Commentary
Of alternative voting methods, instant runoff is best
by Rob RichieMathematician Donald Saari last week criticized ranked choice voting (also known as instant runoff voting), which will be used this November in Minneapolis municipal elections.
He prefers a method known as the Borda Count, in which voters also rank candidates in order of choice, but their first choice receives a certain number of points, their second choice a smaller number of points, and so on.
Professor Saari is not alone in favoring a particular voting method. Princeton's Eric Maskin zealously advocates Condorcet voting. NYU's Steven Brams promotes approval voting, and so on. That scholars fail to agree on the meaning of the same facts reflects the truth that there is no "perfect" voting system.
There is no objective means to weigh competing values in areas such as rewarding breadth of support or core support, or the balance between upholding majority rule and achieving consensus.
Most politically engaged reformers focus on instant runoff voting (IRV) instead of other alternatives, because it improves the status quo in ways that are easy to see and is more reflective of voters' expectations of fairness.
Political and civic leaders like President Obama, Sen. John McCain, consumer advocate Ralph Nader and the League of Women Voters of Minnesota agree on the value of IRV. Compared to plurality voting, IRV diminishes the "spoiler" impact of third party and independent candidates. Compared to traditional two-round runoffs, IRV saves taxpayer dollars, reduces the influence of money in politics and maximizes turnout.
IRV also nicely balances the value of first choice support with the value of broad support. Plurality voting rewards only core support.
In contrast, the Borda Count tilts toward rewarding broad support by allowing votes to count for more than one candidate at a time. As a result, Borda can elect candidates who might come in dead last in our current system, but be many voters' second or third choice.
Borda also could defeat a candidate who under today's rules would earn an absolute majority of 51 percent of the vote. Many voters would react skeptically to such results.
Worse, Borda makes it easy for voters to game elections. Nicolaus Tideman, author of "Collective Decisions and Voting," includes Borda in his list of voting methods that "have defects that are so serious as to disqualify them from consideration." The system's inventor, Jean-Charles de Borda himself, stated that the system should be used only for situations where voters are disinterested judges.
Consider Borda and a famous example of the "spoiler dynamic," the 2000 presidential election in which Republican George W. Bush defeated Democrat Al Gore by 537 votes in the halted Florida recount, even as more than 97,488 people voted for Green Party nominee Ralph Nader. Given that more Nader backers supported Gore than Bush, it's clear that under instant runoff voting, Gore would have won Florida and the presidency.
But there's no way to know what would have happened with Borda. With Borda, ranking a second choice counts directly against the chances of your first choice. Some Bush voters might tactically have ranked Nader and other independent candidates ahead of Gore. If more Gore voters honestly chose to rank Bush as one of their top choices, then Bush would have won simply due to his voters engaging in better tactical voting.
Instant runoff voting avoids this problem because your ballot never counts for more than one candidate at a time. Unlike Borda and most other alternative voting methods, ranking additional choices never hurts your first choice. As with every voting method, there are cases where candidates might benefit from insincere voting, but with IRV such situations are too convoluted to be an effective tactic.
That IRV has fair results and isn't prone to gaming explains its successful track record in thousands of hotly contested elections -- from nearly a century of elections for Australia's House of Representatives to decades of elections in Irish presidential races. Professor Saari's examples of its theoretical flaws may interest fellow mathematicians, but they are irrelevant for those wanting to improve American democracy.
----
Rob Richie is executive director of FairVote, a national nonprofit organization that says it works "to transform our elections to achieve universal access to participation, a full spectrum of meaningful ballot choices and majority rule with fair representation for all."
Comments (26)
This piece demonstrates Rob Richie's typical tactic of misleading rhetoric. While academics may fall into different camps as to what the "best" voting method is, there is a broad consensus that Instant Runoff Voting is almost the worst. Some even think that it is worse than ordinary plurality voting, when you consider that it increases ballot spoilage rates and increases susceptibility to fraud.
Richie ironically points to the conclusions of Nicolaus Tideman, while failing to mention that Tideman also calls *IRV* unsupportable. More important is the simple fact that Tideman's analysis contained severe flaws, largely due to the fact that he chose arbitrary criteria for his "strategy resistance" measure, with equal weight. A Princeton math Ph.D. named Warren Smith used Tideman's data, while correcting these flaws, and got completely different results. As a quick "sanity check" of Tideman's findings, Smith notes, "Tideman's book also came to the conclusion somehow that approval was more subject to strategic manipulation than plurality voting, contrary to the apparently unanimous opinion of every other expert."
And Richie really shouldn't talk about strategic voting, when IRV is also highly susceptible to it, and effectively degrades to plurality voting when enough voters are strategic. A concrete example is in the last mayoral election in Burlington, Vermont. A faction of voters who preferred GOP>Democrat>Progressive could have gotten the Democrat instead of the Progressive if they had insincerely top-ranked the Democrat instead of the Republican. Well, you can bet that after using IRV for another election cycle or two, they will realize that in a liberal area like Burlington, it is strategically unwise to waste your vote on a Republican. So instead you should top-rank your favorite between the Progressive and the Democrat. When that happens, IRV won't really accomplish much, aside from occasionally preventing a Nader-type spoiler situation. That small benefit is approximately offset by IRV's flaws.
