R-U-N-N-O-F-T

As Peter Bagge notes in his November feature, many third-party candidates champion instant runoff voting for its potential to loosen the duopoly's stranglehold. Like Bagge, I have no idea whether IRV will work as advertised, but I'm happy to give it a shot, and happier that San Francisco is instituting it for local voting on Tuesday. Thanks to Starchild, perennial Libertarian candidate for something and the LP stalwart who holds it all together in the city of St. Francis, you can get a look at how instant runoff voting works in this mock vote for the SF School Board. Of course, while you're there (sneeze, sneeze), you can show your support for Starchild, the one candidate (running on his record as a public school graduate!) who can really work out the kinks in the school system.

Update: In the comments, Ironchef points out that the School Board mock voting page is an example of Approval Voting, not Instant Runoff. Here's an explanation of IRV, with an endorsement from 1980 footnote John Anderson. I don't have enough of that thar booklarnin' to comment on the pros and cons, I just know a vote for Starchild is a vote for the future!

Comments

[Abiola Lapite]

Instant runoff voting can also lead to rather unexpected and undesirable results, and in particular, it does nothing to stop spoilers from determining the outcomes of elections. I think the LP would be much better off with approval voting (also see here), which would at least guarantee that wackos like Badnarik never get elected in the future.

[Ironchef]

Thanks for the link, Abiola. But your criticism is not terrible.

IRV, also known as Ranked Choice Voting (RCV) seems to be the best way for me to express whom I would like to see in office.

Any other choices involving "tactical voting" or "strategic voting" go against the intents and desires of my voice. I wouldn't want whackos like Gore ("sticks and stones". . . ) to whatsoever, and would be happy to place him as my 29th choice in a field of 10. (just above gnat shit)

Regardless, does anyone else notice that the SF school board mock election is not IRV, but Approval voting? Am I getting this wrong?

Just

[Chuck]

Any voting system can lead to unexpected and/or undesirable results (see Arrow's Theorem). Instant runoff voting would at least enable people to vote for those whom they really want, without worrying about "wasting" their vote. I'd bet that the number of votes the LP candidate gets as first choice would at least double if this system were adopted.

On the other hand, we have a large number of voters who can't understand the ballots we have now--just imagine voters in Palm Beach County trying to make sense of an instant runoff ballot...

[Randy]

Chuck, it's so sad when we have to cater to the most stupid. That is just no excuse to not implement Instant Runoff.

[Greg]

Sigh, the bad idea that never dies. It was once called "Single Transferable Vote" until it was thoroughly discredited. I guess it's now being rehabilitated and sold as "Instant Runoff". But it's damaged goods.

In STV or "Instant Run-off" it's possible for a candidate to lose an election because he got too many votes. That's actually worse than the "spoiler" effect in a straight plurality election.

That is, say you're deciding between voting between Kerry or Nader. In the current system your worst fear is that by voting for Nader you might make Kerry lose the election to Bush. In instant run-off it's possible that in voting for Nader you would make Nader himself lose the election to Bush.

That's a violation of the single most basic criterion for an election system. That a vote for a candidate should never hurt the candidate's results.

There's no need to choose an election system based on vague explanations or hunches. There's actual mathematical analyses of election systems.

[Bill G]

The Wacky Left in San Francisco sees IRV as the key to their wresting complete control of City politics away from the other end of the local political spectrum, the Machine Democrat Left. No one to the right of Ted Kennedy could get more than their family's votes in San Francisco.

I doubt that it will lead to any great differences in this election vs. what a traditional runoff system would have produced.

Anyway, the Wacky Left is already arguing that San Francisco isn't implementing true IRV [which would allow *every* candidate to be ranked] since voters will only have the option to rank their top 3 choices.

This is reportedly because of technical limitations of our local voting technology [machine read "connect the arrows" paper ballots], but the Wacky Left smells a conspiracy.

[Abiola Lapite]

"Any other choices involving "tactical voting" or "strategic voting" go against the intents and desires of my voice."

I hope you're aware that there's a mathematical theorem that shows that no electoral system short of dictatorship can prevent tactical voting. IRV certainly won't do the job.

"Instant runoff voting would at least enable people to vote for those whom they really want, without worrying about "wasting" their vote."

So would approval voting, and it has two additional merits:

1 - It's simple and transparent to administer, unlike IRV (look how long it's taken to tally Australia's Senate election results).

2 - It can't lead to freakish outcomes like that in which Le Pen squared off against Chirac, despite the majority of the French election leaning to the left.

Of course, you'd know all this if you'd bothered to follow and read the links I provided ...

[thoreau]

OK, the mathematics of voting is one of my favorite hobbies. I'm working on a theorem concerning strategic voting, and if I ever finish it I'll have a nice little publication that will be ignored by countless real politicians around the world 😉

Instant Runoff Voting does indeed give incentives to vote insincerely. So does our current system (where those of us who support, say, the LP have an incentive to insincerely vote for our lesser evil amongst the 2 major parties). The incentives aren't as common as in our current system, but they're present. There's a mathematical theorem called the Gibbard-Satterthwaite Theorem, which states that any ranked voting system (system where you indicate your ranking of the candidates) is susceptible to strategic voting.

