A Response to Richard Anderson-Connolly

In response to this comment, by UPS sociology professor Richard Anderson-Connolly

Richard,

It's ironic that you begin by calling other, non-IRV, voting methods "complex", especially since one of the biggest advantages of Range Voting over IRV is its simplicity. In Australia, for instance, voting is compulsory, and a rank ordering of all the candidates is required on each ballot. That is significantly more work for voters than using Range Voting, where they could just score the few candidates they know or care enough about, and ignore the rest. With Range Voting, this isn't particularly harmful.

Consider a few places where Range Voting is used on a massive scale: YouTube, Hot-or-Not, Amazon.com. Range Voting doesn't seem to have been too complicated for the millions of visitors those sites draw. Millions of customers at Amazon.com seem to have no problem scoring the books they've read with a quick click of the mouse. Olympic judges seem to have no problem using Range Voting for gymnastics, ice skating, snow boarding, etc. I don't think they'd want to use IRV however.

When it comes to the actual administration of the elections (ballot layout, processing, and mathematical aspects of tabulation), Range Voting dominates IRV. You just calculate the average score for each candidate. With IRV, the process isn't rocket science, but it's definitely more complex. You have to successively remove the candidate with the least number of votes, and then reassess, until a candidate has a "majority" (although that's a misleading term, since it doesn't necessarily imply a Condorcet majority). See this page for a more thorough look at the respective complexity of Range Voting versus IRV.

Regarding Condorcet methods, what is your basis for claiming that "Condorcet voting is probably the most preferred by experts"? Consider that with Range Voting, if voters are strategic (willing to selfishly vote insincerely, to get a better result, from their perspective) the end result is the same as we'd get from honest Condorcet voting. Let me repeat in layman's terms: the worst result that Range Voting can achieve, due to selfish strategizing, is the same as the best, 100% honest, result that we would expect from Condorcet! But if voters are strategic using Condorcet, the results are the stuff of nightmares. Consider the DH3 pathology, in which strategy under Condorcet yields the candidate universally seen to be the worst. It is not a stretch to say that Range Voting is, under reasonably strategic voting, a better Condorcet method than real Condorcet methods. But the real argument ender is the fact that Range Voting produces substantially lower Bayesian regret than Condorcet methods. Here's a complete head-to-head analysis between Range Voting and Condorcet methods. Simply put, if these unnamed "experts" say that Condorcet methods are better than Range Voting, they're wrong, in every reasonable objective way we can evaluate that contention.

Next, you claim that you are more "practical about the issue" than a "character" like me. What do you consider to be practical about endorsing an expensive, complex alternative to our current voting system, which Bayesian regret calculations show can actually bring about lower voter satisfaction (under realistically strategic electorates) than even plurality? Do you find it practical for the government to spend a lot more taxpayer dollars to implement a system that is, at the very best, only a very modest improvement over our current system? Do you think it is practical to implement a system that will not work on many voting machines, when Range Voting can be used on any machine that can do plurality elections? Does IRV's near-century of producing two-party domination in every country where it has been used not impress you? Does the fact that it has been used, and then later abandoned, in around two dozen U.S. cities not impress you?

Your faith that IRV will be a large incremental step forward, that may well lead to continued improvement in the future, simply does not fit with the evidence. If anything, IRV's failure to deliver on its promises, especially for alternative parties, will leave voters disenfranchised, bitter, and weary of similar election reform in the future. It's like trying to convert U.S. road signs to metric, by first converting to from miles to feet.

Your mention of Arrow's theorem is irrelevant to a discussion of Range Voting. Range Voting passes all of Arrow's criteria, but Arrow's definition of a "voting method" excludes it. The lesson here is that Arrow used a silly definition of "voting system". Arrow's theorem is commonly misunderstood, misquoted, and misused. For more information, check out the Center for Range Voting's page on Arrow's theorem

It is also wrong to say that supporting a switch to IRV would be no more suicidal to third parties than sticking with plurality (while supporting Range Voting). If all the effort and money they expended, in order to get IRV implemented, could have gotten Range Voting implemented—in which case they'd actually have a chance of winning elections—then their work would have indeed been suicidal. And if IRV is implemented, people are going to be much less willing to go through the laborious process of implementing and learning to use a new system all over again. So in a decade, when those people have seen the failure of IRV (which history says they will), and they've decided that they need something better (i.e. Range Voting), it will be like the little boy who cried "wolf". Voters won't be very receptive to their message. They'll say, "Yeah, we heard this line about getting a better voting method before..and you're still not happy? Tough luck. We're not going through that ordeal again."

