Centerfield

Hey, welcome to the front page, John!

I'm glad to see some success with IRV. But I think that change of this magnitude should be undertaken slowly. Let local communities scattered in different states around the country try it and get used to it and work out any kinks first. Then let one or two states try it. See what effects the new system actually has on the outcomes of elections before any wide-scale implementation.

A 2-party, winner-take-all voting system has served us very well for over 200 years. We are the most stable democracy in the history of humanity. Making fundamental changes to our electoral practices should be done slowly and with great care.

With that caveat, I like the concept of IRV, and I think it could help bring about a little more moderation. Time will tell, though. Reality often turns out differently than theory predicts.

Posted by: PatHMV at March 7, 2006 11:59 PM

I'm very suspicious of the idea of using a different scoring rubric, but I'd be willing to see what the "research" shows. My main worry is that such mechanisms wouldn't always be well understood. This leaves open the possibility, no, actually the strong likelihood IMO that in some instances people would not feel the outcome was just. Or trustworthy.

I do have to say that I'm not at all convinced that any of the alternative rubrics is somehow inherently more fair. For example, it's clear that the rubric you are advocating will have the effect of allowing voters to feel safe about voting for fringe candidates. That's a pretty fundamental change, and one that I'm almost certain the majority of Americans views as either undesirable or very risky.

So it's a good thing that various municipalities can act as guinea pigs for a substantial amount of time.

Posted by: bk at March 8, 2006 09:07 AM

"I can't think of any real drawback to IRV over the current system."

What about counting the ballots? That is, when a candidate gets dropped at each stage, how do you assess what that candidate's voters' next preferences are? There are essentially two options, neither of which are good. One is to keep all voter level data at the precinct level and at each stage have the central Board of Elections (BOE) contact all precincts with who was eliminated and request that candidate's transfer votes. The other is to transfer the data from each individual ballot to the BOE. Both of these become much harder to implement as the number of voters and area covered increases.

As far as IRV helping moderates. Consider a hypothetical election with a far left, far right and left (right) of center candidate. Now if the leftward vote splits between far left and left of center while the rightward vote remains in a bloc, then if more leftward voters go far left, the left of center candidate would be first eliminated, even if the leftward voters unanimously agree that left of center is preferable to far right.

There is a voting method that solves both these problems. Under majority rule voting, voters cast their votes almost exactly like they do under IRV. However, unlike under IRV, instead of counting first choice votes, majority rule pairs each candidate against every other candidate, thus counting how they do relative to each other. Thus the data transfer requirement between the precincts and the BOE is one number per precinct per pairing of candidates. Further, in the far left/left of center/far right matchup, rightward voters would prefer left-of-center over far-left and vice versa so that any amount of voters ranking left of center first would secure a left of center victory.

As for the difference for the voter, it occurs for the voter who gives more than one candidate the same rank. Under IRV, that voter's ballot is void. Under majority rule, the only part of that ballot that's ignored is any pairing involving candidates with the same rank. Can't think of any advantage to IRV on that score.

Posted by: Scott Smith at March 8, 2006 09:28 AM

We attempted to adopt the Louisiana Primary in Washington State to replace our open primary that was thrown out by the courts. Moderates were for it, but the two parties argued successfully that it was a violation of the "right to associate."

I like the "Cajun" primary better than I do any closed primary, but how does it allow voters to get their first choice? It seems to me that a lot of people don't actually vote for the candidate they prefer because of the strenghth of the two major parties, and I for one would like to see a system that allows for inclusion of a more diverse range of views. IRV seems to be one way and I am glad to see it tested in Vermont. I don't know if it is the best way, but I think it is something that should be looked into and addressed further.

There is a voting method that solves both these problems. Under majority rule voting, voters cast their votes almost exactly like they do under IRV. However, unlike under IRV, instead of counting first choice votes, majority rule pairs each candidate against every other candidate, thus counting how they do relative to each other. Thus the data transfer requirement between the precincts and the BOE is one number per precinct per pairing of candidates. Further, in the far left/left of center/far right matchup, rightward voters would prefer left-of-center over far-left and vice versa so that any amount of voters ranking left of center first would secure a left of center victory.

