Why IRV is the Best Single Winner Method for Public Elections

Comments currently closed for Why IRV is the Best Single Winner Method for Public Elections

• I don't understand how you can read me like that. I accuse you of misinformation on the line to dishonesty, and you say "fine, start your own advocacy organisation"? You should take a look at recent events. Gordon Brown's reforms could never have been proposed if it wasn't for "debating societies" like the Hansard society. They seem to have a great culture for that in Britain - you can look up the projects of the late Chris Lightfoot for some other shining examples of effective working for reform. Seeing how well they work, do we really need the kind of advocacy organizations that have a cynical attitude to informing the public? One thing I'm pretty sure of: If you pretend to have not only the "what" questions answered, but all the "how" questions as well, with no objections tolerated ("so we're cherrypicking voting critieria? Well, found your own advocacy organization!"), you need good luck in finding supporters.

  Posted by Harald Korneliussen, 2007-07-04 04:50:32 (6 years ago)

• Harald: What I don???t like about this site is the cynical way it ignores and dismisses other methods and criteria. ... You also argue that IRV is the most realistic option, and we should therefore ditch our true preferences if they are otherwise. Isn???t that sort of thing exactly what we want to avoid? I can see two ways to read Harald's complaint. (1) FairVote's proposals are no good. My proposals are better, and here's why. Okay. Do what IRV and choice voting activists do. Educate people. Persuade legislators to introduce bills and lobby to get them passed. Mobilize grassroots support. After your proposal gets adopted, help make implementation go smoothly. If IRV supporters can do these things, you can too. It's not magic or rocket science. It's just hard (but rewarding) work. You are under no obligation to limit your activities to complaining about IRV. (2) FairVote's emphasis is on adoption and implementation of one proposal, but what we really need is ongoing -- perhaps even endless -- discussion and debate among the proponents of numerous proposals. FairVote's problem is that it's not a debating society. That's true, FairVote is not a debating society. There are debating societies out there. One that I participate in is the Election Methods Mailing List. Such forums can be extremely valuable to activists as well as theorists, But they are not advocacy organizations. You should only "ditch your true preferences" if you decide that IRV is good enough to be worth the effort. Otherwise, you should work toward adoption of a voting method that you think is good enough to be worth the effort. If, in your opinion, there's no such method, then I hope you enjoy endless debates.

  Posted by Bob Richard, 2007-06-30 18:47:03 (6 years ago)

• Voting systems specialists will have this debate for years to come. A few key points, however: IRV is particularly well suited to a country where runoff elections and clogged partisan primaries put a premium on core support. Americans like a candidate with an agenda. Note we are interested in primaries and runoffs - that is, public elections - not a 10-candidate race with an electorate so small that no voters share a first choice favorite. Sometimes people call FairVote about running elections like that, and we chat frankly about the options. But our advocacy of IRV is not blind; we genuinely believe it to be the most appropriate choice (there is a large menu) for public elections in this country. Two, FairVote gets a lot of flack because, after 15 years of this, we've become the point organization for voting systems reform. Some advocates of range voting for public elections - and approval voting before them, and ranked pairs before them - launch particularly scathing attacks at us (websites canonizing us and allies, individually, as deliberately disingenuous agents of election defrauders, for example). To the extent that FairVote is perceived as a giant, I can understand the incentive to use such a tone. But the sheer virulence of the attacks creates a zero-sum game. The advocates I'm talking about do not approach public officials, writers and others interested in IRV with balanced, level-headed, comprehensive critiques. Instead, they relentlessly post stock text on a mathematical anomaly to any site even mentioning IRV. They place repeated, late-night phone calls. They post to the Internet ad hominem demonization of election reform advocates. They (some - by no means all) approach charter commissions and rail against IRV with a level of zeal far exceeding their own promotion of their own favorite systems. They have created a hostile environment for discussion. Your objection, Harald, should not be leveled at FairVote. It should be leveled at the fanaticism that some have brought to a formerly collegial table. If you want a netural discussion of single-winner election methods, spend three days reading up on social choice theory in a library. What you will find is what has always been true: 1) There is no perfect voting system. 2) The voting system you choose should depend on your normative aspirations for politics. 3) Some voting systems are better demonstrated than others.

  Posted by Jack, 2007-06-27 15:37:06 (6 years ago)

