[EM] Approval Voting vs Instant Runoff Voting

[MIKE OSSIPOFF]

Reply continued:

I'd said:

>> >by voting someone else over him. IRVies say you can safely rank your
>> >favorite 1st. They forget to mention that that's only true if your
>> >favorite is a sure loser.
>>

Craig replied:

>>To a certain extent, it's also true of a favourite & compromise vote in
>>Approval.

No, because in IRV you take a risk if you don't vote Middle over
Favorite. You never have any incentive to do so in Approval.


>>It only works if your favourite is a probable loser.

You mean it could only be advisable or necessary if you expect your
favorite to not win. Sure, but it works just as well when you do it.
When you vote for Nader & Gore in Approval, you're fully supporting
Gore. The only difference is that you aren't giving it all away to
the extent that you must, when you need to protect Middle in IRV.



>>If your
>>favourite is neck and neck with a less favoured, but not the worst, 
>>popular
>>candidate, it's almost impossible to work out how to vote.  The 
>>consequence
>>is that everyone uses their vote in different ways, and the result is
>>largely random.

Hardly random, Craig. Each person votes according to his feeling about
what's best, or else his calculation about what's best, based on his
feeling about how he rates the candidates and about the Pij, or
the win-probabilities from which he calculates Pij, or about the polls
from which he calculates Pij.

By "random", you mean based on strategy considerations & ratings
& predictive estimates rather than simply on sincere rankings.
But IRV voting isn't based on sincere rankings either. In Australia,
based on your own posting, small party voters commonly vote a
big-2 lesser-evil in 1st place, over their favorite, saying that they
do so in order to avoid "wasting their vote".

As I said, you're berating Approval for not being the same as an
ideal rank method. Condorcet is as close as you can get to an ideal
rank method, due to compliance with SFC & GSFC. For practical purposes
it's ideal. But all you're really saying against Approval is that it
isn't Condorcet.

Sure you'll need strategic judgements with Approval that you won't
need with Condorcet. That's one reason why I prefer Condorcet, if it
were up to me to choose the voting system. Various rank-advocates
keep pointing out that Approval doesn't let you vote all of your 
preferences. Neither do nearly all rank methods, when we consider
the drastically insincere strategies that they force on the voter.

You still need defensive strategy with all rank methods except for
Condorcet. And IRV's defensive strategy is incomparably more complicated
than that of Approval or Plurality. Rank balloting usually accomplishes
nothing other than to complicate the strategy, and doesn't get rid of it.
Additionally, nearly all rank methods, including IRV, require more
drastic strategy than Approval does. In other words, rank balloting
usually just makes the situation worse, and IRV and Borda are the
best examples of that.

>>See my example of opinion poll ratings & expected utility
>>outcomes and try to decide how to cast your Approval vote.

As I said, we don't have Pij estimates. In a real election someone
would be motivated to calculate them. Maybe someone will here, for
your example. Maybe I will, but if you're considering figuring it
out, don't wait for me instead, because it might be some time before
I calculate it.

But, as I also said, it isn't necessary to use mathematical strategy.

I consider A so much better than B & C that I'd vote only for A in
Approval. Especially since his win scenarios are only marginally outside
the 90% reliability zone. If you're saying that it's a strategy
judgement, of course it is. You have to use strategy in Approval if
you're an instrumental voter. But I'm more interested in what you
_don't_ have to do, like vote someone over your favorite. You're
concentrating on the need for strategy, and I'm saying that what's
important is that there's no need for drastic giveaway strategy
that dumps your favorite.


>>
>> >In IRV, a CW can be favorite of more people than any other candidate is,
>> >and still lose. That's a very plausible IRV scenario, and is likely
>>
>>As demonstrated, this can also happen in Approval.  The fact that it is
>>somewhat less likely isn't particularly comforting.

Then advocate Condorcet to replace the Alternative Vote.

If the CW is the Plurality winner, there's a strong chance that his
voters suspect that. Voters who suspect that their 1st choice is CW
have good reason to not vote for anyone else in Approval. With 3
candidates, the supporters of the middle candidate shouldn't vote for
anyone but their favorite. Approval depends on estimates. Sometimes
IRV voters can rank sincerely, sometimes they'll be sorry if they do.
They need strategic judgement--but IRV's defensive strategy is more
drastic.

>>
>> >Not reassuring when your crucial 2nd choice compromise isn't there
>> >anymore because he got eliminated while your single traveling vote
>> >was on your favorite. Your vote will not be dropped? Your preference
>> >for Middle over Worst won't get counted. At least Approval reliably
>> >counts every preference that you vote.
>>

>>This is only true because you're forced in such a situation by Approval to
>>only express one preference - either favourite over middle or middle over
>>worst.

No, Condorcet counts your preferences much more reliably than IRV
without limiting their expression.

But no, Approval doesn't just allow you to express one preference.

You can vote Middle over Worst, and Favorite over Worst, or you
can vote Favorite over Middle, and Favorite over Worst.

Note that you can express 2/3 of your preferences in that example.

Typically, voters will be voting about half of their pairwise preferences in 
Approval.


>>IRV counts one of these preferences, Approval counts one of these
>>preferences.  I don't see how you can say one is worse than the other in
>>this regard.

As I said, Approval counts the ones that I want counted. If you
want to ensure that Middle>Worst gets counted in IRV, you often
need to do something that you never need to do in Approval: Vote
someone over your favorite. If you refuse to do that in IRV, then
IRV, not you, decides which of your preferences will be counted.

Mike Ossipoff

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com