Buy My Book Here

Fox News Ticker

Please check out my new books, "Prosecutors Gone Wild: The Inside Story of the Trial of Chuck Panici, John Gliottoni, and Louise Marshall" and also, "The Definitive Dossier of PTSD in Whistleblowers"

Saturday, August 22, 2009

Deconstructing Range Voting

A couple weeks ago, I did a story on a referendum in the Tacoma area on Ranked Choice Voting. In RCV, a voter ranks the candidates in the order of their preference. Then, if one candidate gets 50.1% or more they win. If not, the bottom candidate is dropped off and the second choice of everyone that chose that candidate first gets their votes. This process continues until someone finally does get to fifty percent or more. This is a way to encourage more third party participation.

In response to that story, I was contacted by Dale Shelton. Shelton is a proponent of range voting and he introduced me to that concept. In range voting, each candidate is ranked numerically, for instance 1-9, and then the candidate with the highest overall score wins.

RCV is supposed to solve the problem of the "spoiler". By this I mean the Nader example in Florida in 2000. In that case, almost all of the people that voted for Nader would have voted for Gore. In so doing they took votes away from Gore and Bush won. Had Florida had RCV, all Nader voters would have chosen Gore and he would have gotten those votes in the second ballot.

Shelton says that while that is fine, the spoiler system still applies once you get three or more strong candidates. Now, you get all sorts of potential combinations that could lead to a spoiler. Shelton explains it with a detailed example in this post.

Shelton also points to Arrow's Impossible Theorem as further evidence that range voting is best.

In social choice theory, Arrow’s impossibility theorem, or Arrow’s paradox, demonstrates that no voting system can convert the ranked preferences of individuals into a community-wide ranking while also meeting a certain set of reasonable criteria with three or more discrete options to choose from. These criteria are called unrestricted domain, non-imposition, non-dictatorship, Pareto efficiency, and independence of irrelevant alternatives. The theorem is often cited in discussions of election theory as it is further interpreted by the Gibbard–Satterthwaite
theorem
.

Arrow's impossible theorem won it's purveyor, Kenneth Arrow, a nobel peace prize. Arrow believed that there is no perfect voting system in which candidates are ranked, like in RCV. It's important to note that range voting doesn't rank anyone. You can vote for all equally, differently, or in any combination. The key, in Shelton's view, is that in range voting everyone's ranked separately from everyone else.

In this, the dynamic of the election would change entirely. We've all heard that a candidate has voters energized. We'd expect that this candidate would get the maximum score from said voters. On the other hand, other candidates have a more lukewarm electorate. In such a case, that candidate would get more marginal scores from said voters. The way that candidates approach the base, the middle and the opposition would change entirely. In our current voting system, you either get someone's vote or you don't. In range voting, you are always trying to woo voters because most people aren't overwhelmingly impressed with any candidate from the beginning.

Currently, no elections have range voting though Shelton is working on changing that. He says that range voting was used for years in the world's oldest democracy of Venice and in Sparta. Finally, Dr. Warren Smith, of Temple University, published a study of the benefits of range voting.

8 comments:

Greg said...

No public elections have Range Voting for good reason, largely due to its extreme susceptibility to strategic voting. 1) Range systems fail the Later No Harm criterion, which means voting for a second choice under Range hurts the chances of electing your first choice. This creates a huge incentive to dishonestly bullet vote. 2) Range gives voters have an incentive to dishonestly exaggerate their preferences. 3) Range suffers from the Burr dilemma.

Also, your Gore/Bush/Nader example shows how the spoiler problem doesn't go away with Range Voting. Nader's entrance into the race would probably cause some of his supporters to score Gore lower, thereby potentially throwing the election to Bush. That would not be an issue with RCV/IRV.

These reasons help explain why Approval Voting, a form of Range Voting, was abandoned by the IEEE and the Dartmouth Alumni Association.

Shelton is wrong that a "spoiler" scenario, at least as the term spoiler is traditionally understood, can happen under RCV. I suspect he's referring to the far rarer and more obscure problem called the "center squeeze" --- a problem that opponents of RCV will often try to conflate, with the spoiler scenario.