• There are hordes of voting criteria that all sound "reasonable" if you don't know better. If you pick and choose criteria, you can make any method seem best. What I don't like about this site is the cynical way it ignores and dismisses other methods and criteria. You are way too quick about it, giving no justification for your choices beyond what is needed to convince the uninformed. You also argue that IRV is the most realistic option, and we should therefore ditch our true preferences if they are otherwise. Isn't that sort of thing exactly what we want to avoid? Some may say this is revealing about the attitude of IRV's supporters. For an example of cherry-picking criteria, you claim that there should be at least one person who has the winner as his favourite. Why? If, in a small election with ten voters, everyone places a different canidate in first place, but the same candidate in second place, surely the one everyone agrees is the second best should win? It's a criterion that sounds reasonable until you take a critical look at it (something you apparently hope people won't do!) And tactical voting concerns are just dismissed out of hand, although we see plenty of where it IRV is used, in that parties publicize tactical "place the other candidates in this order"-lists, to minimize the very real risk of spoiler candidates. Also, no mention that IRV has bad summability properties, meaning that counting can not be decentralized efficiently - an important democratic safeguard. But you manage to claim that Condorcet methods are harder to count! You cite Tideman as an authority, but makes no mention of Tideman's Ranked Pairs method, a method which is a true Condorcet method, quite simple to explain, and far more resistant to strategy than IRV. Strategic voting is useless if there is a Condorcet winner, and immensely impractical if there is not: for it to work, you need to overcome two hurdles: 1. You'd need to have very accurate information about how people would vote, otherwise your strategy might very well backfire 2. You'd need to tell your supporters how to vote without revealing it to your opponents! Needless to say, that's infeasible in all but the smallest elections (how small must a group be to keep a secret?) IRV artifically empowers extremists (although not as much as plurality), making it subject to tactical voting. Borda artificially empowers centrists, making it subject to tactical nomination (even worse!) Your critiques in the draft of approval and Condorcet are shallow and downright dishonest. I urge everyone who reads this to seek out information on those methods elsewhere.

  Posted by Harald Korneliussen, 2007-06-27 04:32:02 (6 years ago)

• RealDemocracy: ... If one candidate wins by both methods, she is elected. If there are different winners, we use Condorcet to determine the winner from those two. This suggestion is good for opening up academic discussion but I don't think it would work for real-world elections. Here are two problems. (1) Transparency. Would voters understand what is going on, and the general reasoning behind it? (2) Decisiveness. Would losing candidates and their supporters always accept the results as legitimate, or would there be endless partisan wrangling (and lawsuits) about who really won? This question isn't original with me. I'm pretty sure I got it from Michael Dummett's writing, but I can't find the specific reference. By the way, I think both issues plague some of the many Condorcet-completion methods out there.

  Posted by Bob Richard, 2007-06-22 12:28:06 (6 years ago)

• "If by chance it is still a tie, the two candidates share the office as co-Representatives, or whatever." Try getting that passed. Two governors. Two mayors. Two members of the U.S. House -- from the same district. Then try and blame the one-person-one-vote violation on Condorcet cycles.

  Posted by Realist, 2007-06-20 20:54:17 (6 years ago)

• It seems that we share some common ground here, and that is the awareness that our current voting method is antiquated, and just about the worst possible way to go about them. So, we should begin from there. There is obviously some controversy about which method is more acceptable in its worst moments, based on which presumption we begin with (i.e. social utility efficiency or the majority rule criterion.) Perhaps we should work together to either blend the two methods or create a new one that will take both of these criteria as vital. While I am no mathematician, I can imagine that if we have voters rate candidates as they would in Range voting, we can imply their rankings for an IRV count from those ratings and compare results. If one candidate wins by both methods, she is elected. If there are different winners, we use Condorcet to determine the winner from those two. If by chance it is still a tie, the two candidates share the office as co-Representatives, or whatever. Would such a method not solve much of this argument?

  Posted by realdemocracy, 2007-06-20 19:02:53 (6 years ago)

• Of course it???s your subjective opinion that ???social utility efficiency is the ultimate voting system criterion.??? It's not a subjective opinion, it's the fundamental point of voting. An election is a group choice, where every voter is trying to maximize the satisfaction of his preferences, so the best election method then is, by definition, the one that best satisfies the voters. That's social utility efficiency. I prefer the majority rule criterion. 1) The point of a good voting method isn't to satisfy your preferences, but those of society as a whole (social utility). Range Voting is vastly better at this than IRV. 2) The majority axiom was refuted by Arrow (and certainly others before him). The axiom is, simply put, "If more voters prefer A to B, then A is a better choice for society". A "better winner" in other words. But Condorcet cycles are possible, in which A is preferred to B is preferred to C is preferred to A - refuting the axiom which underlies your majority concept. Range Voting allows a candidate opposed by 99% of voters to win. This rather vague and misleading statement is addressed here: http://RangeVoting.org/MajCrit.html And remember, Range Voting has a much higher voter satisfaction index than IRV, so it makes more voters more satisfied - by a large margin. That???s what you really mean when you say it???s good for 3rd parties, and that IRV does violence to them. No, that's not what we mean. What we mean is that IRV makes voters strategically forced not to top-rank their sincere favorites, but instead to top-rank their favorite front-runner. That is, with IRV you would never want to top-rank "Nader", because you might get "Bush". Range Voting gets rid of this insanity, and so third parties get fair treatement, because voters can vote their hopes, and not their fears.

  Posted by Clay Shentrup, 2007-05-05 21:33:04 (6 years ago)

• The RangeVoting.org website includes a page about Richie's argument (refuting essentially every point he makes) here: http://RangeVoting.org/RichieRV.html Warren D. Smith (co-founder of rangevoting.org and math PhD.)