(Before anybody says "Whoa, I thought Arrow's Theorem proved that!", Arrow actually proved that a different property can't be satisfied. Gibbard and Satterthwaite were the ones who addressed strategy. The logical foundations of both theorems are closely related, however.)

It is also true that in Instant Runoff Voting one can lose by getting too much support. Consider this very simple example with 3 candidates named A, B, and C.

40 voters have the preference A>B>C
31 voters have the preference B>C>A
29 voters have the preference C>A>B

C has the fewest first-place votes and is eliminated. All of C's votes are transferred to A, and A wins. Now, if 3 or more of the people with the preference B>C>A decided that they really do like A, and instead listed the ranking A>B>C, B would have the fewest first place votes. B would be eliminated and the runoff would be between A and C. C would win. So A getting more support caused A to lose.

(Yes, I know, this is a simplistic example, but the same phenomenon can occur in more complicated situations as well. I just wanted to illustrate the point.)

Anyway, I personally prefer Approval Voting (www.approvalvoting.org), but unlike most Approval Voting supporters I don't think IRV is the end of the world. IRV would still give third parties more chances to compete, even if it isn't perfect. So, I'll always prefer Approval Voting to IRV, but I don't go around bashing IRV.

Also, the "Single Transferable Vote" is the name given to a version of IRV that can be used to elect multiple candidates. It's more complicated, but it is used to elect (among other things) the Australian Senate and one of the chambers of the Irish Parliament. I'm not endorsing it, but it is incorrect to say that STV was discredited.

[SR]

Having read the links you posted, Abiola, this sounds like an interesting system, but I find it difficult to reconcile with the "one-man, one-vote" principal, since someone who marks every choice on a slate of five candidates effectively gets five votes, while a voter who hates one of the candidates and doesn't want to see him elected at any cost only gets four votes. Also, although the first site claimed one of the advantages of approval voting is that it would reduce negative campaigning, to me it seems that this system would lead to the filthiest campaigns imaginable. Here's why:

Under winner-take-all, a candidate merely has to convince voters to choose himself, since they have only one vote. Under approval voting, a candidate must also convince voters to *not* vote for his opponents, otherwise the votes will cancel each other out.

[SR]

Sorry, that should be "principle". Where's an edit button when you need one?

[Abiola Lapite]

I see that Thoreau's already gone into the issues I mentioned at greater length. The Gibbard-Satterthwaite Theorem was the very result I had in mind when I said strategic voting was unavoidable.

"Under approval voting, a candidate must also convince voters to *not* vote for his opponents, otherwise the votes will cancel each other out."

Why is that necessarily a *bad* thing? Don't you want candidates to engage in robust criticism of each other?

[kwais]

Gregg,

Can you explain this;

"In STV or "Instant Run-off" it's possible for a candidate to lose an election because he got too many votes."

How do you lose an election because you got too many votes in a runoff system?

[thoreau]

OK, in regard to Approval Voting and negative campaigns:

Every election reform in the history of the world has promised to reduce negative campaigning. Approval Voting advocates claim it. IRV advocates claim it. Fans of esoteric methods like Borda and Condorcet even claim it. They also claim that their methods can slice and dice, remove stains instantly, and groom your dog. But wait, there's more!.... :->

Seriously, though, the main idea of any alternative system is to make it possible for people to express more information on the ballot than just one vote. In the case of IRV the goal is to express an entire ranking. In the case of Approval Voting the goal is to express approval or disapproval of each candidate (I'll get to one-man-one-vote in a moment), not just a vote for a single candidate. In both cases, the idea is that with more information more parties will be able to compete in the marketplace of ideas. If people still support the D's and R's then so be it. But at least we'll know it's because of the ideas, not because we have a system that gives strategic incentives to quickly narrow down the field.

More information, more competition. That combination should have some libertarian appeal.

As to one-man-one-vote: I've been in these arguments before. I'm on a mailing list of election method afficionados. If you take the most literal definitions of those words then Approval Voting (hereafter denoted AV to save me typing) flunks it. But there are counter-arguments:

1) Many people find the idea much more appealing if I frame it as "vote yes or no for each candidate, and the candidate with the most yeses wins." The goal of the method is to find out who has the greatest overall approval. If you decide to approve more than one person then do so. If you think it's advantageous to support multiple candidates then instead of complaining that somebody else supported multiple candidates you can do the same. Or, if you think it's advantageous to only support one candidate, then you have no reason to complain if somebody else does something disadvantageous and supports multiple candidates. Under this system the choice is yours.

2) In many local elections we already use systems that allow multiple votes. Here in Santa Barbara, if there are 3 school board seats up for election you usually get to vote for up to 3 candidates if you want. Those who want to support as many as possible can vote for 3. Those who are passionate about a single candidate can vote just for him, to avoid supporting any of his competitors.

3) The spirit of one-man-one-vote is equity and impartiality. Every person's vote is treated the same as any other's in AV. AV doesn't have anything to do with weighting some people's votes more than others. Your vote for a given candidate counts the same as anybody else's vote for that candidate.

