A case for instant runoff voting

by John Quiggin on September 19, 2005

This NYT article[1] discusses the problems New York Democrats are having with their primary system. If they use first-past-the-post, given a large field, they end up with candidates supported by only a minority of voters, who in turn are an even smaller minority of Democrat voters. So they have had a runoff system when no candidate gets 40 per cent of the votes, but this has caused divisions and delays.

The solution is obvious: adopt the instant runoff/single transferable vote/optional preferential system, listing favored candidates in order of preference and omitting those for whom you don’t want to indicate a preference.

Obvious as it is, this idea is almost certainly unsaleable in the US context. For reasons I don’t fully understand, the US, which was once a pioneer in all kinds of institutional innovation ( the primary system itself, for example, or decimal currency) is now intensely conservative about such things. The NY Dems have been radical, by US standards, in going as far as a runoff, and according to Wikipedia only outliers like Ann Arbor and San Francisco have been willing to try the instant runoff system.

More generally, in looking at the US, I’m struck by the fact that, with so many independent jurisdictions at all levels, there isn’t more institutional variation and experimentation. For example, AFAIK, all 50 states share the Federal model of a bicameral legislature and separately elected executive governor, even though there’s no requirement for this. I assume this is an example of institutional isomorphism, but I don’t know if there is any literature on how the process works in this case.

fn1. The NYT has just gone pay-per-view on its Op-Ed pages, and there does not appear to be a workaround for blogs. For me as a blogger, there’s no point in paying for something if you can’t link to it. So I probably won’t be linking to the NYT as much in the future, and only to news stories.

{ 59 comments }

1

nick 09.19.05 at 12:57 am

Nebraska has a unicameral state legislature. And numbering candidates in order of preference is a tricky sell when you have an already apathetic electorate.

2

abb1 09.19.05 at 1:52 am

…numbering candidates in order of preference is a tricky sell when you have an already apathetic electorate…

More tricky than a separate runoff election?

3

Ben Alpers 09.19.05 at 2:09 am

I’m not so sure that this would be a hard sell. It’s just that it isn’t currently being sold by either major party or the media, which among them tend to set our nation’s rather narrow political agenda.

In fact, lots of states and localities have runoff primaries (including my state of Oklahoma). To my knowledge only Lousiana has runoff general elections (at least for federal offices). Any of these states and localities could save significant amounts of time and money by going to IRV.

And IRV general elections end the problem of minor parties or independents “spoiling” elections. So you’d think, in these times of often razor thin margins of victory, one or both of the major parties would grab on to this idea. It’s certainly one of the electoral reforms most desired by the Green Party (who believe that more folks would vote for them if they didn’t feel that they’d be helping the right win by so doing).

Any CTers wise in the history of NY elections know why they go with 40% instead of 50%? If you’re going to have runoffs, why wouldn’t you insist on a majority to win in round one?

4

Randolph Fritz 09.19.05 at 2:59 am

San Francisco isn’t all that different from NYC, politically. But NYC is much larger, and it seems a poor place to experiment with a new voting system; any failures will show on the grand scale.

5

Adam 09.19.05 at 3:03 am

IRV has major theoretical flaws, and is relatively complicated. Better would be approval voting (http://en.wikipedia.org/wiki/Approval_voting), which would have basically the same effect in terms of getting rid of spoilers, is simpler, closer to the system that Americans are already used to, and lacking in some of the theoretical problems that IRV has.

6

Aidan Kehoe 09.19.05 at 3:49 am

“Experiment?” Okay, now you put it like that, there’s a good chance that the levels of high-fructose corn syrup in the diet of people in the US will show up new problems with this voting system that haven’t been apparent in the many, many industrialised democracies it has been working in for decades.

7

pdf23ds 09.19.05 at 5:36 am

I’ve read in places (like here) that Condorcet is preferable, and that IRV leads more often to less preferable outcomes. While it seems IRV would be more of an improvement over FPTP than Condorcet would be over IRV, and IRV seems to be more politically viable, still, I’m not sure.

8

Patrick Nielsen Hayden 09.19.05 at 6:15 am

“NYC is much larger, and it seems a poor place to experiment with a new voting system; any failures will show on the grand scale.”

What a strange thing to say. In fact New York City has been the venue for quite a lot of “experimenting”; a lot of the New Deal amounted to FDR putting the same people to work at the Federal level who’d previously been successful reformers in New York city and state.

Not all experiments work, of course, and New York City certainly has its problems; but one of the saving graces of its political culture has always been a relative openness to new approaches. To Randolph Fritz’s fretful worry, one could just as easily respond that in fact New York City is big and diverse enough that the results of its “experiments” have a better chance of scaling up.