  Posted by Warren D Smith, 2007-02-18 12:43:39 (6 years ago)

• Here's some further responses to more of Rob's claims. http://groups.google.com/group/rangevoting/web/2-haloscan-response By the way Rob, did you know that in an election with around 10 or more candidates, Range Voting more often picks the Condorcet ("majority rule") winner than IRV? I didn't even realize that myself until just tonight.

  Posted by Clay Shentrup, 2007-02-07 04:50:11 (6 years ago)

• Correction: The average satisfactions are: I summed them, rather than averaged them. You can do it either way, I just said the wrong thing.

  Posted by Clay Shentrup, 2007-02-07 01:42:12 (6 years ago)

• With respect, what do you mean by social utility efficiency? Expected satisfaction voters will get from the results of elections. Similar to the economics concept of "expected value". Say that we could read voters' minds, and express their like or dislike for the candidates in terms of utility - how much satisfaction they'd feel if that candidate won. How much they want that candidate to win. Here's a hypothetical example: McCain, Gore, Dean Rob 0, 3, 4 Clay -2, 5, 6 Jack 0, 7, -5 The average satisfactions are: McCain: -2 Gore: 15 Dean: 5 The utility produced by the "ideal" winner, Gore, is 15 "happiness units" (yes, this is like using "cubits", but we'll divide out the units later, to get a unit-less ratio). The average utility of the candidates is 6. With plurality voting, Rob and I would pick Dean, and you'd be quite unhappy. Social utility efficiency is calculated as (U - R) � (I - R) where U is the utility produced by the voting method in question, R is the expected utility produced by random selection (in other words, the average utility of all candidates), and I is the utility of the ideal candidate. So in this case, plurality achieves a social utility efficiency of (5 - 6) � (15 - 6) = -11% That is, plurality achieves a worse result than what would be expected by randomly pulling a name out of a hat. That's what a negative utility efficiency means. In this election, honest Range Voting would have elected Gore, achieving a social utility efficiency of 100%. Now do hundreds of millions of these simulations, with different utility generators, different strategies, different numbers of voters and candidates, different proportions of honest vs. "gaming" voters within the electorate, etc. etc., and look at the average results. That's how you calculate social utility efficiency. In this example, a novice might say that Dean should win, since a "majority" prefers him. Maybe Rob and I would stomp our feet and say the same thing. But, if we use a method that produces that effect in this election, we will find that it will, over the course of many elections, leave us much worse off than we'd be with Range Voting. Any good economist would, once informed of this fact, very much want to have Range Voting, not plurality or IRV (or Condorcet, or Borda). ???Condorcet-ness??? is about whom most voters prefer when the field is restricted to two. Well, no. There can be a Condorcet winner regardless of how many candidates there are. But I think what you mean is that Rob was specifically talking about a two-candidate scenario. In this example two-candidate scenario, digit-Rob and digital-Clay may claim that we are a "majority", and should therefore get to have our way. But if Gore is elected, we are both just a tiny bit less happy, whereas your happiness increases substantially. Since we may not be in the majority in the future (even within the primary for our party), it is stupid for us to care more about electing the "majority" winner than about using the method that gives us the most happiness the most often. But say this still doesn't satisfy you, just as it didn't satisfy me for a long time (I insisted that a majority was the paramount decider), think about this Range Voting scenario, using a 0-10 range: McCain, Gore, Dean Rob 0, 0, 10 Clay 0, 0, 10 Jack 0, 10, 0 With Range Voting, a majority can always bullet vote like this, and get still get their way. IRV is about preventing elimination by being enough voters??? first pick. That's how the theory goes at least. But as I've pointed out again, and again, and again, IRV has serious flaws that make it unsuitable for use in political elections. Regards, Clay

  Posted by Clay Shentrup, 2007-02-07 01:40:07 (6 years ago)

• With respect, what do you mean by social utility efficiency? "Condorcet-ness" is about whom most voters prefer when the field is restricted to two. IRV is about preventing elimination by being enough voters' first pick.

  Posted by Jack, 2007-02-05 21:59:42 (6 years ago)

• Jack makes a good point here that range voting is very interesting for something like this presidential poll, although even here only if people vote sincerely and ???one side??? doesn???t figure out it gets a boost by giving a zero to its chief competitor. Wrong. As I have explained several times, Range Voting even with 100% strategic electorates still outperforms other methods - in fact it outperforms plurality and IRV with 100% honest voters. I will note that Warren Smith has just added a new voting method to his simulations (there are like 50 methods now), where you do approval voting, and then hold a plurality election between the top two winners from the approval process, and it seems to do better than Range Voting. We'll see how that holds up to scrutiny. But I still prefer the one-step simplicity of Range Voting. About your "if one side gets it" comment, Warren also recently (finally) tested scenarios where only one side of the political spectrum is strategic. Yawn. Range still dominates. Clay, our criteria aren???t arbitrary. They???re central to a system['s] being able to win politically. Really, try to explain to a charter commission or state legislative committee that your system regularly would allow a candidate who would lose by a landslide in two-candidate race in plurality to have a very real chance to win with range voting. 1) I completely understand your point, but we have to educate the public about why social utility efficiency is a better metric than Condorcet-ness. You seem to be resigning yourself to the idea that we can't change public perception, so why not try to sell them the best method that satisfies their rather flawed notions of what criteria are necessary for a voting method to obey. Given the ENORMOUS social utility benefit that would arise from Range Voting (we're potentially talking about the success or failure of humanity, because our voting methods are terrible), I think it's worth the good fight. I think it would be do-able if organizations such as yours would 100% support Range Voting. 2) IRV can easily have situations where the centrist loses, even though he is preferred to the left and right candidates by a 66% super-majority. But such a situation in Range is less likely i think, or at least less "bad". Clay

