Instant runoff voting – the system by which the voter can rank choices by preference in an election with more than two candidates, thus eliminating the third-party “spoiler” problem – just had its first major trial run in the U.S., the San Francisco municipal elections. There were a few glitches, but on the whole, IRV seems to have worked out just fine. From the San Francisco Chronicle, http://www.sfgate.com/cgi-bin/article.cgi?file=/chronicle/archive/2004/11/03/BAGPN9KOG41.DTL. Congressman Jesse Jackson, Jr. (D-Ill.) has introduced a bill to conduct all federal elections by IRV – http://www.fairvote.org/irv/jacksonirvbill.htm. Will the idea spread?
Not to the federal level. The parties would have too much to lose. Things have not changed since last year when you started this thread on the same subject.
It better not. IRV is even worse with regards to viable third-party candidates than the current system is. I agree that we definitely need a new voting system, but the approval and Condorcet methods are the only ones we should be considering.
I’ve never heard that – what’s the reasoning?
We used this system to elect the leader of out political party (NDP, in Canada), and worked out great. Fewer divisions, and we got a leader most of the party likes and can stand behind – first one who really inspired us for ages.
I don’t know how well the system would work on a very large scale, but I’m glad to see in working well in San Francisco.
I’m kinda pressed for time right now, so I’ll fill in the details later, but in short, it doesn’t eliminate the need for strategic voting, it allows a winning candidate to lose by receiving more votes, and it fails the Condorcet criterion, which is just slightly technical but very important.
Here’s a hypothetical scenario: There’s a Democrat, a Republican, and a Libertarian running for some office. Polls are showing that the first choice is pretty evenly split. You prefer the Libertarian to the Democrat, the Libertarian to the Republican, and the Democrat to the Republican. Which way should you vote?
Of course, in this hypothetical scenario, IRV is in place.
OK, I have a little more time now.
There’s all kinds of stuff you can read about how bad IRV is–here are a few articles: an overview of why IRV is bad, a couple responses to IRV advocates by Condorcet advocates, and some specific examples of how IRV can fail to elect the candidates that most people prefer.
If for some reason you don’t like that site, you can try this short article, although you’ll note it links to the first article above.
That’s just a pretty general look at IRV, though. This page describes some of the mathematical criteria that are used to evaluate voting systems, and IRV fails to have every single property mentioned, some of which are pretty fundamental.
Consider the monotonicity criterion, which says that moving a winning candidate’s ranking higher should never cause that candidate to lose. IRV is non-monotonic, which means that sometimes a losing candidate could win by getting votes thrown out. You think Bush v. Gore was bad…
Condorcet voting is the only voting method that satisfies the Condorcet criterion. The Condorcet criterion just says that if there’s one candidate who can beat every other candidate in a one-on-one race, that candidate should win the overall race. I’d say that’s pretty important, and I think most people would agree.
Approval voting doesn’t satisfy the Condorcet criterion, but it does have some other nice properties that IRV and the plurality method don’t, and it’s dirt-simple to boot. If you’re going to push for voting system reform, push for one of those two.
I’ve never seen the Condorcet system before, but it looks like it has a lot of merit. Thanks for the information, ultrafilter.
Here’s the concrete example of IRV’s problems that I like to use:
- There are three candidates in an Instant Runoff election: a
Republican, a Democrat, and a Libertarian. The votes are as follows:
6 people vote #1 Republican, #2 Libertarian, #3 Democrat.
2 people vote #1 Libertarian, #2 Republican, #3 Democrat.
5 people vote #1 Democrat, #2 Republican, #3 Libertarian.
4 people vote #1 Libertarian, #2 Democrat, #3 Republican.
The Democrat has the fewest #1 votes, so he is eliminated. His five
votes are given to the Republican, who now wins, 11 votes to 6.
- Now suppose those two L,R,D voters had decided to rank the
Republican in first place instead, joining the larger group of R,L,D
voters. The votes would be as follows:
8 people vote #1 Republican, #2 Libertarian, #3 Democrat.
5 people vote #1 Democrat, #2 Republican, #3 Libertarian.
4 people vote #1 Libertarian, #2 Democrat, #3 Republican.
This time, the Libertarian has the fewest #1 votes, so he is
eliminated. His four votes are given to the Democrat, who now wins, 9
votes to 8.
See what happened? By ranking the Republican higher, those two voters
caused him to lose.