The charge against instant-runoff voting is that in multiple-candidate elections, it has a tendency to deal out victory to the candidate who’s least unacceptable to the largest number of voters. Having run several small instant-runoff elections and (more to the point) spent thirty years in a subculture where the system is widely understood and used, I tend to think of this as a feature rather than a bug. Of course there are no multiple-candidate voting systems that can’t be jiggered by the determined into producing bizarre results, but that’s true of first-past-the-post as well.

9

Alex R 09.19.05 at 6:22 am

“Alternative” voting methods such as IRV or Condorcet have become become associated in many people’s minds with the aspirations of third-party, independent, or fringe candidates. This poses a political problem, I think, for the adoption of such methods even in something like a party primary, where adopting IRV would be unlikely to strengthen third parties. The other perceived flaw of IRV is the complexity of the ballot.

A good middle ground, then, might be a simplified IRV — something that looks as much as possible like the existing system.

To replicate exactly a runoff between the top two candidates would still require the voters to come up with a full list of preferences on election day, so that every voter would have had the opportunity to express a preference between the to two candidates.

But it would be pretty easy, both for the election officials and the voters, to have a single “second-choice” vote column. This vote would be used in the event that the first vote was not for one of the two top candidates and the second vote was. There are still problems with this, but it might be a foot in the door to a full preference system like IRV or Condorcet. In the recent NY mayoral primary, it certainly would have given Ferrer — and possibly Wiener — 40% of the vote.

10

Brian 09.19.05 at 6:22 am

The charge against instant-runoff voting is that in multiple-candidate elections, it has a tendency to deal out victory to the candidate who’s least unacceptable to the largest number of voters.

I agree with Patrick that in many cases this is a feature not a bug. But it should also be noted that it isn’t always the case. Take a fairly realistic case, somewhat modelled on Australian politics in the 1980s.

The (traditional) left wing party gets 45% of the vote.
The (traditional) right wing party gets 45% of the vote.
A centrist splinter group (splintered from moderates on both sides) gets 10%.

Even if by anyone’s standards the centrists are the ‘least unacceptable to the largest number’, they ain’t going to win under IRV. In practice in a major election where there aren’t that many viable candidates, you need to be the first choice of 25 to 30 percent of the field to have a chance. This isn’t true in all cases, and may not be true in primaries, but in practice this feature of IRV rarely has the negative features that get oddly associated with it. (If those features did crop up in practice and not just mathematical models, you’d expect to see people mention real-life cases of them, since IRV has been used in so many real world elections.)

11

John Quiggin 09.19.05 at 6:33 am

Brian is correct. In general, the winner under IRV will be someone with reasonable primary support, and who is not unacceptable to a majority of voters. This seems pretty good to me.

The exception, given the proportions listed above, is the case when one of the major party candidates is so odious to the supporters of the other that they vote tactically in support of the centrist (of course this outcome can also arise with FPP in cases of this kind).

12

Steve LaBonne 09.19.05 at 7:27 am

For reasons I don’t fully understand, the US, which was once a pioneer in all kinds of institutional innovation (the primary system itself, for example, or decimal currency) is now intensely conservative about such things.
I’m glad you noticed that; it’s something that’s been bugging the hell out of me for a long time.I don’t get it either, but the remarkable ignorance of much of the “educated” US population about the rest of the world, or for that matter about much of anything besides reality TV, SUVs and football, may have something to do with it.

13

Patrick Nielsen Hayden 09.19.05 at 7:38 am

Another problem with instant-runoff voting is that it arguably gives too much power to the voters who start by voting for the least popular candidate.

Huey, Dewey, Louie, and Fred are all running for dogcatcher. Huey gets 43%; Dewey gets 11%; Louie gets 30%; and Fred gets 16%.

Since Dewey came in last, the first thing that happens here is that his votes get redistributed, according to their second-place choice, between Huey, Louie, and Fred. If enough of Dewey’s voters chose Huey in second place, Huey gets past 50% and wins. If not, a further runoff ensues and it’s not inevitable that Huey will ultimately win.

It can be surprisingly difficult to make a strong case for why, at the start of the automatic-runoff process, Dewey’s voters should have so much more “kingmaker” power than (for instance) Fred’s.

However, again, no electoral system is free of this sort of problem, including those actually in use in America today.

14

Kieran Healy 09.19.05 at 7:42 am

numbering candidates in order of preference is a tricky sell when you have an already apathetic electorate.

Except you don’t have to: if you like, you can treat your vote as a FPTP ballot and just vote for one person. There’s no requirement that you rank-order all the candidates.

15

nik 09.19.05 at 8:11 am

…the winner under IRV will be someone with reasonable primary support, and who is not unacceptable to a majority of voters. This seems pretty good to me.

Why is reasonable primary support desirable?

From the point-of-view of Condorcet: the winner could have reasonable primary support, but there could still be another candidate that the majority of the voters would prefer, who would not get elected. That’s the situation that Brian points out.

45% L>C>R
45% R>C>L
10% C>L>R
Here the winner is clearly C (a 55% majority prefer him to either L and R). But L gets elected under IRV.