  Posted by Clay Shentrup, 2007-02-02 15:04:13 (6 years ago)

• What???s socially useful in one decision-making situation might not be socially useful in another. Nope. Our social utility efficiency calculations included scenarios where utility values were picked randomly, such that they don't specifically model a political ideology distribution. Those utility values could represent how much the voters liked different kinds of soups that their job site cafeteria would carry, for instance. They could represent the utilities from any kind of decision. Range Voting impressively beats other systems though. There are good places to use range voting, and there are good places to use instant runoff. This was an intriguing use of RV, for example. I do not know of a single example of an election that would be better off with IRV than with Range Voting. Can you name one?

  Posted by Clay Shentrup, 2007-02-02 14:45:26 (6 years ago)

• Of course it???s your subjective opinion that ???social utility efficiency is the ultimate voting system criterion.??? Nope. You yourself will be happier with the results of a voting method with a higher social utility efficiency. I prefer the majority rule criterion. Ultimately, the only sensible thing "majority rule criterion" can mean is "Condorcet criterion" (a candidate who would win head-to-head against every other, must win) So what about this example: 35% A > B > C 33% B > C > A 32% C > A > B All three candidates are beaten by a landslide by another candidate. Who is the "majority" winner here? This was Arrow's famous reductio ad absurdum, which shows that "majority rule" is simply a broken concept, that sometimes doesn't even have a meaning. Furthermore, these "majority rule" Condorcet voting methods can absolutely fall apart under strategic voting. Read about the Dark-Horse-plus-3 pathology: http://RangeVoting.org/DH3.html Range Voting allows a candidate opposed by 99% of voters to win. False. It allows a candidate who is not the favorite of 99% of voters to win. That's completely different than being "opposed"; and it only does this if the winner is only slightly less preferred by that majority, and hugely preferred by the 1% minority. I explained this above in a previous post, but it seems that none of my reponses are acknowledged. The same arguments are simply repeated. Hmm. That???s what you really mean when you say it???s good for 3rd parties, and that IRV does violence to them. No, this has nothing to do with why Range Voting is more fair for alternative parties. It is better for alternative parties because it passes the following criteria: * Independence of irrelevant alternatives (means you never get spoilers). IRV fails this. * Favorite betrayal criterion. Never an incentive to strategically "betray" your favorite candidate. So if Nader is your favorite, you have no reason to give any other candidate a higher rating. IRV fails this, as I have demonstrated above once or twice. That is, Nader is your favorite, but if you say that, you get Bush, so you lie and say Gore is your favorite. Poor Nader - with IRV, he's a goner. But with Range Voting, there's no reason to vote strategically except for the few candidates perceived to be serious contenders. So what we see voters do, in exit poll experiments, is maybe give Gore 9 or 10, and give Bush a 0 or a 1, but Gore voters will honestly give Nader maybe a 4 or an 8, because they don't think he's going to win anyway. This helps fledgling candidates get realistic representation of their support. In subsequent elections, this momentum might cause them to have a realistic shot at winning, and then they have to face the effects of strategy. But the "nursery effect" from Range Voting helps them enrormously. And with no fear of "splitting the votes", there's never a reason for an independent or alternative party candidate not to run. So now you know the real reason that Range Voting is good for minor parties. You can read more at http://RangeVoting.org/NurseryEffect.html Regards, Clay Shentrup

  Posted by Clay Shentrup, 2007-02-02 14:26:33 (6 years ago)

• Jack makes a good point here that range voting is very interesting for somethign like this presidential poll, although even here only if people vote sincerely and "one side" doesn't figure out it gets a boost by giving a zero to its chief competitor. Clay, our criteria aren't arbitrary. They're central to a system being able to win politically. Really, try to explain to a charter commission or state legislative committee that your system regularly would allow a candidate who would lose by a landslide in two-candidate race in plurality to have a very real chance to win with range voting.

  Posted by Rob Richie, 2007-02-01 11:43:42 (6 years ago)

• What's socially useful in one decision-making situation might not be socially useful in another. There are good places to use range voting, and there are good places to use instant runoff. This was an intriguing use of RV, for example.

