How to vote in Approval

Forest Simmons fsimmons at
Thu Mar 28 13:18:40 PST 2002

On Wed, 27 Mar 2002, Dave Ketchum wrote:

>       I like Condorcet because I CAN say the two things I care most about:
>            Approve of the candidate I MOST like.
>            Get in on deciding between the candidates most likely to win.

>From my experience, Dave's top two requirements for a voting system
reflect what most voters would rate as top priorities.

Perhaps the simplest method that satisfies these two criteria could be
called Approval Completed Majority: You approve down to the preferred of
the two front-runners, and circle (or otherwise identify in a machine
readable way) your favorite. If it turns out that candidate X is the
favorite of more than fifty percent of the voters, then candidate X wins,
else the candidate with greatest Approval wins.

Under this method you can safely vote your favorite above all other
candidates, even if it doesn't technically satisfy the Favorite Betrayal

> Item of interest is that, among Approval/IRV/Condorcet, paper ballots can
> be identical - a box beside each candidate's name for a number or an X.

Since most people cannot be trusted to print machine readable numerals,
there should be a supply of numerals printed on stickers for applying to
the blanks of this type of ballot (for use with CR or Ranking).

Approval can be done with existing ballots, existing machines, and
existing programs (by using the "at large" option).

> In anything new, counting should be by computer - which can be programmed
> for any of these (have the IRV people solved consolidating among a state's
> precincts with 9 candidates for governor (NY had 9 people in 12 slots
> (fusion) in NY in 1998 - likely as many this year for we have 8 parties
> trying to hang on, and Libertarians and others wanting in).  Approval and
> Condorcet have manageable amounts of data to consolidate.
>       Adding to the above - how does IRV tell people what their precinct
> or county did to contribute to state totals?  Again, Approval and
> Condorcet have manageable, understandable, statistics.
> A puzzle:
>       We are told IRV has been widely used for a long time.
>       We are convinced IRV can result in real life horror stories.
>       Have any of these actually happened?

Because IRV is not effectively summable, they have a good excuse for not
revealing the data that would bring to light these horror stories.

> If there are horror stories we should be warning the media that they had
> best not get caught promoting a sick horse.
> I can see promoting Approval as better than IRV, for lack of horror story
> potential and for being easy to describe.  Seems to me that Condorcet
> should get mention at the SAME time as being even better for giving the
> voter more power, even though the description is more complex.

One way to introduce Condorcet to the public would be through Proxy
Condorcet: your favorite becomes your proxy in a round robin contest.

After observing several of these contests the voters would see that it
made little difference whether the proxies voted on every pair of
candidates or just let their preferences be inferred from their rankings
or ratings.

The next logical step would be for the voters to fill out their own
preference (or CR) ballots.

It would make sense to keep the "above line" proxy option for those voters
that didn't want to be bothered with the rankings or ratings.

Finally, neither Approval nor Condorcet is uniformly better than the other
in every respect.  The two ballot types have overlapping information, but
neither is sufficient for reconstructing the other, i.e. each contains
information that the other lacks.  For that reason, the tiny extra effort
of indicating an approval cutoff should pay dividends if used properly.
(That's the genius of Demorep's ACMA.)

In ratings, the approval cutoff can be made (by convention) at the
middle rate, i.e. half way between the highest and lowest possible
ratings, so no extra mark or virtual candidate (like NOTB) is needed.


More information about the Election-Methods mailing list