[EM] The worst about each system; Approval Preferential

[Abd ul-Rahman Lomax]

At 10:39 PM 5/26/2010, Kathy Dopp wrote:
>Yes. I agree with your scenario of not needing a runoff Abd ul.

You get the basic system you suggested simply by changing the line in 
the election code that prohibits counting more votes for an office 
than there are office-holders to be elected. One vote per voter per 
candidate (maximum!) is Approval. And so then you simply use Approval 
voting in a top-two runoff election. It does no harm and sometimes it 
will find a majority. Costs nothing extra to count except for the 
actual extra votes, which won't be a lot, usually. Enough to fix a 
spoiler effect, sometimes. A bit of a push towards a majority.

Bucklin is better because it allows people to express a first 
preference and gives that first preference a chance to win before 
additional approvals are brought in.

I believe that the fact is has an honorable history in the U.S. may 
help get implementations. It's precinct summable, no problem. You 
just count it as if it were three elections, same candidates. First 
choice election, second choice election, third choice election. And 
it's just sum of votes that you need in each rank. So three-rank 
Bucklin sums three ranks for each candidate. Theoretically, you 
wouldn't have to count the lower ranks until you know you need them, 
but I think it's rude to ask people to vote and then not count all their votes!

But the ballots themselves can also be used for Condorcet analysis, 
and even more so if a higher-resolution Range ballot is used. (For 
the Bucklin part, same idea as Bucklin, you add in the votes in each 
rating, starting at the top, until you have found a majority, or you 
have counted all the votes.)

Original Bucklin allowed, most implementations, the voter to rank as 
many candidates as desired in the third rank, but only one in first 
and second rank. Most voters wouldn't use it, but I can see no good 
reason to prohibit overvotes in those first two ranks. The only 
problematic kind of overvoting in Bucklin is where the voter votes 
for one candidate in more than one rank: those extra votes must be 
eliminated or the voter would have cast two votes for a candidate!

Original Bucklin counted them, I think, in the highest expressed 
rank. There is another option I think is slightly better, it actually 
improves the method a little, allowing more flexibility of 
expression. But I gotta go....