Jameson Quinn – Reflections on Voting Alternatives

Rob R: Thanks for including me on this mail. I know our opinions differ – even, sometimes, to the point where we are forced to doubt each others' good faith. Yet I know that we have a common enemy in plurality voting.

Robert S: I presume you're involved with the “politics and electoral reform” working group of occupy wall street (or some other “occupy” group?). I'm sorry, reading back through this thread, it's hard to tell exactly what your role is. But as far as I can tell, the intent here is to try to find some “consensus language” about voting reform that the various sides can all agree with.

I very much sympathize with this goal. I myself have tried to promote something similar, and the resulting “consensus statement” is posted here and here (same statement, different groups of signatures; over 15 signatures ove-rall, and to my knowledge ALL of those signers have advanced degrees and/or years of experience with the mathematical analysis of voting systems).

The basic thrust of that statement is:

1. Plurality is bad.
2. We can all agree to endorse 4 good systems: Approval, Condorcet (several variants), Majority Judgment, and Range.
3. Some of us endorse IRV as a clear improvement over plurality, but others have serious reservations about it.
4. Some of us favor a newer system called SODA voting, but others feel that it is too new and untried.

Rob Richie has stated that he refuses to sign this statement, as the lack of an endorsement for IRV is unacceptable.

I have also contacted the principal academic advocates for the four systems endorsed in the statement, as I see them from my knowledge of the academic literature. That is:

-Approval: Steven Brams (author of an editorial in Science endorsing Approval).

-Condorcet: Markus Schulze, creator of the Schulze method, which has been used by far more real voters than any other Condorcet method.

-Majority Judgment: Michel Balinski and Rida Laraki, the method's inventors (although it is similar to several older methods, including one used for a few years in 18th-century Geneva).

-Range Voting: Warren Smith, author of the original paper which found that Range came out consistently superior when measuring voter satisfaction in simulated elections (he was pioneer of this technique, which he calls “Bayesian Regret”).

Of these, only Smith would consent to sign the statement. All of the others responded with some variant of, “I like what you're doing, and I wish you luck with your statement; but [my system] is clearly better than other systems, so I'm afraid I can't sign your statement.”

Several of them also said that the statement should take a stronger position against IRV; this, in spite of the fact that the statement does not endorse IRV, and explains why it doesn't endorse by mentioning several of IRV's objective flaws.

So, as you can see, it is not easy to get consensus.

I personally would sign/support any statement which endorsed any subset of the 6 systems mentioned above (Approval, Condorcet, IRV, Majority Judgment, Range, and/or SODA voting). Despite what Rob says about “serious” reformers, though, I think that you would get the broadest consensus from voting reformers on Approval, Condorcet, Majority Judgment, and Range. That's why I included those four in the consensus statement linked above. Note that those are not exactly my four favorite methods. My personal favorite method is SODA.

Why is IRV not on that list? Certainly, there is no denying that IRV has more “mindshare” among voting reformers, primarily because of Fairvote. I'm quite aware that, as lobbying outfits go, Fairvote is small potatoes; but still, I have little doubt that over it's more than 10 years of existence under various names, Fairvote has spent more than a million dollars promoting IRV. That kind of money, and the things it buys — full-time staff, access to lawmakers (I don't mean bribes, just travel budget and tickets to events and such), professional polling — simply dwarf the resources that promoters of other systems have had (not to mention that Balinski and Laraki's research, which probably counts as some of the most expensive promotion of other methods, happened in France).

Yet IRV has serious flaws. This is not simply bias or closed-mindedness talking. In fact, IRV was the first voting reform proposal I heard of and supported. But as I learned more about voting theory, I came to appreciate its serious shortcomings; and now I consider it to be only slightly better than plurality. I'm sure many other voting reform advocates could tell the same story of initially supporting IRV, yet coming to view it with suspicion. Probably the Burlington 2009 election was a watershed for many, highlighting as it did the problems with IRV (and leading to IRV's repeal in Burlington itself).

What are IRV's flaws? Basically, there are two.

1. In some cases, IRV can lead to spoiled results, just like Plurality. The fact that these spoiled results are less common than under plurality is little help; as Burlington showed, they are very real, and so the fact that they strike unexpectedly is little comfort. (See here or here for some pictures demonstrating the craziness. The pictures at the first link for approval voting might be painting a slightly unfairly rosy picture, but there is nothing unfair about the hideous IRV pictures.)

2. Because it is not “summable”, IRV can only be counted by transferring all the ballots (either physically or electronically) to a central location. This is fundamentally more logistically-difficult AND more insecure than if an initial local precinct count is possible. (It also makes using IRV for US presidential elections totally impossible; unlike a summable system, which could be implemented through an interstate compact like the National Popular Vote movement.)

