Political experiment, take 2

[triadruid]

Thanks to input from lysana and others, I have revamped the Condorcet voting poll I posted a few days ago. As before, feel free to cross-post this to your own journal or appropriate communities to increase the sample size; despite being self-selected, I hope it will be useful to show any flaws in the Condorcet voting system. I am particularly interested in how unusual voting patterns such as ties and partial ballots affect the total result.

The Condorcet method is a system that utilizes preferential ranking, rather than a single choice. The aim is to eliminate "strategic voting" or "lesser of two evils" voting, as well as provide a meaningful way to support third-party candidates. The winner of a Condorcet election is that person who wins all pairwise matchups, i.e. Anderson vs. Baker, Baker vs. Connor, Connor vs. Anderson, etc. This poll is designed to test that theory, when it comes to the 2004 Presidential election. *All candidates on a statistically meaningful number of state ballots (i.e. they would actually have a chance of winning, based on the 2000 election returns) have been included; as much as I'd like to include the right to vote for Marilyn Chambers, she is running-mate (pun fully intended) on only a handful of ballots). There was one exception to the inclusions: The Socialist Workers Party selected a slate of candidates who are ineligible to hold the offices they are running for (one is a naturalized citizen, the other is too young). The other major Socialist Party was included, however, as were the Greens. Ralph Nader has been classified as (primarily) an Independent, despite using Reform Party ballot access in some states.

The candidates are listed in alphabetical order by party; this seemed the fairest option. You may rank candidates as ties if you wish, or refuse to rank certain candidates at all (thereby assigning them a rank of '8', effectively). If you choose one and only one candidate, your vote will have exactly the same effect as the current single-choice system; I don't want to discourage this, as I am curious as to what effect it will have. At the conclusion of the poll (sometime before November 1st) I will tabulate the results of the pairings and the eventual winner.

::edit:: Results of the quiz (I've left a link up in case you want to run your own analysis) can now be found here:
http://www.livejournal.com/users/triadruid/84347.html

Comments

[publius-ovidius.livejournal.com]

Before I vote, I just want to ensure that I understand the rules:

* '1' is high and '7' is low?
* I can have ties
* I can refuse to vote for a candidate, thus ensuring that he/she receives an '8' (which is why I assume that '7' is low)

Other than that, everything seems clear.

Correct

[triadruid.livejournal.com]

1 is high, 7 is low, 8 is off the scale (as in abyssmal).

Ties are fine. Condorcet's pairwise rankings either give a win to both/all candidates, or no win to any, with respect to ties, so it doesn't statistically affect the field (they still beat everyone below them and lose to those above them). The only way it will make a difference is in the unlikely circumstance of a tie for first, but there are methods to resolve it: see a technical explanation, with tie-breaker.

Re: Correct

[omegatetris.livejournal.com]

Whoops, then I voted precisely the opposite of how I wanted to. I suppose this illustrates that any such voting system would have to have explicit instructions, because I'm sure there are people even stupider than I.

Re: Correct

[triadruid.livejournal.com]

Edited the post to clarify this; I had thought that 1 to 7 was a natural ranking system, with 8 being the default for not voting, but now it is explicit. Feel free to edit your votes, please.

Re: Correct

[elisabeth.livejournal.com]

*sigh* And I screwed the pooch here too. The trouble with editing your votes is that it seems impossible to remove a choice to make an option an 8. Ah, well. I guess a 7 will have to do...

Re: Correct

[triadruid.livejournal.com]

That's okay; functionally, a 7 loses to anything (except another 7), so it's equivalent to an 8. Some Condorcet systems count ties as double-wins; I'd prefer to set them as double-losses, although I'm willing to try both methods.

[belovedcrown.livejournal.com]

so let me understand. say 10 people vote. if 5 people say GWB is ist choice and libertarian is 2nd choice, and the other 5 people say kerry is 1st and libertarian is 2nd choice, the libertarian will win because he has 10 votes whereas the others only have 5?

