[EM] Response to critique of MJ (MCA subtype) procedure by Felsenthal & Machover

[Jameson Quinn]

Researching Majority Judgement book led me to a critique of this procedure:

http://eprints.lse.ac.uk/24213/1/The_Majority_Judgement_voting_procedure_%28LSERO%29.pdf

Basically, they give two problems with MJ that they consider serious:
1. Strategy is possible.
2. Seriously nonmajoritarian results are possible if most preferences
between two candidates do not straddle the median. For instance:
49: A10, B7
2:    A4, B5
49:  A3, B0
Although 98 voters prefer A by 3 levels, 2 voters elect B.

Their second objection is easily addressed by offering only 3 rating levels.
I think that avoiding this objection is worth the reduced expressivity.

The first objection is tougher to resolve. While F&M do not analyze the
probability or the nature of possible strategies, I believe that the only
serious practical problem is with truncation between two near clones. As
I've recently argued on this list, I believe that an asset-like procedure to
resolve ties can acceptably minimize the bad effects of this possibility.
Furthermore, I feel that even the basic Majority Judgement system, or
another MCA system which doesn't specifically address this issue, is
preferable to most alternatives. Specifically, I feel that MJ would give
better Bayesian regret, and is in most cases simpler, than any Condorcet
system; is more expressive than approval; and there is less of a strategic
imperative / nonstrategic penalty than with range. (Note that the latter two
systems share the near-clone truncation problem).

Jameson Quinn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20110301/07018f10/attachment-0003.htm>