It seems to me that this effect is a bug and not a feature. People who dismiss it are dismissing majoritarianism. I think the IRV proponents have to give a much more concrete justification for this than I’ve heard so far. Why should a candidate not prefered by the majority of the voters be elected over one who is?

16

Ben Alpers 09.19.05 at 8:12 am

“Alternative” voting methods such as IRV or Condorcet have become become associated in many people’s minds with the aspirations of third-party, independent, or fringe candidates. This poses a political problem, I think, for the adoption of such methods even in something like a party primary, where adopting IRV would be unlikely to strengthen third parties.

Become associated by whom? I honestly think that 80% of the American electorate has never heard of these alternate voting schemes, and has no idea what any third party in this country really wants. You may be right that the most politically active American voters have this association, but do such people really embrace a beggar-your-neighbor attitude toward the desires of third parties?

In fact, there is enormous hostility to real electoral reform from the leadership of both major parties, which have, over the last century or so, made a system that already favors two parties into one that virtually locks in the Democrats and Republicans. But I don’t think that average voters in my state (Oklahoma) would be nearly so hostile to, e.g., more open ballot access laws or write-ins (which are illegal in OK) as our elected officials are.

17

Michael H. 09.19.05 at 8:12 am

It is an interesting topic and one that should have been settled by now. It is clear that when there are more than two serious candidate, plurality voting can lead to terrible results. Plurality voting with runoff might lead to the very worst two candidates facing each other and, in fact, exactly this thing happened in Lousiana about 15 years ago, (of course, arguably it always happens in Lousiana).

What I think is odd is that no one has considered including both positive and negative votes. A negative vote would negate someone else’s positive vote. The idea is simple: you may not know who you like best but you do know who you like least. Isn’t this typically the case? Why now allow people to vote that way.

The biggest theoretical problem the the transferable vote is that if you really only want to kill off candidate number 3, you need to guess which candidate, 1 or 2 is more likely to succeed. If a lot of voters hate 3 but split on 1 and 2, 3 might still win. If you can vote against 3, then your primary vote kills 3 and secondary vote decides between 1 and 2.

Let me give a simple example. Candidate three is very polarizing. 40% love him, 10% like him better than 1, 10% like him better than 2 and 40% really really hate him (think George Bush with a bit of David Duke). Candidate 1 and 2 are bland. 31% like 1, 29% like 2 but no one hates either one. In the transferable vote, 3 will win. But with the negative vote, 1 will win. Negative votes help rid politics of really unlikable politicians.

18

Ben Alpers 09.19.05 at 8:31 am

The problem with Condorcet is transparency. Simply put, it’s too difficult to understand how the results are arrived at. Though marginally better than IRV at measuring voter preferences, a Condourcet system would say, in effect, “trust us, these are your preferences.”

Condorcet proponents respond to this problem by saying “trust us, we’ll be able to explain it to people.” But don’t listen to me, listen to them:

The complexity of Condorcet voting comes in two parts. The first part is in interpreting the rankings on each ballot as an equivalent pairwise matrix; the second is in resolving cyclical ambiguities when necessary. The determination of pairwise matrices is very basic and should be understandable with moderate effort by eighth-graders. The resolution of cyclical ambiguities is slightly more complicated, but it should be understandable with very moderate effort by average high school juniors. If that level of intellectual effort is now too much to ask for something as fundamental as a superior election method, democracy is in trouble.

Does that sound like something that the public will come to understand? Perhaps realizing that the “really it’s simple” argument doesn’t quite fly, these Condorcet proponents essentially go on to argue that the best way to get the public to accept it is to first decide not to consider the problem of public acceptance at all:

The implementation strategy, therefore, requires at least two major steps. First, a consensus must be developed, among those interested in election methods, that Condorcet voting is indeed a superior method in principle, regardless of how difficult it will be to sell to the general public. After that consensus is reached, the job of selling it to the general public can be effective. (emphasis in original)

Sorry. As a person “interested in election methods,” I care a lot more about getting a better system in place than on insisting upon the “superior method in principle,” regardless of what the public may think.

Even if Condorcet could be imposed, any system that makes election results difficult to understand threatens the principle of open government. And it would encourage the kind of fraud and gamesmanship we’ve seen in the last couple U.S. elections (since it would be that much more difficult to document such things in a way that the public could understand).

Condorcet’s marginal gains in preference-measuring relative to IRV are not worth the loss in transparency.

19

pdf23ds 09.19.05 at 8:56 am

Ben, that’s an odd definition of “transparency” and “open government” you’re using. I usually take the terms to mean that voters have access to information about the legislative processes and other information that lets them be aware of and evaluate the actions of their elected officials and other aspects of their democracy, not that the processes are straightforward or simple to understand. There’s nothing keeping a potential voter from understanding how Condorcet works besides their own lack of motivation or intelligence. While there is genuine debate over what exactly the pros/cons of various voting systems are, this debate in itself doesn’t put Condorcet at a disadvantage relative to IRV or other systems (though the results of that debate might).