So why does Richie make these claims if IRV doesn't really deliver on them? Simple. His organization, FairVote, has a long-term goal of enacting a method call STV, which leads to proportional representation (i.e. a party that has 20% of the vote gets about 20% of the seats). FairVote used to be called "Citizens for Proportional Representation" in fact. IRV is the single-winner form of STV, and FairVote knows that by getting the machinery in place to do IRV, they are making a more viable path to STV. They think of IRV as a "stepping stone" to STV.
So it doesn't really even matter whether they believe IRV is good. They want STV, and see IRV as a means to an end. Spend some time reading about Rob Richie the person, and you will find this to be the undeniable truth.
Instant Runoff Voting has been used and proven in many jurisdictions to elect majority winners and eliminate the spoiler effect. It is constitutional, as the MN Supreme Court found earlier this year and it is based on the one-person one-vote norm for our elections. There is no opportunity for strategic voting because the voter has to rank their selection on one ballot. And no evidence of any increase in susceptibility to fraud because the same rules apply to the voter for proper registration as is currently used in our two round plurality election processes. Additionally, there is no increase in incorrectly completed ballots from our current voting process.
I'm a visual learner and hearing how votes get transferred from 1st to 2nd if they lost, and recounted, makes my head swim. I have to SEE how it works to understand it.
These YouTube videos really helped me to understand the intricacies of Instant Runoff Voting:
www.tinyurl.com/IRVvideoOneVote
www.tinyurl.com/IRVinLB
www.tinyurl.com/IRVtwo
Shentrup is full of accusations and personal attacks. That is quite unfortunate. All of the statements Richie made about IRV are true, and Fairvote and its allies such as Californians for Electoral Reform, which I am active in, believe that IRV, on its own, whether or not it ever leads to STV, is a very important democratic advance. I personally believe that Condorcet would also be fine for single winner public elections, while Borda and Approval are fatally flawed for that type of election. Voters in Minnesota are not stupid, they know what they are doing, and those in Minneapolis clearly picked wisely when they chose IRV.
What planet is Clay Shentrup on? Most of what he says is simply untrue. There is not "a broad conses that IRV is the worst".
In some very rare and special cases, a majority system could cause voting for one's first choice to cause the voter's second choice to loose and if most of the second choice voters vote for the third choice then the third choice would win. However it could be that the second choice votes for the second choice candidates swing the majority to the voters first choice. The voter has no way of knowing if this could be true when he votes. So any attempt to vote strategically could easily backfire causing the first choice to lose when they could have won.
To simplify, if your vote causes you second choice to not make the runoff, then your third choice may win.
This is true of any majority system whether or not there is a separate runoff or if the runoff is done instantly.
This will happen so rarely that it hardly seems worth mentioning. It certainly doesn't justify not using a majority system especially when there is no other system that provides a majority winner that doesn't occasionally have this flaw.
Shentrip has told me personally that he favors approval voting. Using this system, any vote for anyone other than ones first choice helps defeat the first choice. It is much worse than IRV. I have yet to see him recommend any voting system that is better than a majority wins system. He always compares IRV to Condorcet but that's comparing apples and oranges. Condorcet is a system to choose a broad base support or the candidate in the middle while IRV and other voting types select narrow strong first choice support. Decide which objective is desired then select the system that gives that result. But don't compare two systems designed to meet different objectives to each other.
Shentrup, a mathematician should know better! Perhaps he is too enamored with his math to see the forest from the trees.
Chuck O'Neil
It seems there is alot of buyers remorse pout there for Instant Runoff Voting. Aspen has it on their ballot to get rid of it and so does Pierce County, WA.
This is the language of the Repeal Ordinance:
WHAREAS the cost of running the IRV portion of the 2008 General Election was $1,692,163 and
WHEREAS the IRV portion of the ballot proved to be expensive, complicated and confusing, and the results of the IRV election were not available until weeks following the election date, and
WHEREAS 66% of the 90,738 responding to the Authority's survey indicated they did not like IRV
These are pretty strong condemnations from cities who bought into this scheme. Buyer Beware!
Clay Shentrup cites an example of how Republican voters with leanings Republican>Democrat>Progressive in Burlington Vermont, a city with liberal leanings, would likely "game the system" by insincerely top ranking "Democrat" in their IRV vote, in order to keep the Progressives out of office. But the results of doing this would be exactly the same as if they voted their true preferences. That's the point! That's the beauty of IRV!
Great article. A good summary of the serious shortcomings with Borda and Approval voting.
As Richie said, there is no "perfect" voting system. But when it comes to electing a winner in a single-seat election, IRV is the best choice. It elects a majority winner with higher voter turnout, all while eliminating spoilers and vote splitting effects.
Mr. Shentrup and others who attack ranked choice voting/IRV using "strategic voting" examples always fail omit a critical piece of information: The factions that might change an election result must know its results before they can know how to game the election.
Every example these critics offer starts with the final numbers in an election. They have to know that outcome before they can calculate how strategic voting would have changed it. They never tell how to do that before an election.
Rob Richie is right. There are only three single-winner voting systems used with any regularity in the world: Plurality, Two-Round Runoff, and Instant Runoff. Of those, Instant Runoff is clearly the best. These other alternative systems haven't been sufficient vetted in real settings.
