Mr. Hill,

I read your recent response to my IRV letter.  You're right.  I was not aware that Oakland elections are non-partisan.  In fact, I just learned this week about non-partisan elections, as one of the co-founders, Jan Kok, is from Fort Collins, Colorado, where he says they use (or have used) a non-partisan system.  While that might take away two-party domination in the literal sense, I don't think it would prevent the same effect.  There would still be a sort of "center-squeeze" as exemplified in the example I showed.  As long as voters know about where the candidates sit in the political spectrum, that achieves the same effect I would think.  I'm forwarding this email to Warren Smith and Jan Kok, the two founders of the Center for Range Voting, to see what their take on the issue is.  Warren's a Princeton math doctorate, so I think he'll have some valuable feedback.  And I do appreciate your bringing this to my attention.

You said:

The benefits of IRV in Oakland have nothing
to do with third parties, spoiling major party candidates, or
anything like that. He has set up a "strawman" that does not apply.

Here are some direct quotes from
Maybe you ought to tell those at the Oakland IRV site that they are making false claims about the benefits of IRV.  And I'm curious what you think the benefits of IRV are, if not to prevent spoilers and what not.

I'm not sure what planet or parallel universe would see Ralph Nader
winning 37% of the vote in a three way race with Al Gore and John

So you admit that, with IRV, third party candidates (whether they go by party names or not) have no chance to win with IRV?  Great.  That's something a lot of IRV proponents won't admit.  Instead, they sell IRV on the myth that it's good for third party candidates.

In any case, elections exhibiting the problem I brought up with that hypothetical scenario have occured in the real world.  Peru 2006 and Chile 1970 are but two examples.  Here's another hypothetical election scenario exhibiting the same problem:

Voters -- Their ballot orderings (say, after the candidates have been whittled down to three contenders)
28% Dean > Kerry > Brownback
27% Kerry > Dean > Brownback
45% Brownback > Kerry > Dean

I don't think you can call this scenario unrealistic.  With IRV, Dean wins, even though 72% of voters prefer Kerry.

That's the interesting thing about all of
these critics and advocates of other methods like IRV, they always
propose these mathematical "paradoxes" that, while in theory are
interesting for mathematicians to doodle around with on their sketch
pads, in fact have no basis in reality. In the real world, these
sorts of paradoxes rarely if ever manifest themselves.

That's incorrect.  Here's the response from

15. Q. Do you really expect me to believe that non-monotonic IRV elections, "no-show paradox" nightmares, and near-tie IRV nightmares really are going to happen in real elections? Because I've never heard of an example. I bet examples are unbelieveably rare and will never really occur.

A. "Yes," we really expect you to believe this, and "no," such examples are not unbelievably rare.

In most IRV elections that have been held throughout history, the votes were simply never made public, preventing anybody from knowing about whatever non-monotonicity occurred. However, Debian has employed ranked-ballot elections for their leader each year since 1999 and published the votes in the most recent 5 of these 7 cases, i.e. 2001-2005. See this discussion of the Debian elections, including why they may be the only real and consequential moderately large ranked ballot elections with publicly available full vote sets in the entire world. (If you know of any others, please tell me about them!) Of these 5 elections, one exhibited an exact tie in an IRV round, and another exhibited a 1-off near-tie. The 2003 Debian election, with 4 candidates and over 450 voters, was a complete nightmare for IRV since it involved at least two different IRV near-ties, severe-nonmonotonicity examples, and no-show-paradox examples each!

Then, to really put the nail in IRV's (and plurality-with-delayed-runoff's) coffin, in the Peru election of 2006 Garcia won despite the fact that, pre-election polls indicated Lourdes Flores was pairwise-preferred over both him and Humala! (More examples, e.g. Chile 1970.)