What exactly is Condorcet voting? That is, what are its mechanics? Approval voting I understand easily enough – you just check the name of every candidate acceptable to you; it’s like IRV but simpler, with no rank-ordering. (I still feel a certain preference for IRV just because having a chance to rank one’s choices is more emotionally satisfying to the voter, but I might be dissuaded.)
From a voter’s perspective, Condorcet works the same way as IRV. You rank the candidates in the order you prefer them, possibly leaving out certain candidates if you don’t have a preference about them.
Each ordered ballot is converted (behind the scenes) into a matrix representing pairwise matchups between each combination of two candidates, and all these matrices are added together to produce the final vote count. After applying some math, you can determine the winner, unless there’s a genuine contradiction in voters’ preferences. ElectionMethods.org has a more detailed explanation.
In that case – since they are the same from the voter’s perspective – I will continue to tout “IRV,” if you don’t mind, with the implication that that includes the Condorcet system as a possible alternative vote-counting mechanism; IRV is a term that has a head start in the publicity war, and it’s easier to explain.
The problem is that when you say IRV, people hear IRV. Anyone who has heard of IRV will think “oh, you vote for a list of candidates, and if the first guy loses, your ballot goes to the second guy”. IRV isn’t just a ballot with boxes to write in a number for each candidate, it’s also a system to count those ballots, and that part of it is horribly flawed.
If you must tout something simple that’s already in voters’ minds, go with approval voting: it works the same way as our current system, except voters can choose as many candidates as they like.
If you really think the difference is that important, you ought to join the Center for Voting and Democracy (http://www.fairvote.org/), and bring your concern to the attention of Rob Richie, the national director.
ultrafilter, interesting links about IRV. I’ve always thought it would be best suited for multi-seat elections, but now I see that is not the case, either. Thank you very much for the link. (I’ve been a fan of approval voting for presidential elections no matter what your site says so there 
)
I have a paper you might like reading about Condorcet methods and democracy. If you want, I can email it to you. You can DL it here: “Epistemic Democracy: Generalizing the Condorcet Jury Theorem” by List and Goodin.
it is mildly off-topic so I won’t bother quoting anything here.
No – for multi-seat elections, that is, elections to a multimember policymaking body such as Congress, the best method would be some form of proportional representation, aka full representation. See http://www.fairvote.org/ and http://en.wikipedia.org/wiki/Proportional_representation.
But they’re not the same from the voter’s perspective except for when they’re filling out the ballot. Condorcet voting is strategy-free, and the standard form of IRV is not.
I agree that approval voting’s not all that great. Consider a scenario with four candidates (A, B, C, and D) and 12 voters, where A, B, and C are equally favored as first choices, and half the population favors D as a second choice, while the other half doesn’t like anybody other than their first choice. Here’s what you get:
A: 2
A, D: 2
B: 2
B, D: 2
C: 2
C, D: 2
Each of A, B, and C got 4 votes, and D got 6 votes, so D is the winner, even though s/he was nobody’s first choice.
But like I said before, approval voting is better than the plurality method, and it’s much simpler to explain than Condorcet or instant runoff voting. That makes it attractive as an immediate goal for vote reform.
From your Wikipedia link:
Bolding mine.
Well, BrainGlutton, I guess that depends on what you want out of your representatives. I generally like a race to the middle because I hope it means something approximating fiscal responsibility and social liberation (or at least not going backwards), but I also would like the dialogues that extra parties could bring to the table. I love libertarianism so long as, you know, no one actually puts it into practice, and I’m a big fan of universal health care but right now there’s simply no intelligent discussion about it (excepting maybe here at the dope). I want to see libertarians at debates. I’d like to see someone like Nader getting more press if he’d butt out of the presidential election. The advantage extra parties offers, to me, is not the chance of better representation per se but rather that they’d open up the dialogue a lot more. I’m sort of from the frame of mind that I don’t care who is in the government if they represent my interests, but I don’t want them to tell me what my interests are, I want to tell them. Realistically, I think open dialogue serves my interests vastly better than simply tossing more opinions in Congress by creating five parties or something, since I already know they aren’t really going to listen to me to form their opinions.
Additionally, even a legislative body requires a selection criterion, and in the case of the US it is a simple majority (excepting amendments to the constitution). This means representatives will further have to compromise… meaning that even if you manage to get a green party candidate in, you’re still not being represented the way you might think from the result of voting alone.
Something to chew on.