I think the track record of IRV carries a lot of weight. I used IRV when I lived in San Francisco in 2004 & 2006, and it was easy to use and nice not having to come back for a runoff in December. The best part is that it eliminated the spoiler effect, so I could freely vote for who I really want, even if the guy I really like has no chance of winning, without it being a "wasted" vote.
Chuck O'Neill, I am a mathematician and actually I do know better. Clay Shentrup's comments are spot on. And I have met and talked with Nicolaus Tideman. He is not in favor of IRV in any way.
Tideman currently favors a combination of Score Voting and a robust Condorcet method, so using his results to imply support for IRV is disingenuous.
Saari has been pushing Borda for years, despite the fact that it is one of the easiest methods to manipulate strategically. This manipulation has been borne out several times in, for example, dance or skating competitions.
Approval Voting is indeed very simple to implement, and can lead to a better choice than single vote plurality. But in close elections, it can cause problems for voters who are unsure where to draw the line for their approval cutoff, even if they are sincere. However it is favored by mathematical organizations for electing officers, and performs well in situations where sincere voting is the rule.
Instant Runoff Voting is indeed unstable. It follows the tendency of Single Transferable Vote to choose the candidate closest to the center of the largest faction. I don't think this is the goal we want to strive for -- I don't think this would decrease divisive politics.
Besides that, any form of runoff, whether instant or top-two, tends to segment the solution space and constrain your solution away from the global optimum.
Proportional Representation is a worthy goal, and it should be pursued for the settings where it is most appropriate: multi-winner elections for legislative districts. This would ensure a more diverse group of voices in proposing new laws and regulations. STV does a good job of ensuring that most portions of the population have some representation, and it would be a vast improvement over 51% majority rule.
However, for executive and senatorial positions, we should have some form of centrism-biased aggregation, to apply selection to the legislative options in the most broadly supported way possible.
The voting method that performs best in selecting the centroid of the population is Score Voting. The only way for a ranked voting system to find the candidate closest to the centroid of the population is if the majority of the population is clustered around the center. If the population is polarized, IRV and other ranked methods will sustain the polarization.
Score voting is so pervasive that most people are not even aware that it is being used.
The valedictorian of the class is the student with the highest grade point average. That means she has received high grades from each of her teachers. Substitute voters for teachers and candidates for students and you have Score Voting.
Bees decide where to find honey using score voting. The site that has the strongest average score from forager scouts is the one that the hive directs its efforts towards.
Incidentally, Approval voting is just Score Voting with Pass/Fail grades.
A shame that so many are blinded by the promise of IRV that they cannot see that Clay is in fact correct.
IRV is riddled with flaws including falling to the spoiler effect (caused by all ranked systems failing IIA). For those that say it doesn't happen, look at the CRV's detailed analysis of the Burlington, VT election. This election exhibited the spoiler effect, the no-show pathology, and worst of all, the non-monotonicity pathology. IRV also elected a different winner than any voting system other than itself or plurality would have elected. Similar problems can be found in many other elections using IRV. Additionally, these kinds of problems can be found wherever IRV is used, including Australia 2007, Ireland 1990, and Peru 2006.
Many commenters failed to see how Clay's example worked. The Republican would normally vote R>D>P, and get the Progressive, but by switching to D>R>P, he can elect the Democrat, a better choice in his view. As to the fact that people must know the totals ahead of time to make a strategic decision, it is easy to know rough support and figure the best choice in most cases.
IRV does in fact produce significantly higher spoilage rates. Which is easier, pick one choice, or put each candidate in order, with no skipping numbers? Even one number off would invalidate it.
Although it is true that Borda also has many problems, it is not true that Approval does. Approval is highly effective for a simple solution, being the easiest system possible. It also prevents problems found in all Ranked methods (Plurality, IRV, Borda, Condorcet) because it is a rating system. Even better than Approval is it's big brother, Range. Range allows for far larger freedom of expression that IRV, can be done on all plurality voting machines, and is far, far easier to use.
Furthermore, there is a very strong, objective measurement that can be used to evaluate voting systems. Bayesian Regret, the difference between the maximum Utility and the expected utility for an election, has shown Range voting to be the best voting system available, being very near to the best with honest voters. What about strategic voters? Range voting (and Approval), when employed strategically, beats out IRV with honest voters, in terms of lowest regret. There is also data supporting the idea that most (75%) voters do not vote strategically with Range, making it hugely better.
One problem with the war against IRV is that you guys should be our allies. We all agree that Plurality sucks, and should be replaced. You just happen to have picked IRV, which isn't much of an improvement, simply because it is closest to Plurality. There have been more Range votes conducted than every other voting system combined. It is used all over the internet (Youtube, IMDB), used in the Olympics (Truncated), and used by Honey Bees.
When people go to IRV, they quickly realize that with the pathologies, the costs (no voting machines are certified for it in the US, and Plurality machines can't do it), the complications, they switch back. People don't switch away from Range, because it is the best system in nearly all cases.
IRV supporters, consider learning the facts from the CRV website (Google range voting or score voting), and joining with us for a system that is verifiably better than the other options.
Also, sorry for the wordiness, I just had quite a bit to say. Not much longer than Clay's though.
From What Bob T. states, there is not a good track record.
San Francisco has not elected a majority candidate in 10 of it's 11 elections.
http://www.youtube.com/watch?v=Ja9FtQqBYeY
That's not what was promised to the voters when they voted to change to IRV and might even violate their Charter.
