Dec 20, 2010

Mea Culpa: I was wrong about Instant Runoff Voting!

This is a reposting from A New Kind of Party.

Over at Daily Kos, I posted an entry that admits that my principal objection against the use of IRV in bigger elections is wrong. This post came about after an email exchange on Saturday with the executive director of Fair Vote, Rob Richie, where he expressed his conviction that I had been bashing IRV in a way that didn't help to win his support for my own ideas.

This got me to rethink things and I saw that I was wrong and that he was right. And I also had to admit that I've been lapsing into the ugly infighting common among proponents for different alternatives to first-past-the-post elections. A las, I forget sometimes that it's more important to relate than to be right...; for if we truly relate with each other then what is right will tend to emerge with less ugliness or infighting.

I also realized that I needed to take a hiatus from posting about electoral reform. As such, this is my last post for a while. But I'm glad I started posting more after the November election. The idea of Strategic Election Reform is out there. And, Rob Richie does support the use of 3-seated election rules. But I can't expect to talk smack about his main election rule and then get him to bring up my idea.

So, hopefully, he'll come around to seeing that 3-seated Hare LR (3HLR) is superior to 3-seated Single-Transfer Voting (3STV). Here are three arguments in favor of 3HLR, just in case Rob or some other advocate of ranked choice voting happens to read this blog-entry ;-)

1. 3HLR is simple and more easy to explain to voters accustomed to FPTP elections. This is because FPTP is equivalent to 1HLR.

2. 3HLR helps minorities to have a better chance to be decisive in more elections. 3STV is over-committed to the notion of proportionality to help minority groups. 3STV transfers the votes of those in majority to their second choices, which keeps those in majority more likely to be influential than minority groups in the determination of who wins the third seat.

3. 3HLR is more biased towards smaller parties than 3STV. Yet this is not a bad thing, for so long as we continue to use single-seated elections that are biased in favor of bigger parties, one can argue that the different biases would tend to cancel each other out. In which case, the use of an election rule that tries harder to ensure proportionality might end up hurting minority groups. For we can't (almost) always guarantee that the majority rules and commit ourselves to the protection of ethnic/economic/ideological minority rights!

And with that thought, I'll bid my readers adieu and a happy holidays to you!
dlw
ps, if I believed that voters had cardinal preferences over all candidates and that random noise captures the essence of low-info voters then I would agree with Approval or Score Voting advocates like Broken Ladder or Dale that their election rules are objectively better than IRV. But I think these models fall short in capturing the dynamics of voting behavior (for complicated "reality-based" reasons I don't want to get into now) and that objectively the playing field among alternatives to First-Past-the-Post is relatively level. As such, IRV is worthy of our (strategic) support for a wide variety of single-seated elections, simply because it has been marketed well by Rob Richie and FairVote et al and is more-or-less understood by more voters.
dlw

15 comments:

Dale Sheldon-Hess said...

I call this the "But it's doing nothing NOW!" argument.

The thing is, IRV is still doing (almost) nothing.

DLW said...

That's not true. A half of loaf in hand is better than a pie in the sky, especially if political jujitsu is the only game in town. IRV moves the political center to the left and helps third party candidates to get more voice.

The selling pitch for AV rests too much on assumptions of voter-rationality that are not realistic. If low-info voters have correlated random shocks to their preferences(like due to manipulative campaigns), such that the shocks look more like signal than noise, the strength of AV would become a liability.

But I'd still prefer to use a form of AV in the first round of an IRV election, as a way to goad voters into studying about more candidates and to make it difficult for Sarah Palin's with a limited but enthusiastic base of support to get elected.

But the root problem is both our country's nearly exclusive use of single-seated elections and the cultural wars. The former enables the latter to crowd out too many other issues and provides cover for $peech to be more aggressive in tilting our polity towards utter kleptocracy and away from a tenuous balance between popular democracy and kleptocracy.

AV advocates arrest reforms when they trash IRV. IRV is the best marketed election reform in the US and will likely be a gateway to more comprehensive and varied electoral reforms.
dlw

broken ladder said...

