Dartmouth alumni oust approval voting 82% to 18%
That said, the zealous behavior of some advocates of approval voting and its companions system range voting (or "score voting") justifies explaining why we focus on instant runoff voting for reforming elections for a single-winner offices. Today the Dartmouth Alumni Association announced that Dartmouth alumni had voted by a four-to-one margin -- 82% to 18% -- in favor of an amendment to its constitution to replace approval voting with traditional runoffs in elections to the powerful Dartmouth College Board of Trustees. This overwhelming rejection of approval voting after a series of controversial hotly contested elections with the system provides an opportunity to remind people why we focus our research, communications and advocacy on instant runoff voting for reforming single-winner offices.
For newcomers to this topic, approval voting is a method of voting where you cast one vote for as many people as you want. The candidate with the most votes wins. It's a simple system to describe and can be implemented on some of the voting equipment currently used in the United States.
When you probe deeper, however, you find that it can generate odd results. For example:
* Approval voting basically throws out our current understanding of majority rule. Suppose in a three-person race, our current plurality system would have a result that provides for a relatively comfortable, majority win for one candidate:: 52% for Janet Garcia, 40% for Bill Jones and 8% for Robin Hayes. But in approval voting, the results might be: 54% for Garcia, 56% for Jones and 18% for Hayes. That is, the same set of voters might now elect a candidate who would have lost 52% to 40% in a "vote for one" election.
* Approval voting rewards easy-to-anticipate tactical voting. In our example above, Jones defeated Garcia with approval voting. One reason that he might have succeeded with approval voting is that a number of his backers realized that if they approved of Garcia as well as Jones, that would hurt Jones. So they instead cast a "bullet vote." However, too many Garcia voters "didn't get the memo" -- they sincerely said, "well, I kind of like Jones too" and gave him an approval vote. By doing so, they lost out. Bottom-line: the insiders will be trying "to get the memo" to their backers to bullet vote while outwardly pretending to be inclusive in order to draw approval votes from backers of other candidates. (In comparison, with instant runoff voting the Garcia voters could rank Jones second without fear of that second choice coming back to hurt Garcia --it pays off to be sincerely inclusive with instant runoff voting.)
As it turns out, tactical voting is exactly what was happening in the Dartmouth alumni elections: some voters were gaming the system better than others. The Dartmouth newspaper editorialized last month that:
"As they are currently run, trustee elections can give an unfair advantage to candidates elected by petition, who have traditionally been supported by a vocal alumni minority. When the alumni electorate fails to take advantage of the approval voting process, the three required Alumni Council candidates tend to split the majority vote, giving petition candidates an advantage. By reducing the number of Alumni Council candidates, and instituting a more traditional one-person, one-vote system, trustee elections will become more democratic - and will more accurately reflect the desires of our alumni base."
Approval voting's record is not promising in the limited number of contested elections where it has been used. With the Dartmouth repeal, I'm not sure where it's used in elections of import that have hotly contested races with more than two candidates. The largest elections using approval voting were those held by the Institute of Electrical and Electronics Engineers, but the IEEE dropped it as well after a few elections.
Meanwhile, instant runoff voting is recommended by Robert's Rules of Order for elections by mail and has been used for literally thousands of hotly contested governmental elections and thousands more hotly contested private organization elections -- including at least 51 American colleges and universities and dozens of significant organizations for contested elections. The record demonstrates that it does not lead to tactical voting and regularly generates majority winners -- and certainly would never fail to elect a candidate who 52% of voters saw as their first choice.
I would be fine with approval voting being tried out in more places, but would hope its advocates would accept that there are legitimate reasons to prefer instant runoff voting - and legitimate arguments for IRV being a better system.
Comments currently closed for Dartmouth alumni oust approval voting 82% to 18%
-
Posted by JEB, 2009-06-17 08:49:49 (4 years ago)
-
[...] voting continues to have a best mixed experience in real-life elections, being repealed for Dartmouth Alumnni trustee elections this year by a vote of 82% to 18% and being repealed by its other most significant user, the [...]