  Posted by Jack, 2007-02-01 11:22:03 (6 years ago)

• Of course it's your subjective opinion that "social utility efficiency is the ultimate voting system criterion." I prefer the majority rule criterion. Range Voting allows a candidate opposed by 99% of voters to win. That's what you really mean when you say it's good for 3rd parties, and that IRV does violence to them.

  Posted by Boggled by CRV, 2007-02-01 11:10:43 (6 years ago)

• Social utility efficiency is the ultimate voting system criterion. It is the aggregate of all other criteria. It is, literally, voters' expected satisfaction. Since the whole point of a choice is to derive as much net satisfaction as possible, and an election is just a group choice, the "best" voting method is the one causes the most satisfaction for the most people. In fact, any good economist, once properly educated about utility efficiency, would definitely want to use Range Voting over any other method, whether in a public election, or to choose which catering company his office would use for an employee banquet. If you don't agree, and you insist that this is just my "subjective opinion", I challenge you to name any sensible alternative definition of "best winner".

  Posted by Clay Shentrup, 2007-01-31 12:25:05 (6 years ago)

• What amazes me is how the factions cite "voting systems criteria" as though they're gospel. There is no perfect voting system. The one you choose depends on the criteria you favor. Thus I laugh when people like Shentrup start throwing around the words "better" and "worse winner."

  Posted by Boggled by CRV, 2007-01-31 10:48:16 (6 years ago)

• Rob Richie: In my previous response, I explained why your three arbitrary criteria are not good measures of the quality of a voting system. I then named three common sense criteria that are much more widely regarded in voting theory - so much that they even have specific names - that Range Voting passes, but IRV does not: * Monotonicity criterion - Increasing support for a candidate can help but never hurt that candidate's chances of winning. Example: Percent of voters - their vote 33% Bush > Gore > Dean > Nader 29% Gore > Bush > Dean > Nader 24% Dean > Gore > Bush > Nader 14% Nader > Dean > Gore > Bush -- ?! Bush wins. But if the Nader voters had instead dishonestly voted Bush > Nader > Dean > Gore, then Gore would have won (which they'd prefer to the old winner Bush) despite the fact this just *raised* their opinion of Bush from last to first place. That is a severe example of non-monotonicity, and an open invitation for strategic voting. * Favorite betrayal criterion - Those Nader voters above had a strong incentive to strategically place Nader second ("favorite betrayal"). This is a large part of the reason that IRV is lethal to third parties, as demonstrated by Australia's House of Representatives, and Ireland's Presidential post (which is actually closer to a monopoly than a duopoly). * Independence of irrelevant alternatives - In the above example, Gore beats every other candidate head-to-head, so it seems he should be the rightful winner. And indeed, if Dean had not run, Gore would have gotten a majority in the first round. By running, Dean does not win, yet he changes the outcome of the race - he is a spoiler. Why should Dean's participation affect the race between Gore and Bush? Why should there be spoilers? With Range Voting, this paradox is solved. These serious problems and paradoxes make IRV a poor choice for use in public elections. Range Voting solves these problems. You point out that Range Voting can fail to elect a winner who is the first choice of many. But this is misleading. When Range Voting does this, it is usually because it picks a *better* candidate, whereas with IRV it almost always elects a *worse* candidate. For example: Voters - Their score (0-10) Voter 1 - A:10 B:9 Voter 2 - A:9 B:8 Voter 3 - A:2 B:9 Voter 4 - A:10 B:8 Voter 5 - A:9 B:7 A's average = 6.2 B's average = 8.2 While A may be the first choice of 80% of the voters, electing B makes voter 3 extremely satisfied instead of miserable, and only decreases the satisfaction of the other voters by a small amount. So Range Voting gives the greatest overall satisfaction. For anyone who studies economics or gambling, you should recognize that this means getting your highest expected value with Range Voting - and expected value is the bottom line. With Range Voting, voters can expect to be happier with results than with any other common voting method. Since the entire point of making a choice is to be as satisfied as possible, this is a clear sign that Range Voting is objectively the best voting method. This is true whether voters are 100% honest or 100% strategic, so it is ironic that you bring up strategic voting. Range Voting is objectively much less susceptible to strategic voting than IRV is. For example, had every voter in the example above been strategic, A would have won the election, which is what plurality or IRV would have achieved anyway. But look how badly IRV behaves in the face of strategic voting. The following scenario shows how voters would have rated the candidates on a 0-10 scale. 35% Bush:2 Gore:1 Dean:0 32% Gore:10 Bush:1 Dean:0 33% Dean:10 Gore:9 Bush:0 Bush average = 1.02 Gore average = 6.52 Dean average = 3.3 Here Gore would command an enormous lead, in terms of overall happiness. But with IRV, Gore would be eliminated in the first round, giving Bush a 67% landslide victory. Again and again, deep analysis reveals what oversimplification hides away. IRV is riddled with problems, and is not a good method for public elections. I agree with you that it is a slight improvement over plurality voting. But if IRV is taking the time and reform energy of electoral reform activists, that could be much better spent on Range Voting, then it is a net loss. Also, it is not fair to use a system like IRV that leads to entrenched two-party domination. Alternative parties deserve a chance to break stifling two-party domination, which they could get from Range Voting. This is why the Libertarian Reform Caucus has already gotten behind this innovative idea. IRV has been around for 137 years, and continually waxed and waned in popularity in America. There is every sign that it has stalled, and has no long-term momentum. Range Voting, on the other hand, is new. The Center for Range Voting was founded in 2005, yet Range Voting has already gained traction and popularity, because it succeeds where IRV fails. I encourage you to work for progress, and make the most productive use of the reform momentum that exists in America right now. We can do better than IRV. Range Voting is the key. Regards, Clay

