[EM] [CES #8441] "true expressivities" of voting methods

[Abd ul-Rahman Lomax]

At 11:53 AM 5/27/2013, Jameson Quinn wrote:
>Interesting, that you can usually calculate the median using 1.5 
>bits per grade. That would seem to indicate that a 3-level Bucklin 
>system such as MCA uses approximately all the info on the ballot.

As to Bucklin, historically, it usually did. I recall looking at some 
elections where the second and/or third ranks were not amalgamated.

(Rank information, within the approved ranks, was used with each 
rank, but the ranks then are subsumed or collapsed. The information 
from first rank is used to determine if there is a first rank winner, 
and then, if not, the second rank votes are elevated to be equal to 
first rank, etc. So each rank is considered, in turn, before the 
lower ranks are addressed and used.)

Following the principal of Count All the Votes, I'd strongly 
recommend that in a Bucklin election, all the ranks would be counted 
and reported, even if they are not used to determine the winner. As 
my readers may have noticed, I also recommend using a Range ballot 
(which can be a Bucklin ballot with an additional unapproved rank) 
with pairwise analysis, which would, then, also be reported.

There is value to that information for purposes other than 
determining a winner, and it's a courtesy to voters to count the 
votes they have cast.