Truncated ballots

[Rob Lanphier]

On Tue, 7 Apr 1998, DEMOREP1 wrote:
> Regarding Mr. Eppley's posting of Mon, Apr 6, 1998 3:29 PM EDT---
> 
> I mention some more from the 1996-1997 discussion-
> 
> The range of approval on a choice is from plus 100 percent (total approval) to
> minus 100 percent (total opposition).
> 
> Some voters (perhaps even a majority of the voters) just might think that
> candidates such as Hitler, Stalin, Saddam, etc. are ABSOLUTELY unacceptable,
> regardless of their relative evil.

I largely agree with Steve on this issue, but I think I see a little bit
of merit in this argument, and I think I understand why you're dug in on
this issue.

Though I think by far and away relative preference is the *most*
interesting and fair thing to measure in an election, there may be ways of
simultaneously measuring absolute preference to the level of granularity
that DEMOREP seems to want, using any arbitrary ranked ballot system.

(Note: this idea is not original; Steve or Mike came up with it ages ago.
I'm just expanding on it).

Basically, use the ballot that DEMREP is suggesting, ranking all
candidates on a scale of -100 to 100 (or whatever).  Treat the numbers
-100 through 100 all as "candidates", and factor them into the tally using
a ranked ballot method (if you ask me, Smith//Condorcet).  The highest
number that the candidate "beats" is their rating.

This metric could actually be useful data, especially if there is some
semantics tied to the number.  For instance, a candidate who wins and
gets a 70 or higher could get a 6 year term, while a candidate who only
gets a positive score under 20 could get only a 2 year term.  Furthermore,
a winning candidate who gets a negative score might actually be only
elected "interim governor" for a 6 month term with limited powers while a
new election is being held.  Perhaps a dreadfully low score of -50 in a
"winning" candidate could be a plea for U.N. intervention or a
constitutional convention to change the wacky voting system.

The great part about using something like Condorcet for assigning the
numbers is that it reduces the incentive to insincerely rank the
candidates low.  If, for instance, without your vote, a particular
candidate would have gotten a 56, then your vote is no different whether
you've given that candidate a 55 or -100.  Essentially, your vote has the
effect of either raising the candidate's number a fixed amount over the
previous consensus, keeping it the same, or lowering it a fixed amount,
but without having to know what the consensus of everyone else is
beforehand.

Even without the semantics tied to the number, the data would still be
useful.  One could say that "geez, Smith//Condorcet is already complicated
enough without this addition", and I would agree.  This method wouldn't
stand a snowball's chance in hell of being accepted as a method for use in
general elections in the near future.  However, I think that there is
other applicability of such a method is very powerful, and could be used
in different settings, such as computer programs "voting" in parallel
computing applications (artificial intelligence applications using
competing evolutionary program snippits leap to mind), and in
political/technical settings such as standards bodies where the
participants don't instantly recoil at solutions that sound complex.  More
on this in a bit.

Rob Lanphier
robla at eskimo.com
http://www.eskimo.com/~robla