  Posted by Clay Shentrup, 2007-01-29 23:31:11 (6 years ago)

• Rob Richie: In my previous response, I explained why your three arbitrary criteria are not good measures of the quality of a voting system. I then named three common sense criteria that are much more widely regarded in voting theory - so much that they even have specific names - that Range Voting passes, but IRV does not: * Monotonicity criterion - Increasing support for a candidate can help but never hurt that candidate's chances of winning. Example: Percent of voters - their vote 33% Bush > Gore > Dean > Nader 29% Gore > Bush > Dean > Nader 24% Dean > Gore > Bush > Nader 14% Nader > Dean > Gore > Bush Nader > Dean > Gore, then Gore would have won (which they'd prefer to the old winner Bush) despite the fact this just *raised* their opinion of Bush from last to first place. That is a severe example of non-monotonicity, and an open invitation for strategic voting. * Favorite betrayal criterion - Those Nader voters above had a strong incentive to strategically place Nader second ("favorite betrayal"). This is a large part of the reason that IRV is lethal to third parties, as demonstrated by Australia's House of Representatives, and Ireland's Presidential post (which is actually closer to a monopoly than a duopoly). * Independence of irrelevant alternatives - In the above example, Gore beats every other candidate head-to-head, so it seems he should be the rightful winner. And indeed, if Dean had not run, Gore would have gotten a majority in the first round. By running, Dean does not win, yet he changes the outcome of the race - he is a spoiler. Why should Dean's participation affect the race between Gore and Bush? Why should there be spoilers? With Range Voting, this paradox is solved. These serious problems and paradoxes make IRV a poor choice for use in public elections. Range Voting solves these problems. You point out that Range Voting can fail to elect a winner who is the first choice of many. But this is misleading. When Range Voting does this, it is usually because it picks a *better* candidate, whereas with IRV it almost always elects a *worse* candidate. For example: Voters - Their score (0-10) Voter 1 - A:10 B:9 Voter 2 - A:9 B:8 Voter 3 - A:2 B:9 Voter 4 - A:10 B:8 Voter 5 - A:9 B:7 A's average = 6.2 B's average = 8.2 While A may be the first choice of 80% of the voters, electing B makes voter 3 extremely satisfied instead of miserable, and only decreases the satisfaction of the other voters by a small amount. So Range Voting gives the greatest overall satisfaction. For anyone who studies economics or gambling, you should recognize that this means getting your highest expected value with Range Voting - and expected value is the bottom line. With Range Voting, voters can expect to be happier with results than with any other common voting method. Since the entire point of making a choice is to be as satisfied as possible, this is a clear sign that Range Voting is objectively the best voting method. This is true whether voters are 100% honest or 100% strategic, so it is ironic that you bring up strategic voting. Range Voting is objectively much less susceptible to strategic voting than IRV is. For example, had every voter in the example above been strategic, A would have won the election, which is what plurality or IRV would have achieved anyway. But look how badly IRV behaves in the face of strategic voting. The following scenario shows how voters would have rated the candidates on a 0-10 scale. 35% Bush:2 Gore:1 Dean:0 32% Gore:10 Bush:1 Dean:0 33% Dean:10 Gore:9 Bush:0 Bush average = 1.02 Gore average = 6.52 Dean average = 3.3 Here Gore would command an enormous lead, in terms of overall happiness. But with IRV, Gore would be eliminated in the first round, giving Bush a 67% landslide victory. Again and again, deep analysis reveals what oversimplification hides away. IRV is riddled with problems, and is not a good method for public elections. I agree with you that it is a slight improvement over plurality voting. But if IRV is taking the time and reform energy of electoral reform activists, that could be much better spent on Range Voting, then it is a net loss. Also, it is not fair to use a system like IRV that leads to entrenched two-party domination. Alternative parties deserve a chance to break stifling two-party domination, which they could get from Range Voting. This is why the Libertarian Reform Caucus has already gotten behind this innovative idea. IRV has been around for 137 years, and continually waxed and waned in popularity in America. There is every sign that it has stalled, and has no long-term momentum. Range Voting, on the other hand, is new. The Center for Range Voting was founded in 2005, yet Range Voting has already gained traction and popularity, because it succeeds where IRV fails. I encourage you to work for progress, and make the most productive use of the reform momentum that exists in America right now. We can do better than IRV. Range Voting is the key. Regards, Clay

  Posted by Clay Shentrup, 2007-01-29 23:29:29 (6 years ago)

• Clay misses the point that range voting has absolutely no viability given the flaws we lay out in this draft analysis -- e.g., allowing a candidate to win who would lose 98% to 2% in a plurality system, allowing a candidate to win who would get 0% of the vote in a plurality election and immediately and constantly establish incentives for strategic voting. His regular flaming of IRV around the internet serves only one potential purpose: preservation of the status quo, which even he will admit is improved by IRV.