I question the "bird in the had" reference. Considering that IRV has generally replaced Top-Two Runoff, not Plurality Voting, and considering the increased spoilage and decreased transparency associated with IRV, it is questionable whether IRV is an improvement at all. To expend reform manpower on adopting a reform of questionable value, instead of a substantive reform like Score Voting or Approval Voting, is a stunning example of lost opportunity cost.

I see no evidence that IRV moves the political center to the left. Perhaps you could discuss your justification for that statement. And in any case, the point of altruistic election reform should be to increase the representativeness of government. That means moving the government toward whatever direction the electorate moves, not necessarily to the left. I say this as a generally progressive person, concerned about climate change and a host of other stereotypically liberal issues.

The selling pitch for Approval Voting does not rest on any assumptions about voter rationality. We already know what voter do under Plurality Voting, which is just Approval Voting with a 1-candidate-per-race limit. And we have extensive evidence about voter behavior from exit polling using Approval Voting, and from use of Approval Voting in meaningful competitive elections at various large organizations.

If low-info voters have correlated random shocks to their preferences(like due to manipulative campaigns), such that the shocks look more like signal than noise, the strength of AV would become a liability.

This argument does not hold any less true for Instant Runoff Voting, or any other system. So it's not a differentiator in the context of a debate on the respective merits of Approval Voting vs. IRV et al.

On top of that, extensive Bayesian regret calculations using all sorts of variation in voter ignorance levels still show Score Voting and Approval Voting behaving extremely well -- far better than IRV or Plurality.

broken ladder said...

But the root problem is both our country's nearly exclusive use of single-seated elections and the cultural wars.

Single-seat elections are not the "real problem". We have solid evidence that an improvement to our single-winner voting methods could roughly double the representativeness of our democracy. And lots of elections are single-seat by definition, e.g mayor, governor, senator, president. A comprehensive look at the democratic-ness of countries around the world, with an emphasis on their use or lack of proportional representation does not show any conclusive evidence that an upgrade to PR could compete with an upgrade to Approval or Score Voting, in terms of increase in voter representativeness. See, for instance, the Economist's "democracy index".

And upgrading to PR, which requires multi-member districts, is politically much more difficult than a change to our single-winner districts. Multi-member districts are illegal for Congress, for instance. That will not change unless we first upgrade to a single-winner voting method which breaks free of duopoly, as a prerequisite.

I would highly advise you read this page on Proportional Representation. It discusses Cynthia McKinney's efforts to change the law to allow multi-member districts.
ScoreVoting.net/PropRep.html

The "culture wars" you mention are largely a symptom of the voting method. Because we have two-party duopoly, the "left vs. right" (good vs. evil) polarization is self-reinforcing.

Your fears about "$peech" would also be largely addressed by the adoption of a voting method which passes the Favorite Betrayal Criterion, which IRV does not, but Score/Approval do.
ScoreVoting.net/RL2parties.html#cash

Lastly, please cite any evidence whatsoever that IRV will be a "gateway" to better reforms. So far, I have not seen any IRV/RCV municipalities upgrade from IRV to Score or Approval Voting, but I have seen several repeal RCV back to what they had before.

P.S. I would be wary of Rob Richie. Even if you feel a sense of agreement with him on some issue, it is noteworthy that there are numerous documented instances in which he has made misleading and even outright false claims about voting issues, e.g. saying that a voter's best tactic with IRV is to vote sincerely.

DLW said...

Hi BL/Clay,

1. You conflate "reality" with the Bayesian Regret model of it. Models are like swim-suits, what they reveal is interesting, but what they conceal is often vital.

2. Bayesian Regret analysis doesn't determine the extent to which voters vote strategically. It takes the two extremes: no strategic voting and all strategic voting. It is true that IRV doesn't do any better than FPTP when there's a lot of strategic voting. But it does improve things somewhat when there isn't any strategic voting. As such, IRV is predicted to improve upon FPP to the extent that it coincides with less strategic voting. I believe that in "reality" there is less strategic voting with IRV and so it does improve things.

If it didn't improve things then why would conservatives or others who benefit from the status quo oppose the adoption of IRV?

