Spread spectrum analysis: IR vs Condorcet

Mike Ossipoff dfb at bbs.cruzio.com
Sat Oct 26 13:06:32 PDT 1996


Yes, so IRO's fault _does_ make a difference in the reasonable
scenario of a 2-peaked distribution of voters on the political
spectrum.

One thing that I pointed out, when the subject was being debated
on ER,  the time when IRO doesn't have a problem, in a
3-candidate race (between Middle, Worst, & Favorite) is when
MIddle doesn't have fewest 1st choice votes. I then pointed out that
in that situation, Plurality doesn't have a strategy dilemma either.

That's because if Middle is bigger than Worst, then he doesn't
need your vote. If Middle is bigger than Favorite, and you're sure
of that, then Favorite isn't going to win, and Middle may need
your vote to beat Worst, so you have no practical dilemma: You
get your best possible outcome by voting for Middle.

So only if Middle might be smallest is there a strategy dilemma
in Plurality. So IRO isn't better than Plurality in regards to
strategy dilemma.

Then I point out that there's another condition that can avoid
strategy dilemma. If Favorite is big enough. How big does Favorite
have to be? In Pluralty, you need only be sure that Favorite is
bigger than Worst, to be free of strategy dilemma.

In IRO, that isn't enough. Favorite would have to have an outright
majority. That's if you don't know which way Middle's voters will
go in their 2nd choices. 

But say you knew Middle voters' 2nd choices. If you knew that
their 2nd choice is Favorite, then you've no strategy dilemma
at all. But what's the catch? If you knew their favorite were
Worsts, then you'd have to know that Favorite has a majority in
order to not have strategy dilemma. And the thing is that one
side or the other is going to have strategy dilemma, won't they.
Whether or not Middle voters' 2nd choice is known.

So that's why I say that IRO is worse than Plurality, when it
comes to strategy dilemma. Plurality is free of strategy dilemma
when IRO is, but IRO has strategy dilemma when Plurality doesn't.

***

Though that's only a 3-candidate example, the conclusion seems
valid when there are more candidates. Or at least the conclusion
that IRO isn't better than Plurality seems to hold. Since it's
been quite a while since we debated that on ER, I'm not, right
now, prepared to start a discussion about more candidates, but if
someone else wants to  start that debate, I'll join it.

***

That's why I call IRO a token symbolic reform. In terms of strategy
dilemma, it's no better than Plurality.

***

IRO advocates share the same standards & goals as Condorcet
advocates. For instance all the IRO promotional literature I've
seen claims that IRO gets rid of the LO2E problem. To show that
it does that, the IRO literature points out that a sure-loser
favorite will transfer your vote to your more winnable 2nd choice,
after the sure-loser is dependably eliminated immediately. Oh
thank you. IRO doesn't have a problem as long as your favorite
hasn't got a chance. But weren't we hoping that your favorite
might eventually have a chance? Would you buy a vacuum cleaner
guarantaeed to break down just before it finished the floor?

***

Some IRO advocates also claim that IRO is ok by majority rule, because,
after you eliminate the candidate who has a majority over everyone,
then, among the remaining two, IRO will pick the one preferred to
the other by a majority !! :-)

But even Plurality does that when there are just 2 candidates. 
Hacking it down to 2 candidates, and then becoming aware of majority
rule is what the IRO advocate does.

Say I took you to a restaurant and handed you the menu, and said
"You can have whatever you want on this menu". You point to something,
and I say. "You can't have that. Which of these other ones do you
want?"  Am I giving you your choice of whatever you want on that
menu? If I am, then IRO honors majority rule.

***

One IRO advocate rae-defined the LO2E problem so that, if a voter
doesn't mind the election of Worst, due to his failure to insincerely
rank Middle in 1st place, and if that voter therefore votes sincerely,
then the method doesn't have a LO2E problem, by that person's definition.
He asked me for "hard evidence" that IRO has a problem, and disdained
any "theoretical" demonstration of a problem. Sure, by his definition,
empirical evidence would be needed. 

Well say voters didn't care about Plurality's LO2E problem, and
so Dole won because no one insincerely voted for Clinton next month.

Is that ok? I guess if voters don't care then anything can be
called ok :-)  Someone said: I don't have a drinking problem.
I drink. I get drunk. I fall down. No problem.

True, and your neighborhood has no crime probem if you don't mind
getting mugged, just like, by that IRO advocate's definition
IRO might not have an LO2E problem.

What an effective & powerful way to eliminate any problem.

That same IRO advocate also said that the voter would resent
it if Middle won with few 1st choice votes (we're talking about
the voter who likes Favoriate best, awith Middle his 2nd
choice, and who hates Worst. This IRO advocate claimed that
that voter would rather have the hated Worst win, because the
election of Middle with fewest votes would offend that voter.

What does that voter's ranking mean? Does it not mean that the
voter would rather Middle win than Worst win? 

I told that IRO advocate: Ok then, let's hold a Presidential
election by rank-balloting, and we'll count the ballots by
Condorcet & also by IRO. Then we'll hold a 2nd election between
the winner by Condorcet & the winner by IRO. Let the people
choose. Guess which one the people will choose?

The IRO advocate never replied to that, just as he never replied
when I pointed out that IRO is the only rank-balloting method that
can fail to elect a BeatsAll winner even if everyone ranked 
him/her in 1st or 2nd place.

But then, when this list, which at that time was probably
just called the "technical list", was going to vote on 
how to deal with the splitting of the ER list, we didn't have
a voting system by which to vote among the alternative splitting
solutions. 

I suggested that we count the ranked ballots by Condorcet &
by IRO (then known as "MPV"), and then take a vote between
the winners by those 2 methods.

Well, that same IRO advocate complained that I was trying to
sneak something by, because I was proposing something that
was equivalent to just using Condorcet. He was pretty much
right about my proposal being equivalent to just using Condorcet.
Letting the voters choose was sneaking in something too much
like Condorcet as opposed to IRO.


Mike





-- 




More information about the Election-Methods mailing list