[EM] falsifying voters' rankings--no.

[Michael Rouse]

----- Original Message -----
From: "MIKE OSSIPOFF" <nkklrp at hotmail.com>
To: <election-methods-list at eskimo.com>
Sent: Tuesday, April 09, 2002 7:37 PM
Subject: [EM] falsifying voters' rankings--no.
> Rouse continues:
>
> 3. We get a net gain in social utility if we replace truncations and ties
> with the order preferred by the majority of voters who expressed a
> preference.
>
> I reply:
>
> You don't replace my ranking, the way I voted it, with anything.

Why? If you double-vote in a plurality election your vote is completely
silenced, and if you undervote your "truncation" is tossed out.  If you make
a rule that any truncations or ties on an ordered ballot shall be ranked  by
majority preference, that violates your vote far less than completely
throwing out your vote if you accidently rank two people the same -- if you
are an elderly Floridian accidently ranking Gore and Buchanan the same on a
ballot, you would probably be happier  knowing that your "tie" still ranked
Gore ranked above Buchanan.

>
> You're suggesting modifying someone's ballot without their permission,
> counting them as having ranked Smith over Jones, when they've ranked
> neither. It isn't very democratic to count preferences that someone
> didn't vote.

If a hundred people prefer candidate A and two hundred don't care which
candidate is elected, candidate A should be elected -- it really doesn't
matter if the polls show one hundred votes for him or three hundred. If I go
to the polls but refuse to vote between two candidates, I should not
complain when one is chosen above another. Giving a tie or a truncated vote
more influence in an election than a fully-ranked ballot isn't very
democratic, since it punishes voters for having complete preferences.

Just as an example where one person's order can take precedence over two
people with no preference, suppose there are three people planning a party
and they want to offer two types of ice cream for dessert. So far, they have
narrowed it down to three flavors: Vanilla, Strawberry, and Chocolate.
Voter 1 likes Chocolate and Vanilla the best, and doesn't care for
Strawberry.
Voter 2 likes Vanilla the best, Strawberry second, and Chocolate least.
Voter 3 likes Strawberry first, and has no preference between Vanilla and
Chocolate.

We thus have the following ordering:
(C,V) S
V S C
S (V, C)

Now, what two flavors should we serve? Well, two people have expressed no
preference between Chocolate and Vanilla, but one person likes Vanilla a
whole lot more. It makes sense that choosing Vanilla over Chocolate will
make one person happier without affecting the other two. We can then rank
them like the following:
V C S
V S C
S V C

So we see that Vanilla and Strawberry should be the two flavors we choose,
whereas it would have been quite easy to fall into a circular tie otherwise.

>
> Of course, with Condorcet(margins) there'd be no reason not to to
> what you suggest. But if you did it with Condorcet(wv), you'd
> reduce Condorcet(wv) to the the merit-level of Condorcet(margins).
>
> Mike Ossipoff

I'll need to look closer at Condorcet (wv) to form an opinion, though I
don't doubt it's an excellent method (I generally like Condorcet methods
anyway, though Approval is probably as complicated as the electorate can
handle). Still, I will say that rewarding ties, truncations, and masking of
voter preferences over full and complete preference lists would seem to be a
flaw in an election method. Better to punish for a lack of relevant
information rather than too much.

Michael Rouse
mrouse at cdsnet.net