And when you vote for a candidate that has no chance of winning, withouut it being a wasted vote, what does that mean? You vote still goes to the popular candidate and you favorite thrown to the curb, his vote not even counted, yet you feel good? 48% of the SF voters didn't even know they were supposed to rank the voters. Some system you have their after spending millions. You can keep it. Aspen's scraping it, so is Pierce County, and other will not implement it.
I agree with Clay Shentrup that IRV is not the best way to go.
Still, plurality is not nearly good enough to deserve our continued suffering with it, and the way IRV allows voters to rank candidates is worthy for letting them indicate their major preferences easily. IRV's trouble is that it does not notice all that they may say.
Many races are simple - only one or two candidates deserving serious consideration and, thus, the method matters little. But, sometimes many voters would like to vote for more than one, would like to indicate which of such are best liked, and would like the rules the counting will use be understandable. Condorcet lets them use ranking, reads ALL that they vote, and uses that to see what they have said means as to each pair of candidates (like them equally or prefer one). Borda is similar, but its way of counting encourages attempts at strategy.
IRV can do such as, given 33A, 34B>C, 33C, 2A>C, it will never notice that twice as many voters prefer C as A, for its first step is to discard the 33C and see 35A>34B. Condorcet will see 67C>35A and 35C>34B, meaning C properly wins.
Some argue for Runoffs - valuable recovery for plurality where voters cannot fully express their desires - and flunks if offered two to pick from when truly best-liked was third.
Several of the responses here support my case that IRV is supported in large part because of myths and misinformation, that are unfortunately perpetuated by IRV advocates such as Rob Richie.
For instance, some comments repeated the claim that IRV is proven to "elect majority winners and eliminate the spoiler effect" as well as to eliminate the "wasted vote" problem. All of these notions are simply inaccurate. It is a proven fact that IRV can elect candidate X even when candidate Y is preferred to X by a majority of voters, and got more first-place votes (which Richie often refers to as "core support"). Here's a simplified example:
http://scorevoting.net/CoreSuppPocket.html
And in the recent mayoral election in Burlington, Vermont, the Republican acted as a "spoiler" candidate. If you reticulate the exact same ballots, but with him removed, then the Democrat wins instead of the Progressive (and note that a majority of voters in fact preferred the Democrat to the Progressive, again refuting the "IRV elects a majority" myth). Here's a thorough analysis of this election by Warren D. Smith, a Princeton math Ph.D. who served somewhat as the protagonist in the book "Gaming the Vote", by William Poundstone.
http://scorevoting.net/Burlington.html
Roan Cramer commented that "There is no opportunity for strategic voting [with IRV]". But as I mentioned in my first comment, there was a faction of voters in Burlington who preferred the Republican over the Democrat over the Republican, and could have caused the Democrat to win instead of the Progressive by insincerely/tactically top-ranking the Democrat. Another commenter, Lois Brain, inexplicably said:
"..the results of doing this [strategically/insincerely top-ranking the Democrat] would be exactly the same as if they voted their true preferences. That's the point! That's the beauty of IRV!"
This is simply false. As I stated, if this faction of voters had insincerely top-ranked the Democrat, then the Democrat would have won instead of the Progressive - an improvement from the point of view of this group of voters. It is a mathematical fact that every deterministic voting method is susceptible to strategic voting. Check out the Gibbard-Satterthwaite theorem.
Check O'Neil says my claims are "simply untrue", yet I've provided proof for every concrete claim I've made. IRV does not eliminate spoilers, and does not elect true "majority winners", and does not prevent strategic voting. These are proven facts, and we have full ballot data from elections like the one in Vermont, which prove it.
Chuck O'Neil and Ken Bearman also have the common misconception that such strategic voting will be impractical, because voters will have no way of knowing that such a strategy will be useful in advance. There are a couple of reasons why that's unrealistic. For one, if Burlington voters consistently see that the Republicans lose to the more liberal candidate (whether the Progressive or the Democrat), regardless of which liberal candidates makes it to the second round, then they will eventually have empirical evidence that says "voting Republican is a waste of your vote". Then they can effectively use the strategy of top-ranking their favorite between the Democrat and the Progressive, regardless of who their overall favorite is. This is the same basic idea that voters have in today's vote-for-one elections, when they know a Green Party candidate probably won't win, so they don't want to "waste" their vote on one. This is also based on pre-election polls. Chuck, don't you remember the last presidential election, in which polling matched up various candidates one-to-one? Say Burlington voters see a poll that says the Republican will lose to any candidate who has a viable chance to making it to the final round of voting? That's all they need to know to use the strategy I described. It's not rocket science. If you really just can't believe that such polls will still exist in an IRV world, just read about Australian politics, where they've used IRV since the early 1900's. Strategic voting is widespread, and they even have "How to Vote" cards that make it easier.
