[EM] [CES #8791] Upper-Bucklin naming (was: Median systems, branding....)

[Abd ul-Rahman Lomax]

At 01:51 PM 6/17/2013, Benjamin Grant wrote:

>Well, that sounds a lot like the system we have be talking about in 
>the other thread.  DAT sounds confusing to me in this context. One 
>of the Cumulatives makes the most sense instinctually to me as (if I 
>understand this correctly) we keep adding in more ranks until we get 
>enough to answer the question.  IRAV makes it seem like a flavor of 
>IRV, which in my full lack of experience seems wrong (Buckley seems 
>unlike IRV), so I guess I would vote something like this:

Some facts should be known. First of all, the system as described is 
*almost* identical to the system called Bucklin that was widely used 
in the U.S. That traditional system was different in two ways (as 
generally implemented, there were variations):

1. It only allowed equal ranking in the third rank. With our modern 
perspective, we see little reason to require exclusive ranking at the 
top rank, and less reason to require it in the second rank. It merely 
makes the system less flexible for the voter. First rank exclusion 
*might* be required because of ballot placement rules and public 
campaign funding, but there are better ways to handle this, we could 
suspect. (The basic cost of requiring exclusive ranking is that some 
votes will be "spoiled" and some voters who have low preference 
strength will be *forced to choose.* Even if that is difficult.)

2. The method *is* "instant runoff approval." That is, it simulates a 
series of repeated approval elections. In the earlier elections, 
voters may bullet vote, just for their favorite. But as it becomes 
obvious that this will not complete the election, voters will start 
to add approvals. They will do this according to an "internal 
descending approval cutoff." With a series of elections, this is a 
powerful method. The single Bucklin ballot really does simulate a 
short series, essentially three such elections with three-rank 
Approval. In a more sophisticated version, the Bucklin ballot is the 
first poll in a repeated election, and my theory is that this can 
find a *true majority* almost always in two ballots. The second 
ballot has the *huge advantage* that the voters get another look, 
more motivate voters may show up to vote, and, if the elimination 
involved in listing candidates on the second ballot is fair, there is 
less that voters need to look at.

IRV does simulate runoff voting, but a different kind of runoff 
voting, called sequential elimination, where the candidate with the 
least votes is eliminated from the ballot with each round. It's also 
called "exhaustive runoff." So the name "instant runoff" is fair for 
both methods, but IRV is "instant runoff plurality," whereas this 
method is "instant runoff approval."

The behavior is *far better,* because the behavior of approval is 
better than that of plurality.

The name could be IRA, instead of IRAV, i.e., "instant runoff approval."

IRA actually should do, better, what IRV pretends to do, find 
majorities. When if fails, IRA is *honest about the failure,* it does 
not pretend to find a majority by setting aside and not counting all 
those ballots with votes only *against* the top two remaining.

So, yes, IRA might bring up negative associations with IRV, but there 
are also a lot of positive associations, and runoff voting is the 
most common advanced voting system in use, and Bucklin balloting and 
amalgamation *improves* runoff voting instead of trying to kill it. 
It should *reduce* runoffs. How much it will do that, we are not 
certain. But it is a cheap method to amalgamate, it's just the sum of 
votes in each rank, and those can be added up precinct by precinct. 
(IRV gets *very complicated*, and a single mistake in some precinct 
can require recounting *all the other precincts.*)

So, without being thoroughly aware of these conditions, Benjamin, 
your opinions are still valuable as to "first impressions." The name 
of "Approval voting" we have already decided to promote, and it's 
been on the table for many years as a major proposal, with some 
implementations in organizations. The Bucklin method is also very 
old, in fact, going back to Condorcet himself, around 1800. Bucklin 
is named after James Bucklin, who promoted and saw applied his method 
in Grand Junction, Colorado, in 1909, and it became all the rage, 
seeing something like ninety implementations around the U.S.

The method described is being considered as a suggested *second 
reform*, after approval is adopted. This may well already be in a 
runoff environment.

Under some conditions, it might be a first reform. It *is* an old 
method. I was tried and it worked, and it was not ended because it 
did not work. A proper study of why Bucklin was disadopted has never 
been done, but it's obvious from my research.