A lack of transparency in election methods would have to be something like prohibiting public viewing of the ballot counting, or not allowing neutral third parties to review the code of electronic voting systems, especially those without a paper trail available for recounts. The US does have these problems, but Condorcet wouldn’t make them any better or worse. These are separate, (and in my opinion, more urgent,) reforms.

20

Matt McIrvin 09.19.05 at 8:59 am

John said: For reasons I don’t fully understand, the US, which was once a pioneer in all kinds of institutional innovation ( the primary system itself, for example, or decimal currency) is now intensely conservative about such things.

Steve Labonne suggested that this is just because Americans are all pig-ignorant (note, though, that reality TV came from Europe). I think it’s more the early-adopter problem: we’re locked into version 1.0 of the modern representative republic plus a ragged assortment of code patches, and the system on most levels was designed to be very hard to modify because the people who put it together were justifiably paranoid about the second-system effect. Some other countries are on version 3 or 4 by now and have some interesting new features to play with.

Also, many of the radical innovations at the beginning of the US were put in place by a tiny super-educated elite at a time when government here was not really all that democratic (and the whole population was relatively small compared to today). There’s probably more inertia arising from popular ignorance with the current setup.

21

Michael H. 09.19.05 at 9:26 am

I think commenters here are missing the real application of using transferable voting. This will not be used in general elections: it would require a constitutional amendment and it won’t happen. It would first be used in primaries. It could be used in any party’s primary in the next election. And it would be tempting for Democrats, who have been losing a lot of general elections recently, to think about using such a system to improve the quality of the candidate.

But I think it needs repeating: the candidate who should be first eliminated is not the least popular, it is the most unpopular. Parties need to get rid of the candidates that go into the general election with large negatives. Parties need to think about candidates who are at least acceptable to most voters.

22

Ben Alpers 09.19.05 at 9:45 am

pdf23ds,

If one sets up a system, such as Condorcet, under which many or most voters would have a hard time understanding how their raw votes translate into an electoral result, I would say that that amounts to a problem of transparency.

michael h.,

No federal constitutional amendment would be required to put IRV in place for any general election. For example, states are quite free to select their methods for choosing their presidential electors. Nothing would prevent a state from adopting IRV.

As for Democratic presidential primaries, since 1972, the powers-that-be in that party have essentially tried to substitute their own opinions (through various devices such as superdelegates and a carefully tailored primary calendar) for the opinions of the grassroots. Why would such leaders choose a method of candidate selection that better reflects grassroots opinions?

23

Uncle Kvetch 09.19.05 at 10:07 am

No one’s mentioned fusion voting, which is legal in eight states and, here in New York at least, has shown some real potential for broadening and enriching the political discourse. (It can also be something of a joke: viz, NY State’s “Liberal Party,” which was essentially a one-man patronage machine.)

The fate of fusion at the national level speaks to the institutional inertia that many have referred to above. The Supreme Court rejected a suit by the New Party in Minnesota to force their state to offer fusion, arguing that political stability is “best served through a healthy two-party system.” (Put to one side, for the sake of argument, the fact that it is to our ostensibly “healthy” two-party system that the Justices owe their positions on the Court.) Most galling was the opinion by, IIRC, Sandra Day O’Connor, which spoke to the “confusion” that might result if voters were faced with anything more complicated than Column A vs. Column B.

Whether or not you think American voters are too apathetic &/or just plain dumb to get their heads around fusion voting (and I certainly don’t), the people in charge definitely do, and they’re calling the shots.

24

pdf23ds 09.19.05 at 10:12 am

Ben,

I don’t think that any person of normal intelligence who read a well-written article explaining Condorcet would have any serious problems comprehending it. That this is more than most people would do is unfortunate, but I don’t think that much effort is a barrier high enough to justify calling the method opaque.

On the other hand, a well-written article contains many times more information than can be fit into a political ad or that can keep a non-interested person’s attention, so the complexity is definitely a problem for promoting Condorcet.

25

jlw 09.19.05 at 10:20 am

To me, the pissing match above between the Condorcet supporter and the IRV supporter demostrates why reform isn’t coming anytime soon. Even if the average, apathetic voter is stirred enough to investigate potential changes to the voting system, he’s gonna get bombarded with various factions bad-mouthing all reformes save their preferred one.

JPF or PFJ? Fuck it, lets watch some gladiators at the Colosseum.

Me, I’m suspicious of anyone asking voters to count higher than three or keep track of everyone they might approve or disapprove of. (Indeed, I think that the biggest cause of voter apathy in the U.S. is the ridiculously long ballots that expect voters to have an opinion on candidates for minor local courts and state agricultural commissions.) But I’m willing to sign on to any improvement over FPTP, be it IRV or multimember PR or whatever. And I wish that supporters of all the various “perfect” systems” would as well.

