Monday, 7 February 2011

Why vegetarians are bad for AV

Fairness is often subjective, at least when it comes to electoral systems, so I am taken aback at the unquestioning ease with which the more ‘enthusiastic’ proponents of the alternative vote (AV) have appropriated the term. Ever contrary, I adapted an earlier example of how AV can lead to an unfair result, by using the dinner analogy popular with such support.

Imagine you are in a party of 21 and there are four restaurants within reach for a work-time lunch; in alphabetical order: Lentil Heaven, Liver Lounge, Pizza Palace and Thai Temptation.

One strange individual plumps for Liver Lounge and puts down Thai Temptation as his second preference. Nine people vote for Thai Temptation as their first choice; it subsequently turns out their second choice isn’t relevant - likewise for the seven people who choose Pizza Palace. The four people who remain are vegetarians, they’re cool; they don’t really care where they eat so long as they can avoid the liver.

That’s fair enough, and the best way to achieve this is a feature of AV that isn’t afforded by our current first-past-the-post (FPTP) system; the ability to vote against one option – by voting for all the others. In our example the vegetarians do exactly this, and since they don’t have a preference with the other options they rank them according to the order in which they are listed. This is called a donkey vote and whilst it is most common where full preferential voting is required, it applies equally here.

With the vote in, the count begins, and after the first round Liver Lounge, having the fewest votes, is eliminated. The next preference of that singular person is now added to the vote to give us the following: just over 47% of you would like a Thai, with around 33% opting for Pizza. It’s still not quite a majority so Lentil Heaven is eliminated… and something interesting happens. The ‘second preferences’ of the vegetarians are added to the count; those would be the people whose only real preference was to avoid the liver and who consequently voted for the other options in the order that they appeared - despite having significantly less of the ‘positive’ vote, Pizza Palace wins through having a superior position in the alphabet.

What this illustration and others show is how AV can work; the result - fair or otherwise - depends on how the example is framed. What this particular example shows is that under AV it's possible to have a result that many people would consider unfair; it's a system not quite as simple as some would have us believe.


  1. Having written an original example that I think you refer to:

    I can say that you have simply pushed the analogy too far.

    AV for general elections involves thousands of votes, not just a handful - trying to work *detailed* examples of AV with a handful of people like you have doesn't work and isn't representative.

    Use restaurants to demonstrate splits by all means, but don't start attaching detailed figures and expect to get anything sensible out.

  2. That's not necessarily how people will vote under AV!

    I have a Lab MP that i don't like in my constituency. I would vote against her like this:


    Now that goes against what you just said. I would give me real preferences. I would much prefer SNP over Conservative even though on a ballot paper your suggesting i would vote Conservative ahead of SNP.
    There was a debate on this in the HoC & i'm sure they made sure that the order on the ballot paper is mixed up so that doesn't happen.

  3. @pop - I do (kind of) agree with you. I wanted to demonstrate that the dinner party analogy works both ways; it can show AV in a -ve as well as +ve light. To suggest it can only be used with percentages suggests we need a different analogy? Perhaps one along the lines of three candidates we don't particularly care for and an extreme right wing candidate - I can imagine/hope people would come out - in large numbers - to vote against the extremist even when they don't particularly care for the others?!

    @Nic - You're absolutely right, it isn't necessarily how people will vote, but it *could* be. However, my example is't saying you'd vote "Conservative" - they're the "Liver" option in my example! Taking the order you've shown - is it a vote for the "Lib Dems" or is it a vote against the "Conservatives"? If I didn't give a hoot for the "Lib Dems" or "SNP" but really disliked the "Conversatives" I would vote *exactly* the way you describe - which gives the "Lib Dems" an unfair advantage don't you think? Unless of course there order is mixed up between batches of ballot papers - which is how they mitigate it in some parts of Australia.

  4. ====
    I posted this at

    Phil, you present a valid issue (at last someone does!! :) )

    If people really don't care between a number of lower options, but want to rank them so they can put someone *last* what would really happen?

    If they all 'happened' to put their "don't really cares" in the same order, it could potentially give a result as if a lot of people really do care between those lower order prefs...

    The questions are then:

    - would this really happen and influence a result?
    - If not then meh.

    - But if it could/would then is it a worse side-effect than FPTPs' side effects?

    - If so can it be mitigated to the extend that it doesn't?

    I'll think on it - do you have any thoughts already?

    (My firs thought is that I'd suspect this is unlikely to matter, because it is only lower order prefs - which are very unlikely to be counted - would have to still be in after higher pref went out etc...)