Voting Stuff

[danaeris]

Went to a seminar on voting technology. Wow are we screwed.

The Diebold machines and others of that breed by different companies are TOTALLY untrustworthy. If you can vote on paper ballot, do it.

The summary I can give is:
The whole process is a mess. Even if it were attacked with careful consideration, accountability, etc., there are no perfect solutions, only educated choices which are compromises of the lesser evils.

The good news is that in California by 2006 the electronic voting machines have to print out receipts. I'm not clear, however, as to whether or not the electronically recorded vote will be the vote, or the printout will be the vote. If the former, the innovation is useless and there is no accountability.

Anyone know if there are groups lobbying for runoff voting? If I got the name wrong, what I am looking for here is voting wherein you rank your choices and if your first choice gets too few votes to be in the running, your vote is cast in favor of your second choice instead. That's probably a simplified description of the algorithm, but you get the idea. I really feel like it would solve a lot of the political problems in this country and I could really get behind such a drive.

Comments

[ebenezer.livejournal.com]

Instant Runoff Voting (IRV) is in fact a more reasonable alternative to plurality voting in a lot of ways. Though it still can be manipulated (cf. Arrow's theorem), such manipulation is more complex to carry out. Also, it does not necessarily stabilize into a two-party system the way plurality voting does. Yes, I agree that it will solve a lot of problems with voting in the US.

Re:

[auros]

No, IRV is not the more reasonable alternative to Plurality.

Approval is a good alternative to Plurality.

And in the long run, a true Condorcet method would be ideal; my preference is to use Condorcet, with an approval bar as one of the items ranked; the approval data is used to break cycles (and potentially to introduce some subtle variations that favor consensus, which I can explain in detail if you like).

I also recently came up with a way to use Condorcet/Approval ballots to do a form of proportional representation that I think is resistant to both splitting and cloning, which, if I'm thinking things through correctly, would be pretty cool.

Re:

[angledge.livejournal.com]

Is Condorcet voting the system that's used in Cambridge, MA? (http://www.ci.cambridge.ma.us/~Election/prop-voting.html)

Re:

[auros]

No. They're using Proportional Representation, which is a system in which you vote for a whole bunch of offices at once -- in theory the idea is that you can have the whole electorate vote for every slot, while still giving a minority the ability to at least elect one or two of the people who wins.

The method Cambridge is using appears to be an IRV variation, which is vulnerable to both cloning (in which a large voting block nominates several candidates, but lists them as independent or coming from different parties, then ranks them all at the top, so they take every seat) and splitting (in which "similar" candidates don't manage to consolidate first choices, and thus get eliminated earlier than they should). IRV is safer for PR than for a single-office election, but it just generally sucks. Runoff systems were devised by politicians who didn't understand the math underlying elections. They lead to situations like the recent thing the French presidential election, where too many people voted for fringe-lefties, and as a result the center-left candidate was eliminated early, leading to a runoff between the center-right guy and a far-right guy. It's not certain the same thing would've happened in IRV, but if enough of the people voting fringe-left actually voted

{random arrangement of various fringe-lefties, center-left, center-right, far-right}

rather than ranking the center-leftie second, IRV would've replicated the error.

Note that many pollsters believe that in a one-on-one match, the center-left candidate would've beaten the guy who won with ~55%.

Just to make a point...

[auros]

The classic example that demonstrate's the problem with IRV...

Current situation:

35 Republicans voting (Bush, Gore, Nader)
32 Democrats voting (Gore, Bush, Nader)
31 Greens voting (Nader, Gore, Bush)

Gore is winning under IRV. Note that Gore would win a ~2/3 landslide against either opponent.

Now two more Greens come along.

35 Republicans voting (Bush, Gore, Nader)
32 Democrats voting (Gore, Bush, Nader)
33 Greens voting (Nader, Gore, Bush)

Now Gore is eliminated first, the non-Nader-supporting Dems' votes transfer to Bush, and Bush wins.

By showing up and voting their true preferences, those two Greens 1. took the victory from the Condorcet winner (the guy who beats all other candidates in a one-on-one match) and 2. made things worse, according to their stated prefs.

Any system that can simultaneously fail both the Condorcet criterion and the Participation criterion should never, ever be used in practice.

Note that places that do use it have ended up with wacky strategy issues. Look up "how to vote card" on Google some time. (It's an Aussie thing. They delegate the strategizing to their political parties.)

Re: Just to make a point...

[admiralthrawn.livejournal.com]

Well, it does fail both of those criterion, but only in rather extreme cases (in the example, you have three evenly split parties, which reflects a lot of ambiguity in the electorate; in the US, we have a lot of reasons why we'll only ever see two strong parties, which eliminates most of the failure modes).

You also had to posit a rather extreme asymetry whereby gore supporters don't like nader and nader supporters do like gore; I suspect in practice that candidates are much more likely to pair off with A saying "vote me first, with B second" and B saying "vote me first, with A second". In fact, if you look at Cambridge (which does have IRV), we have city council people running on slates -- "vote me, then any of the other six people who support this issue/party". One guy even sent me a letter saying "I am endorsed by the following N groups; this one also endorsed A, B, C; that one endorsed B,C,D; ...".

Also note that our current system applied to your example elects bush in both cases, failing condorcet. I'm not aware of any cases where it fails participation, though. So, one could argue that IRV is less likely to fail condorcet, and only fails participation if you have three almost-balanced parties, and thus is for all practical purposes better than plurality.

Re: Just to make a point...

[auros]

Sure the example is simplified, but it describes a variety of situation that does occur, as in the French presidential election, which I mentioned above.

If you would like to break the duopoly, why would you use a system that fails as soon as you break the duopoly? IRV only works as long as the 3rd-parties are certain to lose. I want it to be possible for multiple parties to flourish and come into roughly even balance; the duopoly, and particularly the relentless rightward drift of both parties, is, I think, a big problem.

The fact that you're relying on the officials to help you strategize emphasizes that there's a problem with IRV. That's not a good thing. A voter ought to be able to decide his or her prefs on the merits of the candidates, without aid from the people on the ballot.

And the current system does not elect Bush if some of the Nader voters opt for strategy over sincerity. IRV gives them the illusion that it's safe to vote sincerely. The only way in which IRV is better than Plurality is that, as long as the phenomenon people think it fixes doesn't get fixed, it prevents spoilage. Once that problem actually starts to go away, IRV becomes a disaster.

Re: Just to make a point...

[auros]

Incidentally, I'd also note that the simplified prefs I described do in fact describe the majority sentiment among the people who would've sincerely first-ranked the candidates under Plurality. Conservative dems and independent swing voters who went for Gore almost universally regarded Nader as way-too-far-left. I personally would've said "Gore, Nader, Bush" as my sincere prefs, but most people who ranked Nader above Bush also ranked Nader above Gore (in their sincere prefs -- many of them still voted strategically for Gore).

[nathanjw]

Voting reform of any kind usually works to the disadvantage of those already in power - or, at least, those in power can't tell if they'd benefit from the change or not. So it tends to be opposed by anyone who is already elected or has a stake in the current elected officials.

By the way...

[auros]

In regard to your original question: Yes, there are some (misguided) groups lobbying for IRV. The recently-formed Citizens for Approval Voting has started trying to get organized to lobby for Approval, and to educate IRV supporters as to why they should shift support to either Approval or Condorcet.

You will never find somebody who is both familiar with the field of game theory that covers elections ("social choice theory") and also supports IRV.