[EM] [CES #6984] Re: Wow: new, simple Bucklin motivation for CMJ. So renaming to Graduated MJ.

[Abd ul-Rahman Lomax]

At 02:31 PM 1/8/2013, robert bristow-johnson wrote:
>On 1/8/13 1:03 PM, Warren Smith wrote:
>>So should this realization by Jameson Quinn tell us that all 
>>previous historical
>>examples of Bucklin voting should be regarded as examples of the
>>"Majority-Judgment" median-based system,
>>and hence can be used to help evaluate how the latter behaves in practice?
>>
>>Unfortunately I think not because I think Bucklin voters historically
>>were urged to provide rank-orderings not ratings.
>
>well, isn't that how it works for Bucklin?  the inclusion of the 
>2nd-choice or 3rd-choice votes in the total vote count is a sorta 
>discrete decision.  then only inequalities exist for the decisions 
>made by Bucklin.  if that's the case, the only information that 
>matters in rating candidates are the relative ranks.
>
>i dunno.

Equal ranking in classic Bucklin was allowed in the third rank. It's 
a bit of a mystery why they did not allow it in the first two ranks, 
but I suspect it was merely to make it more familiar to people 
accustomed simply to "vote for your favorite." The third rank equal 
ranking possibility made the system even more clearly an Approval 
system, and when there were many candidates, it was common for the 
ranks to be completely collapsed, so it really did end up as an 
Approval election, with the standard Approval phenomenon of candidate 
totals exceeding the number of voters.

It was also allowed to skip ranks. That is characteristic of Range, 
and not of ranked systems, and the meaning of a skipped rank in 
Bucklin is quite the same as skipped ratings in Range: it indicates a 
stronger gap in preference strength.

Bucklin really was Instant Runoff Approval, quite the same as a 
simulated series of Approval elections (but, unlike a real series of 
elections, voters in subsequent rounds have, with plurality-electing 
Bucklin, no additional information. They are guessing, more or less.) 
Imagine that voters have a range-ballot-in-mind, and are with each 
round, lowering the approval cutoff a bit.

Voting reform in the U.S. has neglected Bucklin, to its cost.

And voting systems experts have frequently neglected Top-Two runoff, 
the most widely-implemented voting reform. TTR has some obvious 
flaws, but the *concept* of iterated voting has tremendous power. The 
flaws can easily be fixed by using a more sophisticated primary 
system, and with a better ballot, a runoff, if needed, could handle 
two candidates plus write-ins, without a spoiler effect there. With 
improved ballot analysis, which could still be simple, a runoff 
system could be made Condorcet compliant *and* SU optimizing.

That last one has been controversial, but it works by testing 
absolute preferences with a runoff. Low absolute preference 
difference between runoff candidates equals low turnout from those 
with low preference. When there is a pairwise beaten Range winner, 
the preference strength is necessarily low for that pairwise winner 
(otherwise this would be the Range winner!).

Real runoffs produce "comeback elections," in about one-third of the 
cases (for nonpartisan elections), where the runner-up in the primary 
wins the runoff. Runoff electorate is not the same as primary 
electorate, in addition to vote transfers. (In non partisan 
elections, simple vote transfers will generally not shift relative 
position; that is why IRV fails to produce this comeback effect in 
nonpartisan elections. It does only in partisan elections, where vote 
transfers are strongly correlated with the first preference.)