Point by point examination of proposal of IRV for Denver

(Return to main range voting page)

The proposal of Instant Runoff Voting for Denver is available (as a CRV local copy) in this pdf file. We are now going to go through that proposal (and some associated remarks made by fairvote.org founder Rob Richie), point by point in detail correcting erroneous, false, or misleading points.

Unfortunately IRV advocates, including fairvote.org, have a long history of making erroneous, false, or misleading statements.

Misleading presentation of one-sided admiring picture of San Francisco:

Fairvote.org, in its greatest achievement to date, got San Francisco to adopt IRV voting for its 2004 elections, and it is repeatedly mentioned in the Denver proposal as an example of a "success." This was misleading because they "forgot" to mention:

1. San Francisco's voting software screwed up and was unable to call the four nontrivial IRV races, whose result-announcements had to be delayed for weeks. (Fortunately, SF had mandated a "paper trail" to allow the ballots to be recounted manually if necessary.)

2. San Francisco's new system not only had trouble producing winners, it also confused voters to a much greater degree than plain plurality voting. The result was a seven times greater rate of "spoiled" ballots in SF's IRV versus its plurality races. The proposal described this as "87% of those San Franciscans polled understood ranked choice voting." It may be just us, but we fail to appreciate why it is a "success" to have a voting system that 13% of the citizens do not understand. Also do not know how fairvote.org decided the 87% "understood" IRV anyhow. For example, suppose we had asked these voters: "In IRV voting, can raising your ranking of a candidate from bottom to top, actually cause that candidate to lose instead of win?" The correct answer, sadly, is "yes" as this election example (see fact #5 there) demonstrates, but we rather doubt that 87% of San Franciscans would have known that.

False claim about IRV #1:

The proposal on page 6 ("how ranked choice voting works") says "it's best to use all three of your rankings." Unfortunately this statement, while appealing and making IRV look good and logical, is simply mathematically false. The same 4-candidate counterexample IRV election as before also demonstrates that (see fact #2 there). Specifically: it can be best to rank only one candidate and not the top 3 (which since there are 4 candidates, in this example is equivalent to ranking all 4). Specifically: if the "F>N>G>B" voters foolishly rank all the candidates, then in that example election their most-hated candidate B wins, but if they rank only one ("F" only) then G would have won (which those voters prefer as a lesser evil). We do not see how electing your most-hated candidate is "best" for you.

False claim about IRV #2 (actually two false claims):

One page 3 of their proposal is

[IRV assures] that a "spoiler effect" will not result in undemocratic outcomes. IRV allows all voters to vote for their favorite candidate without fear of helping elect their least favorite candidate, and it ensures that the winner enjoys true support from a majority of the voters.
Again, this sounds simple and great for IRV, but again, sadly, it is simply mathematically false. In fact, it took advantage of the fact it had two sentences to use, to make two false claims:

1. Actually, IRV can and does exhibit "spoiler effect." To review, in the Bush-Gore-Nader 2000 presidential race, Nader was a "spoiler" and the act of 97488 Florida voters of voting for Nader, caused both their favorite Nader, and their second-favorite (in most cases) Gore, to lose. If these voters had (dishonestly) voted instead for their second choice, then Gore would have won. Thus those Nader votes "spoiled" it. Now consider the following (hypothetical) IRV election:
#voters Their IRV Vote
45 Bush>Gore>Nader
10 Gore>Bush>Nader
10 Gore>Nader>Bush
35 Nader>Gore>Bush
In the above election, by the rules of the IRV system, Bush wins. But if the 35 Nader-voters in the last line were to dishonestly vote "Gore>Bush>Nader" or "Gore>Nader>Bush", then Gore would have won. So again, these voters' act of honestly voting Nader caused both Nader and their second-favorite Gore to lose – again, Nader was a "spoiler" even though we are now using the IRV system. The claim that "IRV allows all voters to vote for their favorite candidate without fear of helping elect their least favorite candidate" was simply false.

