[EM] Fwd: Re: Approval vote

[Rob LeGrand]

Doug Woodard has kindly granted me permission to post my reply to an e-mail he
sent me.  I thought it might be of some interest to list members.  Of course,
comments are welcome.

Doug wrote:
> In a message to the election-methods list a few days ago, you asked for
> comments on your support for approval voting in your campaign for election
> to the Texas state legislature.

Well, more precisely, I asked for comments on the section of my campaign
webpage that argues for Approval Voting and against IRV.  However, it's always
helpful to receive divergent views on which voting reform is best, so I thank
you for writing.  (I regret that you couldn't watch my recent TV appearance in
which I explained the advantages of Approval Voting and the defects (some
academic, some serious) of IRV.  It went very well.  The host of the show,
previously an IRV supporter, seemed convinced.)

> As Donald G. Saari has pointed out in his recent writings, the approval
> vote while (like the Borda count) very attractive if people are honest,
> is vulnerable because it sets up a direct competition between one's
> favourite candidate and less preferred choices.

I'm not exactly sure what you mean by "direct competition".  (I'd be happy to
see an election example that illustrates the problem.)  I agree that the Borda
system fails miserably when voters are free to vote strategically, but it
turns out that Approval Voting doesn't fail in this way.  One of Approval
Voting's strengths is that it elects good compromises whether voters are
sincere or strategic, unlike Borda or IRV.

> I note that this seems
> to be a feature of all voting systems which allow for more than one choice
> but do not use a transferable vote or cumulative voting.

I disagree.  In fact, all systems that use a transferable vote have similar
defects stemming from the fact that they consider only a small part of the
ballot information before they eliminate a candidate.  Very often compromise
candidates are eliminated relatively early in the running.  Here's a simple
example using IRV:

40% vote D over L over R
25% vote L over R over D
35% vote R over L over D

This example is simplified but not contrived.  It's perfectly reasonable that
Democratic and Republican voters would vote a Libertarian second as a
compromise, and most Libertarians prefer Republicans to Democrats.  Using IRV,
L is eliminated first because he doesn't have enough *first-place* votes, and
R wins the election over D.  Is this a fair result?  I don't think so.  Take R
out of the election and L beats D 60% to 40%.  Take D out of the election and
L beats R 65% to 35%; in effect, IRV ignores the D voters' preference of L
over R.  So L beats every other candidate one-on-one (and so is called the
Condorcet winner).  I'd say that's a strong reason for an election method to
choose L in this case.  An even stronger reason is this:  IRV's choosing of R
makes the D voters (40% of the electorate!) regret their vote.  If only they
had compromised by voting L in first place, they would have ended up with L,
their second choice, instead of R, their last.  An election system that
chooses L as the winner in this example doesn't cause any voter to regret his
vote.

Cumulative voting, the system that allows some number of votes per voter to be
distributed among the candidates, has been shown to be equivalent to plurality
(a.k.a. FPTP).

> Where voters are
> likely to have a gradation of strong feelings about candidates, the
> approval vote is likely to degenerate into first-past-the-post.

I encounter this criticism of Approval frequently.  I have shown it to be
false using computer simulations, assuming several different kinds of voter
strategy.  Voters who try to gain an advantage by stubbornly voting for only
one candidate tend to do poorly.  When voters tend to use non-stupid
strategies, Approval Voting quickly homes in on the Condorcet candidate.  In
the example above, the stable Approval outcome would be

40% vote only D
25% vote only L
35% vote R and L

and L would win with 60% approval.  Even when voters stick to their guns,
Approval (by definition) elects a broadly-acceptable candidate.  Approval
Voting has shown to be incredibly good at finding the best compromise
candidate whether voters are smart, sincere or anywhere in between.

> The alternative vote (instant runoff) while not perfect is probably
> the most practical improvement on plurality (FPTP) voting for the
> filling of one place. It is criticised by theoreticians because it
> can prefer extremists to a generally acceptable moderate candidate,
> but I think that in practice this weakness is seldom shown because the
> extremes are rarely symmetrical.

I disagree.  My biggest criticism of IRV, besides the fact that it would
require expensive overhauls of voting equipment and re-education of voters, is
that voters will still *often* face a lesser-of-two-evils dilemma when there
are more than two candidates who have a chance to win.  As long as only two
candidates have a chance, voters can vote sincerely without fear of regret. 
But as soon as a third-party candidate gains enough support to challenge the
frontrunners, voters must decide whether to risk voting sincerely, which could
easily end up eliminating a compromise choice on the way to electing their
last choice.  In the end, IRV does not encourage a truly multi-party system;
the lesser-of-two-evils and spoiler problems remain, even if they're not as
obvious as under FPTP.  (See Mike Ossipoff's writings for more detail on this
point.)  If a voting system uses ranked ballots, it must use the information
on those ballots more completely to insure that compromise candidates aren't
discarded capriciously, encouraging insincere strategic voting.

> Usually there is a candidate straddling
> the centre assymetrically but competing effectively with one extreme, who
> would be elected under AV but defeated under FPTP, as we saw in the U.S.
> presidential election in 2000 (assuming it was honest which may be a
> considerable assumption).

AV (which I call IRV to minimize confusion with Approval Voting) does often
happen to elect a good compromise, but it's certainly worse at this than
Approval.

> The Coombs version of AV may be preferable but I am not sure how usable it
> is where one cannot count on all candidates being ranked, although I must
> admit that I don't know much about the apparently sparse literature on the
> Coombs count. It seems to eliminate the theoretical excluded-moderate
> problem of AV.

Coombs does well when candidates are arranged on a single-dimensional
political spectrum.  Unfortunately, it's not much better than IRV at finding
compromise candidates when the situation is more complicated, and Coombs turns
out to be even more prone to manipulation than IRV.

> I am surprised that you are not more interested in proportional
> representation. It seems to me the likeliest way to get a Libertarian
> voice in a legislature.

I have studied PR also, though not as much as single-winner systems.  In my
opinion, the best PR method is SNTV (*not* STV), but I've never come across a
PR method without serious flaws.  Keeping in mind that the best single-winner
methods are much better than the best multiwinner elections, I think as long
as district lines are drawn fairly and without regard to politics,
single-member districts are good enough.  I certainly haven't closed my mind
to PR, but its disadvantages discourage my support of it.

> I think it is much more democratic than FPTP or
> any system using single-seat constituencies, produces better public
> policy results, and in its PR-STV form deals effectively with the
> problem of attack ads and negative campaigning which is such an
> unfortunate feature of American politics.

What do you mean by "better public policy results"?  Would a legislature
elected through PR be more or less likely to get mired in deadlocks than one
elected in single-member districts?

I would say that Approval Voting is the ultimate in discouraging negative
campaigning, especially when there are many candidates.  This is due to the
independence of choices; a voter's voting for a given candidate doesn't
restrict him in his voting for the other candidates in any way.  Why would a
candidate go on the attack when trying to convince voters to approve him?

Thanks for your message; you certainly brought up some interesting points. 
May I have your permission to post my reply to the election-methods list?

> I wish you luck in your campaign.

Thank you!

=====
Rob LeGrand
honky98 at aggies.org
http://www.onr.com/user/honky98/campaign.html

__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em