Getting Past First Past the Post: Instant Runoff vs. Range Voting and Approval Voting

At Attack of the Machine Elves, Maikeru's excellent series of posts on voting reform continues today with an article on alternatives to plurality voting. Maikeru has argued we need to "get past first past the post":
If there's one thing that many pro-reform advocates seem to agree on, it's that the voting regime we're using right now isn't working. Currently in America, the vast majority of elections are decided using the first-past-the-post voting method, in which each person is given one vote and the candidate that receives the most votes is declared the winner. . . . The shortcomings of the first-past-the-post system are numerous and glaring. In the case of plurality voting, an election can in extreme cases be won by a candidate who falls far short of a majority and of achieving the mandate that goes with it. . . . This often leads to tactical voting-- . . . rather than voting for the candidate that best reflects their political beliefs, most voters who don't prefer one of the two major-party candidates nonetheless hold their noses and vote for what they perceive as the lesser of two evils. . . . Another often-cited deficiency of the first-past-the-post voting system is that it tends to result in a large number of "wasted" votes, or votes that are cast for either losing candidates or for winning candidates beyond the threshold needed for victory. . . . Another criticism leveled against first-past-the-post voting is that it tends to encourage gerrymandering due to the relatively high number of "wasted" votes the system produces. . . . .
Today, Maikeru compares instant runoff voting with range voting and approval voting, which is essentially a simplified variant of range voting. Maikeru writes:
in addition to first-past-the-post voting there are two other major categories of voting methods that are used within single-winner systems: preferential voting methods, in which people are allowed to vote for multiple candidates and do so by ranking them in order of preference; and range voting methods, in which people are also allowed to vote for multiple candidates but are not required to rank them against one another. Over the last couple days, I've been discussing the various types of preferential voting methods and have offered the observation that instant-runoff voting seems clearly superior to first-past-the-post in terms of efficiency, providing for a runoff in cases where no candidate receives a majority of the vote while avoiding the expense and hassle of holding a separate second round of runoff voting, as well as in terms of providing for a more open and representative political system. . . . Today I conclude my look at the various voting systems with a focus on range and approval voting, which offer the most intriguing reform alternatives to IRV.
Read the whole thing. In the end, Maikeru comes down on the side of approval voting:
Ultimately, I think that any of the three voting systems discussed in depth here is better than first-past-the-post voting and would reflect a significant improvement over the status quo. But taking into consideration how easy it would be to implement these systems, to administer them once they are in place, and what kind of results they would produce, approval voting appears to be the best option available for reform advocates to get behind in attempting to promote change within the context of the single-winner system that predominates in this country.

14 comments:

Maikeru said...

Thanks for linking to my articles. Before I wrote them, I was actually more strongly in favor of IRV and wasn't even really all that familiar with range or approval voting until I began my research for what I ended up writing. I do think that all three would be better than what we have right now, but there are also other ideas we should be considering as well, and I hope to discuss those in the near future. At some point in the near future I plan on talking about proportional representation and how we might go about implementing something like that without screwing around too much with the federalist compromises that the founders made when they created our government.

d.eris said...
This comment has been removed by the author.
d.eris said...

[That deleted comment was one of my own, mis-clicked.]

I think the case for range voting in general is very strong. But, unfortunately, very few people seem to be familiar with it. The question is, how do we change that. Pieces like this one, Maikeru, are a big step in the right direction.

broken ladder said...

Score Voting, and its simplified variant called Approval Voting, are basically the two best alternatives. You can see that objectively quantified via Bayesian Regret metrics here.

http://scorevoting.net/UniqBest.html

IRV may appear to be advantageous at first glance, but it has numerous problems once you analyze it a little more deeply. For one thing, it basically degrades to plurality-style voting when there are a lot of tactical voters. (This is a criticism that IRV proponents often make of Score/Approval Voting, but they are quite mistaken.)

As for efficiency, it may be cheaper to conduct an IRV election than to have an election followed by a runoff election. But that doesn't take into account that runoff elections aren't always necessary; whereas the cost to implement IRV (e.g. voter education materials, more advanced voting machines) is incurred on every election. By contrast, Score/Approval can be done on normal "dumb totaling" plurality vote-counting machines with no upgrades. Here's a report about the cost of IRV from the University of Vermont.

There are numerous other problems with IRV. For instance, it makes ties (and near-ties, which can lead to costly recounts and court battles) more likely. It also cannot be sub-totaled in precincts (it must be centrally counted), which is concerning from an election integrity standpoint.