3. IRV moves the center to the left to the extent that any increase in the population of "voters" tends to move the center to the left on average. With FPTP, folks who vote for third party candidates on principle are essentially nonvoters and so IRV makes them voters again. It is not a coincidence that most advocates for IRV and other electoral reforms are lefties...

4. AV's lack of use in a high-stakes general election is still a liability. FPTP is also IRV with a 1-candidate-per-race limit or 1-seated Hare Largest Remainder. This is why I've proffered that you push first for the modification of IRV so that the first stage uses a limited form of Approval Voting by disregarding the rankings and tallying up the number of times candidates got ranked by voters to determine the three finalists.

5. If low info-voters have correlated random shocks to their preferences then strength of AV would become a liability. In BR models, AV or SV does better than other election rules, like IRV, when you introduce noise into preferences because it's better at picking out the signal. But what if the noise starts to look like signal? Then they would respond more to the noise than other election rules and their strength would become a liability.

I doubt that BR calculations have been made when one models correlated random shocks to voter-preferences... but show me the model and we'll talk.

broken ladder said...

DLW,

Bayesian regret models use the exact tabulation algorithm of whatever voting method they are testing. So they are reality.

There are a handful of environmental factors that they must model. Such as:

- The ratio of strategic to honest voters

- The ignorance level of voters (difference between a voter's assessment of a candidate, vs. the assessment he would make if perfectly informed about that candidate)

- Preference distribution (the distribution of voters into different realms of ideological space)

- The number of candidates

- The number of voters

What we find is that it doesn't matter what the real-world values for these things are. You can tune them in every remotely plausible way, and Score Voting still does the best.

I've explained this to you multiple times.

Bayesian Regret analysis doesn't determine the extent to which voters vote strategically. It takes the two extremes: no strategic voting and all strategic voting.

That is absolutely false. Please don't just make things up.

The "strategy knob" was tuned in gradual increments all the way from 0% to 100%. You can see this in some of the primary data here.

And here's a page showing figures for numerous different voting methods, both for 100% honest, and 50%-honest-to-50%-strategic.

Again, I have pointed this out to you before.

It is true that IRV doesn't do any better than FPTP when there's a lot of strategic voting. But it does improve things somewhat when there isn't any strategic voting.

That's true. But...

1) IRV doesn't usually replace FPTP. It usually replaces TTR (top-two-runoff), which is FPTP followed by a runoff if no one gets a majority. And as my previous link shows, TTR is approximately as good as IRV, and a lot more simple and transparent.

2) You ignore the "opportunity cost", of spending election reform energy to get a tiny improvement, instead of the enormous improvement that could be gotten by Score or Approval Voting. That can easily make the adoption of IRV a net negative.

As such, IRV is predicted to improve upon FPP to the extent that it coincides with less strategic voting.

No, not "to the extent that it coincides with less strategic voting". Purely from a Bayesian regret perspective (leaving cost and transparency out for a moment), IRV is an improvement over FPTP, even if exactly the same percentage of voters will be strategic with either system. Actually, the adoption of IRV could even coincide with a small increase in strategic voting, and still outperform FPTP.

I believe that in "reality" there is less strategic voting with IRV and so it does improve things.

Compared to FPTP it may result in somewhat less strategic voting. Compared to TTR, I think it results in more strategic voting. That might explain why TTR has correlated with having three or more parties in most of the 27 or so countries where it has been used, whereas IRV maintains two-party domination, e.g. in Australia.

If it didn't improve things then why would conservatives or others who benefit from the status quo oppose the adoption of IRV?

Because they don't understand voting theory. It's really that simple. Also, by their very definition, conservatives are generally opposed to major changes in traditional ways of doing things.

Again, look at Australia, where they have used IRV since 1918. Their IRV-using House of Representatives is approximately as two-party dominated as our Congress. And that's in spite of the fact that their proportional Senate has several seats occupied by e.g. Greens.

broken ladder said...

IRV moves the center to the left to the extent that any increase in the population of "voters" tends to move the center to the left on average.

I assume that you mean, each new generation of voters is generally a little more progressive than the previous one, therefore population growth generally moves the population to the left.

