Correspondence with THE NATION on Instant Runoff Voting and Electoral Reform

The Nation Magazine ran two pieces on these topics in late 2006, one by "bass drummer" Krist Novoselic, the other by editor Katrina van den Heuvel. (Actually my musician friends inform me there is no such thing as a bass drummer, but that's how they'd billed him. He used to play for Nirvana.) Both contained false statements. I was moved to respond, submitting the following letter. It was published (but with the colored words deleted without asking my permission) in the 25 Dec 2006 issue, page 1.

To the editor.

K.v.d.Heuvel is mainly correct in her suggestions to "bring democracy home." But her suggestions for "ending the duopoly" are wrong. Instant-Runoff Voting (IRV) will not do so: every one of the four countries that has used IRV for 5 or more years is 2-party dominated in IRV seats (exactly why the Australian third parties are trying to get rid of IRV). "Fusion" will not do so: it has been used in New York State since forever, and recently the "great white hope" Kevin Zeese ran for the Maryland Senate under the "revolutionary" new "fusion unity strategy" (Libertarian, Green, and Populist) collecting... 1% of the vote. KvdH's third suggestion (proportional representation) would indeed work, but its constitutionality is questionable to say the least, and if it is not constitutional, then it is politically impossible to get, even if the entire democratic party vote for it.

I wish, just once, somebody making these sort of suggestions would actually be an expert on voting systems instead of revisiting the same hoary old myths that have been dismissed by real political scientists decades ago. The Nation recently published a piece (full of untruths) on IRV by, of all things, a Bass Drummer.

I recommend the simpler and better "range voting" system: your vote is a score on an 0-9 scale for every candidate - highest average wins. It works on every USA voting machine right now (no reprogramming or modifications needed); it should genuinely break the duopoly; it eliminates the "spoiler" effect; voting is never worse for you than not voting; and raising your score for somebody cannot hurt them. IRV enjoys none of those properties.

Warren D. Smith
math PhD and co-founder of (this web site backs up the facts claimed here)

The editor Katrina van den Heuvel then wrote a response (same issue pages 22 and 36) to this and other letters. Here is the part of her response dismissing my letter above (I have numbered her sentences for easy reference):

  1. W.D. Smith is wrong to argue that instant-runoff voting will not end the duopoly.
  2. IRV gives third parties political space to breathe and build support and, yes, if major parties run flawed candidates or become unrepresentative, win.
  3. (Witness Mary Robinson's victory in the Irish presidential race in 1990 or Ken Livingstone's win in the first London mayoral race.)
  4. Smith is inaccurate in arguing that PR is constitutionally questionable.
  5. Many experts believe that range voting would create a strategic mess; there are good reasons it hasn't been adopted for any election of any kind.

We conclude with a long and detailed response to that:

RESPONSE by Warren D Smith (math PhD and co-founder of, two facts THE NATION deleted from my letter)

1. Nope. I repeat. Every IRV country became 2-party dominated in IRV seats. Australia is the country that holds by far the most IRV elections (its House of Representatives and statehouses; has used IRV for over 80 years). Here are the statistics: In all Australian federal, state and territorial legislative bodies that use IRV there are 564 seats. Only one of those seats is a third-party member (from the One Nation party). Where are the Greens, Democrats, etc? This is similar to the USA's statistics but more striking because Australian third parties actually are powerful because of the Senate (elected using PR) and because Australia is nonpresidential and comparatively ungerrymandered – but they still can't win IRV elections.

The country that used IRV the second-most, was Fiji (just overthrown by a coup; adopted IRV in 1997-8). Two-party domination then gradually set in. By the 2006 election the stats were as follows:

Of the 71 seats, 36+31 were won by the top 2 parties, and two were won by members of a minor party. Probably if this experiment had been allowed to continue the minors would soon have been zeroed out. In contrast: it took a lot longer than 9 years for the USA to settle into 2-party domination mode.

In both Fiji and Australia there also are independent seat-holders, most of whom (again like in the US: Lieberman and Jeffords) are major party candidates who split with their parties. These constitute 2-5%, thus outweighing the minor parties by a lot (again as in the US).

2. "space to breathe and build support"? What does that actually mean? I have actual evidence that range voting leads to tremendously higher vote counts relative to Bush & Kerry than our voting system. I am talking factors of over 50 higher for Nader & Cobb (Greens), factors of over 1000 higher for socialists. That isn't a vague statement about space to breathe. It is exit poll data.