Some research would help that point if it is in fact correct. My experience in Washington is that under majority rule voting, the grassroots have the advantage; hence, the more conservative and liberal candidates win. One obvious solution is for centrists to be more organized, but I don't know how possible that is is we are split up amongst two parties. I have no problems with ballots that have the same ranking for two candidates being thrown out. It would seem to me that it wouldn't be too difficulat to prefer one candidate over the other, but again, more work and research is needed.

Posted by: Mathew at March 8, 2006 12:02 PM

Mathew, it's a mixed bag down here. There's a lot of gamesmanship and theorizing that goes on around election times about casting various votes to "manipulate" who winds up in the run-off.

What it mostly rewards is unity. If there's only one relatively moderate candidate in the race, flanked by two extremists, that moderate is probably going to get in the run-off with a mixture of R and D votes. But if there are 2 or 3 moderates running, along with 2 candidates with strong appeal to the "base", then the 2 base guys are in the run-off. But if you have 2 "base" candidates on one side splitting that vote (often by having a white and a black Democrat both on the ballot), then it could be a moderate and the opposite wing in the run-off.

My own rule of thumb is that 20% of the vote is guaranteed to the right-most candidate and 20% to the left-most candidate. The 60% in the middle of that is the biggest contested area. If you can carry your 20% base plus pick up just 8 to 10% of the rest, you're almost guaranteed to be in the run-off if there's 4 or 5 non-fringe candidates running.

A whole lot of it comes down to who decides to run, and it certainly does result in less party cohesion. The state party structure is almost immaterial in elections down here, in fact. (As a result, I believe, our legislature is nowhere near as viciously partisan as Congress and most of the rest of the states.) It's frustrating when a good Republican has to "vote for the crook" because a former KKK wizard has managed to get just over 20% in a primary election which had several more moderate options who split the vote amongst themselves. Acting independently in a separate primary, Republicans in La. would never have elected David Duke as their candidate for governor.

Posted by: PatHMV at March 8, 2006 12:23 PM

Duke's point of view was certainly, well, diverse.

Posted by: Mathew at March 8, 2006 12:41 PM

Matthew:

My experience in Washington is that under majority rule voting, the grassroots have the advantage; hence, the more conservative and liberal candidates win.

It is unlikely that what you're refering to is majority rule. Majority rule basically means that the winning candidate is preferred by a majority over any other candidate in a one-on-one matchup. Thus, if the conservatives and liberals both turnout, they'll cancel each other allowing moderates to give the victory to the centrist candidate.

Some research would help that point if it is in fact correct.

There is a body of research on the topic, albeit theoretical because no political entity has adopted such a method. For an introduction, check out The Fairest Vote of All in the March 2004 issue of Scientific American. You can find more information on wikipedia under the subject heading "Condorcet."

I have no problems with ballots that have the same ranking for two candidates being thrown out.

So a voter who prefers both A and B to C but is indifferent between A and B should have his preferences of A to C and of B to C disregarded? It's one thing if there was no way to meaningfully count the vote, but to do so when there is a method which would capture the clearly expressed preference of the voter?

Pat:

What it mostly rewards is unity. If there's only one relatively moderate candidate in the race, flanked by two extremists, that moderate is probably going to get in the run-off with a mixture of R and D votes.

Under majority rule, all of that becomes irrelevant. A moderate voter would be able to rank all moderate candidates above all non-moderate ones, a conservative voter would be able to rank all conservative candidates above all non-conservative ones, etc. It wouldn't matter how the first choice votes would distributed within those groups, just how the candidates with fare relative to those outside.

Posted by: Scott Smith at March 8, 2006 03:01 PM

I can't think of any real drawback to IRV over the current system.

