[EM] 2-step CR: A proposal

[Forest Simmons]

I would like to add a slightly different alternative to Joe's suggestion
at the very bottom ...

On Wed, 22 May 2002, Joe Weinstein wrote:

> 2-step CR?
>
> Alex’ 2-step CR proposal seems like a generalization of  Majority Choice
> Approval to the case of a higher-res version of CR (Approval = CR with
> resolution=2, i.e. just two available grades).  For that reason I really
> liked it at first.  However, there is a fly in the ointment: a key if subtle
> respect in which the proposal does not accomplish what Majority Choice
> Approval does.
>
> Consider Majority Choice Approval, for the case (which will often occur)
> where no candidate is - or in advance looks like - a majority-favorite.  For
> that case, there is no instrumental difference between top grade
> (‘preferred’ : two checks) and the next marked grade (‘accepted’, one
> check).  Both grades yield  the same degree of instrumental support.  As a
> result, an instrumental strategist need have no regret giving top grade to
> his true favorite, and a lesser grade to his unfavorite but acceptable
> compromise.
>
> That is, as Alex earlier correctly noted and emphasized, Majority Choice
> Approval has the merit of giving little or no incentive toward insincerely
> ranking your lesser-evil compromise equal to your true favorite.
>
> However, under the new 2-step CR proposal, in case there is no majority
> favorite, top grade and next-to-top grade will count differently.  The
> difference may seem small to some, but it will exist and be serious for
> others. If you are serious about optimal instrumentally focused strategy,
> you will still face considerable pressure to insincerely rank compromise
> equal to favorite.
>
> That is, you will face the same old strategist’s quandary that now exists
> with Lone Mark plurality (and is our main reason for seeking election method
> reform):  you can NOT simultaneously address two major voting goals: (1)
> expressively indicate your true favorite as strictly preferred, and (2) give
> maximum instrumental support, when needed, to your acceptable (‘lesser
> evil’) compromise.
>
> The advantage and deep insight of both the original Five-Slot Approval (or
> other X-slot Approval) and the new Majority-Choice Approval schemes is that
> with them you CAN simultaneously address both these goals.
>
> But, compared with Majority Choice Approval, the 2-step CR proposal does
> offer the benefit of extra expressive flexibility.  There is a good way to
> get this benefit while retaining the essential merits of Five (or X) -Slot
> Approval and Majority-Choice Approval.
>
> Namely (for instance), use a Majority Choice version of Five (or X)-slot
> Approval.  That is, if no candidate is a majority favorite - i.e. gets a
> top-grade (A) vote from the majority of voters - then count each of the
> higher votes (A, B or C) as approved, and each of the lower votes (D or F =
> blank) as not approved.

Alternately, if no candidate gets a fifty percent plus style majority of
top-grade (A) votes, then use the rules of CR, but count Top and Next to
Top equally, i.e. A's and B's both count as three in the Cardinal Ratings
step, so that they are indistinguishable after the the majority check
step.

Forest

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em