Posted by FairVote.blog » Blog Archive » Voting method debates go way back: Electing the Pope, 2009-06-06 12:32:39 (4 years ago) -
My issues w. the mathematical approaches that Clay appeals to are more epistemic. I don't think people have well defined rankings of candidates as most people operate with both low levels of information and often vote using deeply imperfect rules of thumb and candidates have been known to "flip-flop" following elections. This is why I hold to the mantra, "more choices for more local elections", as I believe that it'd be easier for us to become better informed about all the options in local elections if they were to become competitive thanks to election reform. I also think that even if the majority of political scientists and mathematicians prefer AV over RCV to prevent strategic voting, that doesn't mean that they prefer AV to RVC in the overall sense. The potential complications of strategic voting are one facet of elections or "tending our own democracies". I'd say it's pretty much an issue that's hard to settle, since we don't observe voter preferences over candidates, and strategic voting by groups isn't per se bad, if it were to move the major candidates closer to the de facto political center. I also think that, with respect to the fact that we are pretty much stuck with using a first-past-the-post system in bringing about election reforms in the US, the optimal strategy is to go our separate ways in advocating the use of IRV, AV or what-not in more local elections, and if another group gets momentum, we should lend them support or vote strategically for them over the status quo. Then, we'll have a larger sample size for to make comparisons between the different approaches, as to their strengths and weaknesses in practice, and be able to use such information to continue to tend our own democracies! dlw
Posted by dlw, 2009-05-29 12:57:42 (4 years ago) -
Maybe the answer is to mandate the use of either Approval or IRV to select alternative voting arrangements? So long as everyone gets one vote, it'll be too easy to divide and conquer different approaches to election reform. I myself believe that there is no one system for elections and it depends on the type of election and that changing people's voting habits is just as if not more important than changing the election system. But I agree w. Rich that IRV seems to be a little more structured for single-member elections and voters used to fpp elections.... dlw
Posted by dlw, 2009-05-28 17:31:56 (4 years ago) -
I'm curious as to why you would think that "the same set of voters might now elect a candidate who would have lost in a plurality election" is a valid criticism of approval voting. This is exactly the same reason why you *support* IRV!
Posted by PB, 2009-05-27 03:00:24 (4 years ago) -
Jack S., The data isn't "made up". It is *calculated*. And the source code to the program that does the calculations is freely available on Warren Smith's web site. That's all clearly explain at the following link: http://scorevoting.net/BayRegDum.html Yes, Warren D. Smith is an advocate of score voting. So there are two primary ways of looking at this scenario: 1) Smith wrote the program in an objective unbiased way, and was so impressed with the performance of score voting that he became an advocate for the system. 2) Smith became an advocate for the system, and then wrote the computer program with some bias for score voting, with the intent of making score voting look good rather than getting an objective result. Smith maintains the former is the case. By his account, the computer program was written with the expectation that score voting would do well in some scenarios (e.g. lots of honest voters) but poorly in others (e.g. lots of strategic voters). He claims he was actually surprised when score voting excelled in ALL 720 combinations of settings for the 5 basic "knobs" (one such "knob" allows you to tune the ratio of strategic-to-sincere voters from 0% to 100%, for instance). I would encourage you to read the above link and get a sense of how the calculations actually work, and how the utility distributions are actually derived. Give the source code a look (or have a programmer friend do it for you). You'll find the code to be correct and not biased.
Posted by Clay Shentrup, 2009-05-15 00:35:19 (4 years ago) -
In other words, made-up data are better than real-world data because the real-world data are allegedly impossible to get... ...especially when the guy making up the data is an advocate of the voting system he's supposed to "study."
Posted by Jack S., 2009-05-14 08:56:54 (4 years ago) -
Jack S., Your recent comment links to pages at ScoreVoting.net that rebut you as well as I could here. For instance, the second link (WhyNoHumans.html) explains precisely why Bayesian regret simulations are superior to empirical data. (Simply put, because you can't read real voters' minds.) This also addresses the bizarre comment by Rob Richie that he has proof of his position from "thousands upon thousands of IRV elections". How exactly could he tell a sincere IRV ballot from an insincere one, aside from some magical voter-mind-reading device? Obviously he can't. But he has this habit of making claims like this, as if they are facts. As for Tideman, you also cite a page at ScoreVoting.net which details some of his massive flaws in measuring vulnerability to strategy. And in any case, Tideman also calls IRV "unsupportable". My experience is that the vast majority of credentialed mathematicians who study voting methods support my general perspective on the relative merits of IRV and approval voting. That should come as no surprise, since this is just based on objective and provable facts. I admit I do not have a scientific survey on this. Here is a page that touches on this topic of expert consensus, in a somewhat more objective way: http://scorevoting.net/Consensus.html
Posted by Clay Shentrup, 2009-05-14 01:34:24 (4 years ago) -
Clay, you have no evidential basis for making the claim that "virtually all election methods experts" prefer approval voting to IRV for the strategic incentives it supplies. If we add Nicolaus Tideman to your supposedly representative sample of "election experts" with "math/politics degrees," only 75% prefer approval to IRV. According to the Center for Range Voting, Tideman isn't a valid "election methods expert" because his opinion is supposedly flawed. Moreover, we need to distinguish the "math" research approach from the "politics" one. Choice-theoretic formal modeling is not the same as empirical analysis of institutions of political representation. In his so-called evaluation of voting methods, Warren Smith uses formal models built on computer parameterization of Bayesian regret because quasi-experimental research designs using actual voters, measuring interesting dependent variables like democratic stability or the content of public policy, would not lead to politically useful findings. While I am sympathetic to the use of game theory in institutional analysis, game-theoretic findings hold a lot more water when supplemented by consonant findings from the real world. Being a math expert does not make one an elections expert. More to the point, "virtually all election methods experts" do not in fact prefer approval (or range). You say Steve Brams. I say Donald Horowitz. You say Robert Norman. I say Benjamin Reilly. You say Warren Smith. I say Andy Reynolds. STV and IRV have plenty of advocates who are nonetheless cognizant of STV/IRV's real-world limitations.