It would be nice not to have to critique a voting method that so many reformers are convinced could be a really positive reform against plurality voting. But a fuller analysis of Instant Runoff Voting shows it to be a very poor system, that in some ways is actually worse than our current system.

Clay Shentrup
San Francisco, CA

Jack said...

I am glad this conversation is happening, but I think there is a reason that Instant Runoff has had so much more political and private group use than some of these alternatives. You help your first choice as long as that candidate has a chance to win, then your 2nd choice and so on. Your vote only counts for 1 candidate at a time.

If you use Approval, your vote counts for more than one candidate at a time. That means majority winners can lose and voters can be confused.

I say try it out more in private groups before thinking it's the way to go.

broken ladder said...

Jack said:

If you use Approval, your vote counts for more than one candidate at a time. That means majority winners can lose and voters can be confused.

IRV is confused about "majority winners". In the last mayoral election in Burlington VT, the Democrat was preferred by a majority to both the Progressive and the Republican, yet the Progressive won. This is because IRV erroneously eliminated the Democrat and put the Progressive and the Republican into the runoff. That is a result of a serious flaw in IRV, which you pointed out:

Your vote only counts for 1 candidate at a time.

This is a severe flaw of IRV, which causes it to ignore most ballot data, and makes it extremely susceptible to strategic voting, as I will explain below.

You help your first choice as long as that candidate has a chance to win, then your 2nd choice and so on.

Let's look at a fairly recent IRV election that I previously mentioned, from the 2008 mayoral election in Burlington, Vermont. The Progressive won, but there were people who preferred Republican over Democrat over Progressive who would have gotten the Democrat instead of the Progressive if they had (insincerely) top-ranked the Democrat instead of the Republican.

Maybe you would say that your above statement was correct, and that the Republican didn't "have a chance to win"; so those voters shouldn't have ranked him first. But then you are admitting that IRV maintains one of the biggest fundamental flaws of our current plurality voting system: voters can't fearlessly support their favorite candidate unless that candidate appears to be "electable". But the media and voters determine electability largely on the basis of:

1) Being in a major party.
2) How much money a candidate has raised.

This helps explain why IRV has maintained two-party duopoly everywhere it has been used (e.g. Australia since around 1918). It also suggests IRV does little to reduce the importance of cash in elections. (Looking at Burlington, it's important to note that it is politically shifted to the left, so that the Progressives and the Democrats effectively are the major parties there; so this logic still holds true there.)

Moreover, your understanding of strategy is simplistic and/or vague. You talk about whether a candidate has a "chance to win". But you don't specify exactly what that even means in terms of probability. Does a 10% probability of winning constitute a "chance"? How about 20%?

A better (more accurate) way to think about it is in terms of whether an insincere vote is more likely to help or hurt. Basically, if it is more likely to help than hurt, then it is advisable. So let's imagine your sincere ballot looks like this:

Minor > MajorX > MajorY

You prefer the Minor party candidate over the two major party candidates. Now you consider whether a switch to the following (insincere) ballot is more likely to help or harm you.

MajorX > Minor > MajorY

Situation 1:
It hurts you if the Minor candidate was going to win, but by changing your ballot, you cause MajorX to win.

Situation 2:
It helps you if MajorY was going to win, but by changing your ballot, you cause MajorX to win.

It is mathematically/statistically obvious that situation 2 is more likely (in fact it is what would have occurred in Burlington). Thus the "favorite betrayal" strategy is effective, thus IRV leads to two-party domination and effectively degrades to plurality voting.

Score/Approval Voting don't have this problem, since tactical voters have no reason not to support candidates they prefer to their favorite front-runner.

Liberal Arts Dude said...

I've seen the range vs. IRV voting debate go on in the Internet for quite some time now. To be frank it's driving me up the wall how rancorous the debate between these two camps can get.

I'd love to see range voting tried in practice in some real-life situations and see how it performs.

I'm partial to IRV right now because it seems to have momentum behind it and it is actually being used in real-life elections in some cities and counties. I'd like to see range voting advocates take that step as well so we can see how range voting performs in a side by side, real life comparison.

broken ladder said...

Liberal Arts Dude,

You can't test how well a voting method will "perform" with real life usage, since you can't read voters minds. More real-world usage certainly wouldn't hurt, but the evidence from Bayesian Regret calculations, and essentially every other measure or type of analysis you can name says that Score Voting is far superior to IRV.