Finally, approval voting is a far superior system to IRV. It is much simpler, it reduces the rate of spoiled ballots instead of increasing them, it can be totaled in precincts (unlike IRV) so it is more resistant to fraud, and it is much more representative according to an objective measure called Bayesian regret, explained here (lower Bayesian regret means better):
http://scorevoting.net/UniqBest.html
The criticism that people often make of approval voting (which is the simplest form of score voting), is that voting for a candidate that is not your favorite can cause that candidate to defeat your favorite. The implication is that voters will thusly "bullet vote" for only their favorite candidate, and then approval voting will effectively degrade to plurality voting, like we already have. The irony of this criticism is astounding, because that's actually the problem with IRV, *not* with approval voting. This is explained at length here:
http://groups.google.com/group/scorevoting/web/degrade-plurality
The simplest way to see the unsoundness of this argument is to realize it's the same as saying Nader supporters back in 2000 wouldn't have wanted to vote for Gore, since that would help Gore defeat Nader. It's also effectively the same as arguing that plurality voting is free from strategy, in practice - even though one of the common criticisms IRV proponents wage against plurality voting is that it gives people an incentive to betray their favorite candidate for their favorite front-runner - oh the irony).
One final note regarding the comment by Roan Cramer:
She correctly noted that ballot spoilage was less of a problem with Burlington's IRV election than in most other IRV elections. But here in my home of San Francisco, and in other locations, ballot spoilage has been about seven times as severe with IRV. Perhaps Burlington could teach us some lessons about better implementation of IRV, to reduce ballot spoilage (although a vastly superior option is to just use approval voting or score voting).
See http://scorevoting.net/SPRates.html
While I disagree that Range Voting is a reasonable alternative, it's clear that IRV can lead to some funky results: http://elections.cognitivesandbox.com/#ignored
In the linked example, the majority prefer one candidate over another, but IRV elects the latter. Does't seem right, does it?
An excellent video on single winner voting systems can be found here:
http://www.teach12.com/ttcx/VotingFreeLecture.aspx?ai=32018&WT.mc_id=FLSNI20081103&ev=s
For the record: I consider IRV to be one of the worst voting methods. (It fails many criteria, scores very poorly in terms of baysean regret, looks like a drunkard decide who gets to win on Yee pictures and is not summable by precincts which should make it a hugely unpopular for election integrity advocates who favor having publicly observed counts of votes at the precinct level.)
Condorcet methods, approval voting and range voting are all improvements over plurality in my opinion. I most favor approval voting and low-grained range voting.
Earlier in his career, Rob Richie often just told blatant falsehoods as part of his quest to push IRV. But I am glad to see he is improving. For example, in this piece he correctly says "IRV diminishes the 'spoiler' [effect]" whereas the old Richie would have falsely said it "eliminates" it (as commenter Roann Cramer unfortunately does).
Unfortunately, in many cases the "improvement" consists of now saying things which technically are true, but whose wording has been carefully honed over years to try to neverthless create a misleading impression. That's better, I guess, but still not people should really want.
Richie mislead #1: "Ranking additional choices (under IRV) never hurts your first choice." This statement is true but nevertheless carefully designed to mislead. Ranking additional choices under IRV can be a strategic mistake for you (the voter) since it will worsen the election result (in your view). Note my statement here is subtly different than Richie's and addresses the question you (as a voter wondering whether to 'rank additional choices') actually want to know -- not a wrong question carefully designed to sound similar to the right question.
In fact, IRV elections can and do exist where you, by ranking fully, cause somebody you dislike, to win; and by not ranking fully, you get a better winner. Don't see why? (Indeed, it is not so obvious why. You'll have to work for a while with paper and pencil.) In that case, since I know how to game IRV and you do not, I will have an advantage over you at the polls.
Richie mislead #2: "IRV avoids this problem because your ballot never counts for more than one candidate at a time." Well... "one at a time"? Kind of a silly criterion. And why concentrate on avoiding supporting "more than one"? Nothing should be wrong with supporting more than one. The problem here is Richie is stating a criterion which naively sounds mildly reasonable... and using it as the foundation of the whole design of IRV... but actually, when you think about it, it's a silly and wrong criterion.
What CAN (disturbingly) happen with IRV is your ballot says you prefer Gandhi over Hitler, BUT this preference is NOT COUNTED at the "time" (as Richie puts it) when the decision is made over who to eliminate -- Gandhi or Hitler. But meanwhile, some Hitlerite who ranked Hitler top, IS counted. I.e. some votes, but not others, get counted. Hence Hitler ends up prevailing, even though a huge majority prefer Gandhi over Hitler. This phenomenon not only can happen, it has happened, many times, with IRV, including in the Burlington VT 2009 mayor election some posters cited: http://rangevoting.org/Burlington.html . Burlington's adoption of IRV was one of the early proclaimed "successes" of Richie's organization. But they got major pathologies in only its second election.
Richie mislead #3: "As with every voting method, there are cases where candidates might benefit from insincere voting, but with IRV such situations are too convoluted to be an effective tactic." (Offers zero evidence.) In fact, what has happened in every IRV country so far, is that, after an initial adjustment period, they ended up with solid 2-party domination in all IRV seats. The foremost example is Australia which elected zero third-party members to any central-govt IRV seat in both of the last two election cycles combined. Meanwhile with other voting systems -- including the non-IRV system Australia itself uses to elect their senate -- more than 2 parties can and do win. Why? Well, a plausible explanation is that with IRV, voters insincerely vote one major party top, the other bottom or nearly, and all the other candidates (usually about 7 run) in the middle. This happens over 90% of the time in Australia. But how can this behavior possibly be "sincere"? Obviously, it is not. It is strategic. And it is a mathematical fact that, with IRV and voters who behave that way, third-parties cannot win. So the facts on the ground are: strategic insincere voting with IRV is massive and has huge effects.