Well, the voting public struggled with the butterfly ballot so... Posted by: c3 at March 8, 2006 03:42 PM

Well, my basic reason for starting this post with such enthusiasm about IRV was that it allows candidates who have a different agenda that combines the positives of the left-right agenda into one ticket.

Come on, how often to we more centrist-minded cooperative people wish we could see a strong candididate taken seriously who fits the description of the ideal centrist fleshed-out on the Centrist Coalition page and many middle-road advocate books? Someone who'll be progressive on social agenda but be a strong fiscal conservative interested in shaping up the budget priorities?

Posted by: John at March 8, 2006 04:23 PM

I favor the IRV, myself, as a start. The current system is rigged for a 2-party system. This kind of system would allow somone to vote their true choice first and then, who their second choice would be.

Our voting system has gone through many changes. Heck, it was not even a secret ballot until the late 1800's. If we are going to start using new technology to vote, then new ways of voting should be the next step. IRV could come closer to chosing a candidate that the majority of people would support over the current system. Maybe even allow a higher caliber of candidate. [Ok, that may be a reach]

Personally, I would like to see the elimination of primary vote systems and replace it with a single IRV type system. Cut down the time and cost of campaign season and could allow more people direct choice of our leaders over todays fractured primary system. Will it happen, not very likely; but I would be in favor of it.

Posted by: Jim M at March 8, 2006 06:40 PM

Here's the problem with Condorcet, from an article at Wikipedia:

Some argue that Condorcet methods and approval voting are better at selecting compromise candidates and at reducing the spoiler effect. Yet Condorcet methods may not yield a winner (see Condorcet paradox), in which case one must resort to another voting method (i.e. Borda count, Instant-runoff), or restrict the field of candidates (see Condorcet_method#Resolving_circular_ambiguities).

How is restricting the field of candidates fairer than IRV? How is reverting to IRV fairer than using IRV?

Posted by: The fairest system? at March 8, 2006 07:19 PM

I think there are good arguments for change, but I'm not willing to make major, country-wide changes based solely on theory of outcomes from various voting systems. I want to see several of the possible systems tried out at the local level, and then the state level, with real candidates and real voters and real consequences to see whether the system works in fact the same way predicted in theory.

Posted by: PatHMV at March 8, 2006 08:26 PM

Personally, I would like to see the elimination of primary vote systems and replace it with a single IRV type system. Cut down the time and cost of campaign season and could allow more people direct choice of our leaders over todays fractured primary system.

I'm all for working towards that ultimately. However, starting with that as an objective might raise unnecessary opposition. One of the advantages of majority rule is that if there are two candidates who are "entitled" to be in the race, say by virtue of winning respective primaries, then the results of the contest between those two in the general election would clearly be unaffected by the other entrants because there would be a result for the pairing between those two. The result is that over time, the public would see the primary as nothing more than an endorsement and can decide what to do based on that.

Also, you forgot to mention the lower turnout in primaries, another reason to ultimately replace primaries with a fair one-stage election.

Posted by: Scott Smith at March 9, 2006 09:55 AM

Yet Condorcet methods may not yield a winner (see Condorcet paradox), in which case one must resort to another voting method ...

Condorcet paradoxes are a possibility, but overblown as a problem. One thing that Dasgupta and Maskin show is that when there are linear preferences, ok something which is not exactly like our real electorate, the chance for Condorcet paradoxes vanishes. To illustrate a linear preference, consider asking someone for a favorite number between 1 and 10. Under linear preferences, if that person said 5, you would have no idea how that person ranks 3 relative to 6-10, but you would know that that person prefers 7 to 9. With such preferences on a left to right scale, there would be no need to worry about alternative methods.

However, even if a Condorcet paradox were to occur, there is a majority rule consistent method to resolve the paradox--successively ignore the pairing with the smallest margin until there is a Condorcet winner.

How is restricting the field of candidates fairer than IRV? How is reverting to IRV fairer than using IRV?

