[EM] Two simple alternative voting methods that are fairer than IRV/STV and lack most IRV/STV flaws

[Abd ul-Rahman Lomax]

At 09:32 PM 1/13/2010, Kathy Dopp wrote:

> > This sounds like a variation on Borda count, but with an 
> incentive to vote on fewer candidates.
>
>Yes perhaps, but normalized to give a value of one in total to all
>ballots since Borda was rejected by the MN Supreme court as violating
>one-person/one-vote.

No. Borda was never considered by the MN Supreme Court. You are 
thinking of Bucklin, which was simply called Preferential Voting 
then, as I recall, after the ballot. FairVote created some propaganda 
that claimed that the rejection was based on one-person, one-vote, 
but that's not supported by a full reading of the decision. And Brown 
v. Smallwood wasn't confirmed anywhere else, it was disliked even 
within Minnesota, bucking the general opinion of the legal 
profession. You can read Brown v. Smallwood on rangevoting.org, there 
is a copy there. There was an appeal for reconsideration, be sure to 
read it all.

The decision was quite remarkable. It quoted another decision, with 
approval, that it was the number of voters supporting a candidate 
that mattered, not the number of votes. And then, next breath, it 
counted the number of votes and noted that there were more votes cast 
than voters.

The principle was correct: it's the number of voters that count, and 
a voting system like Approval or Bucklin simply finds ways to allow a 
majority of voters to assemble through making compromise choices on 
the ballot. In the end, if only one full vote is effective, or none, 
should the voter not have supported the winner at all, then we have 
one person, one vote. Because we have counted the number of *voters* 
supporting the result, not the number of votes, per se.

Bucklin, with a majority requirement, simulates what happens in 
repeated single-vote elections, which is standard democratic process, 
as does Approval. Each election, if the voters want to move toward 
resolution, they will lower their approval cutoff to include more 
candidates. With Bucklin, the method does this for them, allowing 
them to participate in a limited series of such elections. If they 
want to. They can just vote for one, if they want. It depends on what 
they would prefer to see happen: completion or a runoff.

It's like IRV, in that way, but without the top-N eliminations, which 
are what cause the trouble with IRV. There are no eliminations in 
standard repeated-election, majority-required elections, there are 
only voluntary withdrawals (which can't happen with Bucklin, 
presumably, there isn't the time provided and it would do harm if 
done mid-counting) or presumably increased voter compromise and 
respect for an appearing majority and possible willingness to accept 
it and terminate the process.

That multiple votes are cast simultaneously is confused with one 
person one vote violation. They wouldn't have to be counted 
simultaneously, there could be a way to count Bucklin votes so that 
only one vote is counted at a time, it would be an iterative process. 
But why do an iterative process to just count one vote at a time, 
when you'd get the same result by counting them all at once?

(Okay, I'll describe the algorithm: just consider each pairwise 
election, and only count votes, in any round, for each pair of 
candidates. Count them as votes are presently counted, where 
overvotes void the ballot -- but these votes will be counted later, 
if needed. Is there a candidate who beats all others? Consider this 
the tentative winner, or just the winner, period, if a majority isn't 
required. If a majority is required, count all the votes up to the 
final round, for the winner. If no majority of valid ballots, then 
move to the next round of counting and repeat. If no majority after 
the last round, follow runoff rules.)

(If a method violates one-person, one-vote, surely it would produce a 
different result when only one vote is considered at a time! With 
IRV, voters cast more than one vote at a time, but only one vote from 
each voter is considered at a time. Or none. Same as Bucklin. The 
difference is in how the votes are counted; IRV is counting different 
ranks on different ballots, at once, based on having eliminated the 
higher ranked candidates on some of the ballots. And the result is 
that some votes, cast by a voter, are not effective and are passed 
over, whereas had the voter voted for another candidate in that exact 
same position, it would be counted.

An argument can indeed be made that IRV violates basic voting 
principles of equality. Bucklin doesn't, in spite of Brown v. 
Smallwood, which made its argument defectively, and, as written then, 
clearly would have applied to IRV as well as Bucklin. We know that 
some very smart lawyers were on the Bucklin side (not to mention the 
political scientists who generally loved Bucklin), but they were 
unable to prevail. And they did not have the political clout to 
follow the Supreme Court's advice: if you want to do this, get the 
constitution changed to allow it. Politics as usual, folks, it has 
little to do with what the best voting methods are.