[EM] Obvious Approval advantages. SODA. Approval-Runoff.

[Abd ul-Rahman Lomax]

At 12:57 PM 3/8/2012, MIKE OSSIPOFF wrote:
>It's well known that Approval is precinct-summable. Approval also 
>has the easiest, simplest, and least expensive hand-count, of any
>method other than Plurality. That means that Approval can have the 
>security of a handcount. Even if it's somehow possible to make
>a computer-count secure, the security of a handcount is easy, 
>already well-established and well-known.

Yup. Count All the Votes. Why did we ever think we shouldn't?

(Answer: Habit. Tradition. Our traditions, not world tradition. 
Approval was used for a very long time by some successful societies.)

>Of course Approval would be the obvious thing to replace Plurality 
>with, in state and national elections.

Forget national. There aren't any. Probably shouldn't be, my opinion. 
Want national presidential election reform? Return the Electoral 
College to its original intention. Electors should be on the ballot, 
not the presidential candidates. It's tricky to do it. It's not where 
I suggest starting. Start locally.

>What about local municipal elections, where Runoff is what is 
>currently in use? Plainly the simplest, most modest change would
>be to replace Runoff with Approval-Runoff. Just do Approval (instead 
>of vote-for-1) in the 1st balloting of a Runoff election, and then 
>hold the runoff as
>usual, between the top two votegetters.

If there is no majority. With Approval, there is a small chance you 
won't need a runoff that would otherwise be necessary.

>That would almost surely violate FBC. Voting for favorite, F, in 
>addition to compromise, C, in the 1st election could cause F to edge
>C out of the runoff. But maybe F can't win the runoff, but C would 
>have. So someone worse than C wins.

The final election, the voter fears, would be between F and this 
worse candidate. Of course, if write-ins are allowed in the runoff, 
it's possible to fix the situation, and if overvotes are allowed in 
the runoff too (why not), it could be harmless to vote for a write-in 
(i.e., C) and F.

I see why you call this Favorite Betrayal now, Mike. You are saying 
that you would *suppress* your vote for F in order to help C win the 
primary, or at least to get into the runoff. I'd like you to notice 
the weirdness. You think that C might not make it into the runoff, 
but you are afraid that F might. Yet you think that F can't win 
against Bad Guy. But C can. What is wrong with this picture?

Sure, it's possible, but I'd really like to see the underlying 
preference profiles (utilities, to be sure) that would set this up.

If you are afraid that F might make it into the runoff, you are quite 
likely to think that F could win. Runoff elections reverse preference 
about one-third the time. So you could be casting a vote that we 
could call, technically, Really Stupid, by not approving F in 
addition to the other compromise. I think that the idea that this is 
forced strategically is basically made up.

Sure, it's possible. But that doesn't make it a significant possibility.

>So Approval-Runoff is no good as a _destination_ method. But it's 
>still acceptable as a
>_transitional_ method: It could bring some public attention to 
>Approval, give Approval some precedent, which might enable
>its adoption for state &/or national elections. Or maybe that would 
>have to happen via the mechanism of Approval-Runoff being
>replaced by Approval in municipal elections, where the municipal 
>Approval is what provides precedent for state and national
>Approval.

I don't see TTR going back to simple Approval. People are not going 
to care about technical criterion compliance. They will care about 
real performance. What I do see is a next reform, after Count All the 
Votes, is allowing and using Ranking. Probably Bucklin, from the 
history and performance.

Local elections are the place to start, such as municipal elections, 
which are usually nonpartisan. IRV performs very badly with 
nonpartisan elections, i.e., if they are two-major-candidates, IRV 
generally produces results the same as Plurality (whereas TTR 
reverses preference in the runoff about a third of the time). IRV is 
expensive to canvass. IRV breaks down if there are three major 
parties or three major candidates. Bucklin doesn't.

Bucklin is much more efficient than IRV in collecting majorities, 
because Bucklin doesn't discard votes, IRV does. Bucklin either finds 
a majority in an early round, or it ends up Counting All the Votes.

>Not that this is very important, because Approval-Runoff, failing 
>FBC as it does, is only a transitional proposal, but the
>options available for Approval could also be offered in Runoff's 1st 
>balloting in which Approval replaces vote-for-one. But of course
>I wouldn't suggest bothering to even mention options for 
>Approval-Runoff, since it's no good as a destination method, due to its
>FBC failure.

That's your isolated opinion, Mike. That "failure" is extremely 
unlikely, as a reality. What it amounts to, though, is that *if it's 
true,* i.e., if there is a significant possibility of an FBC scenario 
presenting, the method simply defaults to what happens with ordinary 
Top Two Runoff. Count All the Votes has not made things worse.