- Fewer spoiled ballots.
- Makes cash and "electability" far less important because
- Negative exaggeration doesn't equal duopoly
- Lower risk of ties/near-ties.
- Usable on standard voting machines with no upgrades.
- Can be sub-totaled in precincts (is "additive").

I don't know how exactly you can think that the verdict isn't in, and that we somehow need more real-world usage. Please tell me, what exactly would more real election tell us that we do not already know about the difference between these methods?

As for momentum, what does it matter how much "momentum" IRV has if it fundamentally changes very little (behaves almost identically to plurality voting once there are enough strategic voters), and actually is worse than plurality voting in several ways (more spoiled ballots, less auditability, greater probability of ties/near-ties, etc.)? Voting reform is a very limited and precious resource, and if it is wasted on a non-solution (IRV) when it could be used on Score/Approval Voting, that is very very bad for our chances of significantly improving our democracy.

I think the debate gets so rancorous mainly because the IRV camp is fueled largely by sheer misinformation and myth. Almost every major claim of the IRV proponents is demonstrably untrue:

- Elects "majority winners"
- Eliminates spoilers
- Incentivizes voters to cast sincere votes
- Significantly opens the door to greater participation from third parties and independents

But no matter how many times we show most IRV proponents proof to the contrary, they just keep repeating these untrue statements. Maybe you could help to explain that.

Liberal Arts Dude said...

Hello Broken Ladder

It's one thing to demonstrate something to be superior and correct in theory. It's quite another to muster the necessary support from people, election officials, activists, and politicians and implement a voting system in real life situations.

I applaud BOTH range voting and IRV advocates for being passionate about the cause of electoral reform. I am against winner take all plurality voting and any solution that seeks to solve problems inherent in that system I am all for it.

I don't think it is unreasonable for me to ask range voting advocates to take the step of actually getting the system implemented somewhere in real-life elections so we can see how it performs.

More importantly I'd like to see how the public reacts to it (will they be in favor of keeping it, will they vote to implement it widely, will it become popular and spread like wildfire, etc.).

broken ladder said...

Liberal Arts Dude,

It's one thing to demonstrate something to be superior and correct in theory. It's quite another to muster the necessary support from people, election officials, activists, and politicians and implement a voting system in real life situations.

Whether a voting system is superior (thus, should be implemented) and whether there is the support to implement it are two entirely different issues. Score Voting is superior (e.g. to IRV and the others). That is true regardless of whether there is support to implement it.

I don't think it is unreasonable for me to ask range voting advocates to take the step of actually getting the system implemented somewhere in real-life elections so we can see how it performs.

Again, there is no way to see how well a voting method performs based on real-world elections, because you can't read voters' minds. So you have to use Bayesian Regret calculations, and those show that Score Voting performs far better than IRV or plurality.

It is completely wrong to think SV is in need of some kind of further study before we can know that it performs well enough. The evidence is already overwhelming that it does.

More importantly I'd like to see how the public reacts to it (will they be in favor of keeping it, will they vote to implement it widely, will it become popular and spread like wildfire, etc.).

Well, that is a relavant issue of course. Even if SV is superior, if lay voters aren't savvy enough about mathematics then they may end up thinking they got an unfair result, even if that's not true. Then they could end up rejecting the system, which is bad whether it's right or wrong. So voters should work to implement it in some city, so that they can have the benefits. I believe the evidence shows it to be more than worth the risk.

d.eris said...

Broken Ladder and Liberal Arts Dude, I just came across a comment at Third Party and Independent Daily stating that the Texas Libertarian Party will apparently be utilizing approval voting:

"for electing party officers and nominees for public office beginning with the 2010 LP Conventions. As there will be a lot of multi-candidate races during the March County Conventions and the June State Convention (including a 5 way race for Governor), there will be plenty of chances to try out this important experiment to improve democracy."

Given the wide array of LP candidates in Texas this year, there will likely be significant feedback on the method following the convention.

Rock said...

Just to be clear, Approval Voting is now an available option in the Texas LP. The old way of voting (multiple ballots with bottom candidate drop off) is still the default mechanism. Since AV is a new option, we have our own education campaign to do to get Texas LP delegates comfortable with it. My guess is that at first AV will be relegated to races where delegates don't want to spend time on multiple ballots. We shall see.

d.eris said...

Thanks for the clarification, Rock.

broken ladder said...

Yeah, thanks indeed for that clarification.

Here's a Score Voting exit poll I did in Beaumont back in 2006, if anyone's interested.

http://scorevoting.net/Beaumont.html

 
http://www.wikio.com