[EM] "Reweighted" methods are second rate (medium article)

Kristofer Munsterhjelm km_elmet at t-online.de
Mon Jul 2 02:00:58 PDT 2018

On 2018-06-30 16:24, Jameson Quinn wrote:
> I just published "Reweighted" methods are second rate 
> <https://medium.com/@jameson.quinn/reweighted-voting-methods-are-second-rate-50d9c7814c35>. 
> This is of course my own personal opinion only, but I'd welcome comments 
> either here or on Medium.

To be a bit nitpicky, I don't think it's true that good proportional 
methods ignore considerable amounts of the ballot data. Consider Schulze 

One could argue that every subset contest (e.g. ABCD vs ABCE) only takes 
some of the ballot information into account, but the same is true of 
single-winner Schulze (the A vs B contest only considers how many voters 
preferred A to B), and I would say Schulze uses the full information 
available to it.

My strategy-hardened variant of BTV 
also considers more than just a fixed quota of the ballots each time a 
candidate is to be elected; and it does so for the purpose of reducing 
free riding. When it considers whether A should be elected, it maximizes 
the number of voters who can be counted towards A subject to that every 
candidate already elected has the support of more than a Droop quota.

I think the reason that RRV doesn't work is because RV itself doesn't 
obey Majority. This lets information from the downweighted ballots bleed 
through to the no-longer-downweighted ballots in a way that makes free 
riding particularly easy. The reweighted party list methods that use 
only Plurality ballots (e.g. Sainte-Laguë, D'Hondt) don't have that 
problem because there's no possibility of such a bleedthrough, and 
Plurality passes the Majority criterion.

Does PAV pass Droop proportionality according to the submitted ballots? 
Approval passes Majority according to the submitted ballots (i.e. not 
necessarily the preferences leading to the ballots, but the yes-or-no 
preferences expressed on them).


More information about the Election-Methods mailing list