So why does FairVote support IRV in spite of those flaws? I would guess that it's some combination of the following three factors:

1. IRV is more friendly to candidates/incumbents than many other systems. This makes it much easier to get implemented. Fairvote expresses this in terms of a property called “later no harm”, which only IRV (or very similar, elimination-based systems) can satisfy. But basically, this is about the worry that a centrist could win by being everybody's second choice, even if they were weak in first-choice support. Such a win may or may not be favored by voters [1], but it would definitely be opposed by incumbents who'd won under the current system. Under a system like Approval or Range, candidates would therefore pressure their supporters to vote for them exclusively, which could reduce the benefits of the system, or even – if the voters foolishly succumbed to such pressure – lead to “spoiled” results. But such spoiled results under Approval or Range are entirely avoidable by the voters. So to me, the valid part of this argument is not the (silly and wrong) idea that Approval would reduce to plurality, but the idea that Approval, Range, or Condorcet have essentially only downsides for current incumbent candidates, and are thus harder to get implemented. (That doesn't hold for Majority Judgment or SODA, even though they do not meet “Later No Harm”; but both of these systems are newer and, as far as I know, FairVote hasn't made any statements on them yet.)

2. IRV is the single-winner form of STV, which is one of the primary systems for proportional representation. PR is an important goal, and FairVote supports its STV version. Unlike the single-winner case, the huge improvement from what we have (FPTP single-member districts) to any PR system is much bigger than the difference between the best and worst PR system, so the multi-winner domain doesn't see as many theoretical fights as the single-winner one.

3. The above two reasons are legitimate ones, although to me they don't outweigh IRV's flaws. However, I can't help but think that FairVote's IRV support has an element of irrational attachment. If they admitted IRV's flaws now, they'd lose face; so they don't, even as those flaws become more apparent. This extends to exaggerating IRV's benefits or approval's flaws, something which infuriates approval supporters.

[1] A “weak centrist” winner under Condorcet is, by definition, favored by a majority against each other candidate; but they might actually lose in a separate runoff, as voters took a closer look and decided they didn't like the candidate after all. On the other hand, they might maintain that advantage in a runoff, in which case they deserve to win. As far as I know, the only systems which might be able to correctly distinguish between these two cases, without actually having a separate runoff (with the accompanying costs and drop in turnout), are Majority Judgment (as experimentally shown by Balinski and Laraki), and SODA (purely my own hypothesis, though justified logically).

….

OK, that's a lot of inside baseball. And probably more information than you were looking for. But I think it's all germane to the point here: why it is so hard to find a meaningful consensus among voting reformers and/or theorists, even though we all agree that plurality voting is a horrible system.

So, what's the best consensus attainable? Well, it's important to realize that the consensus process which OWS uses is different from the kind of consensus that I was looking for in my “consensus statement” linked above. I needed to make a statement that people would actually choose to sign, as opposed to simply passively not signing. But “people's mic” consensus process is looking for something that everyone can assent to, as opposed to actively blocking. That allows a crucial extra bit of room to include some options that not everybody might support.

So, ….

… Oh. Sorry. I'd read the OWS statement before, and thought this was about that. Now I just followed the links to the “Electoral Reform Act of 2012”, and it's much more specific.

Which makes it much harder to get consensus support.

The “electoral reform” act materials are ambiguous. The 2-page summary refers to “Instant Round-Robin Voting” (a synonym for Condorcet voting; in fact, one that I myself coined – I could dig up on the wayback machine a web page I made back in 1998 – 1999, if you want proof). But the act itself talks about Approval for single-winner contests, and open-list or multimember-district PR for multi-winner.

I think that these are all good systems, but that it is too soon to pick universally “best” systems (“all elections must use…”) at this time.

So, I'd just say:

All national, state, and local elections must begin, and if possible complete, the process of phasing out plurality voting and adopting a better voting system. An improved voting system should collect more information from voters using improved ballot design, and use that information to accomplish at least two of the following three goals: (1) to better satisfy more voters if all voters vote sincerely; (2) to satisfy more voters if all voters are strategic; (3) to encourage voter sincerity. Well-understood systems which satisfy two of those goals include Approval Voting, Condorcet voting, Instant Runoff Voting (IRV), and Range Voting. Newer systems which may satisfy all three goals include Majority Judgment, Simple Optionally-Delegated Approval (SODA) voting, and Proportional Representation (except closed-list forms). All ballot counting must be done using publicly-witnessed precinct-based Hand-Counted Paper Ballots (HCPB) Election Day must fall on a Sunday. Every citizen 18 or older, regardless of condition or transient status, must be able to vote easily. Early Voting must be universal.

The “better whether all voters sincere or all are strategic” language is a way of allowing all good voting systems like those listed, while keeping out dangerous yet superficially-attractive ideas like Borda voting or other fixed-point-value systems.  Here's the implicit table:

Good luck. Please let me know if this screed was helpful or if, instead, it made no sense.

Jameson