[EM] Re: Total Approval Ranked Pairs

[Ted Stern]

On 14 Mar 2005 at 22:02 PST, Jobst Heitzig wrote:

> Dear Forest, Russ, and Ted!
>
> I suggest that we call the method we discussed under various names
> in the last days ARC (Approval Runoff Condorcet) and continue to
> study its properties, especially its anti-strategy properties.
>
> I agree with Russ that it is perhaps a very nice first public
> proposal, especially because it may be a nice compromise between
> IRV- and Condorcet-supporters: Both methods are Runoffs which delete
> the single candidate with the least points until there is among the
> rest a candidate with a special property!
>
> Yours, Jobst

I do agree that Approval Runoff Condorcet (ARC) finds the same winner
as Approval-seeded Bubble Sort (ABS).

But I happen to think eliminating candidates through Runoff is one of
IRV's weakest points.  Its only appeal is familiarity.  It seems to me
that the key reason for eliminating primaries is to keep candidates in
as long as possible, to enable voters to coalesce around the one whose
views are closest to the majority.

However, if you think that it can be a useful argument, then go ahead.
I just worry that you would be letting the IRV advocates frame the
debate in a way they think they can win.

I admire of the compact description of ABS.  I think it is the fastest
possible way to reveal the social ordering of this method.  Another
point I would like to make is that although successively eliminating
lowest approved candidates may be easier to describe -- easy to
describe in the same way IRV is easy to describe -- every time you
eliminate a candidate, you have to look for undefeated candidates
again.  I think it is just as much work as the Bubble Sort.

Back to elimination again: I was thinking about how elimination of
candidates tends to narrow voter choices, and I thought of this
analogy.

Say Democrats vote along a North-South axis, and Republicans vote
along an East-West axis.  But most of the voters are sitting somewere
in the South-West.

In the primary, the core partisans of each camp will choose a
candidate somewhere along their axis.  Democrats will probably not
choose the South-leaning candidate who has drifted slightly to the
west, and Republicans will probably not choose the West-leaning
candidate who has drifted slightly South.

The voters end up with a choice between a too-far-north Democrat
with little Western tendencies and a too-far-East Republican with
little Southern tendencies, neither of whom is particularly close to
what they feel is truly important.

When the voters can specify exactly how close they feel to each
candidate without having to channel their choices, the winner will
more likely be the WSW or the SWS candidate.  Or even the true SW
candidate who isn't a member of either party.

Ted
-- 
 Ted Stern                               
   Change reply address to tedstern at u dot washington dot edu