With FPTP, folks who vote for third party candidates on principle are essentially nonvoters and so IRV makes them voters again.

That doesn't make sense. If the population as a whole becomes more liberal, then the major parties will inevitably nominate slightly more liberal candidates, without any change in voting method.

It is not a coincidence that most advocates for IRV and other electoral reforms are lefties...

Of course it's not. Liberals are, by definition, more open to change. That is not in any way evidence that IRV will elect more liberals.

One plausible way it could elect more liberals is if liberal minor party candidates run more often than conservative ones, causing liberals to suffer more often from the spoiler effect than conservatives. That actually might be true, but I'd have to see some evidence.

In any case, that is a different argument than the one you're making, which I see to be woefully lacking in evidence.

AV's lack of use in a high-stakes general election is still a liability.

A liability? What does that even mean?

And we have seen how AV behaves in numerous competitive elections with clearly passionate voters. It does exactly what we would theoretically expect it to.

FPTP is also IRV with a 1-candidate-per-race limit or 1-seated Hare Largest Remainder. This is why I've proffered that you push first for the modification of IRV so that the first stage uses a limited form of Approval Voting by disregarding the rankings and tallying up the number of times candidates got ranked by voters to determine the three finalists.

This makes no sense. Why would I support the use of Approval Voting to pick 3 finalists for an IRV election, when I could just use Approval Voting, which is vastly superior and simpler than IRV?

If low info-voters have correlated random shocks to their preferences then strength of AV would become a liability. In BR models, AV or SV does better than other election rules, like IRV, when you introduce noise into preferences because it's better at picking out the signal. But what if the noise starts to look like signal? Then they would respond more to the noise than other election rules and their strength would become a liability.

It seems like the point you're trying to make is that, if a voter is so uninformed that his vote contains more noise than signal, then Score/Approval Voting will, by listening to all that noise, make a worse decision than a voting method which ignores more ballot data, e.g. IRV.

That is incorrect. Voting methods aren't able to differentiate noise from signal, so if voting method X throws away twice as much data as voting method Y, the total ratio of signal to noise will be greater with voting method Y.

Case in point:

A ballot is 9 parts noise to 6 parts signal. Say AV ignores 1/3 of ballot data (turns it into noise), while IRV ignore 2/3 of ballot data. Then with AV you get 11 parts noise to 4 parts signal. Whereas with IRV you get 13 parts noise to 2 parts signal.

This is all a crude simplification, but I think it addresses the point you were getting at.

I doubt that BR calculations have been made when one models correlated random shocks to voter-preferences... but show me the model and we'll talk.

The ignorance model means voters randomly are assigned some amount of ignorance, which distorts their preference away from what it really should be (i.e. you think a candidate is great, whereas if you knew him and/or the issues better, you'd support his opponent).

DLW said...

All models simplify reality and take as exogenous something that could be modelled differently as endogenous.

If voters are randomly assigned "ignorance", but the ignorances are correlated across voters then things wd be different than with the standard ignorance model. All of which is to say, it ain't necessarily so...

And if a half a loaf has a better chance of success than a full loaf, let's go for the half of loaf that most US_American progressives already understand and appreciate, instead of insisting on getting them to rally around the "best" rule.

For, like I've argued before, if IRV is flawed, including its tendency to reelect incumbents that could work to make it easier to get existing politicians who are likely to be incumbents in the next election to adopt it. For politics is the art of the possible, not the ideal...
dlw

DLW said...

My apologies for not adequately responding to Broken Ladder's well written and comprehensive responses.

1. Bayesian Regret models take the percent of strategic vs honest voters as exogenous. My point essentially was that IRV's benefits relative to FPTP are greater than BR analysis models would predict, holding all of its exogenous controls like the percent of strategic voters constant. If the amount of strategic voting is endogenous and IRV reduces the amount of strategic voting then IRV would have more oomph than BR models predict.

2. I was not aware of the extent that IRV has been adopted in place of a TTR election rule. I agree that that muddies the waters somewhat about the benefits of IRV. But to be constructive, I'd rather push for a version of IRV that essentially uses a limited form of Approval Voting in the first round to determine the three finalists. This lets FairVote, our country's de facto leader in marketing electoral reforms, save face over the unduly strong stand it has taken in favor of IRV.