But I do agree that it's not a good idea to exaggerate the importance of meeting all of these criteria that are the center of so much debate between supporters of rival voting methods.  We should be looking at one aggregate measurement: Bayesian regret.  I made a point about this in my letter at /IRVLetter.html

Of course, we can compare the individual properties of voting systems ad infinitum, but that's a bit like comparing the engines, tires, and aerodynamics of two race cars.  The ultimate metric is simply, when the rubber meets the road, which one gets the best time?  The analogous test for a voting method is called "Bayesian regret".  In lay language, it is simply the avoidable voter dissatisfaction produced by an election process.  A theoretical process that could read the voters' minds, and choose the candidate who would bring about the greatest average happiness, would have a Bayesian regret of zero, by definition.  Rigorous experimentation has shown that Range Voting ( produces about 20% as much Bayesian regret (or five times as much voter satisfaction) as IRV or plurality, even when voters are extremely strategic instead of honest.  This also shows us that Range Voting gives us as much improvement over plurality and IRV as either of those methods gives over non-democratic random selection of the winner.  This means that Range Voting effectively doubles the happiness brought about by the invention of democracy!  This also means that using Range Voting would produce a far greater improvement to democracy than the total eradication of fraud (in any country where fraud is not so bad that it produces worse results than random selection would).  Incidentally, in highly strategic electorates, IRV tends to produce as much Bayesian regret (as much voter dissatisfaction) as plurality.  That should be the nail in the coffin for IRV, for anyone who sees the significance of Bayesian regret, and knows how poor plurality voting is.

It's also possible that a meteorite will strike the Earth and wipe out life as
we know it -- though not probably likely for a few more million years.

It's also possible that it could rain next month in Seattle.

Your assessment of the "money saved" with IRV is flawed.  For one thing, you are comparing this with a traditional plurality plus runoff election, not with a straight plurality election, where the expense of a runoff election doesn't play a part (there is no runoff).  Further, IRV and top-two runoff aren't two different ways of doing the same election; IRV and TTR literally produce different results.  As my letter mentions, many (possibly all) of the at least 26 countries that use TTR have broken out of two-party duopoly, whereas the main four to have used IRV have achieved two-party domination.  Your estimate doesn't take into account the regret cost associated with using IRV.  You also don't consider that a runoff election won't be required every year.  So to be honest and fair, you'd have to consider the real cost difference by comparing the average cost with TTR (factoring in all the years where no runoff election is needed) to the costs of IRV.

I grant you that if increased costs can produce a substantially lower Bayesian regret, then it might well be worth it to spend the money.  But IRV produces just slightly lower average Bayesian regret than plurality, so it's questionable whether it's worth the cost.  Range Voting, on the other hand, would be significantly cheaper and easier to implement, tabulate, etc., and it produces substantially lower Bayesian regret (improved voter happiness) than IRV.  My point isn't so much to knock IRV (which is probably an improvement to plurality, overall), as to point out what a disservice we do for society by pushing for IRV instead of much better methods, like Approval Voting or Range Voting.  These systems also fit in with a non-primary-election world, so your cost savings argument, even if it was true, wouldn't refute the argument that Range Voting and Approval Voting are superior to IRV.

And just a quick note.  Warren Smith is a math doctorate, not a student.  I believe he got his Ph.D. in the 80's.

On and on and on, this critic's post is filled with substantial
distortions and misinformation. His opinion is unfounded on anything
we would call facts or reality.

That's the pot calling the kettle black.  IRV organizations repeat egregious falsehoods frequently.  They say that IRV prevents spoiler candidates, and only elects "majority" (Condorcet?!) winners, and prevents strategy, etc.  I've contacted many of these organizations, like the Oakland Measure O web campaign, and they simply refuse to correct these errors.

Recently an IRV advocate told me that my anti-IRV letter was misleading, because it stated that two-dozen U.S. cities repealed IRV.  Well, they repealed STV, which is IRV when applied to single-winner elections, so my claim wasn't false, like a lot of these IRV claims are.  But just in the spirit of honesty, I edited my letter to make this clear.  I would be a lot kinder to IRV advocates if they could be this honest.  Some are, and probably have hope of eventually abandoning IRV, once they finally realize how flawed it is, relative to better systems like Approval Voting or Range Voting.  Some like Rob Richie, however, will probably never embrace that spirit of honesty, and will continue to make the false claims about IRV.