See the alternative I presented, which neither restricts the field of candidates nor reverts to IRV. Further, IRV would allow a candidate who is not in the Smith set, the set of candidates who would defeat any candidate outside the set in a head-to-head matchup, to win. Now explain this. If in a two-man race A beats B, how would be fair if C were to enter, take first-choice votes from A and allow B to win?

IRV also creates more incentive for insincere voting since it is relatively easy to determine if someone is likely to be eliminated early. Voters could thus change their votes to eliminate early someone who would be strong in later stages in order to help a prefered candidate. In extreme cases, it would be possible for a group of voters whose real preferences rank one candidate first to get that candidate elected by changing their votes to rank that candidate last (alright, that's pathological, but so's a Condorcet paradox). While there are chances for insincere voting under majority rule, they all revolve around Condorcet paradoxes, and it is a lot harder to know that there will be a Condorcet paradox reducing the urge to vote insincerely. So why should a special status be accorded to (possibly insincere) first-choice votes?

Posted by: Scott Smith at March 9, 2006 10:31 AM

Yeah, people who had trouble understanding the butterfly ballot will have no problem with that.

Posted by: PatHMV at March 9, 2006 12:04 PM

The thing that is forgotten is that the butterfly ballot folks also have trouble with every other kind of system too. I heard they could not figure out the new touch screen systems, either.

The butterfly ballot was not that difficult.

Posted by: Jim M at March 9, 2006 12:58 PM

Yeah, people who had trouble understanding the butterfly ballot will have no problem with that.

If that was the case, then places experimenting with IRV would have had an increase in vote spoilage. Since there has been observed increase in vote spoilage, and it is harder to spoil a ballot under majority rule, majority rule should not introduce any more spoiled ballots. It should actually reduce ballot spoilage because overvotes would not be fully discarded.

Posted by: Scott Smith at March 9, 2006 03:19 PM

Scott, you continue to make fine theoretical arguments. I want actual examples and statistics of how it works in real life with real American voters.

"It is harder to spoil a majority ballot." That sounds fine, and I understand the theory that says so. But that's no guarantee that it is no more difficult for the voter to actually make and mark the choices they THINK they are making. Maybe the ballots aren't technically "spoiled" as an overvote ballot is under the chad system, but that's not the same as saying the ballot accurately reflects the intention of the voter.

Posted by: PatHMV at March 9, 2006 05:50 PM

But that's no guarantee that it is no more difficult for the voter to actually make and mark the choices they THINK they are making.

Can we ever know that? The only way to really know is to sample a few ballots, go to the voters who cast them and ask if how the pollworkers interpreted the ballots is what the voters meant. The question is, is there any reason to believe that any alternative system would introduce a complication that would make the voter more likely to cast a vote contrary to his or her intentions than under the current system?

Posted by: Scott Smith at March 10, 2006 09:41 AM

Scott, if you haven't noticed, I haven't been arguing against IRV or Condorcet or any of the alternative proposed voting methods.

The way we determine how they work is to try them out on a very small scale, here and there across the country FIRST. Don't jump whole hog and adopt it all of a sudden for all federal elections. Don't sell it as the end-all-be-all replacement. Sell it as having a good chance of being better, of a concept which needs real-world testing to be proven.

The real test does not come from surveys of voters. The real test is whether the electorate as a whole is willing to accept the results after they've seen it in action. Does it produce candidates which history later shows to have been the right people to meet the needs of the times? Maybe it would produce nothing but dull, bland, inoffensive people who do very little because they just want to be liked by everybody. I don't know. You don't know. Because it hasn't been tried out in the real world in American elections to any significant extent at all.

Go find some communities with young, energetic, far-sighted mayors or councilmen. Have them adopt the process for their local elections. See how it works. If they try it and there's a consensus that it really does help select better candidates, then adopt it in a few states and try it there. Then we can look at taking it nationwide.

Posted by: PatHMV at March 10, 2006 01:17 PM