Response to Klarreich's Science News article on voting systems

Klarreich's article was comparatively good. (Science News week of 2 Nov. 2002; Vol. 162,18 p.280; comments by various readers were appended to the online version of that article)

However, this article unfortunately did not mention RANGE VOTING, which is explained on the Center For Range Voting home page. We are going to go through Klarreich's article showing how, for every point she makes, range voting would have addressed that point very nicely.

Range voting: Basically your vote is a numerical rating of each candidate in an 0-9 range (or other such scale) and the candidate with the highest average score wins. Kind of like Olympics judging.

Let's now go through the Klarreich article and see how range voting would have done if she'd also discussed it.

1. Nader in FL 2000 – range voting also would have made Gore win, and range voting has the property that "favorite betrayal" is never strategically forced, i.e. there is always a best-strategy vote which ranks your honest favorite highest.

In contrast, in Plurality, Borda, and Instant Runoff, unfortunately favorite betrayal quite often can be the uniquely strategically best vote.

Hence honest Nader-favorers could easily find themselves faced, in any of those systems, with a "be dishonest or be an idiot" Sophie's choice.

2. Observe that RANGE VOTING actually accomplishes what Arrow said was impossible! One reason is because Arrow assumed that voting systems involved rank-order ballots – and range voting does not use them. So range voting (or for this purpose I prefer to call it "honest utility voting"...) jumps through a "loophole" in Arrow's theorem.

Range voting also jumps partially through the same loophole in the Gibbard-Satterthwaite impossibility theorem. The GS theorem says that in any voting system, there always exists a 3-candidate election scenario, in which strategically best voting is dishonest, i.e. voting A<B when you honestly feel A>B. Meanwhile in range voting, strategically best votes never rank A<B if your honest feeling is A>B (but the problem still occurs in 4-candidate range voting elections, albeit rarely – and even then, if you know everybody else's vote, you can always choose a strategically-best range vote which never ranks A<B in any case where your honest feeling is B<A).

3. Klarreich complained about Instant Runoff exhibiting "nonmonotonicity."
Well, Range Voting is monotonic!

4. Klarreich then discussed Borda and Saari:

Saari argues that the way to identify the best voting procedure is to consider which scenarios should result in ties. If three voters have what researchers call cyclic preferences – one prefers A-B-C, one B-C-A, one C-A-B – there should be a tie, he says. Likewise, if two voters have exactly opposite preferences – one prefers A-B-C, say, and the other, C-B-A – their votes should cancel. The only common voting procedure that would give a tie to both of these cases is the Borda count, which gives two points to a voter's top choice and one point to his second choice in a three-candidate election.

However, Saari is wrong. E.g. if the A>B>C voter prefers A more strongly over B than B over C, then that should outweigh the other voters' expressions if they do not feel strongly about any preference. Range voting is more expressive than Borda since it can express intensity of preference. Therefore, Range voting does better than Borda voting even in Saari's pet pro-Borda scenario. (And if all preferences are equally intense, RV will yield a tie just as Saari wanted.)


Like the instant runoff, the Borda count gives weight to a voter's entire preference ranking.

This Klarreich quote is actually deceptive in two ways. First, there are situations where instant runoff ignores a good chunk of your preference ranking and never even looks at it. Thus this CRV web page (click the link) gives an election example where IRV elects a winner D who would lose by approximately a 2:1 landslide versus the "losing" candidate C. The reason IRV does that is because it ignores and never examines the D versus C preferences of most of the voters.

Second, Borda can leave the voter kind of powerless. In FL 2000, a voter who preferred Gore over Bush would be strategically forced to vote Gore "top" and Bush "bottom" (in the 7-candidate election) since the honest choice of, say, Gore top, Bush 3rd, would be incredibly ineffectual and stupid, basically like having only 2/6 of a vote.

Simplify it to 3 candidates Gore, Bush, Nader – then voting anything other than Gore=top, Bush=bottom (or the reverse) is strategically unwise, like having half a vote. Now as soon as you've done that (i.e. not been stupid) the Borda voter has no choice about Nader – who by the Borda rules has to go in the middle slot, and the voter is unable to express any opinion about Nader. The fact that Borda then "takes account" of the voter's non-opinion about Nader is then not greatly reassuring!

6. Finally, Klarreich discusses "approval voting" which is actually a degenerate form of range voting. She did not mention that the (first) modern inventor of approval voting was Guy Ottewell. Brams is much more famous, but he (re)invented it considerably later. (Also Approval had been used in the Republic of Venice for over 500 years...) Anyhow, Ottewell has now endorsed Range Voting. Brams also told me privately that he is unsure which (Approval or Range) is the better system, since he can see advantages for both (e.g. Range is more expressive but Approval is simpler). Because Approval is a degenerate form of Range it is easy to switch from one to the other.

So we hope you can see by this point-by-point examination of the Klarreich article that with respect to every point there, range voting is as good or superior.