Essentially, yes

[triadruid.livejournal.com]

Given a perfect voting result like that, no system is going to be able to come up with a clear winner. The libertarian has 5 wins in the pairing against the republican, and 5 wins in the pairing against the democrat. The republican has 5 wins against the democrat and 5 wins against the libertarian, and the democrat has 5 wins against the republican and 5 wins against the liberarian. It's a perfectly balanced "cycle". Note that our current "single choice" system would be unable to define a winner either: 5 for Bush, 5 for Kerry. Instant Runoff Voting eliminates the liberarian, but is then stuck.

You might try different numbers at the calculator listed at the bottom of my entry, to see what difference it makes.

[naruvonwilkins.livejournal.com]

Do note that this would never happen, actually, because the libertarian wouldn't come into play. The House would break the tie with one vote per state.

[triadruid.livejournal.com]

True, the House would resolve the matter, but it's facetious to say "the libertarian wouldn't come into play" - perhaps the House would hold Libertarian Representatives by that point.

[naruvonwilkins.livejournal.com]

It would almost definitely be a tiebreaker between D and R - since neither of them had "less than 50%". Or, that's how the law would probably be written specifically for this eventuality.

[triadruid.livejournal.com]

It would depend a lot on whether we were still using the Electoral College or not. The Constitution doesn't actually say anything about what type of voting system each state may use for selecting Electors.

[naruvonwilkins.livejournal.com]

I should hope we wouldn't be using the Electoral College - it's rather incompatible.

[publius-ovidius.livejournal.com]

Interesting results. I think it's pretty clear that most people don't have a clue what the Constitution party is since many are voting them as a '2'. This party makes GW Bush look like a tree-hugging, hippy liberal.

[naruvonwilkins.livejournal.com]

Yeah, and you'd have to give out more information - like "don't rank candidates you don't know about"

[triadruid.livejournal.com]

Yeah, I'm going to be very curious what the result of ties and blank rankings is going to be, and whether it makes any significant difference.

[naruvonwilkins.livejournal.com]

Well, there's no difference made - there's never a tie when you're dealing with a hundred million votes.

[triadruid.livejournal.com]

Sorry, I wasn't clear; I meant the 'tie rankings', such as listing both Party B and Party C as equally ranked, but inferior to Party A.

[naruvonwilkins.livejournal.com]

ahh. That's a good question - but, again, it wouldn't happen in practice.

[triadruid.livejournal.com]

How do you mean? You don't think tie ranks should be allowed, or you don't think *anyone* will do so?

[naruvonwilkins.livejournal.com]

I think we'd need to have an eventuality for them, but when you're talking about groups in the millions (or, for local races, sometimes tens of thousands), I think it'll rarely, if ever, happen.

I must not be being clear

[triadruid.livejournal.com]

I'm speaking of an *individual* voting to score Party A as their first choice, then Party B or Party C as their next choice, then the rest on down the line. Not the aggregate totals; I agree that those are statistically unlikely to show up as a tie.

Re: I must not be being clear

[naruvonwilkins.livejournal.com]

Oh, I see.

I recognize that people might want to do that - but they don't have that choice now. Would that choice be important to a lot of people? It seems that picking one party over another is the entire basis for our system. :)

Re: I must not be being clear

[christtrekker.livejournal.com]

It's important to me. I find I get to a certain point of "odiousness" in a candidate where I can't really say one is better than another any longer. I'll tie them all for last. And other times I may not be fully informed on several candidates beyond a vague feeling. I might tie them somewhere in the middle of my range. And theoretically we could bypass the primary system. There might be two candidates I'd be equally happy with, so I'd tie them for first. These situations could very well happen.

Re: I must not be being clear

[naruvonwilkins.livejournal.com]

Also, tying for last is easy - just don't rank them.

Re: I must not be being clear

[naruvonwilkins.livejournal.com]

My first response on that line didn't go through.

I don't think it's a good idea to get rid of the primaries.

What do you do now if there are two candidates you're equally happy with?

Re: I must not be being clear

[christtrekker.livejournal.com]

I don't think it's necessarily a good idea to eliminate primaries, just saying that in theory you probably could. Having just one candidate from a party is less confusing to voters, since party affiliation is a useful summary of a candidate's views.

