[EM] Anyone got a good analysis on limitations of approval and range voting?

[robert bristow-johnson]

On Nov 7, 2009, at 8:14 PM, Warren Smith wrote:

>
> --well... there is the whole rangevoting.org website...
> my more-recent papers at
> math.temple.edu/~wds/homepage/works.html
> discuss range voting including some ways it is provably better than  
> every
> rank-order voting system for either honest or strategic voters...

okay...

> --but those are not exactly "succinct"...
>
> OK Let me try:
> 1. Range for 100% honest voters behaves better than IRV, Borda,
> Condorcet and it is pretty intuitively clear why -- strength of
> preference info used, not discarded.
>
> 2. For 100% strategic voters Borda is a total disaster as is also
> pretty obvious... far worse than approval voting...  but range voting
> just degenerates basically to approval voting, which still works
> pretty well since, e.g. it obeys the "favorite betrayal criterion."
> If the strategic voters use "I'll exaggerate on the top two" naive
> strategy (which in fact, in the real world, they pretty much do) then
> Condorcet and IRV both degenerate
> to strategic plurality voting, which is pretty obviously worse than
> approval voting,
> so range beats them.  For Condorcet systems with ranking-equalities  
> allowed,
> they might behave better with strategic voters, though.  I've posted
> on that topic before.

... so *not* always "better than every rank-order system for ...  
strategic voters".

now the relationship that citizens of a community or nation have with  
each other is not the same kind of trusting and sometimes blind  
relationships we have with members of our families and other partners  
(like business partners) that we end up trusting by default,  
essentially in our own enlightened self-interest.  in general, we  
have contracts and laws essentially intended in the best  
circumstances to "keep honest people honest".  we have locked doors,  
secret credit card numbers and passwords that we don't share with  
even people we normally trust (they have no legitimate need for  
them).  essentially we rely on a system that does *not* rely on  
individual honesty in completely opaque, private, or confidential  
situations.  what if, instead of passing the offering plate around in  
full view, everybody gets to go into a private booth with the open  
plate (and not a secure drop box with a slot) and "add" their offering?

the guarantee that you and i have (at least in the US) is to vote in  
complete secrecy and security.  one *big* difference between any US  
election and what i have read for the UK is that there are no serial  
numbers on our ballots just in case there may be a legal ruling that  
someone was mistakenly allowed to vote in an election they weren't  
eligible to vote in that turned out to be critically close and  
contested.  i believe that i read, in the UK they can actually run  
the ballots through the scanning machine and yank an errant ballot  
and uncount it.  in Saddam's Iraq, there was an election in the 1990s  
where Saddam Hussein was re-elected with 99% of the vote, but the  
ballots there were also serialized and associated with the voter when  
issued and i'm surprized that even 1% felt it safe to vote against him.

but you can't do that in the US, and there *was* a case that went to  
the US Supreme Court that even Scalia voted the correct way on (and  
interestingly, his lapdog Thomas did not vote likewise).  in this  
case someone was discovered to live on the other side of the district  
or town line in some election.  when this was discovered, the person  
(and his wife, i think) was brought to court, everyone understood and  
acknowledged that these persons weren't supposed to vote in that  
election and the judge asked the guy what his vote was so they could  
uncount it and decide the election.  to the guy's credit of  
integrity, instead of lying and telling the judge exactly opposite  
(guaranteeing that his choice is elected), he tells the judge that he  
voted by secret ballot and refuses to say how, even though he was  
ordered to by the judge.  (i dunno what his wife did.)  anyway, the  
SCOTUS upheld that right.

every place i have lived and voted in the US (ND, IL, ME, NJ, NH, VT)  
had ballots with no serial number nor distinguishing marks on it when  
it ultimately went into the ballot box.  in IL i worked as a poll  
watcher, and watched the precinct officials open absentee envelopes,  
decide on their inclusion, and if included tossed an interior  
unmarked envelope onto a pile with others.  at that point, the  
inclusion or the ballot was committed because of its anonymity.  we  
have totally confidentiality with the content of our vote, but not  
with fact of whether we voted or not.

so Warren, you're gonna sell your system on the hope that in this  
place of absolute privacy, we'll all be sincere.  that i can count on  
the sincerity of voters who are political opponents just as they can  
count on my sincerity.

i want a system that offers nothing for the voter's political  
interest for voting insincerely.  i s'pose Arrow tells us that this  
cannot be completely counted on, but, save for the case of a  
Condorcet cycle, i don't see how voting insincerely in a Condorcet- 
decided race ever serves the political interest of the voter.  now,  
you and others have suggested that having a Condorcet-decided  
election will cause voters to vote in such a way to *cause* a cycle  
(as if these voters could somehow control the resolution of the cycle  
to favor their guy?) in a manner similar to how we *know* that FPTP  
favors an adversarial 2-party system (we end up voting for the lessor  
of evils of the two major-party candidates).  i find that to be so  
unlikely (in comparison to 2-party compromising in FPTP) that it's  
unpersuasive.

> 3. The main perceived flaw in range voting is for strategic+honest
> voter MIXTURES
> and under the worrying assumption that who decides to be strategic, is
> CORRELATED with the politics of that voter.  Thus for example,
> strategic Bush voters could beat unstrategic Gore voters.  Problem
> isn't really much of a problem if the Bush strategy-fraction is the
> same as the Gore strategy fraction (a claim backed up by computer
> sims).

just like the issue of drawing random ballots in multi-winner PR STV  
and counting on that as "representative", i wonder why counting on  
computer simulation to support this balance of tactical voters would  
be as trustworthy.

> It is only if they differ.

in other words, if in some election, one side has more politically  
savvy strategists, advocates, and voters?  isn't that often how one  
side wins and the other side loses?  one side has a Karl Rove and the  
other side has John Kerry and one of his 3 serial campaign managers.   
don't count on it not differing.

> 4. I'm unaware of any evidence from the real world that Bushy and
> Gorey voters really are any different strategically with range voting.

Bushy vs. Kerry?

>    However, there is evidence that
> Nader voters are less strategic and more honest.  (Not surprisingly
> since voting Nader
> in the USA *was* unstrategic.)

sure, idealists throwing away their vote.  we want to be able to be  
idealist without having to throw away our vote.  and we want to be  
able to vote effectively without sacrificing or compromising our ideals.

>   However, the evidence from the real
> world is that all political types of range voters are substantially
> honest, i.e.only a small fraction vote approval-style, and this causes
> Nader, despite this relative disadvantage, to do a lot better with
> range voting than he does with approval voting.


Warren, i know you've worked so much more in this whole area than i  
have, your experience is greater than mine, but i just cannot buy  
into your premises.

i want a system that does not *assume* or count on our sincere  
intentions, but *enables* or *supports* our sincere intentions by not  
penalizing our acting according to our sincere intent.

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."