Richie mislead #4: "IRV has fair results and isn't prone to gaming" (both false, a least
in some situations) "explain[ing] its successful track record in thousands of hotly contested elections -- from nearly a century of elections for Australia's House of Representatives to decades of elections in Irish presidential races." Actually, in the vast, vast majority of the elections Richie just mentioned, the set of ballots was kept secret,making it impossible for anybody to tell who would have been elected under different, non-IRV, voting systems. In view of this, any claim IRV has had a "successful track record" is silly. In fact, it for the most part has no track record at all. In the rare cases when all the ballots have been divulged -- or enough of them -- disturbing facts often have surfaced.
Let me concentrate on the cases Richie cited as "success" -- Ireland and Australia. In Ireland, there is only one central govt seat contested by IRV -- the presidency. (And I am unaware of any seats in local sub-governments contested via IRV, but won't swear there are none.) Only one such IRV election in all of Irish history has ever elected somebody different than plain plurality voting would have: 1990. But that election exhibited disturbing pathologies. E.g, Lenihan lost. But if 344,362 new voters had magically appeared, all voting Lenihan WORST, *and* if 80,000 Lenihan-best voters had switched to voting Lenihan WORST -- the combined effect of those two changes, under IRV, would have made Lenihan WIN! Clearly, therefore, either IRV should have elected Lenihan in 1990, or IRV should NOT have elected Lenihan in the thus-altered fake election. Clearly, IRV was electing the wrong winner in at least one of these two elections. Details: http://rangevoting.org/Ireland1990.html
In Australia, I examined the latest (2007) IRV election cycle, comprising 150 elections. I found at least 9 elections exhibiting pathologies similar to the above Ireland & Burlington problems. Details: http://rangevoting.org/Aus07.html
Roann Cramer also falsely claimed "There is no opportunity for [IRV] strategic voting because the voter has to rank their selection on one ballot." And also falsely claimed IRV "elect[s] majority winners." Well, IRV SOMETIMES elects majority winners (such as, if 100% of the voters rank McCain top, he will win), but sometimes refuses. In Burlington 2009, as the cited webpage shows, Montroll was the "majority winner" beating Kiss by 590, Wright by 929, Simpson by 5676, and beating everybody else by even larger margins. But IRV did not elect him! The reason is IRV ignored a lot of Montroll>Kiss preferences. Also, the were plenty of ways the Burlington voters could have voted strategically to change the result in a manner they'd prefer, as that webpage also makes obvious. (IRV's supposed "justification" for electing Kiss was Kiss's 250-vote margin over Wright, which note is smaller than every Montroll margin.)
I recommend you support "range voting" or its simplified form "approval voting." Both are simpler that IRV, less prone to fraud and voter error, less prone to weird pathologies, and easier to count; in computer simulations range voting outperforms all common voting system proposals including IRV
for either honest or strategic voters. A comparison chart among several voting systems is here: http://rangevoting.org/CompChart.html .
Warren D. Smith (PhD)
We've had 2 IRV pilots in my state, Cary and Hendersonville NC in 2007, and just Hendersonville NC this November. Cary tried IRV, had to actually count the votes, and the candidate who won the contest decided by IRV, Don Frantz -led the charge to ditch IRV. Cary held public hearings early this year and ditched IRV. Hendersonville did not hold public hearings nor invite public input, and the city council chose IRV in order to avoid runoff elections. For them this is not about the voters but about their campaign finance.
I am a verified voting activist who opposes IRV because of its negative impact on election integrity as employed in my state of North Carolina.
The compromises North Carolina's IRV pilot forces are highly toxic to election transparency and to create barriers to vulnerable segments of the population. The current guidelines and procedures violate key values of election transparency and weaken the standards of the hard fought for and nationally acclaimed Public Confidence in Elections Act (paper ballot.election audits law) passed in 2005.
IRV is not "additive", so it increases reliance on more complex and bleeding edge technology and requires the central counting of votes. Central counting means hauling ballots away from the polling place to be counted elsewhere at a later time. This is in direct conflict with North Carolina statute § 163-182.2. and opens elections up to the risks of ballot box stuffing or tampering.
The procedures to tabulate IRV in touch-screen jurisdictions cut corners on election transparency. Since there is no federally certified software to tabulate IRV votes, the State Board of Elections has devised a work around. This "workaround" employs a spreadsheet using a 5 page single spaced algorithm to tabulate the votes. The NC Coalition for Verified Voting argues that this uncertified "workaround" is an encroachment on the hard fought for and nationally acclaimed standards of SL 323, The Public Confidence in Elections Law Certified voting systems are required per § 163-165.7.(2) "That the voting system comply with all federal requirements for voting systems." These standards were put into place to protect our elections from the types of failures we suffered in the past.IRV is even more complicated when using touch-screen voting machines.
Hendersonville voters will once again have the disconcerting experience of casting an IRV ballot on touch-screen voting machines. With touchscreens, the voters will see a field of choices, make their selection, and then go to the next page. On the next page they will see the same candidates names - again. Voters may experience confusion wondering if they made a mistake in paging through the ballot.