26

jet 09.19.05 at 10:25 am

Uncle Kvetch,
Fusion voting meets the contested counties of Florida 2000. Now that would be a nightmare. We’d either have to accept a fairly high level of disenfranchisement (by their own stupidity), or stick to the current methods.

Personally, I’m with you and Thomas Jefferson, extreme stupidity doesn’t deserve to vote.

27

nik 09.19.05 at 10:29 am

I think Condorcet’s Method is much simpler than it is often given credit for.

The rule for the Condorcet Winner is to (1) count a higher placement on a ballot as a vote in favour of the higher candidate, and (2) elect the candidate who can beat every other candidate in a one-on-one election. If there is no Condorcet Winner, (3) a victor is chosen by lot among the tied candidates.

That’s a very simple description of a possible system. Can the rules behind IRV be stated this simply? Condorcet gets a reputation for complexity it doesn’t deserve because of its place in electoral mathematics – and all the talk of matrixes, cycles and pairwise comparisons. But the basic idea is simplicity itself. There are plenty of much more complicated systems already in operation (such as the US electoral college, explain how you go from votes to a winner in that!).

28

Matt Austern 09.19.05 at 10:50 am

One problem, I think, is that part of this problem has been settled long ago: it is provable that, if you have more than two candidates and more than one voter, there is no “good” electoral system. That is, there is no electoral system that always satisfies our intuitions for what a reasonable result would be given every possible set of voter preferences. There will always be anomalies and unfair results at least some of the time.

Some people who are aware of this result just give up on the whole idea of improved voting systems. Some people go even further overboard than that, and think this mathematical result is evidence that we should give up on the whole idea of democracy.

Probably a better response is to acknowledge that perfection can’t exist in the real world and then ask which voting systems are less bad than others: which systems are more likely to have anomalous results than others, which kinds of failures are more likely to seem unfair and inexplicable to voters. (This surely matters. If the people don’t accept a purportedly democratic system as legitimate, then it isn’t.)

And now you’ve left the realm of mathematics and entered the messier realm of empirical experience and voter psychology. I don’t think there’s a better answer than experimentation.

29

Eric Scharf 09.19.05 at 10:58 am

I live in the state of Washington, where many voters (supposedly out of general anti-party sentiment) resist any attempt, however benign, at party registration or even one-time at-poll party declaration, with the result that primary elections are open to “cross-over voting” abuse. A few years ago, the Federal courts finally (and properly) struck this system down as violating the parties’ 1st Amendment right to free association. The court’s ruling hasn’t dampened the “open primary” enthusiasm, however, and the state Grange and allied organizations keep casting about for a system that will both pass Constitutional muster while permitting voters to abstain from affiliating with a party for even a moment. The local IRV advocates are already working on yoking the primary election “reform” movement to support for IRV.

I like both IRV and Condorcet and would support either, but only for general elections. Primary elections are taxpayer-funded handouts to the parties, and like most government handouts they foment a corrupting dependence upon the recipients. If the parties want to conduct statewide opinion polls on their candidates, they should do so on their own dime. Otherwise, they should return to nominating conventions whose attendees are restricted to party activists who have put in their time and have helped determine what the party stands for (besides electing candidates).

Real parties don’t hold primaries.

30

David Weman 09.19.05 at 11:11 am

“I think commenters here are missing the real application of using transferable voting. This will not be used in general elections: it would require a constitutional amendment and it won’t happen.”

No. Voting methods is up to the states. The constitution says nothing about FPTP. Louisiana have runoff elections.

31

Uncle Kvetch 09.19.05 at 11:29 am

Uncle Kvetch,Fusion voting meets the contested counties of Florida 2000. Now that would be a nightmare.

Well, if you mean that the same people who brought us the “butterfly ballot” would also be in charge of implementing fusion voting, and more importantly, explaining it to the voters, then yes, a fiasco would be almost guaranteed. But the fault there would lie with the incompetent and/or corrupt people administering the elections, not with fusion per se, no?

I can’t help finding the rationale that fusion would be too “confusing” for voters, when it’s worked very smoothly here in NYS for years, just a wee bit patronizing.

32

Alex R 09.19.05 at 12:00 pm

Ben wrote that most voters haven’t even heard of alternatives to first-past-the-post, but questions whether those who have really are hostile towards third parties.

I’m not sure, but Democratic and Republican party activists are a pretty sizable subset of “well-informed” voters, and as you say, those are the ones most likely to be hostile to third parties.

As for the transparency concerns mentioned by Ben and others, I think they are real and not much better for IRV.