3. The larger point above is that you are neglecting the need to be "political" or diplomatic towards FairVote. FairVote has an important first-mover advantage and absolute and comparative advantages in the marketing of election reforms. If instead of attacking IRV, we helped them to use AV to "fix" IRV then it would have a better chance to produce more light than heat due to rivalry over status and what-not.

4. You state that conservatives don't understand voting theory. I think they understand it well enough, they got a lot of well paid political scientists working for them, and simply don't conflate BR models with reality, which is still your cardinal sin.

5. I wrote, "IRV moves the center to the left to the extent that any increase in the population of "voters" tends to move the center to the left on average."

You wrote, "I assume that you mean, each new generation of voters is generally a little more progressive than the previous one, therefore population growth generally moves the population to the left."

Nope, If third-party dissenters are on average better educated but not terribly wealthy then they can be expected on average to prefer Democratic candidates over Republican candidates. Otherwise, voters who tend to vote for the Democratic party more than the Republican party(if we constrain the choice to these two options, as essentially happens most of the time with FPTP.) are less reliable voters and so anything that generally increases voter-turnout will tend to favor the Democratic party.

6. I wrote, "AV's lack of use in a high-stakes general election is still a liability."

You wrote, "A liability? What does that even mean?"

dlw: Something that hurts rather than helps any marketing campaign to use AV in high-stakes general elections.

to be continued...

DLW said...

I wrote: "If low info-voters have correlated random shocks to their preferences then strength of AV would become a liability. In BR models, AV or SV does better than other election rules, like IRV, when you introduce noise into preferences because it's better at picking out the signal. But what if the noise starts to look like signal? Then they would respond more to the noise than other election rules and their strength would become a liability."

BL:It seems like the point you're trying to make is that, if a voter is so uninformed that his vote contains more noise than signal, then Score/Approval Voting will, by listening to all that noise, make a worse decision than a voting method which ignores more ballot data, e.g. IRV.

dlw: If the noise appears to be signal, due to the consistency with which many angry voters prefer a Sarah Palinesque candidate over a Hillary Clintonesque candidate, then AV will give it more sway on average. AV reflects very well the voter's de facto preferences on election day. But that could be a mixed blessing to the extent that strategic manipulations of voters emotions cloud their judgement.

BL: That is incorrect. Voting methods aren't able to differentiate noise from signal, so if voting method X throws away twice as much data as voting method Y, the total ratio of signal to noise will be greater with voting method Y.

dlw: In total, it does essentially differentiate signal from noise by favoring those who are consistently preferred by many voters. My contention is that your last sentence is not necessarily true if the noise is correlated across voters.

Another line of argument would be that we need to consider both the mean fit of the election outcomes with voter preferences and the variability of the outcomes due to correlated shocks to voter preferences. The variance of election outcomes matter because it takes time to implement significant reforms and if the turnover among parties is too high then it'd be hard to get anything done. We'd pass HC reform in one two-year period and then repeal a good deal of it in the next two-year period. All of which is to say that simply matching election outcomes to whatever de facto voter preferences are on election day does not end the ongoing debate on electoral reform.

You wrote: Case in point:

A ballot is 9 parts noise to 6 parts signal. Say AV ignores 1/3 of ballot data (turns it into noise), while IRV ignore 2/3 of ballot data. Then with AV you get 11 parts noise to 4 parts signal. Whereas with IRV you get 13 parts noise to 2 parts signal.

dlw: I believe it'd be AV wd have 6 parts noise to 4 parts signal and IRV would have 3 parts noise to 2 parts signal. But if the extent of the noise is endogenous, ie because voters do less homework on their less-favored candidates, then the ratio of signal-to-noise for IRV could be higher than it is for AV. And if noise is correlated across voters then the sum of its effects would be nonlinear or the negative effect on election outcomes of twice as much noise would be more than double.

thankyou for your well written replies and passion on this issue. I hope we do get a wider selection of election rules on the menu soon. But I think it would be more strategic for you to advocate initially for the use of AV in primaries or in the first round of an IRV, conceding FairVote its' de facto leadership position in US electoral reform activism.

dlw

broken ladder said...