When you contend that the major two parties hate IRV, I'd like you to call up some politicians in the major two parties of current IRV countries (which are all two-party dominated in their IRV-elected posts), and ask them how much they hate having a system that prevents third parties from competing with them, or even being able to throw the election for them, like Nader did to Gore in 2000, under plurality. The only reason that the current dominant two parties in America might dislike a switch to IRV, is that it is a change, period. Once we were using IRV, they would certainly dislike a change back to plurality more than they disliked the initial change from plurality to IRV.

You claim that the only problem in San Francisco's IRV elections was a "slight delay". But in 2004, the use of IRV there resulted in seven times more spoiled ballots. I didn't bring up any spoiled ballot figures for Burlington, but if you're responding to the Vermont page I linked to, perhaps you could cite our specific claims, as well as verifiable contradictory figures. The Center for Range Voting aims to be as honest an accurate as possible, and we would very much like to correct any errors you may have found. That's more than you can say for IRV organizations, like the Center for Voting and Democracy (the chief proponents of IRV), who routinely publish egregious errors. And while we're on the topic of Burlington, Vermont, I should point out that it is frequently cited as an example of a city in which, in the first year of IRV's use there, a third party candidate, Bob Kiss of the Progressive Party, won the mayoral race. What IRV advocates usually leave out is that Kiss got 39% of the first place votes, and would have won under plurality voting anyway. Some would say that IRV took away the fear of voting for a third party candidate, and that without that effect, Kiss wouldn't have won. But let's also not forget that this was IRV's first year of use in Burlington. Had the conservative voters of Burlington used IRV for more elections by that point, they would have known to strategically vote for the "centrist" of the race, who would have then beaten Kiss, giving those conservatives the "lesser of two evils". Lastly, let's also bear in mind that Burlington is demographically atypical, and this scenario is reminiscent of the 2003 mayoral election in San Francisco, which was between a Democrat and a Green. In a more demographically typical race, with voters who have been using IRV for a few years and know how to be strategic with it, we can basically say that a third party candidate would need to get more than 50% of the first-place votes in order to win.

Finally, your statement about IRV's virtually always picking a "majority" winner (which I can only take to mean "Condorcet" winner) is simply baffling. You state:
The only time IRV doesn't produce a majority is when sufficient people don't rank enough preferences and the race goes into multiple counting rounds.
It's not clear whether you're saying that you think this just happens to have been the case throughout Australia's history, or that it's true as a general rule. The former is unlikely, and would warrant substantial evidence. The latter is patently false. Consider the following generic scenario, where we have three hypothetical candidates representing the left, middle, and right (Left and Right are interchangeable arbitrarily named ideological extrema).

% of Voters How They Voted
49% Left > Middle > Right
25% Middle > Right > Left
26% Right > Middle > Left

This situation produces a paradox, in that the Condorcet winner (Middle) is preferred to the IRV winner (Right) by 74% of the voters. This type of scenario is quite realistic; it has happened several times that we know of (e.g. Chile 1970, Perú 2006). If this hasn't happened in the entire history of Australia (which you haven't cited evidence to prove), that would simply mean that no third party candidate has ever become popular enough that the "opposite" party had an incentive to vote for the middle candidate. It wouldn't negate the point about IRV's "center squeeze" effect however.

As I hope you and others will see, the science of voting methods is quite a complex subject. It is a veritable branch of mathematics unto itself. Society does not benefit from sort of overly optimistic pronouncements, or the hand waving in the face of dissenting evidence, that you demonstrate in your adherence to IRV. If we support IRV, despite the mountains of evidence that tell us exactly what kind of failure we can expect from it, then we're going to pay the price. We're going to get what society always gets when it disregards hard science in favor of easy gut reactions. We all owe it to ourselves to face discomforting realities when the situation calls for it, and to change course accordingly. We're talking here about the very mechanism by which society collectively works to make decisions. If that is broken, then human civilization is broken. This isn't a "political" issue. This is about whether children in a far away country wake up to the sound of our bombs. This is about the fate of our species. If any of that particularly matters to you, then I encourage you to put all your spare time and effort into supporting Range Voting, not IRV.

Regards,
Clay Shentrup

Initial message to which Richard responded

Return to main page