Yes, IRV is simpler to understand than Condorcet, but both depend on the full set of preferences for each voter. This would seem to make it very hard to report on and understand the results of an election — especially a close or contested one. People often look at Florida 2000, for example, and say “with IRV, Gore would probably have won easily with Nader voters putting Gore as their second choice”. But imagine trying to understand and report the results of a large election that is just about as close *with IRV* — you can’t just report on “how many votes” each candidate got, you you also need to be saying things like “There were 1524 voters who put Buchanan first; of those 500 put Nader in second and Bush in third…”, etc, etc. The reults of of a vote with IRV can depend on which candidates get the least number of votes, because their votes are then distributed — you might require recounts for many more situations than FPTP.

With all of these sorts of issues coming up in a close election, how do countries that use IRV actually report thre results of contested elections? How does the media cover such elections?

33

jet 09.19.05 at 12:09 pm

…when it’s worked very smoothly here in NYS for years, just a wee bit patronizing.

How do you know it has been working well? The butterfly ballot cost Gore maybe 4,000 votes in 2000 but cost Dole 16,000 votes in 1996. How do you know that fusion voting doesn’t disenfranchise similar numbers in NY?

The butterfly ballot wasn’t exactly rocket science and look what happened with it.

34

Uncle Kvetch 09.19.05 at 12:23 pm

How do you know that fusion voting doesn’t disenfranchise similar numbers in NY?

Jet, I’m not skilled at proving negatives. When I said it’s been working smoothly, I meant to the best of my knowledge. I haven’t been aware of any allegations of voting irregularities, fraud, disenfranchisement, etc. in this state that can be attributed to fusion voting. If anybody else out there knows otherwise, I’d like to hear about it, and I’ll be more than willing to rethink it if there’s compelling evidence of problems arising from it.

It’s not clear to me why you’re so hostile to the very idea, Jet. How exactly could fusion voting “disenfranchise” a voter in the same way as a poorly designed ballot?

35

Nell 09.19.05 at 1:01 pm

minority of Democrat voters

That’s Democratic voters to you, bub. In the U.S. only Republicans use the noun where the adjective should be. You get a pass this once in case usage is different in Oz.

36

saurabh 09.19.05 at 1:31 pm

I’d like to echo adam and suggest that approval voting is a much more reasonable and salable alternative voting system. Easy to understand, very easy to do, easy to explain (everyone votes for as many candidates as they like) – no messy ordering schemes, etc. Simple ballots, simple results, satisfies most everything you could demand from a voting system.

37

Katherine 09.19.05 at 1:36 pm

with IRV, in order not to give the last place candidates’ voters undue influence over the second, third, fourth last place–why not just set the cut off at 20, 25, or 30% of the vote & distribute everyone’s second place votes accordingly?

38

nick 09.19.05 at 1:46 pm

Something I meant to add earlier on: having seen the apparatus used by my wife to vote in 2004, it appears to me that American voters are already dealing with technology that makes a relatively easy process pretty damn difficult. Now think about Diebolding any form of multiple-preference voting.

39

jet 09.19.05 at 2:36 pm

Uncle Kvetch,
My apologies for not being clearer. When I said this: “Personally, I’m with you and Thomas Jefferson, extreme stupidity doesn’t deserve to vote.” I mean that I was totally for fusion voting, even if there were a few people who couldn’t figure it out. I’m for anything that makes the country more democratic.

My only point was that it didn’t seem likely we could have the same level of franchisement while implementing a relatively radical change.

40

nik 09.19.05 at 2:42 pm

with IRV, in order not to give the last place candidates’ voters undue influence over the second, third, fourth last place—why not just set the cut off at 20, 25, or 30% of the vote & distribute everyone’s second place votes accordingly?

This just changes the manifestation of the problem. Instead of voters for last place candidates getting undue influence, voters for candidates polling below X% of the vote get undue influence. The problem – undue influence – is still there.

I think Condorcet is simpler to understand than IRV. If there is a candidate who can beat all the other candidates, that candidate wins. It’s far harder to verbalise the victory condition for IRV, and even more difficult to justify it morally.

41

Uncle Kvetch 09.19.05 at 2:52 pm

We don’t disagree, Jet. There will always be risks in implementing this kind of change: the risk that compromised public officials will try to bend the changes to their own partisan purposes; that incompetent civil servants will fail to implement the changes effectively and/or explain them adequately to the electorate; and yes, that some people will be just too dim to deal with the same candidate appearing more than once on the ballot, under multiple party affiliations, without their heads exploding. Any of these could conceivably make the changes counterproductive, but that’s hardly an inevitability. I think it’s a risk worth taking insofar as fusion allows for a broader range of political participation without the dreaded “spoiler effect.” It’s by no means a panacea, and it may not be nearly as effective as the other alternatives being bandied about here–which I simply don’t know enough about to comment on.

The first step, of course, is getting a critical mass of public & elite opinion to recognize that the system can and should be changed, and I fear we’re still a long way from that point.

42

Michael H. 09.19.05 at 3:36 pm

Four comments: 1, I don’t see the benifit of a simplistic formula like Approval voting that has most of the problems of the current system.