Further, the world's largest computer simulation study (#56 here) comparing voting systems had the result that by a statistical yardstick called "Bayesian regret", range voting is better than all the alternative systems Klarreich mentioned, plus a lot more, and this is a very clear, robust, and reproducible conclusion.

Now we discuss some of the points raised by the commenters to the Klarreich piece.

Bellman & Malone: Borda and instant runoff are too complicated.

My reply: Approval voting is actually simpler than plurality since it is like plurality but without having to detect (and discard as illegal) "overvotes." Range voting also is pretty simple – who has not rated somebody (of the opposite sex?) on an 0-10 scale? – and (although it may not be immediately obvious) range and approval voting both can be run on every voting machine in the USA, including noncomputerized ones designed for plurality only – without any modification needed at all. Right now.

Also in spite of its complexity, instant runoff voting has been successfully used in Ireland and Australia since well before the computer age. But since neither Instant Runoff nor Borda can be run on many of today's voting machines (USA), you are right they are problematic.

Toxen: "spoiler" third parties merely force the major parties to pay attention to their views.

My reply: I wish. Actually, the Democrats, who could have listened to Nader, or who could have offered him a cabinet post, say, instead started a well documented campaign of dirty tricks and harassment lawsuits (over 20 suits) intended to destroy the Green Party and to prevent Nader from getting on ballot in swing states. This attempt largely succeeded.

Meanwhile there were claims (which might be true) that the GOP actually aided Nader (who might seem to have been more their ideological opponent than not) financially.

Rosen: invents and advocates system he calls "bucket voting."

My response: This is a lot like range voting but not as good because you are strategically forced, in Gore v Nader v Bush FL 2000, (assuming you prefer Gore to Bush; if prefer Bush same holds in reverse) to put all your eggs in the Gore bucket, unless you want to waste part of your vote and thus be strategically foolish. With range voting, you can do that but then still give Nader the maximum possible score. So I think Rosen had the right idea but really wants Range Voting. If Rosen's idea were adopted then Nader and third party candidates still could never win with non-idiot strategic voters – but with range voting they could win.

Everett: Advocates modified Borda voting with rank-ties permitted.

My response: this is very much like range voting – permits more expressivity which plain Borda does not. However, it is not quite as good as Range because of this example Borda election where omitting a clear-loser candidate (who was suddenly realized not to have been eligible to run) reverses the election results for all the remaining candidates! With range voting, omitting some irrelevant candidate has no effect.

Snyder: we know from polls who the top two are, and so plurality works. Screw voters who want to vote for some non-top-2 candidate, they are idiots.

My response: trouble is, that forces 2-party domination and consequently reduced voter choice in all subsequent history, a great diminution in democracy. In fact in the USA we now have one-party domination with 98% of all federal races predictable over 1 year in advance. I.e. almost no voter choice, i.e. almost no democracy.

Luoma: wishes things were simpler...

My response: I think range voting is what you want – simple, easy, intuitive. Who has not been asked to rank things on an 0-10 scale?

Campbell: ... complains about low voter turnout...

My response: I believe range voting would increase turnout because races would no longer be 98% predictable, there would be real competition and real choice going on, and you could express more honest opinions in your votes without being a fool.

And if you do not believe me, then ask – why are most other countries with more competitive elections than the USA, also having higher turnout than the USA?

Harris & Robinson: suggest a different kind of instant runoff system. They did not realize it, but both reinvented the "Coombs" system.

My response: Coombs also has problems, namely it causes the two "frontrunner" candidates to automatically both lose if lots of voters artificially rank them top and bottom to try to maximize their vote's strategic impact.

Lee: Says the main problem is not the voting system, it is the media blackout of third-party candidates. if the media covered them, then we'd elect third-party candidates, even with the present voting system.

My response: I agree somewhat with what you say, but you don't realize that the media blackout is caused and enabled and justified by the voting system we have. See this discussion.

Ossipoff: I agree with his points such as the "participation" criterion (and Ossipoff has since joined the Center for Range Voting and endorsed Range Voting).

Siffert: Wishes Condorcet voting had been discussed.

My response: Condorcet may be a good idea. However it is more complicated than all the systems discussed, it will not run on many voting machines, and it still exhibits "favorite betrayal" and failure of "participation".

Mathewes: wants "none of the above" .

My response: NOTA can be added in many systems, including range voting; this is an independent issue.

Lamb: Basically invents range voting!!!!!

My response: I think I love you.

Hudson: I agree with his points.

Ferko & Bowers: Plurality voting works well enough, so why switch and get more potential for error? Voter incompetence is a major problem.

My response: actually, our studies and the French studies (2nd study) both show range voting and approval voting lead to less human error, i.e. less invalid ballot "spoilage", than plurality. It is pretty obvious why that is true for approval ("overvoting" no longer exists as a cause of spoilage) but less obvious why it is true for range voting. In contrast, Instant runoff in San Francisco caused the ballot spoilage rate to increase by a factor of 7.

Return to main page