In his response under this subject heading Mr. Lomax seemd to think that I was 
advocating Cumulative Voting, then he offered what amounted to a plausibility 
argument for my assertion that Cumulative Voting is strategically equivalent to 
Plurality.   I'm slightly miffed that he would imply that I advocated a method 
equivalent to Plurality.
 
But the main point of this posting is to offer geometric insight into the 
equivalences that I mentioned before.
 
If our ballots are vectors of numbers (to be used additively) representing 
levels of support for candidates, then some limit has to be placed on the 
numbers or the election game reduces to "who can name the biggest number?"
 
One way to limit the vectors is to limit a norm of the vector.  If we limit the 
max norm, then we get range voting, whether or not we allow negative numbers.  
If we limit the L_1 norm, a.k.a. the Hamming norm, which is the sum of the 
absolute values, then we get Cumulative Voting when only positive numbers are 
allowed, but we get something else if negative numbers are allowed.
 
The geometry is this:  In the case of the max norm we get squares, boxes, and 
their higher dimensional analogues no matter whether negatives are allowed or 
not.  The corners of these boxes are the optimal strategies which amount to 
approval strategies.  The only corners that never yield optimal strategies are 
two corners corresponding to vectors with all components equal. (all max or all 
min).
 
In the case of the Hamming norm with negative numbers allowed, we get a 
diamond, an octogon, and higher dimensional versions of these.  The corners of 
the octogon correspond to the  optimal "xor" for xor against strategies.
 
In the case of the Hamming norm with only positive numbers allowed, we get 
triangles, tetrahedra, and simplexes of different dimensions.  The corners 
(except the all zero corner) correspond to optimal Plurality strategies.
 
When we have three candidates, the approval box has two more corners than the 
xor octogon, but the two extra ones, (1,1,1) and its opposite, serve no purpose 
in approval.
 
This is the geometry behind the equivalence of approval and xor in the three 
candidate case, and their equivalence with their respective continuous versions.
 
Forest

<<winmail.dat>>

----
Election-methods mailing list - see http://electorama.com/em for list info

Reply via email to