Voting

[andrewducker]

Following discussions here in a post about Condorcet voting, I thought I'd ask people what voting system people would like for a single-winner voting system.

Let's say you have three parties:
Marmite, Vegemite and Peanut Butter

Vegemite lovers hate Marmite and vice versa. Peanut Butter isn't loved by as many, but isn't hated either.

You have 249 voters. Everyone ranks their preferences in order.

100 people vote marmite 1, peanut butter 2, vegemite 3
99 people vote vegemite 1, peanut butter 2, marmite 3
26 people vote peanut butter 1, vegemite 2, marmite 3
24 people vote peanut butter 1, marmite 2, vegemite 3

FPTP: Marmite wins, because only one vote per person counts, and more people voted marmite than anyone else (100->99->50).
AV: Vegemite wins, because after peanut butter is knocked out of the running, and the second-place votes are transferred, it has 125 votes to marmite's 124.
Condorcet: We compare each option to each other one, and see which one wins. Peanut Butter beats Marmite (149->100), Peanut Butter beats Vegemite (150->99), Vegemite beats Marmite (125->124). As Peanut Butter wins all of its battles, it's the victor, Vegemite is in second place.

[Poll #1637922]

Comments

[andrewducker]

The principle objection to FPTP is that it encourages huge amounts of tactical voting and partisan thinking.
The principle objection to Condorcet is that the least objectionable candidate wins (although whether you see that as bad or good is a personal choice, of course).
AV is generally seen as a wishy-washy "Better than FPTP but not as good as others." approach by people that prefer other voting systems. Nobody seems to love it, but it's not seen as that bad.

[bracknellexile.livejournal.com]

FPTP hates Condorcet, and vice-versa. AV isn't loved by as many but isn't hated either. How very meta :)

ETA: By picking the least objectionable candidate, isn't Condorcet more wishy-washy than AV?

[e-halmac.livejournal.com]

I've seen AV in action and the results can be surprising. I do think AV-type voting systems are more fair.

In other news, I hate vegemite AND marmite, so would always be on the side of peanut butter.

[anef.livejournal.com]

Clearly the answer is to mix marmite, vegemite and peanut butter together in proportions according to first choice popularity. That would satisfy everyone.

[andrewducker]

That's why I said "single-winner" (like a mayor or president).

For large groups I favour either STV or AV+ (the Scottish method seems to work pretty well).

[momentsmusicaux.livejournal.com]

> 100 people vote marmite 1, peanut butter 2, vegemite 3

Which way round does that mean? It's ambiguous.

[andrewducker]

Marmite in first place, then peanut butter, then Vemegite.

Sorry!

[simont]

I found I was mostly influenced in my choice by your gloss on the suggested voting results, which makes it look as if there's some sort of political spectrum of opinion on which M and V are opposed extremist parties and P represents the reasonable moderate centrists. My instinct is always to go with the reasonable moderate in such cases, in the absence of more specific information which might show that one of the extremist parties is actually right. Hence, it seems clear that I would want P to win, and that's what I voted.

However, with a different gloss I might have felt differently. You might have said instead, for example, that there was near-universal consensus that P was terrible, so that all the MPV and VPM voters mentally put P as only just better than whichever one they placed third. Such a situation might arise, for example, if the M and V parties were each going to benefit about half the country at the expense of the other half (say, one rural and one urban), and P were naïve idealists who meant well but were probably going to wreck the economy in practice and not benefit anyone (and the first-choice-P voters were people who had foolishly believed the idealism). That gloss could have described the same set of voting results, but it would have made me want P to be thrown out early and the marginally more popular of M and V to win.

And neither of those pieces of reasoning has anything to do with each voting system's internal rationale for selecting a given winner!

[andrewducker]

Yes. You pretty much have to assume the veil of ignorance over what the actual parties are.

[opusfluke.livejournal.com]

Option #4: Make Voting Mandatory So People Get Off Their Arses And Vote At All.

[ciphergoth.livejournal.com]

That's orthogonal to the choice of voting system, not a separate option.

[andrewducker]

Oh, God, No. I do not want people voting who do not care about it. I want passionate, educated voters.

[bracknellexile.livejournal.com]

No, no, no! Not unless you give me the option to vote "None of the above" on a ballot paper. Lack of turnout isn't always apathy; some of it is "I can't honestly support any of the candidates so I choose not to vote."

[spacelem.livejournal.com]

Problem with Condorcet: what happens when large numbers of people just vote for Marmite, and don't give a 2nd or 3rd preference? Since they hate Vegemite so much, why would they even list it? There's nothing that says you must list a 2nd or 3rd candidate -- only the 1st. This leads to large numbers of battles that just aren't fought.

The only reason to include a 2nd vote is specifically to prevent your least preferred from getting in, sort of an "I'll settle for peanut butter, because it isn't vegemite" option. This isn't a problem for either FPTP or AV.