You have called my statements distortions and misinformation, yet I have cited facts to back them up.  On the other hand, the claims we have made about lies told by IRV advocates, are easy to prove false.  You say that spoiler situations are incredibly rare with IRV, yet history says otherwise, and shows that the effects of IRV's "center squeeze" produce exactly the two-party domination that the math predicts.  Maybe it's time you start supporting something better.  Maybe it's time to consider supporting Range Voting, if you want better democracy.

There are other good single-winner systems out there besides IRV.

Yup.  Borda and Range (and Approval, which is a type of Range Voting) are both better than IRV.

Range voting may be one of them, but it is very untested in public
elections anywhere.

I agree that it hasn't had the kind of public testing that would be helpful for pointing to as a demonstration that it can work well "in the wild", but all of RV's real world use, and a lot of very rigorous mathematical modeling, and some exit poll experiments, say that Range Voting will perform a great deal better than IRV.  Considering that IRV's history of real world use has produced two-party domination and a lot of other ugly election results, and considering how much easier Range Voting would be to implement, we think there's a sound scientific case that RV will do better than plurality or IRV.  Sound enough that we think IRV advocates ought to support Range Voting, instead of IRV.  I don't take this lightly, and it took me a lot of debate to ditch my loyalty to Condorcet, and come aboard the Range Voting club.  But I sincerely believe that if you look at the facts and figures, and really have an objective scientific mindset, you'll see that Range Voting is immensely better than systems like IRV, and at least deserves to be tried in a large public election system.  At this point, it would take disastrous real-world problems to counter the reasonable expectations of RV's performance.  We think that smart people who want better democracy should be supporting Range Voting, or at least working really hard to get it used somewhere, so that we can experimentally prove, on a large scale, that it produces the great things we believe it will.

In any case, I fail to see why certain range
voting advocates apparently think they can advance their preferred
method by engaging in the age old winner-take-all tactic of slinging
mud at what they perceive as their "opponent" -- that is, instant
runoff voting. It's just the same old politics by another name.

I don't think that's a fair argument to make, quite honestly.  "Mudslinging" is a term for ad hominem attacks.  That is, you point out that the candidate you're running against had an affair or something, which is unrelated to most of the issues that matter to voters, except that it produces an emotional reaction that may cause people not to want to vote for him.  I think it's fair to say that Range Voting supporters have been directing our arguments specifically at IRV.

I never said, "Oh, those IRV supporters are all gay, tree-hugging, commies." or some such nonsense.  I said things like, "IRV produces two-party domination" and "IRV produces almost as much Bayesian regret as plurality".  Maybe your argument is that, if we support Range Voting, we should only talk about how good Range Voting is, and not point out the flaws in IRV.  But I don't think that makes sense.  If a lot of people are supporting IRV because they believe it does things that it doesn't, and if we know that democracy would be improved a great deal by supporting Range Voting instead, then it makes perfect sense for us to educate people about the shortcomings of IRV, so that they'll hopefully look for something better.

If you were about to buy a car, and were really leaning toward the styling of car A, and I told you that car B got even better fuel economy, and had a higher safety rating (all positive things), you might still want to go for car A.  But say I told you that car A was known to have a high incidence of roll-over, and tended to require a lot of expensive maintenance, based on objective data.  That's negative, but wouldn't you want to think about going for car B if you knew it didn't have these problems that car A had?

If people are going to make educated, scientific decisions about which voting methods to support, they need to know both the positives and negatives of the various options.  I stand by my decision to call IRV out as a failure of a voting method, compared to other alternatives.  If you understand why that is, I think you should reach the logical conclusion that it makes sense for me to do so.


Previous letter

Return to original letter

Return to main page