3. Ken Livingstone would have been elected by a huge margin without IRV. Indeed, IRV actually decreased his ratio advantage over his nearest opponent Norris. Hence this example is (and intentionally whenever Rob Richie uses it, since I have many times pointed this out to him) misleading. It is a fake example. It says nothing about IRV electing more third party candidates. (Also, Livingstone was not a third-party candidate; he ran as an independent because he did not get the nomination from his major party, the Labour Party, a dispute which was healed by the next election, in which he ran Labour. The situation was exactly analogous to Lieberman's recent victory in Connecticut.)

Mary Robinson is indeed a valid example of IRV electing a third-party candidate. She is however an extreme exception. To learn about how incredibly fluky the circumstances were, see these: /Ireland2002.html, /FunnyElections.html#Robinson or read other sources (for example wikipedia). This apparently was the only time in all of Irish history that any IRV election has ever been won by anybody not from the FF party. I actually do regard this one as a success for IRV, but it is not to be overstated. The sole fluky example in the entirety of a country's history, plus one fake example, does not "ending the duopoly" make. The evidence that IRV leads to 2-party domination seems very massive, comparable to the evidence for the statement that the USA is 2-party dominated.

4. Here is the review of supreme court cases on which my "PR's constitutionality is questionable" opinion had been based. The supreme court has never explicitly ruled on the question of whether PR would be constitutional for federal elections (not surprisingly since it has never been used) but the US congress outlawed PR for federal elections in 1967 (law still stands) and the supremes have issued a very large number of rulings, very often overruling themselves, on related issues, which can be interpreted as both pro-PR and anti-PR. In view of this I consider the constitutionality of PR in federal elections to be risky. If you can get a legal expert to improve my case review, please go ahead, provided the new improved version is made available to me for posting on the website. I myself am not a lawyer and am just looking at all those supreme court cases as a layman. My personal guess is the chances are 60-40 that PR is constitutional, and the chances are over 99% that PR will never be made legal in the USA for federal elections during any time when either the plurality or IRV voting systems are employed (because there will be 2-party domination and the 2 major parties will block PR).

5. What does "a strategic mess" mean exactly? It seems the idea was to accuse range voting of something so vague that it must be true. (Suppose I were to say "many experts agree Katrina vdH is a mess." Well... it would be approximately equivalently non-meaningful and probably actually more true. Basically, this was a matter of finding a dismissive adjective rather than making a case.) For two simple theorems indicating range voting exhibits very mild – most would even say "very pleasant" – behavior in the presence of strategic voters, see /AppCW.html and /PleasantSurprise.html. Nothing comparable to these theorems is available in the literature about any non-range voting method whatever. If you disagree, then I challenge you. Go ahead. Have your "experts" find a single theorem, anywhere in the literature, proving the good behavior [in fact, these two theorems prove optimal behavior in two metrics] of a non-range voting method in the presence of strategic voters, which is in any way comparable to these results.

Now let's compare with IRV. Suppose every voter, to avoid the IRV-spoiler pathology (explained via election examples in /TarrIrv.html and /IRVpartic.html etc, also shown to happen about 20% of the time in computer simulations of 3-candidate IRV elections), ranks Gore top, Bush Bottom, and Nader middle (or the reverse). In that case it is a trivial theorem that IRV will always return the same winner as our current plurality voting system (with its voters also trying to avoid spoiler by voting Gore or Bush). I.e: same voter strategic behaviors imply same election winners in both IRV and plurality voting elections. This theorem is a large part of the reason IRV in practice behaves just like plurality voting and elects 2-party dominated governments every time it has ever been used in any country. (Unlike, say, the 2-round genuine runoff system, which has been used in 27 countries I currently know of and has avoided 2-party domination in 21-23 of them. Some systems encourage duopoly, others discourage it. IRV is in the former class. Delayed runoff – which is only subtly different – is in the latter. These facts are not under dispute by political scientists; see e.g,

Rein Taagepera & Matthew S. Shugart: Seats and votes: the effects and determinants of electoral systems, Yale University Press, 1989. Has similar Duverger-confirming datasets to mine, including as pictures rather than just tables. E.g. see figure 8.1 p.84 and 8.3 p.87; note the latter figure lumps in IRV voting with plurality as additional "two-party-genic" system, agreeing with my view IRV and plurality both obey Duverger's law. [And Duverger himself also stated that delayed runoff (as in France) led to more than 2 parties, as part of his law of 2-party domination.]

Incidentally, Matthew Shugart is a member of's advisory board and he still makes this statement in his book, so I think you can trust this...