[andrewducker]

Would that be a problem? You'd count Marmite as first place, and then count the others as being a "loss" when doing the calculations for who wins.

[skington.livejournal.com]

The thing is, though, that in an AV system it's perfectly reasonable to e.g. vote Green 1, LibDem 2, Labour 3 , SNP 4 and miss out Conservatives entirely, and not necessarily prefer the Greens to win. Perhaps you prefer the LibDems, and expect them to be significant contenders for the seat, but you want to express your support for the Greens, knowing full well that they'll get knocked out fairly quickly and then your vote will go to who you actually want to get in.

[naath.livejournal.com]

If you really care about Vegemite not winning you then have to rank everything so as to rank Vegemite last. On a 3-way election this is quite easy (and merely requires that you understand the voting system); on a 100-way election it is rather more tricky...

If you actually think that if the options were Vegemite or Peanut because you hate them both equally - then you just don't care who wins the VvP fight, and either way you're equally unhappy.

[ciphergoth.livejournal.com]

On a given ballot, unlisted candidates are treated as last equal.

[communicator.livejournal.com]

I'm coming round to FPTP because of all this fiddling and wiggling by political wonks like me running rings round ordinary voters who don't have time or inclination to master the intricacies of this kind of stuff.

[andrewducker]

AV is actually simpler, IMHO, because you remove the need to think about voting tactically.

In a three-way split, where you'd be happy with two of the parties, but hate the third, knowing how to vote in FPTP is very hard. Knowing how to vote in that situation with AV is trivial.

[ciphergoth.livejournal.com]

You have it backwards. FPTP is conceptually simple, but life is complex for the voter and they have to know a lot to vote effectively. Condorcet is complex, and makes life simpler for voters because they can cast an effective vote without understanding the polling in their area.

As a programmer, I'm very used to the choice between implementing something complex to make life simpler for the user, or keeping things simple in the implementation and pushing a job onto the user. See eg worse is better.

[zz]

I don't like any of those, and want raspberry jam on mature cheddar cheese.

[naath.livejournal.com]

So stand for the jam party then :-)

[matgb]

See, you can't just say single winner, no iteration.

All elections are always an iterated series, so behaviour changes over time and adapts to previous voting patterns.

If deciding where to go for a meal, Condorcet wins, with the proviso that people with allergies get to override the peanut butter choice (because, y'know, voting to kill me? Thanks).

But for electing MPs, Condorcet clearly massively strengthens the centrist party, but in a way that's seen as illegitimate. But for mayoral elections, I can see the argument in favour.

Except that those opposed will a) shout about how complicated it is in a populist way and b) whinge about 'losers' getting in, which lets face it is true.

Condorcet would undoubtedly have seen Brian Paddick as Mayor of London. Do you really think, regardless of suitability, that most voters would see that as legitimate or "fair"?

I don't. Hence, while in theory I like it, and would happily use it for internal stuff, you need a sophisticated electorate to understand it, and most people don't care enough to make themselves sophisticated.

[andrewducker]

Yup, my basic objection to Condorcet is that the average person will not understand it, and everyone will be confused why we have the government we do.

[martling.livejournal.com]

I find it interesting how you chose to pose your actual poll question: despite having named the systems, you then chose to identify them in the poll by the winners in that particular situation, despite the question being "Which of these three systems would you rather have". Did you do this deliberately?

It's hard to get people to think purely about what choice of voting system would be most appropriate in principle for a given application; they will always tend to be more interested in what result a given system would give *right now*. Actively posing the questions in this way doesn't help.

Of course, as long as it's possible for the winners to change the rules, it perhaps isn't really unreasonable for people to think this way.

Have you ever played Nomic?

[andrewducker]

I specifically didn't want people saying "But I like AV, so I'll choose that." I wanted them thinking "In this situation, given these numbers, I think that the real winner should be X."

I think I've played Nomic once - I'd like to do it more, but it's not something that appeals to my friends much.

[greenfieldsite.livejournal.com]

I think this is quite a tricky one. Given these preferences the best choice of Marmite, Peanut Butter and Vegemite depends on the relative level of support as well as the order. For example, some people might rate their 1st and 2nd choices quite closely. Others may rate their 1st choice well ahead of the rest.

I don't like FPTP (and I am planning to vote for AV in the referendum), but I think it has one thing going for it. If you are elected under AV then you may be a compromise candidate who has only inspired a small proportion of the electorate to vote for you as a first choice. Your election may be, effectively, as the lesser of evils. To be effective I believe an elected politician needs support and a network of contacts.

Consider two politicians. One has 40% of the electorate actively supporting them and 60% disliking them. The other has 20% of the electorate actively supporting them and 40% who think they are not great but better than the first politician. If the latter reached the AV run-off then they would win, but they would have a smaller pool of enthusiastic people to draw on for active support after being elected.