Doing manual recounts or audits of complex IRV ballots on the long paper trail rolls would be difficult if not impossible, since these printouts do not have a ballot summary. Until the touchscreens print a simple voter verified ballot summary, IRV shouldn't even be considered, or Hendersonville should agree to use paper or optical scan ballots instead. Unfortunately, it seems as if the purpose of IRV IS to unravel the high standards set by the Public Confidence in Elections Act.
Berkely Professor Philip B Stark strongly advises against the use of a spreadsheet to tally IRV saying that "Spreadsheets mix data and programming. It is not possible to tell at a glance whether a cell in a spreadsheet is data or the result of a calculation. As a result, it is quite easy--deliberately or inadvertently--to corrupt a calculation or the data on which it is based. In principle that can be detected, but it requires additional scrutiny--such as clicking each cell and looking at what is displayed. And even that is not foolproof.
"Tom Dahlberg of Dahlberg Business Logic Inc. (his business IS spreadsheets) warns: "How can the state prove, to those who have standing (all voters) consistent with the compelling state interest, that the automation is working properly and not committing fraud? And who has the burden of proof if not the election officials responsible for the integrity of the process?"
The counting of IRV is complex -- the elimination of some candidates at the end of the first round means that second choice votes are transferred to other candidates. If a third round is required the elimination and transfer process continues. The average voter has to place great trust in the reliability of the counting algorithm in a way far beyond what is necessary in plurality voting. So the counting is opaque and non-transparent -- a kind of voting voodoo with election officials in the role of witch doctor producing the magical results. If one believes strongly that the average voter should be able to understand and observe the counting of votes in a democracy, then IRV fails to meet this standard.
Using IRV in US style elections is like trying to put a square tire on a car. There's a reason that IRV has been around for about 100 years yet so few places have adopted it and many have abandoned it.
Voter rights that IRV removes as compared to plurality voting:
1. the right to know the effect of one's vote (i.e. the right to cast a vote that has a positive effect on the candidate(s) of one's choosing)
2. the right to participate (cast a vote) in the final counting round (in IRV whenever there are more candidates than ballot positions many voters' ballots are exhausted prior to the final counting rounds and many voters are excluded from making the final decisions.)
3. the right to change one's mind between rounds or between elections
(eg. in general/top-two runoff elections more often the top vote getter of the general election is not the same winner as the final winner than with IRV)
4. when many candidates are running and time is limited, the right to get to know more about the candidates that a primary/general election or a general/top-two runoff election affords
5. the right to equal treatment of the voter's vote with all other voters' votes
6. the right to verify the integrity and accuracy of the vote counts in a reasonably easy manner
7. the right to have economical, transparent elections that are counted locally rather than centrally.
I don't know about you, but I prefer to keep a voting system that preserves my rights as a voter, or to switch to any one of the alternative mentioned by the experts who've posted here that preserves my rights as a voter rather than to switch to a system (IRV) that removes my rights as a voter and does not solve any of the problems of the existing plurality system but introduces new problems.
It is nice to see some true election methods experts responding to the typical misleading and false claims that Rob Richie spreads. Does anyone really buy Richie's argument that "all voting systems are flawed".... [so therefore adopting the worst of all possible voting methods is a good choice]?
You can know you're getting somewhere with your reform when you get over-the-top resistance from people who prefer the status quo.
I would prefer that the title MPR gave to this piece was different, as it makes me seem like just another Donald Saari-Steve Brams-Warren Smith-type zealot for their preferred ssytem. I of course know that every system has flaws, including IRV. But IRV's flaws are the least objectionable to me because they give no advantage to one set of voters over another even as IRV does a great job at accommodating increased voter choice. (Australians just had a national election in 2007 with an average of seven candidates contenting each seat in its house of representatives -- and every seat was won with a majority.)
That's not true of range voting, approval voting and Borda. All the heated arguments for such systems melt away when they violate the later-no-harm criteria. The first time in a range voting election with two candidates that better strategic voters on one side lead to a candidate with 45% first choice support defeating a candidate with 55% first choice support, it's gone -- and because of that obvious flaw will never be won in the first place. (The fact that we're debating real elections with IRV vs. theory with these alternatives is telling -- they are very hard to win, and their limited experience, such as the Dartmouth College alumni board elections with approval voting that led this year to its repeal by a vote of 82% of of alumni, is not promising.)
I wish advocates of other systems would try to get their systems passed somewhere for a meaningful election. Anywhere. For anything. But that's not been their focus, perhaps because it's a lot easier to bash other reform ideas than win your own.
Jocye McCloy and Kathy Dopp are masters of FUD (Fear Uncertainy and Doubt) who are fine with the status quo, non-majority winners and spoiled elections. Check out their claims as much as they suggest you check out ours. As one example, Cary (NC) didn't "ditch" IRV. It chose not to do the pilot again in 2009, but it may well do so in 2011, particularly after one Cary race looks like it will go to a whole second-round runoff after the frontrunner won 49.97% of votes in the first round. Jurisdictions like Cary are more likely to go to IRV when it's easier to get the count done -- and that's coming.
Meanwhile, the British prime minister is pledging a national referendum on IRV. The Oscars will use it for best picture. The League of Women Voters keeps supporting it after state studies. We're not at 53 colleges and universities using it for student elections. Big cities like Oakland and Memphis will use IRV in upcoming election. I'm confident that we're on the right side of history.
Mr. Richie, why do you label anyone who disagrees with you as a "zealot" or "over-the-top"?