[max power]

The approval voting is not in violation of the one man/one vote principle. In your example, SR, each person is getting to weigh in 5 times, by choosing "yes i approve" or "no i don't approve" for each of the five candidates.

I know it's counterintuitive and I've asked all the questions myself, but approval voting is much better than IRV. More info here:

http://approvalvoting.org/

[SR]

Abiola, I was reacting to one of the sites you linked to, which expressly claimed that an advantage of approval voting is that it would reduce negative campaigning. My post was just pointing out that I didn't think that would be true. I don't particularly mind negative campaigning myself.

[SR]

Thank you all for your thoughts on the one-man, one-vote issue. I see your points but, honestly, if I was a Supreme Court justice, I would strike down such a system as unconstitutional. The system where each voter gets as many votes as there are candidates and can apportion those votes between the candidates, seems both more equitable and more in keeping with the one-man, one-vote principle than approval voting to me, but would appear to accomplish many of the same results.

[thoreau]

if I was a Supreme Court justice, I would strike down such a system as unconstitutional.

SR-

The phrase "one man one vote" is nowhere in the Constitution. What is in the Constitution is the phrase "equal protection of the laws." A law implementing Approval Voting would give every person the right to vote yes on as many candidates as he supports and the right to vote no on as many candidates as he opposes. There's no equal protection violation here.

(And yes, I know, just because something doesn't appear in the Constitution doesn't mean you can't get 5 Supreme Court Justices to rule based on it. I'm was just assuming that SR would be a little more scrupulous than that.)

As to the system that you mention, where you have votes that you can distribute amongst candidates, the best strategy there is to give all of your votes to your favorite among the 2 front-runners. It is strategically equivalent to our current system and substantially different from Approval Voting. Where that system has some merit is when you're electing multiple winners. A small faction can concentrate its votes, a large faction can spread its votes among several candidates, and groups will be represented roughly according to their size (in theory at least, although I am well aware of the drawbacks in this system).

[Warren]

I'm not a fan of IRV. Mostly because I'm afraid it would give real power to the watermelons. Just because we share the same frustrations with the establishment doesn't take away from the fact that they are far far worse.

[thoreau]

Warren-

The most likely outcome of an alternative voting system, be it IRV or Approval, is not the rise of the Greens or the LP. Initially a handful of Greens and Libertarians might be elected just because people will be fascinated by the novelty of a system that lets you support somebody other than a D or R without fear of helping your "greater evil" win.

But the Greens and LP are far from the center. I don't see either party winning more than a handful of races with IRV or Approval. More likely we'd see the rise of new parties that are fairly close to the center but distinct from the D's and R's. A party that's fiscally conservative, socially liberal, and above all else sane (unlike some LP candidates) would probably emerge. A fiscally liberal and socially conservative party would probably also emerge, but with a platform that's less extreme than, say, Buchanan.

The action will remain near the center, but the variations on that theme will increase.

[SR]

thoreau, I recognize that "one-man, one-vote" is not explicitly in the US Constitution. My Supreme Court comment was based principally on the precedents of striking down "at large" voting districts and requiring reapportionment of districts to maintain approximately the same number of voters. Those decisions place a very strong emphasis on the literal understanding of "one-man, one-vote". Approval voting appears contrary to those rulings, as it allows some voters to exercise more influence over the election than others.

[Doug]

"One man, one vote" only refers to equipopulation of legislative districts. That's all . . .

[Abiola Lapite]

"Abiola, I was reacting to one of the sites you linked to, which expressly claimed that an advantage of approval voting is that it would reduce negative campaigning. My post was just pointing out that I didn't think that would be true."

SD,

To be fair, there is some strength to the argument that approval voting might drive down negative campaigning. The key thing to take into account is that attack ads tend to drive up negativity ratings for both the attacker and the attacked. It won't make a difference in a 2-way race, but in more open contests, all it tends to do is give a relative boost to uninvolved parties.

[Allan]

I highly recommend that everyone here check out ElectionMethods.org, and take a look at the pros and cons of all of the various ways of handling voting before picking a favorite. The technical evaluation of various voting methods is particularly informative.

I used to be a supporter of IRV until I read through the material on ElectionMethods. The fundamental problem with IRV is that it discards voter preference data during each runoff. Fortunately, it turns out that there is a similar, but better voting method, known as the Condorcet method. Condorcet works like IRV in terms of how you cast your votes, but the vote tallying is done differently (as pairwise races between all of the candidates). Condorcet avoids many of the problems with IRV. If Condorcet can't be implemented, then Approval voting is a good second choice. Not quite as good as Condorcet theoretically speaking, but still better than IRV.

[Dan]

It won't make a difference in a 2-way race, but in more open contests, all it tends to do is give a relative boost to uninvolved parties.

If I recall correctly, that's the widely accepted explanation for Jesse Ventura's election as governor -- the Democratic and Republican candidates alienated the voters with excessively negative campaigning.

Anyway, it seems to me that approval voting would reduce negative campaigning just by giving candidates a reason to campaign positively. Candidates only have limited campaign resources, after all.