Now, I haven't seen a time when there are two candidates I'm equally happy with. The primary system weeds out "dupes". When I vote in primaries though...those can be some tough decisions. At that level I suppose I vote for whomever can represent his views more effectively, since he still has to "sell" himself in the general election.

Re: I must not be being clear

[naruvonwilkins.livejournal.com]

*nod* That's what I do. I liked Dean and Kucinich in the Democratic primaries, and McCain in the last Republican primaries.

[christtrekker.livejournal.com]

I do have a clue, and that's why I ranked the CP 1, followed by the LP. Maybe you're just flummoxed that people can actually think differently than you?

[publius-ovidius.livejournal.com]

No, I'm not flummoxed by that. If you look at the distribution of votes, the votes for the Constitution party do not seem consistent.

Thanks for being insulting.

[christtrekker.livejournal.com]

That's pretty typical from what I've seen on LJ, though. Either you see Christianity and limited government as perfectly compatible, or you see Christianity (or organized religion in general) as the world's greatest threat. There doesn't seem to be much middle ground.

insulting?

[christtrekker.livejournal.com]

Implying people are clueless for voting CP, and drawing an analogy that makes Bush a tree-hugger (usually not a term of endearment) in comparison, is not insulting?

It was not my intent to be insulting, only to show that differences of opinion do exist.

Re: insulting?

[triadruid.livejournal.com]

I think he's talking about votes like this:
1: Democratic
1: Socialist
2: Constitution
3: Green

or

1: Democratic
2: Constitution
3: Socialist
4: Green

Philosophically, they don't *seem* to make sense. I think this speaks either a) to a weakness in the design of my poll, because Constitution is the first listed of the 'minor' parties, or b) to the CP's marketing skill, in making their name be something people who don't know anything about the party, will be likely to vote for on gut instinct. Who wants to be against the Constitution?

Re: insulting?

[christtrekker.livejournal.com]

Ahh, I see what you're saying now.

By that token, maybe the Libertarians would do better if they were simply the "Liberty" party. No one wants to be against that either.

In some states, minor parties have to fight to get the party name printed by the candidates'. Maybe this explains why.

[naruvonwilkins.livejournal.com]

But... you like Star Trek? I mean, they're socialist.

[christtrekker.livejournal.com]

I don't agree with GR's vision for a utopian future. But I think warp drives are cool, Uhura is hot, and Spock rocks. ;)

[naruvonwilkins.livejournal.com]

Hah. Okay.

[noidd.livejournal.com]

"You don't agree with me therefor you must be ignorant" seems to me what you just said.

Although I think the premise of the Constitution party platform is Lunacy I actually agree with the outcome of a lot of their policies. The Constitutition party is basically Libertarianism + Religous nutjobs.

From my viewpoint:

Smaller Government, Non-Interventionist foreign policy - all good.
Abolition of Abortion, bad.

I could never vote for them, but that doesn't mean that I couldn't rank them above someone I find more abhorent.

[publius-ovidius.livejournal.com]

If you feel that I meant "ignorant" in a pejorative sense, then no, I didn't mean to imply that and I regret if I did. I simply meant that, from what I see in the rankings, I think many people are uninformed about the Constitution party. This is a party that refers to the "abortion holocaust" and the "illegal alien invasion." They launch rather ugly attacks against homosexuals and I doubt one person in a thousand has a clue about their plans for a "debt free and interest free" monetary system or what that even means.

If they had their way, they would turn us into an isolationist, intolerant Christian nation with a disastrous economic plan. Given the vote distribution above, I don't think people meant to vote for that.

[dougo.livejournal.com]

It might be clearer (and easier) if you asked "Who is your first choice?", "Who is your second choice?", etc. with the 7 candidates listed in each question.

[triadruid.livejournal.com]

Yeah, I'd thought about doing it that way; I wanted to err on the side of freedom this time, allowing people to rank candidates as ties, etc.

It certainly would have been simpler the other way, I'm finding out as I compile the data... :P

[ashley-y.livejournal.com]

Which Condorcet method? Ranked Pairs (http://condorcet.org/rp/) looks sensible.

[triadruid.livejournal.com]

Actually, if it becomes necessary, I want to look at both Ranked Pairs and CSSD, but I have a feeling RP meets more criteria.