Posted by Jack S., 2009-05-13 09:39:05 (4 years ago) -
Clay is a vigorous champion of his point of view, but I reject his claims of comparable "failure" of majority rule when approval voting unelects a candidate who wins more than 50% of first choices as compared to instant runoff voting not electing someone who is in third in a three-person race. I also reject that there is anything close being comparable in the insincere behavior of voters in this system -- we have proof of that in thousands upon thousands of IRV elections, not mathematical models. As to Clay's reply on academic support, it is inaccurate. More political scientists support instant runoff voting (and proportional representation, something Clay is skeptical of as well) than approval voting. That's why the association of political scientists installed it for its elections. You can see a partial list of people who gave us permission to use their name associated with this statement here: http://www.fairvote.org/irv/endorsers.htm . Some math-oriented types like approval voting, but I suspect their numbers will decline as they confront the harsh reality of how it has worked out in the limited places it's been tried and rebuffs like the 82%-18% rejection of Dartmouth alumni
Posted by Rob Richie, 2009-05-13 08:36:23 (4 years ago) -
Jack, I said virtually all election methods experts (specifically those with math/politics degrees) understand approval voting to be superior to IRV with regard to strategic voting (and I'll extend that to just being generally superior in all regards, especially tabulation simplicity). This is pretty much common knowledge if you just go around talking to math experts with expertise in this subject. Look up Steve Brams, Robert Z. Norman, Warren Smith, etc. (Also the Bayesian regret figures out there just objectively show this to be the case, so you can be assured than most any math-savvy individual would acknowledge it.)
Posted by Clay Shentrup, 2009-05-13 01:53:56 (4 years ago) -
Rob, You said: Your definition of throwing away majority rule is different than mine. I?m talking about a situation where more than 50% of people prefer someone as a first choice ? but that candidate loses. I reply: Our definitions are actually identical, according to a logically forced principle called Independence of Irrelevant Alternatives. Readers can look that up on a site like Wikipedia, but I'll breiefly describe via example. Say all the candidates but Montroll and Kiss (the Progressive, and actual winner) were to have dropped out prior to the election, with no change in voter preferences. Then your criterion would say that Montroll must win, being preferred to Kiss by an outright majority. And the presence or absence of other challengers, like Republican Kurt Wright, cannot change who is socially preferred between Montroll and Kiss without changing the voters' opinions about Montroll and Kiss (this is logically forced, and simple to see for anyone with even a modest background in mathematics/logic). So to say that we have a "different understanding" is to admit a lack of expertise on one of the most basic facets of election theory. You said: it is __very__ hard to anticipate results in IRV in such a way that it affects casting your true preferences. I reply: Polls on head-to-head match-ups are actually quite common in elections with more than 2 significant candidates. You can actually do a Google search and easily find such match-ups from our recent Presidential election. There were, for instance, polls which attempted to deduce whether Obama or Clinton would do better against their Republican opponent. In the Burlington IRV election, it's the same situation. If were a voter who preferred Wright to Montroll to Kiss, and you saw evidence that Kiss would beat Wright head-to-head, but lose to Montroll, then it would only make sense for you to insincerely rank Montroll as your first choice. That way you work to upgrade from your third choice to your second, since you know your first probably cannot win anyway. In fact in Burlington there were enough voters who voted this way, that they could have used that exact strategy. And I presume they will be more likely to do that in subsequent elections, once they have figured out this simple strategy or gaming IRV. You said: with approval voting, it?s TRANSPARENT how you should vote tactically, any any campaign worth its salt will be telling its insider supporters to bullet vote while telling the outside world to vote inclusively. Well of course the CAMPAIGN will say this, but that doesn't mean it's strategically wise for voters to DO it. Look at Ralph Nader back in 2000. He was telling his supporters to vote for him, not Gore for example. But NES polling data shows that about 90% of the voters who claimed to support Nader actually voted for someone else. The moral here is that bullet voting is often NOT a good strategy, and voters are *demonstrably* aware of that, and not so blind as to just do what politicians tell them. > Any mathematical model telling you otherwise is simply inaccurate. The voting simulations I've referenced aren't telling us how voters will behave. Instead, we use plausible (and widely varied) assumptions, and then see how voting methods handle those behaviors. While there is plenty of randomness to account for the voters who will UN-strategically bullet vote, those simulations primarily focus on more realistic (ACTUALLY-strategic) behavior. The result is, as I said, that approval voting is so far superior to IRV that it performs about as well with 100% STRATEGIC voters as IRV does with 100% honest voters. Please do not try to confuse readers with misrepresentations of how these models actually work. And please, as someone who actually runs an election reform organization, take some time to research real-world voting behavior, rather than assuming that voters will behave like blind sheep. Lastly, please do not equate my attempts to correct your numerous factual inaccuracies with "defending the status quo". I agree with you that plurality voting is a very poor voting method, "unsafe at any speed".