"My point essentially was that IRV's benefits relative to FPTP are greater than BR analysis models would predict, holding all of its exogenous controls like the percent of strategic voters constant."

As far as I can tell, voters are about equally strategic with IRV. For instance, I have called the Australian Green Party, and a phone rep said that one of the questions he gets asked most often is, "why should I waste my vote on the Green Party?" Kind of hilarious in a way, since they've been using IRV since 1918.

And then there's this analysis of Australia.

Also, we did a series of Bayesian regret calculations to test just this very thing. We see the voting methods with 100% sincere voters and then witha 50/50 mixture of sincere and strategic voters. Score Voting with 50/50 does much better than IRV with 100% sincere.
ScoreVoting.net/StratHonMix.html

You still haven't shown me any evidence that our Bayesian regret calculations aren't sufficiently realistic to be reliable.

broken ladder said...

Also, you just don't understand FairVote. We have tried to be "diplomatic" with them. It doesn't work, because they have absolutely no interest in seeking the objective scientifically supported truth.

Their goal is to get proportional representation, via STV, in the USA. And they see the adoption of IRV as a necessarily stepping stone to get there. Complicating it with some Approval Voting modification is of no interest to them whatsoever. They want IRV, so they can get STV. That simple.

Remember, they began as "Citizens for Proportional Representation".

DLW said...

BL:voters are about equally strategic with IRV.

dlw: if there are umpteen parties then many voters, due to their low interest in politics(which is often necessarily a nasty business) and opportunity costs to their time and energy, are going to be strategically low-info wrt many of them, regardless of what election rule is used. All of which is to say, I don't think you can blame voters' strategies strictly on the use of IRV. It also is a flawed inference to presume that evidence of strategic voting disproves my hypothesis that IRV serves to reduce the amount of strategic voting. It simply doesn't follow...

As for the Australian parties always ranking themselves first in the pre-prepared party-line votes. This is strategic vote-ranking by parties in response to the unreasonable requirement that voters rank each and every party for an IRV. That isn't the IRV that is being pushed by FairVote for the US.

BL:"we did a series of Bayesian regret calculations to test just this very thing. We see the voting methods with 100% sincere voters and then witha 50/50 mixture of sincere and strategic voters. Score Voting with 50/50 does much better than IRV with 100% sincere."

dlw: But the focus of your activism is on the use of Approval Voting, not RangeVoting. Approval Voting with a 50/50 mixture has a BR of .2. IRV3 with 100% sincere voting has a BR of .237. Thus, the difference is not that great... The BR of FPP with 50-50 is .644. Thus, if IRV replaced FPTP and thereby succeeded in reducing the amount of strategic voting, it would lower the BR potentially by a great deal.

If RV is so great why do you push so hard for AV? The answer is that you too need to be strategic in your advocacy for electoral reform, less motivated by what theory predicts would be ideal than by what would be feasible. FairVote is the same way, they just have different judgments about what is feasible.

BL: "You still haven't shown me any evidence that our Bayesian regret calculations aren't sufficiently realistic to be reliable."

dlw: Since you're the one trying to overturn what is the status quo understanding (just do a search of how often ""Bayesian Regret" and electoral" pops up in a Google Scholar search. The answer is 18 entries. That is a pittance relative to the 703,000 google scholar hits from electoral.) This implies that the burden of proof is on you to show BR calculations are realistic enuf to be determine what should be the priorities for electoral reform.

DLW said...

BL:voters are about equally strategic with IRV.

dlw: if there are umpteen parties then many voters, due to their low interest in politics(which is often necessarily a nasty business) and opportunity costs to their time and energy, are going to be strategically low-info wrt many of them, regardless of what election rule is used. All of which is to say, I don't think you can blame voters' strategies strictly on the use of IRV. It also is a flawed inference to presume that evidence of strategic voting disproves my hypothesis that IRV serves to reduce the amount of strategic voting. It simply doesn't follow...