So, yeah, no voting system is perfect, including RCV (aka IRV). But RCV is significantly better than the status quo, provides a stepping-stone to proportional representation with STV, and the occurrence of the purported problems with it have been very few and far between.

As for Range, we have very little idea of how it will work in a contemporary competitive public election. There's actually very little data on how well it works in private elections, because so few use it. Personally, I am opposed to using any voting system for public elections, regardless of its theoretically benefits, without it first being tested and proven in increasingly realistic scenarios.

mike volpe said...

Why would strategic voting be bad? Second, how is someone dishonesty voting ever? They vote how they vote. How is that dishonest. Also, what is the Burr dilemma?

Why would anyone score Gore lower? In range voting every vote is separate. No one is forced to score anyone lower because of someone else. Why would someone score Gore lower?

Todd said...

The candidates are competing, so every vote is not separate. Every voter has to weigh the impact of how they vote vote for one candidate on the chances of how they vote for another candidate.

So going into an election, any candidate worth her salt will tell her supporters -- "please max out for me and don't give any points to anyone else." The side that gets this message out better will probably win.

mike volpe said...

I agree but that's not necessarily a bad thing. It's very possible that this sort of election mechanism would favor the fringes which would be bad, but things would need to play out first. I haven't endorsed or dismissed the system. I've simply presented it. we should all know and debate all voting forms. We have a free democracy and should embrace this sort of innovative voting style.

Todd said...

Fair point. I think it's a problem, but would be good to see it tried out in more elections.

One thing going for it is that RCV is quite tested in many settings, and some big fights for it (and to keep it) this year coming up. if you can tell a system by the enemies it has, the people fighting RCV help make the case that it's a good system!

mike volpe said...

I believe that in our thriving democracy we should see as many different forms of voting as possible. I would like to see RCV and Range voting tried in localities all over. That's the purpose of federalism. It's to have states and localities be incubators for ideas. I would support the advancement of both RCV and range voting.

broken ladder said...

Greg commonly repeats these false and misleading criticisms.

Score voting (aka range voting) is actually found to perform better than all commonly proposed alternatives with any number of strategic or honest voters. See http://scorevoting.net/StratHonMix.html

Greg brings up the "later no harm" criterion, which is another misleading red herring. The problem is that IRV proporents naively assume that the "later" means, "later, after you've started by sincerely top-ranking your favorite candidate". But that doesn't happen because IRV fails the Favorite Betrayal Criterion. This prompted me to update this page I wrote about it. http://groups.google.com/group/scorevoting/web/degrade-plurality

It's a little long, but it goes into depth about why the later-no-harm criterion is so deceptive, and why IRV proponents use it anyway. It includes discussion of the infamous 2000 election.

Also I disagree that IRV is a good stepping stone to proportional representation, although that naive belief is what propels FairVote and allies to support IRV regardless of its merits -- they only care about getting PR, so IRV's flaws don't matter to them. That's discussed here

Greg's argument about backsliding at Dartmouth and the IEEE are also disingenuous, and discussed here:
http://www.scorevoting.net/DartmouthBack.html
http://scorevoting.net/FeerstTheory.html

The final argument that score voting is not ready because it has not been sufficiently tested is simply bogus. Score voting achieves extremely low Bayesian regret even if you make unrealistically cynical assumptions about its behavior. That is, even when you assume for the sake of argument that everything will "go wrong" with score voting, it still does better than e.g. IRV. Score voting behaves as well with 100% strategic voters as IRV does with 100% honest voters.

Greg is an anti-scientific partisan when it comes to this issue. That is revealed by reading some of these links. When you see how blatantly inaccurate many of his claims are, it becomes clear that he lacks expertise and objectivity in this subject matter.

Dale Sheldon said...

I'd been busy, so didn't get to read this when it was fresh two weeks ago. But it's Sheldon, with a 'd'; if you don't mind correcting it. Thanks!