Posted by Clay Shentrup, 2009-05-13 01:49:41 (4 years ago) -
On the basis of what evidence, Clay, can you say that "virtually all" political science and mathematics PhDs consider approval to be more strategy-resistant than IRV?
Posted by Jack S., 2009-05-12 09:21:40 (4 years ago) -
Going in reverse order... #3 (Dylan): I'm not suggesting the change of result from plurality is a problem. I'm suggesting it's a problem when an absolute majority of voters want a candidate as the winner, and that candidate loses. #2 (Peter): 12,668 voters. #3 (Clay): Your definition of throwing away majority rule is different than mine. I'm talking about a situation where more than 50% of people prefer someone as a first choice -- but that candidate loses. You're talking about a situation where a candidate in third when the field was reduced to three doesn't win. Note that an Andy Montroll candidate doesn't win in any system used in any governmental election in the world -- "sequential elimination " IRV gives that candidate the best chance possible, yet in your general role of defending the status quo, you criticize IRV far more than your criticize the traditional voting methods. Indeed you express support for traditional runoffs over IRV, even though they are more likely to result in defeat of the Condorcet candidate than with IRV. I'm guessing you'll eagerly support those in Burlington wanting to replace IRV with two-round runoffs with a 40% threshold just because it hurts IRV even though that system will keep transparent spoiler dynamics and allow first round winners opposed by 60% of voters. On tactical voting: it is __very__ hard to anticipate results in IRV in such a way that it affects casting your true preferences. (Note that it's much easier to do so with traditional runoffs, because you can change your vote between rounds, yet you support runoffs over IRV.) But with approval voting, it's TRANSPARENT how you should vote tactically, any any campaign worth its salt will be telling its insider supporters to bullet vote while telling the outside world to vote inclusively. Any mathematical model telling you otherwise is simply inaccurate. It's hard to explain away more than 12,000 Dartmouth alumni with direct experience with approval voting voting 82% against it. Professor Norman might try to do so and I have no problems with him standinroudly stand with the 18% who agree with him.
Posted by Rob Richie, 2009-05-11 07:05:23 (4 years ago) -
I'm curious as to why you would think that "the same set of voters might now elect a candidate who would have lost in a plurality election" is a valid criticism of approval voting. This is exactly the same reason why you *support* IRV!
Posted by Dylan, 2009-05-11 01:49:10 (4 years ago) -
82%-18%, wow. So how many total votes are we talking about?
Posted by peter jackson, 2009-05-10 21:31:16 (4 years ago) -
It should be noted that virtually all election methods experts (academics with degrees in mathematics and/or political science) consider approval voting to be superior to Instant Runoff Voting when it comes to strategy-resistance. This includes Dartmouth professor emeritus of mathematics, Robert Z. Norman. He disagrees with Rob Richie's assessment, and believes that Dartmouth's decision instead had, fittingly enough, political motivations. Here's an assessment of this issue by a Princeton math Ph.D. who specializes in voting theory. http://scorevoting.net/DartmouthBack.html It should also be noted that IRV *also* throws away the notion of majority rule. In the recent mayoral election in Burlington, Vermont, Montroll (D) was not elected, even though he was preferred by a substantial majority to both the Republican and the Progressive. http://scorevoting.net/Burlington.html
Posted by Clay Shentrup, 2009-05-10 19:56:39 (4 years ago)
One might note that in this 82% to 18% vote, it was the supporters of petition candidates (the "outsiders" or "vocal minority") who favored the retention of approval voting, Power Line April 26, 2009; The Dartmouth May 11, 2009. JEB