2, I don’t think it is necessary for all of the voters to understand exactly what would happen in every circumstance. The voting mechanism could be fairly complex, but in general it would work to eliminate one of the n candidates, and continue to do so until 1 remains. People would intuitively understand that in almost every practical circumstance the one eliminated would be the guy with the worst votes (fewest positive and most negative).

3, It really is important in a multi candidate election to have negative votes. It is extremely important that Adolf Hitler is eliminated first off and therefore has no real influence on the outcome. George Washington or someone very popular is bound to win in any system. You need to eliminate the really ones right off.

4, I think it is possible to combine the best eliments of the transferable vote and the Condorcet method. In each stage you look for the “least popular” and you ask, would “she lose in a head-to-head with everyone” If yes, she’s gone, if no, go to the second least popular. If you go through everyone and you cannot eliminate anyone go back to the least popular and you ask: “Is there a candidate D who can beat F and G but Candidate A beats them D and F and G” Again start with the least popular and check. If yes, you eliminate, if no you move on. If again you cannot eliminate anyone then Condorcet fails and you just eliminate the least popular. Then proceed until only one is left. It’s not so very hard.

43

John Quiggin 09.19.05 at 6:02 pm

“we’re locked into version 1.0 of the modern representative republic plus a ragged assortment of code patches, and the system on most levels was designed to be very hard to modify because the people who put it together were justifiably paranoid about the second-system effect”

A nice point, and I hadn’t heard of the second-system effect before.

But I don’t think it’s a fully satisfactory explanation. The Progressives implemented lots of reforms around the turn of C20, and they had to overcome the same obstacles.

44

Randolph Fritz 09.19.05 at 6:14 pm

Sheesh…I’ve supported IRV for years. I still think it would be better to get more experience with the system in places smaller and less politically central than NYC. That said, if anyone with real power in NYC proposed IRV, and I was in a position to support it, I would.

45

Ciaran Quinn 09.19.05 at 6:23 pm

New York city used STV between 1936 and 1947 — see http://www.nyu.edu/gsas/dept/history/public_history/PR/ for details.

46

John Quiggin 09.19.05 at 6:51 pm

“I still think it would be better to get more experience with the system in places smaller and less politically central than NYC.”

Does Queensland count? Geographically, we’re big of course, but you could hardly get less politically central :-).

47

Anthony 09.19.05 at 7:00 pm

While there are issues regarding IRV giving an artifical boost to minor parties in a general election, using IRV in a primary election strikes me as a very sensible way to do things where there are a large number of candidates.

How would the Democratic Party nomination in 2004 have come out if there was one national IRV vote among praimary voters in all 50 states? Would Dean, reinforced with Kucinich’s voters, have come ahead of Edwards? If not, who would Dean voters have preferred? The younger and more charismatic Edwards, or the somewhat more historically leftish Kerry?

48

Tom T. 09.19.05 at 8:03 pm

Re: #43. John, the Progressive reforms were more successful in newer states in the Midwest, Rocky Mountains, and far West, where there was less institutional history. Thus, for instance, most of the original 13 states and much of the South have no popular referendum process. This seems to fit the early-adopter point.

Query for the group: Would the proposed voting reforms have flipped the outcome of the 1992 US presidential election, resulting in the reelection of Bush 41?

49

pedro 09.19.05 at 8:15 pm

Virtually anything is an improvement over plurality. I happen to prefer IRV and Borda count to approval, but between any of those and plurality, the verdict is obvious.

50

Jonathan Lundell 09.20.05 at 12:35 am

I prefer Condorcet to IRV, and either to approval or Borda. Approval in particular throws away a good deal of information present in a voter’s ranking.

It’s been pointed out that the voter experience of IRV and Condorcet is the same–only the counting is different. So once IRV is in place, moving to Condorcet would be relatively easy, technically at least.

In SF, due to voting equipment limitations, the voter can rank only three candidates, something of a shortcoming in SF’s big fields of candidates. That restriction will no doubt be lifted in the relatively near future as SF updates its equipment.

IRV was something of an easier sell in SF because there was already a runoff system in place–the mayor, for example, must get 50%, or there was a runoff election between the top two vote-getters. These runoffs happened pretty much every election; they were expensive; and they had very low turnouts, as a rule.

So IRV (SF calls it “ranked-choice voting”) was an easier sell, as an improvement to the existing not-so-instant runoff scheme.

51

Mike 09.20.05 at 2:16 am

“Instant runoff” and “single transferable vote” are not synonomous. The latter is the method in Tasmanian state elections, the former in most other Australian state elections. If you seriously want to propose one or the other of these, get it right.

52

John Quiggin 09.20.05 at 3:00 am

For the case in question, where a single position is to be filled, the two are synonymous.

53

Syd Webb 09.20.05 at 7:53 am

Another problem with instant-runoff voting is that it arguably gives too much power to the voters who start by voting for the least popular candidate.