Albeit, several well known academics on that advisory board have told me that they have in a dozen years never gone to a meeting, never been asked for advice, nothing, in all their years "on" that board.

Let's also mention the Gibbard-Satterthwaite theorem, which shows that voter dishonesty is strategic in every system which (like IRV) is based on rank-order votes, in 3-candidate elections. This is one of the most important theorems in voting theory. I personally consider it the most important. If your experts do not name this theorem instantly in response to "what are the top 3 most important theorems in voting theory?" then they are not experts. Rob Richie, in the article he co-wrote in Science Magazine 294 (May 2001) 303-306, made the false statement "Used for decades in Australia and Ireland and considered in 13 state legislatures this year, IRV lets voters rank candidates in preference order. A voter's best strategy is to sincerely rank the candidates." This false statement as you can see, directly contradicted the Gibbard-Satterthwaite theorem, which had been proven about 23 years previously and acclaimed universally thereafter. It also is simply false, as you can easily see from any of numerous simple counterexample elections on the web site such as /IrvRevFail.html. This example (Rob Richie, simply flat-out lying about the most important theorem in all of voting theory in one of the most circulated science magazines on the planet) is one reason he has gained a bad reputation throughout academia. (Richie, incidentally, has a BA in philosophy, which is not exactly expertise. This statement was immediately smacked down in Science Magazine by return mail, but he managed to get it printed and has thereafter often referred to this, his one apparently-scientific publication, without mentioning this massive error.)

Anyway, the Gibbard-Satterthwaite theorem does not apply to range voting and its degenerate form "approval voting" because they do not use rank-order ballots. In 3-candidate range or approval elections, casting a vote in which you say A>B when actually you honestly feel B>A, is never strategically good. Never. By a simple theorem.

But for every rank-order vote system (such as IRV), there is a 3-candidate election where such dishonesty is the only strategic vote. Every system. By the most famous theorem in all of voting theory. So range and approval are, in this sense, provably better about strategy than every rank-order voting system. So have your "experts" consider that.

We've now named three ways range voting is provably, by theorem, very well behaved when it comes to strategy. (For another interesting strategy comparison, see puzzle 40.) Is that a "strategic mess"? Or would some other wording have been more accurate?

Please. I beg of you. Stop relying on bass drummers and ultra-biased BAs in philosophy instead of math PhDs for your information about voting theory. I do not understand The Nation's preference for the former.

I offer to write a piece for The Nation on this topic, you name the desired number of words. Consider the number of words you employed just to write your letter-reply on pages 22+36. Or the number your bass drummer used. Unlike you and him, I am an expert on the topic. It is therefore plausible that if I were given that number of words, I would do a better job. It is also plausible I actually deserve more words.

This is important. It is not commonly appreciated by laymen how important, and the Democratic, Republican and Third parties in the US mainly do not understand how much it would help them, but you can read some quantitative analyses/estimates of how important and how much, here:
/LivesSaved.html, /WorldProblems.html, /BaileyNum.html, /BayRegDum.html, /UniqBest.html, /IowaSumm.html, /ForReps.html, /ForDems.html, /ForGreens.html.

Finally, re your "there are good reasons [range voting] hasn't been adopted for any election of any kind." That seems a bit overstated. Approval voting, the maximally-simplified form of range voting with scores 0 and 1 only, is used in many professional societies for their elections, to elect the secretary general of the UN, and it has been employed in elections in Eastern Europe and the Soviet Union during the late 1980s, in (what should be regarded as) an early experiment in democratization. [Theodore Shabad: Soviet to Begin Multi-Candidate Election Experiment in June, New York Times, 15 April 1987, p. A6.] Also, (full) range voting has enjoyed extensive use to select valedictorians, Olympic gold medalists, etc, and hundreds of trillions of range voting elections have been conducted by social insects (honeybees & ants) to make hive relocation decisions. (Oddly enough, Darwinian evolution apparently never invented IRV... there are "good reasons" for that too.) Oh, and there's Sparta and Renaissance Venice. While I suppose each of these examples could be dismissed as not a "real" election, not a country you approve of, or not "really" range voting, I think they are enough to refute the overstrong phrasing "any election of any kind."

Sincerely, Warren D. Smith PhD
(I do not often use my PhD title, but it seemed relevant in this case and required only 3 letters, so I'm surprised to see The Nation just silently deleted it, while for example, discussing Krist Novoselic's bass drummer qualifications in a considerably larger amount of space.)

Return to main page