2. The claim that the IRV winner "enjoys true support from a majority of the voters" is also simply false:
#voters their IRV vote simplified
50 A>B>C>D>E A>B
51 B>A>C>D>E B
100 C>D>B>E>A C>D
53 D>E>C>B>A D
49 E>D>C>B>A E>D
In the above 303-voter example by Mike Ossipoff, the centrist candidate C is the favorite of far more voters than anybody else, and not only would win a head-to-head contest with any single opponent (based on these votes) but in fact would do so by approximately a 2:1 margin. (Details: C beats A by 202 to 101; beats B by 202 to 101; D by 201 to 102; and E by 201 to 102 according to voter preferences.) So the "right" winner here clearly seems to be C, and almost every ranked-ballot voting system would indeed elect C. But not IRV. IRV elects D! (The reason: IRV only examines the preference relations in the "simplified" votes, and ignores the others. In contrast, range voting never ignores any part of your vote.)

We fail to see why it is that the IRV winner D here, as the proposal puts it, "enjoys true support from a majority of the voters." A 2:1 supermajority for C over D would have been a more massive "landslide" than any US presidential election ever, yet IRV in this example tells that supermajority to go jump in a lake!

False claim about IRV #3:

On page 2 of the Denver proposal, it says

Voters can vote for the candidates they really like, which eliminates spoiler candidates and "lesser of two evils" dilemmas.
But this is simply false, as we just demonstrated (countererexample election 1 to false claim #2): the voters favoring Nader could not vote for the candidate (Nader) they "really liked" because that would lead to the (in their view) greater evil (Bush) being elected; they instead were forced to dishonestly vote for the "lesser of two evils" (Gore, in their view) to prevent that. The lesser of two evils dilemma is still there for Nader voters with IRV, exactly the way it was there with plurality voting.

In contrast, with range voting, voting Nader top can never hurt a voter who favors Nader. Range voting genuinely cures the spoiler problem, and genuinely permits voters to vote for their favorite without fear, IRV doesn't.

False claim about IRV #4:

On page 2 of the proposal, they note that IRV is used in "London," and on page 3: "Instant runoff voting (IRV) is a well-tested voting method for single-seat offices, such as Mayor or Governor." Well, London does not have a Governor, but it does have a Mayor, who is the most prominent "single seat office" there. We have no idea what the proposer had in mind about London, but we can tell you that London's Mayor definitely is not elected using the IRV method, although London's system is related. For example, under IRV, if you vote A top but B second-choice, then under the IRV system, A could be eliminated and then your vote would entirely transfer to B. Then if B in turn were eliminated, your vote would entirely transfer to your third choice, and so on. But under London's system, if there was not anybody who got over 50% of the first-choice votes, then your second choice votes are then counted instead of being ignored, except if your first-choice had been for one of the top-two, or if your second choice had not been for one of the top-two, in which case your second choice still is ignored. Then, whoever has the most (non-ignored) votes wins. (If you had any third, fourth, etc, choices, those are always ignored, and there are no further rounds byond the second, unlike IRV.) Thus in the (most recent) 2004 London Mayor election, the first choice vote totals were
candidate vote %
Livingston 36%
Norris 28%
Hughes 15%
Maloney 6%
(others) 12%
so that nobody got over 50%. The second-choice votes if they were for Livingston & Norris and if they were cast by voters who did not vote for Livingston & Norris originally, were then added on at which point Livingston (who originally had 685541 votes) now had 828380 votes and won. This is not the same system as IRV. For more information you could check the wikipedia article, the London elects website, or the BBC news report.

IRV "promotes cost-saving elections"?

That is what it says on page 3 of the fairvote.org proposal. But we repeat, IRV voting does not work with simple non-computerized voting machines (or many computerized kinds either). Meanwhile, plurality and range voting both do. So we fail to understand how that "saves costs." We agree IRV perhaps could save money versus a two-election runoff system (but perhaps not, if the two elections both could be run with today's voting machines but the IRV election could not, and if the second election often was not held) – but any single round system such as range voting would accomplish that.

