Approval vs IRO

Mike Ossipoff dfb at
Wed Nov 13 23:02:01 PST 1996

This a reply to Tom's letter about Approval vs IRO (MPV, the 
Alternative Vote, Preferential Voting, Hare's method).

Sure, admittedly Approval doesn't let you express all of your
preferences. Whereas IRO lets you express them, and then
will disregard them sometimes. How often? When Plurality would
have a problem (more about that later in this letter).

Approval is a frankly non-ranked method, a method that doesn't
hide its strategy need. That doesn't gain it any popularity.
IRO is a vote-for-1 method posing as a preferential method.
The balloting is preferntial, but the count is vote-for-1
at a time, just like Plurality, which is why IRO share's 
Plurality's problem more than any other method except Runoff.

[Though I'm defending Aproval in comparison with IRO, I should
make clear that Approval isn't what I advocate. I advocate
Condorcet & Smith//Condorcet. Approval isn't as good as the
_best_ methods]

I don't agree that your vote for Middle can count against
Favorite in Approval. All it does is, it means that you're
not casting a vote between Middle & Favorite. Sure I agree
that it's better to cast a vote for Favorite over everything,
and an equally reliably-counted vote for Middle over everything
but Favorite. Condorcet allows that, & IRO doesn't. Approval
is a simple modest method. Like a hand-tool as opposed to a
power tool. Use of a hand-tool isn't as easy; it's not as
deluxe as a good power tool. But it won't fail on you. IRO,
in that analogy is like a cheapo power tool, which can
really mess up what you're doing when it predictably goes

As I was saying, in Approval, when you vote for Middle &
Favorite, yoou're not casting a vote for Middle over Favorite.
You're merely not casting a vote between that pair. This
is in dramatic contrast with IRO & Plurality, where, when
you give it to Middle, as you must if you're to protect
Middle's ability to beat Worst, you indeed _are_ casting a vote
for Middle against Favorite. The problem is when Favorite voters
misjudge & give it to Middle, when Favorite otherwise would
have had a win. When that happens, Approval, in comparison with
IRO & Plurality, requires twice as many suckers to give away
the election!

To reply to another statement: Sure Approval needs strategy.
Non-deluxe methods need defensive strategy. So does IRO,
and, from what I've said so far, IRO puts voters in a worse
strategy position. Worse consequences for wanting to reliably
help Middle.

But Approval's strategy isn't to vote for everything that
you Approve, not if you want to optimize your outcome.
You vote for the candidate you're likely to need as a compromise,
the same one you'd vote for in Plurality. The best (from your
viewpoint) whom you can make win, and who needs your vote.
But, unlike Plurality, you also vote for everyone you like
more. With Approval, but not with IRO or Plurality, you
_never_ have to abandon voting for your favorite.

Sometimes can't vote Favorite over Middle? True, but, as I
said, at least you never have to vote Middle over Favorite.

In a 3-candidatre race, Middle voters have no reason to vote
a 2nd choice. If Middle isn't Condorcet winner, then either
one extreme or the other has an outright 1st choice majority,
in which case it doesn't matter what Middle voters do. If
neither extreme has a majority, then Middle is Condorcet
winner, and the rightful winner, and one side or the other
needs Middle to beat someone less-liked. And a victory for
that less-liked alternative would be worse for the opposite
extreme than for Middle voters. The combination of that
better strategic position, and the better ethical position
when the people who know they probably need Middle know that
therefore Middle is the rightful, Condorcet winner, that
adds up to a strong advantage for Middle in that situation.
Middle voters have no reason to vote a 2nd choice.

(Sure, there's the possibility of "principled hold-out",
where Greens could publicize that they'll refuse to vote
for the Clinton, out of principle, but that's getting into
psychological strategy, rather than the immediate outcome
optimization that we're talking about).

The close race you spoke of, where Favorite might be big
enough to eliminate Middle & elect Worst: You said it's
a problem in Approval too.. Alright, now's the time to
talk about that situation:

Why IRO is no better than Plurality:

When Middle might be smallest, either 1 extreme side or
the other has a problem. The IRO advocate will say that
problem only exists if Middle might be smallest. Yes, but
that's the only time when Plurality has a strtategy dilemma.

Why is that? If Middle is known to not be smallest, then that's
because Middle is known to be bigger than Worst, or Middle is
known to be bigger than Favorite.

If Middle is known to be bigger than Worst, then Plurality has
no strategy dilemma, since Middle doesn't need the votes of
the Favorite voters in order to beat Worst. If Middle is known
to be bigger than Favorite, then again there's no strategy
dilemma for Favorite voters. They obviously aren't going to
succeed in electing Favorite, so if Middle might need their
vote to beat Worst, then they give their vote to Middle. No
reason not to (again we mustn't bring principled holdout
into this, since we're talking about immediate outcome 

So when IRO doesn't have a strategy dilemma, then Plurality
doesn't have one either. IRO isn't better than Plurality
when it comes to strategy dilemma.

What makes IRO seem better? The fact that you can safely vote
for Favorite if Favorite is a sure loser. Oh thanks a lot!!
The time when single-winner reform can bring genuine tangible
benefit is when something better than Middle can win. And
that's exactlyk when IRO fails. Would you buy a vacuum cleaner
that is guaranteed to fail just before it finishes the carpet?
If it's either that or a non-electric carpet-sweeper, you'd
get the carpet sweeper (my hand-tool analogy). Better yet,
get a _good_ vacuum, but we're talking about IRO vs Approval.

Unlike IRO, Approval is clearly an obvious improvement over
vote-for-1 Plurality, since you never need abandon your favorite,
and since you never need to cast a vote for anyone against 
your favorite, as you do in Plurality & IRO.

If Approval is better than Plurality and IRO isn't, then
Approval is better than IRO.

Additionally, I've sometimes compared these methods by
naming a "lesser-of-2-evils guarantee" of Approval:

If a full majority of all the voters vote A over B, that
automatically makes it impossible for B to win, without
anyone having to vote a less-liked alternative over a
more liked one.

Approval has that guarantee. IRO has nothing of the kind.


I've discused this comparison in terms of 3 main issues:

1. Comparing both methods to Plurality
2. How many suckers it takes to give away an election
3. Approval's lesser-of-2-evils guarantee

Of course #2 is a consequence of #3.


Small parties here say "vote for us", whether they're winnable
or not. With Approval, people would of course never have reason
not to, even though the compromise voter wouldn't be casting
an effective vote for Favorite over Middle. With Aproval, manyk
small-party voters would be more likely to also vote for 
the Democrat if they knew that their favorite didn't have a
win. Then, later, when their favorite might have a win, sure
they have a dilemma, but the erroneous compromisers won't
do as much harm as they would in a 1-vote method like
IRO or Plurality.


It doesn't seem to me an embarrassment for electoral reformers
to say that STV, while good for PR, is no good for purposes
for which it wasn't designed, like single-winner elections.

Use the analogy of not washing your face with a wire scrub
brush soaked with carburettor cleaner. Use the right tool
for the job.


I like your wording for the Smith Criterion. That wording
is likely to be clearer to people than my separately defining
that criterion and then saying to use Condorcet to choose
among the alternatives whose election wouldn't violate
that criterion. For one thing my wording was a 2-step
wording, and you made it 1-step. For another thing, 
people might be saying "How do you find which ones' election
would violate that criterion; that sounds abstract &
not like a direct instruction." So yes, thanks for lthe
constructive Smith Criterion wording suggestion.




More information about the Election-Methods mailing list