[EM] simple methods' merit limit

[Forest Simmons]

Another non-simple method that is simple enough to use in committees (as
well as in the other contexts we mentioned for non-simple methods):


Each voter submits a Demorep style ballot. [These ballots indicate
both (relative) preference order and "absolute" approval/disapproval.]

The Condorcet Winner wins (if there is one).

If no CW exists, then both the pairwise matrix and the approval scores are
made public. 

In the next round each voter is allowed to adjust his/her approval cutoff
up or down by one position (without changing the preference order).

The candidate with the highest new approval score wins.

[end of description of method]

Comments:

This method is similar to Demorep's ACMA (Approval Condorcet Most
Approval).

The extra step on the end allows the voters to make a strategic adjustment
of their "absolute" approval/disapproval marks to compensate for lack of
accurate polls, etc.

Limiting the size of the adjustment makes it harder to manipulate the
system.

The original Demorep ballot summaries provide enough information about
voter (strength of) preferences to allow voters to adjust their approval
ballots intelligently.

If secrecy is needed, the up/down adjustment is made by specifying "up" or
"down" on a slip of paper bearing some (voter supplied) PIN number to be
matched with the (voter supplied) PIN number on the original ballot.

In large scale applications the ballot totals can be aggregated at the
local level (since both Approval and pairwise methods are effectively
summable). So the PIN numbers would only have to distinguish small
numbers of voters (i.e. they wouldn't have to be globally unique).

Forest