[EM] Janet: What's wrong with IRV (or some of it ).

[MIKE OSSIPOFF]

You asked:

>Could you put in simple, layman's words (not formulas), what is so
>bad about IRV for single winner offices?

First let me tell you why the academics near-universally dislike
& reject IRV: It fails Monotonicity. That means: Say you make
out a ballot, and with your ballot and the others, candidate Smith
would win. But then you decide you want to rank Smith higher, because
he's so good, and so you throw that ballot away, and make out a new
one in which you rank Smith higher (without changing the order of
the other candidates in your ranking). The fact that you ranked him higher 
makes him lose, where he'd otherwise have won.

The voter has a right to assume that the voting system isn't going
to respond oppositely when the voter votes one way instead of another
way. Any method that can do what the previous paragraph describes is
nonsense. How can we take seriously any election result from such a
method?

But IRV doesn't just fail Monotonicity--IRV fails Monotonicity in
an especially extreme & outrageous way: A candidate who was going to
win can lose because some people moved him from last place to 1st
place! Can it get any sillier than that?

For me, a more sinister version of the above failure is: A candidate
is going to lose, but then we find out about a payoff scandal that
he's involved in, and so some of his voters move him from 1st place
to last place, and that makes him win.

But I there's another reason why some of us dislike IRV: It
retains Plurality's "lesser-of-2-evils" problem. You're probably
familiar with that problem. It's why most of your friends are voting
for Gore, though they have no confidence in his honesty, and though
they know that Nader is the honest candidate and is incomparably better.

Not only does IRV retain that problem, but there's reason to believe
that it has that problem worse than Plurality does. IRV is certainly
worse than Plurality in other ways too, and I'll get to those later.

Instead of starting out with numerical examples, let me just tell what
can happen, in terms of this year's Presidential election:

CVD likes to assure us that, with IRV, we can vote Nader in 1st place,
confident that he'll get eliminated and then our vote will transfer to
Gore. Sure, if we're sure that Nader will immediately get eliminated.
But would he? As I was saying before, most people voting for Gore
like Nader better than Gore. In fact I haven't found any exceptions.
So it's reasonable to expect that Nader will surprise you, and get
more 1st choice votes than Gore--maybe lots more. He could therefore
eliminate Gore. He probably would. But then he could lose to Bush,
if Bush has enough support.

So, because you sincerely voted Nader in 1st place, you let Bush win,
where you could have kept Bush from winning if you'd voted Gore in
1st place. If all the Nader people did that, the Nader ballots &
the Gore ballots would combine, making them sufficient to defeat Bush.
But your Nader ballot never had a chance to reach Gore, because Gore
got eliminated while your traveling vote was still on Nader.

There's something sadly familiar about that plausible story: It's our
old lesser-of-2-evils problem. Maybe you've heard the CVD line about
IRV having a track record in Australia. If you've been on this list
for a while, then you know that Australian small-party members are
reluctant to vote their small party favorite in 1st place, because they
want to vote a lesser-evil, one of the big-2, in 1st place. So much
for IRV's track record. Additionally, parties are reluctant to run
more than 1 candidate. Do they fear a potential winner will be
eliminated by another candidate whom they run, throwing the election
to another party?

Since this problem might very well happen in November 2000, with
IRV, there's nothing far-fetched about it.

CVD sometimes claims it would be rare. But, as I was saying the other
day: Say there are 3 candidates, Favorite, Middle, & Worst, and
they're all likely to be _roughly_ the same size, support-wise.

Say Middle is actually close to Worst than to Favorite, so his ballots
are going to transfer to Worst if Middle gets eliminated.

Because the candidates are roughly equal, you know that no one has
a majority. Favorite doesn't have a majority, and he can't get
enough transfers from Middle to win either. Favorite can't win.
So why should you rank Favorite in 1st place? You'd gain nothing by
doing so. But if you rank Middle in 1st place, you might prevent his
elimination, preventing Worst from winning. Voting Middle in 1st
place can give a better result than voting Favorite in 1st place, but
voting Favorite in 1st place can't possibly give a better result than
voting Middle in 1st place.

In other words, you again have the same lesser-of-2-evils need that
you have now with Plurality. And, contrary to what CVD will tell you,
there's nothing unusual or rare or special about that example.

Similarly, if there are lots of candidates, and their support tapers
gradually away from the voter-median position, then we can expect
eliminations from the extremes to send votes inward, and, if the taper
is gradual, then candidates other than the voter median candidate will
thereby accumulate enough votes to eliminate the voter median candidate.

In that way, the candidate who could beat each one of the others in
separate 2-candidate elections gets eliminated, even though he is
favorite to more people than any other candidate is.

To be continued.

Mike Ossipoff

_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.