Another thing: since IRV elections can lead to a tie (or near-tie) in every single round and each such can affect the final election result and hence is very serious, the tie-risk with IRV is far greater than with plain plurality voting. (Meanwhile, switching to range voting would vastly decrease the risk of ties.) As you may recall, every near-tie election leads to a nightmare for election administrators with an expensive and heavily scrutinized recount – and often a legal battle ala Bush-Gore Florida 2000. We fail to see how having vastly more such battles will "save costs."

It should be illuminating to check this press report about the Univ. or Vermont IRV cost study.

Fairvote.org's North Carolina "success" and the "movement" of the USA "toward" IRV:

Rob Richie, commenting on and replying to our preference for range voting on the bulletin board set up for public comments about the Denver voting-reform proposal, said: "Jan may like range voting, but the rest of the nation continues to move toward instant runoff voting. The North Carolina state legislature just passed a bill to use IRV for judicial vacancies (elected with public financing) and test it out in up to 20 cities and counties..."

This was misleading for three reasons.

First, range voting, since it is used for academic "election" of valedictorians (the "candidates" are the students, and the "votes" are their grades in each class, each grader in each class gets one vote, and the student with the highest average range vote, i.e. highest GPA, is elected) has been used for far more "elections" in the USA over the last 200 years, than IRV. The academic world reached a worldwide consensus that range voting was the best voting method to use for this purpose, long before Rob Richie was born. We find it exceedingly dubious that the academic community now, were Rob Richie to present them with the "superior" IRV voting method, would be interested in switching to a technique in which each professor would, as his vote for valedictorian, present a rank-ordering of every student in the school, from best to worst. In fact, we very much doubt that Rob Richie would be able to obtain even one such vote from even one professor at any sizable college or university anywhere in the entire country. And even if each professor were required merely in his vote to select the top three students in the school (in best-to-worst order; this is a simplified degenerate form of IRV which would be illegal in Australia), we think that would still be rejected by every school in the country as clearly unacceptable in comparison to their current range voting system.

Second, the USA has not "continued to move toward instant runoff voting." In fact, more US cities have decided to abandon single-transferable vote (after they tried it for many years) and go back to the plain plurality system, than fairvote.org has managed to get to adopt IRV. The most famous abandonment examples were New York City and Cincinnati. Based on that, it might appear that it would have been more accurate for Richie to say that the nation has "continued to abandon the IRV system as a failure"! What about range voting? Well, no city has ever adopted range voting, to our knowledge, so we unfortunately cannot claim any great national movement as yet. (We would like Denver to begin such a movement.) But we will say this. Among the numerous US organizations that do employ range voting, as far as we know zero have ever abandoned it in favor of any other voting system, and as we mentioned above (the academia/valedictorians example) such abandonment would probably be considered so ludicrous it would not even be considered. That is not the case with IRV; for example the American Mathematical Society abandoned IRV for its internal elections and now employs approval voting (a simplified form of range voting which we "approve" of less). Range and approval voting both also have the advantage of being substantially simpler since they both run on every voting machine in the USA, right now, no modification required. (Try demo.)

Third, the North Carolina bill that Richie is referring to, did not implement IRV, but rather a bastardized version of it which seems far worse than genuine IRV (which in turn is worse than range voting). The North Carolina measure, as described in fairvote.org's own description (local copy), says "In an instant runoff, voters would be allowed to rank their order of preference among the candidates. The first choices of voters initially would be tallied. If the leading candidate failed to win more than 40 percent of the first-choice votes, the top two candidates would advance to the runoff."

The use of "40%" is very peculiar and undesirable. If a candidate gets a 41% plurality, there is no runoff; that candidate wins immediately. (Even if another candidate was close, say 40%.) Now wait a minute. I thought voting reformers agreed pretty much unanimously that Plurality voting is bad. But the North Carolina system acts exactly like Plurality when any candidate has more than 40% of the first-choice votes.