  Posted by Rob Richie, 2007-01-27 14:17:04 (6 years ago)

• ???IRV does not reward candidates who avoid all controversial issues.??? This is such a vague unquantifiable statement as to be hard to prove or disprove.

  Since when does a desideratum not count simply because it's non-quantifiable?
  Posted by IRV is better, 2007-01-23 18:43:38 (6 years ago)

• Clay, your comment was in moderation cue as a first-time poster.

  Posted by FairVote Staff, 2007-01-23 17:34:00 (6 years ago)

• Okay, my sincere apologies. Somehow when I first reloaded this page, I didn't see my comment. Regards, Clay

  Posted by Clay Shentrup, 2007-01-22 21:07:02 (6 years ago)

• I published a thorough response to this posting, which was quite civil and appropriate, and you deleted it shortly after I posted it. Notice we don't censor IRV supports. You are welcome to argue till you're blue in the face at the RangeVoting group at Yahoo Groups. Clay Shentrup Seattle, WA 206.203.6619 http://RangeVoting.org/

  Posted by Clay Shentrup, 2007-01-22 21:04:05 (6 years ago)

• There are numerous reasons why Range Voting is substantially superior to IRV, especially for third parties, since IRV leads to self-reinforcing two-party domination. Your post exhibits the typical myths and misconceptions about IRV, beginning with your very first point. "With IRV, a candidate cannot win an IRV election without being acceptable to a majority of voters." This claim is both false and misleading. See this example: #voters Their Vote 41 Right>Center>Left 10 Center>Right>Left 10 Center>Left>Right 40 Left>Center>Right With IRV, Right wins, 51-50, even though people prefer Center to Right, 60-41. So much for the myth that IRV picks the "majority" winner. Of course, the more logical thing would be for IRV to pick the _utilitarian_ winner - the one that produces the highest social utility; IRV also does a very poor job of that, as demonstrated at http://RangeVoting.org/vsr.html Furthermore, your statement is misleading, because you say that this "majority" will satisfying to a majority. But that is impossible to say, since we have no idea how satisfied voters might be with the candidates. They may hate all of them, and just be picking the lesser evil. Or they may LOVE all of them. IRV doesn't use intensity of preference, like Range Voting does - so it can't tell us anything about how many voters are satisfied. However, Princeton math Ph.D. Warren D. Smith has done social utility efficiency calculations which show that, on average, voters are MUCH less satisfied with IRV results than they would be with the results of better methods, like Approval Voting, or Range Voting. "[IRV] Requires Both Breadth and Strength of Support:" The only objective way to state that is in terms of social utility efficiency. The following SUE figures show that IRV falls short of the results of Range Voting, by a lot. Voting system - Utility efficiency Magically elect optimum winner 100.00% Range (honest voters) 96.71% Range & Approval (strategic voters) 78.99% IRV (honest voters) 78.49% Plurality (honest voters) 67.63% IRV (strategic exaggerating voters) 39.07% Plurality (strategic voters) 39.07% Elect random winner 0.00% "IRV does not reward candidates who avoid all controversial issues." This is such a vague unquantifiable statement as to be hard to prove or disprove. "Ease of Use: Voters simply rank candidates in order of choice." Range Voting is arguably even easier, as it's the very method that millions of people use online every day, to rate movies, books, comments, pictures, etc. on countless web sites. Some might find ranking easier while others might find rating easier; but it's a minor difference, and the most important thing is that the voting system picks the right winner. IRV produces very poor candidates, whereas Range Voting produces substantially more satisfying candidates than plurality OR IRV. IRV is just barely even an improvement over our poor plurality system. "It has been implemented well" Only if you don't count the expense of upgrading voting machines, and the kind of delays in San Francisco, due to some problems they experienced. Range Voting, on the other hand, is a one-round process that can be performed on ordinary plurality voting machines. It's simpler, easier, and cheaper. "exit polls show that large majorities of voters prefer IRV to their prior system." But that is only based upon their subjective feeling of how much they "enjoyed" the voting process; no doubt, IRV is a more interesting process than plurality. But this has nothing to do with the quality of the candidates that IRV selects. I could devise a "fun" voting method that almost always picks the WORST candidate. So statements like this are very unscientific, and misleading. "Cleaner Campaigns: Successful candidates have more incentive to reach out to supporters of other candidates to earn second choice support." It is possible that there is a mild effect like this, but I see no reason why Range Voting wouldn't produce this effect to as great a degree, or to an even greater degree. "No Complex Strategies: Voters in IRV elections almost never face a ???spoiler??? dilemma no matter how many candidates run" On the contrary, spoiler scenarios happen just under 20% of the time in IRV elections. I showed you an example scenario above, where there's an incentive for strategy, to avoid having IRV pick the "wrong winner". You also make a false assumption that the only time to be strategic is when there is a spoiler scenario. Range Voting NEVER exhibits such scenarios, yet there are still strategies to be employed. There is strategy in every voting method, but the issue is how well the voting method behaves under strategic voting. Social utility