>I merely mention Approval-Runoff because it could maybe facilitate 
>earlier adoption of genuine Approval, municipally, &/or state
>or federal.

Sure. I don't consider it optimal either.


>The name SODA refers to Approval elections in which delegation is 
>the only option. I speak of delegation as one of various options
>for Approval elections, most of which are mutually compatible for 
>availability in the same Approval election. I refer to that as
>the delegation option, reserving "SODA" for a method consisting of 
>Approval with no added options other than delegation. I'd prefer
>also offering and making available all the options I've named, such 
>as AOC, AOCBucklin, MTAOC, MCAOC (those of course include the
>options of voting ordinary ABucklin, MTA or MCA ballots too). And 
>the delegation option too.

If any opportunity presents to use Asset techniques in public 
elections -- or in any elections -- it should be seized. This is a 
voting reform that goes way beyond what any single-ballot method 
could even approach.

I've recommended it for NGOs. We had an Election Science Foundation 
election that accomplished, easily, with a single vote-for-one poll, 
what could otherwise have been impossible, electing a 
fully-representative three-person steering committee for the 
organization of the foundation. The election took a few days on-line 
until the voting period ended. We had 17 voters. How do you, with 17 
ballots, get a *fully representative* committee?

It was pretty simple, really. There were four or five candidates who 
got votes. (We could look this up and nail it down). Clay Shentrup, 
who is not an Asset expert, certainly not at that time, set up the 
election, neglecting to specify the rules, or doing so in a confusing 
way. We muddled through anyway.

I got the most votes. I had enough to be considered elected no matter 
what quota one used (i.e, Hare or Droop). I then had extra votes. The 
second vote-getter was Clay, and in third place was Warren Smith. And 
then there were two more candidates with a couple of votes.

I gave enough votes to Warren to ensure that he would win, and then 
sat on the situation for a bit, with some more votes in hand. Because 
we didn't have clear rules, I didn't know how many, but it turned out 
not to matter. It took less than a week. Clay decided to give his 
votes to one of the other candidates, creating him as a winner (at 
least with a Droop quota), and I then completed the election, so that 
all votes were distributed and thus we could have three Hare winners. 
Consider that! Hare quota, all votes used. Warren and I could have 
functioned as a short committee to do anything we decided with 
unanimity, if some problem had appeared. The remaining candidate, I 
think, had one or two votes, and explicitly approved the election.

So everyone approved the complete election, involving 17 voters and 
then a negotiation process among 5 candidates. There was one somewhat 
disgruntled voter who apparently hadn't realized the implications of 
the asset concept, that's all. I think he said, in the end, that it was okay.

I've never seen any election like it. It was effectively a unanimous 
election. It proved that the Asset concept worked, as designed. 
Small, for sure. So what? I don't see why this couldn't work on a 
very large scale. It would simply, if much larger, require some kind 
of facilitating intermediary process, and voluntary delegable proxy 
could work spectacularly. The actual voting, then, in recasting the 
votes, would possibly be done directly by the electors, per what 
their proxies recommend.

Or actual proxies could be used. It's all public, that is crucial. Do 
You Know Where Your Votes Are?

>The delegation option would be for people who want to leave it all 
>to their favorite candidate. It also automatically avoids C/D, if the
>various delegates can negotiate before they use their delegated 
>votes, and after they have the initial ballot-results, and if their 
>negotiated agreements are public
>and binding.

I've written a great deal about why I think that multiple 
designations are a waste of power. But, hey, if multiple designations 
are what it takes to get this implemented, fine! I suggested 
Fractional Approval Asset Voting to handle overvotes. That, then, 
allows the creation of a virtual committee to recast your vote, you 
can split it up, and people can deliberately decide to do that.

Frankly, though, I'd rather have clean and clear representation. I 
can call up my elector and talk with him or her. Sure, if I divided 
up the vote, I could call up each of them and pretend that I voted 
for each of them fully, I suppose. I could lie, in other words. 
What's that going to do with the rapport that makes direct 
communication really work?

What asset does, that is a game-changer, is not only to create fair 
and full representation, it creates a communications network, based 
on trust. It's like delegable proxy in that. It's just delegable 
proxy, perhaps, with a secret ballot stage to assign the initial proxies.

Now, I have someone interested in putting in the time to be active in 
politics. I can talk with this person. Would I want to lessen their 
*public voting power* to spread out the vote? What would I gain by this?

No, I'd *allow* multiple delegation, but not encourage it. 
Complicating the ballot is a poor trade-off, I think. With real 
asset, it's going to be tricky enough. It will be impossible to have 
candidate's names on the ballot (it would be prejudicial). But that's 
a problem that can be solved, there are obvious and cheap solutions 
that would pay for themselves.