Setting aside name-calling to focus on logic and reason, I find your arguments for IRV unpersuasive and Warren Smith et al.'s arguments against it (and in favor of range/score voting) to be highly persuasive.
I share your opposition to plurality voting and your desire to reform the system to be more fair. I too want to give voters the freedom to vote for their favorite candidate without being afraid of actually helping their least favorite candidate to win as a result of that honest vote. That, after all, would seem to be absolutely necessary in order for democracy to work properly. However, I do not agree with you that switching to IRV is the way to achieve this goal.
In the main article, you seem to imply that IRV satisfies this objective because in IRV "ranking additional choices never hurts your first choice." (This is the so-called Later-No-Harm Criterion). Your statement may be true, but it does not actually address the issue. In fact, in IRV, ranking your first choice 1st can cause your last choice to win when he otherwise would have lost if you didn"t vote at all! Thus IRV is no better than plurality in this regard, and is perhaps worse because the reasons why it can happen are confusing and hard to explain, not to mention that IRV supporters such as yourself seem to believe that it cannot happen and spend all your time convincing others to believe the same!
Here"s an example election in which a Democrat would have beaten a Republican, except that 300 Green Party supporters, who didn"t even rank the Republican candidate because they hated him so much, end up causing the Republican to win simply by showing up at the polls to rank the Green 1st and the Democrat 2nd!
http://www.youtube.com/watch?v=fOwDyGCaOFM
And here"s another example where adding two ballots that rank a candidate last end up causing her to win the election:
http://www.youtube.com/watch?v=8Z3zi6hQ9OI
If there is a flaw in the arithmetic in these two examples I urge you to correct it. Otherwise, I challenge you to explain how IRV can possibly be considered "fair" or even sensical, in light of these examples.
Dear Rob Richie,
Perhaps an example from Kenya, the country of President Obama's father could advance the case for IRE by RCV.
Problem Statement:Kenya from 1992 to 2009
a)Under FPTP, President Moi was declared the outright winner with a mere 36% support of the electorate in 1992. (More=36%, Matiba= 26%, Kibaki=20% and Odinga= 18%)
b)Governing against the will of 64% of the electorate was very problematic. Much effort was spent in political survival games instead of socio-economic management. By 2002, economic growth had dropped to than less than 0.5%
c)From 2003, Kibaki government restructured the economic sector and achieved a 7% plus economic growth by 2007. However, Kibaki had won the 2002 election on false election MoU to Make Raila Odinga Prime Minister.
d)The false MoU exacerbated inter - tribal mutual mistrust leading to an inconclusive / polarized / disputable election and hence the 2007 post- election crisis.
e)October 18, 2009 opinion polls by Steadman put Raila Odinga as the leading presidential candidate with 31% ahead of Hon Kalonzo who recorded a mere 14%.
This statistic sends great danger signals to the more "electoral democracy" discerning minds. It could as well mean that 69% of the electorate does not necessarily support the leading candidate!
The political pendulum will swing back, full circle, to 1992!
The solution to Kenya's Political Problems lie in IRE by RCV. Interestingly, had IRE been used in 1992, Kibaki would have easily won the election honestly and restored Kenya's social-economic beacon of hope for Africa ten years ealier.
How I wish that Rob Richie from Takoma Park, MD, could take an interest in Kenya's struggle to achieve electoral reform?
May America Be Blessed Abundantly.May Kenya Be Blessed Through America.
Amen
Nyongesa,Chris Makhanu
cmnyongesah@gmail.com
Any way of voting and counting is right.By the way,the best leaders are never electable.look at the leader of India.Dont worry,be happy
Approval, IRV, and Condorcet-IRV are all excellent methods.
IRV stands out as the method that allows stratgegy-free choice to a mutual majority (MM).
A MM is a set of voters who all prefer the same set of candidates to all of the other candidates.
IRV guarantees that if the members of a MM vote sincerely, then the winner will come from that MM's preferred set. And, with their sincere rankings, that MM have free, sincere choice among that preferred set That gurarantee, that freedom from strategy is unmatched by any other voting system.
Yes, it's only available to a MM. Voters not in a MM can have favorite-burial need.
Whether that's a problem depends on the electorate and the media.
The Greens (GPUS) offer IRV in their platform. What if the Greens were elected, to the presidency and most of Congress? IRV would be the voting system. And any electorate able to accomplish that, with Plurality, would be fully competent to use IRV. Additionally, the media in a Green America would be far more open, free, and particpatory.
Then, IRV's favorite-burial need for non-MM voters woudn't be a problem.
But that problem, a result of IRV's instability when eliminating Condorcet winners (CWs), is avoided by Condorcet-IRV:
X beats Y if more ballots rank X over Y than vice-versa.
If there are one or more unbeaten candidates, do IRV among them. Otherwise just do IRV.
Result: IRV's big advantage, without the instability.
So, shall we start by electing the Greens?
Post a comment
Please be civil, brief and relevant.
E-mail addresses are never displayed but they are required to confirm your comments. All comments are moderated. MPR reserves the right to edit any comments on this site and to read them on the air with attribution. Please read the Terms and Conditions before posting.
You must be 13 or over to submit information to Minnesota Public Radio. The information entered into this form will not be used to send unsolicited e-mail and will not be sold to a third party. For more information see Terms and Conditions and Privacy Policy.