efficiency calculations show that Range Voting dominates IRV, especially when voters are strategic. I should point out that you make another error in implying that "complex" strategies are bad. Strategy incentive is bad; but the more complex the strategies are to use, the harder it is for people to be strategic instead of honest, which is GOOD. So, yay for complex strategies. That's the kind we want. "..which means they can vote their hopes and not their fears." This is one of the most pernicious catch phrases of the IRV movement. It is scientifically meaningless. It is poetry. Say you are a Left fan and you vote sincerely in the IRV election above. You get Right as the winner. But if you vote for your second choice, Center, then Center wins, and you get rewarded for being strategic and "voting your fears not your hopes". Why does the IRV movement love to use these cute phrases instead of actual science? "IRV is essentially resistant to strategy, according to voting experts like Nicolaus Tideman." As I just explained, that is a clear myth. And notice that Range Voting produces higher voter satisfaction with 100% STRATEGIC voters than IRV produces with 100% HONEST voters. Even IF using IRV somehow compels voters to be honest, they still get worse results with IRV than they do with Range Voting even if they are completely STRATEGIC when using Range Voting. "* Potential Taxpayer Savings: Because IRV can replace runoff or primary elections with one general election, many jurisdictions see major savings after adopting IRV." Same is true of Range Voting, and it's even cheaper because you only have to do one round of vote counting, AND it can be done on ordinary totally plurality machines. "Range voting, approval voting and Condorcet and variations of these election methods are not used for elections for any public office in elections around the world, but have their strong advocates." So why not start promoting Range Voting, and make it happen? Why not promote a better voting method, so we CAN get Range Voting used in public elections? "Although these systems may have appropriate applications, IRV is preferable for our elections." What is your evidence for that? I just explained how Range Voting is better than IRV on every single point you mentioned. In light of those facts, you should start changing your tune. "We highlight the following criteria to evaluate a single winner system???s merits and political viability" The only criterion you should be overly concerned with, so long as the system is feasible to implement, and easy enough that voters can use it, is social utility efficiency. That is literally the average expected satisfaction voters will feel over the result. The higher the SUE, the happier the voting method will make the voters. Voters should support Range Voting over IRV in order to promote their own happiness, and the happiness of everyone else in society. As any economist knows, you want to maximize your expected value in any transaction. With voting methods, maximizing your expected value means maximizing your expected satisfaction with the winner. "Does the system meet the common sense principle of majority rule? In an election with two candidates, the candidate preferred by a majority should always win." As I just showed, IRV can produce situations where the clear and obvious majority winner (who would win head-to-head against every other by a substantial margin) actually LOSES. In Range Voting, we elect the winner that makes the most people the most happy, so exceptions to majority rule happen when they are actually more satisfying to voters. IRV, on the other hand, can fail to elect a majority winner, and actually make them LESS satisfied, as the example above shows. So this is another myth from IRV proponents. "Does the system meet the common sense principle of requiring a minimum level of core support? A winner should be at least one voter???s first choice." Why is this an important principle? For instance, say the following sets of ordered preferences are held by four voters: A > B > C > D C > B > D > A D > B > A > C Here, B is no one's favorite choice, yet 2/3 of voters prefer B to C; 2/3 of voters prefer B to D, and 2/3 of voters prefer B to A. How can you claim that B doesn't deserve to win, just because he's no one's first choice? Would it be more fair to give one person his favorite and let TWO people get someone they like even less? Doesn't sound very fair to me. "Does the system meet the common sense principle of rewards for sincere voting? A voter should not likely be punished for voting sincerely under the system???s rules." This is a great argument for Range Voting, since voters will derive much greater average satisfaction with the results of Range Voting elections, whether honest OR strategic. And with Range Voting, there is NEVER a strategic incentive not to vote your sincere first or last place preferences, as there easily can be with IRV. A voting system that obeys this property is said to comply with the Favorite Betrayal criterion. That's ONE major criterion that IRV fails, and Range Voting passes. So your criteria are actually quite weak, and/or turn out to make better arguments for Range Voting than they do for IRV. I would also add that Range Voting passes the monotonicity criterion and the independence of irrelevant alternatives criterion, both of which are seen as significant criteria in voting science, and both of which IRV _fails_. So thank you for your wonderful argument for Range Voting. Now we know why one huge faction of the Libertarian Party has already embraced Range Voting. Regards, Clay Shentrup Seattle, WA 206.203.6619 http://RangeVoting.org/

  Posted by Clay Shentrup, 2007-01-22 19:37:50 (6 years ago)