[EM] IRV's Top Preferences Considered First Feature

[Forest Simmons]

Just to be clear, although approval runoff is uniformly better than (one
of its constrained versions) IRV, approval runoff is inferior to Approval,
because like any method that proceeds by elimination, it is susceptible to
insincere order reversals, while Approval is not.

Of course, Approval's simplicity clinches its superiority over approval
runoff.

Most of us on this EM list agree that no other method as simple as
Approval improves on Approval, and it's hard to find a more complicated
method whose additional complication is justified by some supposed
advantage over Approval.

MCA might barely fulfill the requirement.

Forest

On Fri, 13 Jun 2003, Forest Simmons wrote:

> I'm not advocating approval runoff, only using it as a method that can be
> easily compared with IRV, because IRV can be thought as a special case of
> approval runoff where the ballots are constrained, just as plurality can
> be thought of as a special case of approval in which the voters are
> constrained to mark only one candidate, otherwise resulting in a spoiled
> ballot.
>
> True, Approval doesn't allow you to rank the candidates, but that would be
> impossible with a plurality style ballot, whereas Approval is possible is
> with that style ballot.
>
> There are many versions of approval runoff, some of which do require
> repeated trips to the polls, but the one that I am suggesting is an
> instant runoff like IRV, summable in 2^N pieces of data, and using the
> same kind of ballot as IRV except the ballot cannot be spoiled because the
> voter is unconstrained.
>
> Removing the constraint of one mark per column turns the IRV ballot into a
> (dyadic) CR ballot from which approval runoff can be simulated as well as
> IRV simulates ordinary plurality runoff.
>
> In IRV one row of the ballot looks like this:
>
> Jane Doe   (1st)    (2nd)   (3rd)    (4th)    (5th),  etc.
>
> The voter marks one column to indicate the rank of Jane Doe, or else
> truncates by marking no column.
>
> In the dyadic CR ballot each column has a place value:
>
> Jane Doe   (2^0)     (2^1)    (2^2)    (2^3),  etc.
>
> The voter can mark any number of columns.  The sum of the powers of two is
> the CR score for Jane Doe.
>
> Thus an IRV ballot allowing five candidates to be ranked can be used as a
> dyadic CR ballot with a resolution of 2^5=32 since the corresponding sums
> of powers of two (from 2^0 to 2^4) range from zero to 1+2+...+16=31.
>
> Approval runoff can be simulated by using the convention that a ballot
> approves a candidate if that candidate's CR score is above the mean score
> of the candidates that remain on the ballot.\
>
> The word "mean" can be replaced with "median" or  "midrange" in the above
> explanation to get different versions of approval runoff.
>
> The number of candidates eliminated at each stage can be made to depend on
> how many candidates get above mean approval, as an example of other
> possible variations.
>
> Forest
>
> On Fri, 13 Jun 2003, Dave Ketchum wrote:
>
> >
> > I am not sure exactly what "Approval runoff" is, except here it seems to
> > involve getting voter involvement at each stage, after previous stage
> > votes have been counted.  I see this as unacceptable for public elections,
> > especially for such as mayor or governor, unless it offers some TREMENDOUS
> > advantage.  Here the first stage would be voted by each voter during
> > election day.  The next stage would have to be at some later date.
>
>
> ----
> Election-methods mailing list - see http://electorama.com/em for list info
>