactical voting is possible with imperfect knowledge. The point I make is that, contrary to Roger Mortimore’s assertion, G-S only proves the susceptibility of voting systems to tactical voting in cases where perfect knowledge is available.
I then demonstrated that under FPTP the level of knowledge required to make a sensible tactical choice is minimal, while with AV it is basically impossible to ascertain the information needed. I’ve yet to see a worked example of tactical voting in an AV election which looks plausible from the voter’s perspective.
Whether people vote insincerely ‘out of fear’ is entirely up to them, but that doesn’t mean they have voted tactically. A tactical vote is not merely one which is insincere; it must be both insincere AND more effective than its sincere counterpart.
On your final point, the evidence in Australia (looking at the Green vote, in particular) is that AV definitely affects the way people vote for smaller parties. It’ll also change the way parties campaign in marginals. And in a three-party system, it’ll affect the way seats are distributed nationally. It’s an upgrade rather than an upheaval, but the change would by no means be negligible.
“AV would not prevent that,” true. In fact it would most likely encourage it. Experience with AV in three Canadian provinces and Australia showed that AV exaggerates the tendency of the current system to direct all voters into a choice between two big-tent political parties, and is often less proportional than FPTP:
Wilf, your analysis of Australia is simplistic.
Australia’s society is relatively homogeneous: http://en.wikipedia.org/wiki/Demographics_of_Australia
It is normal in such societies for a two-party system to hold, regardless of what voting system is used. Look at New Zealand, which uses Mixed Member PR and has a two-party system. Or Malta, which uses STV and only has two parties in parliament.
You’ve also overlooked the substantially higher vote the Green Party gets in Australia than in the UK or Canada – this is largely thanks to AV having removed tactical considerations. Indeed, the English and Welsh Green Party overwhelmingly backed AV at their party conference last weekend.
And of course the fact that, strictly speaking, Australia has a two-bloc rather than two-party system – sustainable thanks to AV.
Any system which ends the nonsensical, undemocratic and corrupt FPTP system we are lumbered with is to be welcomed. I suspect AV will be a staging post on the road to Proportional Representation. I also suspect that most die-hard Tories will have to be dragged kicking and screaming into any sort of reform which favours the ordinary voters.