As for the Australian parties always ranking themselves first in the pre-prepared party-line votes. This is strategic vote-ranking by parties in response to the unreasonable requirement that voters rank each and every party for an IRV. That isn't the IRV that is being pushed by FairVote for the US.

BL:"we did a series of Bayesian regret calculations to test just this very thing. We see the voting methods with 100% sincere voters and then witha 50/50 mixture of sincere and strategic voters. Score Voting with 50/50 does much better than IRV with 100% sincere."

dlw: But the focus of your activism is on the use of Approval Voting, not RangeVoting. Approval Voting with a 50/50 mixture has a BR of .2. IRV3 with 100% sincere voting has a BR of .237. Thus, the difference is not that great... The BR of FPP with 50-50 is .644. Thus, if IRV replaced FPTP and thereby succeeded in reducing the amount of strategic voting, it would lower the BR potentially by a great deal.

If RV is so great why do you push so hard for AV? The answer is that you too need to be strategic in your advocacy for electoral reform, less motivated by what theory predicts would be ideal than by what would be feasible. FairVote is the same way, they just have different judgments about what is feasible.

BL: "You still haven't shown me any evidence that our Bayesian regret calculations aren't sufficiently realistic to be reliable."

dlw: Since you're the one trying to overturn what is the status quo understanding (just do a search of how often ""Bayesian Regret" and electoral" pops up in a Google Scholar search. The answer is 18 entries. That is a pittance relative to the 703,000 google scholar hits from electoral.) This implies that the burden of proof is on you to show BR calculations are realistic enuf to be determine what should be the priorities for electoral reform.

DLW said...

BL:"Also, you just don't understand FairVote. We have tried to be "diplomatic" with them. It doesn't work, because they have absolutely no interest in seeking the objective scientifically supported truth."

dlw: If you got more scientific journals to support the use of BR analysis as "objective" that statement would be valid. But it currently is not a valid statement. Instead, IMO, the facts on the ground support that FairVote is right in their focus on strategically marketing specific electoral reforms, rather than trying to under gird analytically a specific reform as optimal based on first principles. Why? BR doesn't command assent from both political scientists and voters at large. Fair Vote (and Fair Vote MN) have chosen to focus on voters' existing frames of what would constitute a "fair" election. They may use marketing slogans that state tendencies as absolutes, but that is justified by the exigencies of marketing to a general population who is largely ignorant of the electoral reform debate.

BL:"Their goal is to get proportional representation, via STV, in the USA. And they see the adoption of IRV as a necessarily stepping stone to get there. Complicating it with some Approval Voting modification is of no interest to them whatsoever. They want IRV, so they can get STV. That simple."

dlw: If you don't try then you'll never succeed. The use of IRV in a statewide election NC took 48 days. That is likely too long for it to be used widely. If you marketed the use of AV for the initial phase as a way to solve this problem then they'd need to adopt it to further their goals. It'd be a win-win but Approval Voting could win more than ranked choice voting if its' use in the first stage led folks eventually to ask how well it'd work if it were used for all stages.

bl: Remember, they began as "Citizens for Proportional Representation".

dlw: Rob Ritchie has told me by email that he likes 3-seated elections. If he recognizes that voters just don't get excited by a switch to a unicameral state legislature then he'd have to accept that our best bet is to push for 3-seated state assembly elections. And if the state legislatures continue to use single-seated state senate elections that are inevitably biased in favor of bigger parties then it'd be overall more proportional in practice to push for 3-seated Hare Largest Remainder, rather than 3-seated PR-STV, for state assembly elections. This is due to how the former is biased in favor of smaller parties. Thus, the biases of the two different rules being used would tend to cancel out more than if we used a 3-seated PR-STV that would be more proportional. (To say nothing of how 3-seated PR-STV is more complicated than 3-seated PR Hare LR, which is the closest 3rd-party friendly form of PR to the FPTP election rule.)

So I believe very much that Rob Richie/Fair Vote may come along in this coming year to endorse the strategy proposed by me via Strategic Election Reform. They may do this so long as I do not join you and other AV-advocates in the denigration of their hard work for the adoption of IRV and what-not.

dlw