It doesn’t. Those who give their first preferences to the least popular candidate never see their candidate win.

Those who give their first preferences to the second least popular candidate can sometimes see their candidate win.

The winning candidate is always the candidate who wins the two party preferred vote. That is, the candidate who is preferred by a majority of voters to the next most popular candidate.

Huey, Dewey, Louie, and Fred are all running for dogcatcher. Huey gets 43%; Dewey gets 11%; Louie gets 30%; and Fred gets 16%.

Since Dewey came in last, the first thing that happens here is that his votes get redistributed, according to their second-place choice, between Huey, Louie, and Fred. If enough of Dewey’s voters chose Huey in second place, Huey gets past 50% and wins.

Which means that more than 50% of voters prefer Huey over Louie.

If not, a further runoff ensues and it’s not inevitable that Huey will ultimately win.

In this particular example only Huey and Louie have a chance of winning. Even if all Dewey’s preferences flow to Fred, Fred still only has 27% of the vote and will be cut-up before Louie.

It can be surprisingly difficult to make a strong case for why, at the start of the automatic-runoff process, Dewey’s voters should have so much more “kingmaker” power than (for instance) Fred’s.

They don’t. If Dewey’s preferences all flow to Louie and Fred’s preferences all flow to Huey then it’s Fred’s preferences that have elected Huey with 59% of the two party preferred vote. In this scenario that’s a great outcome – Huey is preferred over Louie 59% to 41%

54

Syd Webb 09.20.05 at 8:13 am

3, It really is important in a multi candidate election to have negative votes.

No, it’s not. Completely unnecessary, in fact.

It is extremely important that Adolf Hitler is eliminated first off

No, it’s not. All you need to care about is that Hitler’s not elected.

and therefore has no real influence on the outcome.

All voters should have an equal impact on the outcome. So if a minority of voters prefer Hitler, it’s important where their second and subsequent preferences lie.

George Washington or someone very popular is bound to win in any system.

Yes. A popular candidate can win straight away with more than 50% of the first preferences.

You need to eliminate the really ones right off.

Not sure what this means. Perhaps an example is best.

Hitler is prefered by 32% of the electorate. The remaining 68% of voters split their preferences among the other four candidates: Angela, Berthold, Dietrich and Eberhard. As long as the anti-Hitler voters give Adolph their 5th (least) preference he cannot win.

This will be true even if A, B, C and D poll 19%, 18%, 16% and 15% respectively of the votes – a split that would give Hitler the victory in a first-past-the-post election.

[In this scenario neither Dietrich or Hilter can win – Dietrich will be eliminated first and his preferences distributed among A, B and C. Which of A, B or C is ultimately elected will depend on the flow of preferences – mathematically one of these three must be eliminated before Hitler is inevitably eliminated.]

55

james 09.20.05 at 10:54 am

The self limiting aspect of a winner take all system is also a feature and could explain the resistance to other types of voting. The restrictions to the system do limit participation to perceived middle ground candidates. It also allows for more extreme views on core beliefs and core freedoms without those extremes being threatening to the majority. This is a direct result of the extremes being unable to affect the outcome without becoming the majority or a significant minority.

56

Mike 09.20.05 at 7:51 pm

John said: For the case in question, where a single position is to be filled, the two [IRV and STV] are synonymous.
No. STV requires multi-member constituencies. Otherwise the single vote is only transferable as a whole unit and that eliminates the entire point of STV. You can have IRV/preferential systems in multi-member districts but not STV.

57

John Quiggin 09.20.05 at 8:01 pm

The Wikipedia entry on instant runoff voting begins “When the Single Transferable Vote voting system is applied to a single-winner election it is sometimes called instant-runoff voting (IRV)”.

Some people regard STV as applying only to multi-member electorates and others don’t. Let’s agree to differ on this one.

58

Mike 09.21.05 at 1:16 am

If you wish. This not simply semantics, STV is a proportional system, IRV is not. (The Wikipedia article on STV outlines the difference.)
Currently, here in Canada, there is a great deal of talk about alternate voting systems. In BC, we just finished a referendum on STV (I voted “No”) which won a majority but not enough to pass. We will vote again in 2008. I am much friendlier to the concept of IRV (and one-member per riding) than to STV (and multi-member ridings). It’s been on my mind a lot and I apologize if I seem a bit tetchy on the subject.

59

Syd Webb 09.21.05 at 3:27 am

If you wish. This not simply semantics, STV is a proportional system, IRV is not.

Mike’s talking outcomes, John’s talking mechanics. Of course with a single position to be filled, some of the best features of STV – distribution of surpluses and fractional votes – don’t come into play. But yes, IRV is just an instance of STV with the quota being 50.01% The voter still has a single vote that can be transfered among candidates in a downward order of preference.

As John says, let’s agree to disagree.

Comments on this entry are closed.