For a real-life example, consider Florida 2000. Under IRV, the preferences were roughly:
Candidate vote
Bush 49%
Gore 49%
Nader 2%
where the Nader voters preferred Gore over Bush.

So it's pretty clear that Gore would have won under regular IRV and almost every other voting system ever proposed (other than the one actually used). Regular IRV does fine in that kind of scenario.

But, with the North Carolina mutant so-called-IRV, the Plurality winner (Bush) had more than 40% of the votes, so Bush would have been chosen as the winner without any sort of runoff. Don't you think this sort of situation is exactly the sort in which a runoff is maximally desirable?

Examining our table of historical national presidential elections, we observe that in Romania 2004, Senegal 2000, Cyprus(Greek) 1998, Estonia 1992, Cyprus(Greek) 1993, and Portugal 1986, the North Carolina system would in all cases have elected the same winner as plain plurality system; but in all of these cases a top-two runoff was then held (those countries employed the same rules as Denver now uses) and elected somebody else that the voters wanted more. Thus, in our seven historical examples, fairvote.org's North Carolinian pseudo-IRV system would have elected a winner the voters wanted less than the winner that Denver's present system would have given them! We consider, therefore, this example of a fairvote.org "success" to be, compared to the system Denver has now, a "large step backwards."

Fairvote.org's North Carolina "success" (Richie reply):

Rob Richie replied as follows:

The North Carolina bill was a remarkable success story, where advocates in the legislature did a tremendous job getting it through. Your attack shows the kind of political naivete that I can find so irritating. There isn't a prayer in the world that they could have won with range voting or your other objects of enthusiasm, and they should be applauded for doing something really hard and really important... You should understand that traditional runoffs are already used for state and congressional primaries in the North Carolina. The state already has a 40% threshold that triggers these primary runoffs, set in part due to concerns relating to voting rights and opportunities for African American candidates to win in white-majority primary electorates. The North Carolina bill accepts that threshold for some offices, but does allow a 50% threshold for localities that want to try out IRV. I know you enjoy the math of all this, but there's a political world we live in.
To that we reply: we agree progress was made in those instances (if any – we are not aware of any actual instance) where the 40% has been replaced by 50%. However, in the cases (which were the only ones mentioned in the Associated Press article about it) where 40% has stayed 40%, the only "progress" that has been made has been to replace a possibly-2-round system by a single-round system, thus hopefully saving some time and money – but at the same time unfortunately requiring in some cases new voting machines and allowing new dishonest and damaging voter strategies, which may be enough to cancel out the cost savings and/or decrease the winner-quality. (Note: every IRV country is 2-party dominated, just like the USA, because the IRV voting system discriminates against "third parties" just like plurality. However, in contrast, plurality+top-2-runoff countries such as France, Romania, Chile, etc have more than 2 parties – as was recognized by Duverger. This illustrates the fact that the "instant" and "genuine separate" runoff systems are not equivalent, even though they naively sound similar.) So in those 40→40 cases we do not consider this to be a "remarkable success story," but rather "a tale told by an idiot, full of sound and fury, signifying nothing."

To be a little less dramatic (?) about it: the reason we are in the voting reform business is to improve fundamental flaws in our voting system, which cause undemocratic phenomena like "voting for your favorite hurts your cause" and "more candidates on the side of Good actually makes it more likely that Bad will win" and "a candidate preferred over everybody else by the voters, still loses." Range voting fixes all three of these flaws; IRV fixes only the middle one. We are not in the voting reform business because we want to re-issue essentially the same voting system as before, but done in a way which hopefully will save a little money. That could be nice, but there are far more impactful ways to save taxpayer money than that, so if that were your goal it would hardly be worth bothering.

We invite Rob Richie to apply his formidable political skills to pushing a voting system which is simpler, cheaper, more reliable, and less likely to lead to a tie, than IRV, and produces better winners on average with fewer undemocratic pathologies – range voting. We believe Denver could be an excellent opportunity for that, and the result would be something that everybody would agree was an improvement, even the bees.


Return to main page