[EM] Extremism

[John B. Hodges]

>Date: Wed, 17 Sep 2003 23:03:19 -0700 (PDT)
>From: "Alex Small" <asmall at physics.ucsb.edu>
>Subject: [EM] Extremism
>
>It is sometimes asserted in various places (including, occasionally, this
>forum) that PR promotes extremism and single member districts (SMD)
>promote moderation.  There's no denying that PR would give representation
>to small extremist factions.  But small extremist factions would still
>have small caucuses while large centrist groups would have large caucuses.
>(snip)
>The result of this incumbent protection is that legislative elections are
>now decided in the party primaries.  Party loyalists are of course more
>extreme than the rest of us. (snip)
>
>If we had PR, and we were liberated from the tyranny of the party
>loyalists, (snip)

(JBH) Stephen Hill talks about this in FIXING ELECTIONS. But, Whether 
PR would be better depends on the type of PR. PR of any kind would 
limit the representation of extremists to their actual percentage of 
support among the electorate, but any sort of "party list" PR (i.e. 
the most common kind) will tend to have each party list reflect the 
party activists. Whether that is a disadvantage under PR is open to 
debate; perhaps it is a good thing that each party have a clearly 
defined and distinct position. But assuming that it is bad, then this 
argument would tend to favor forms of PR that do not depend on 
parties, and which give advantage to candidates who can attract 
second-choice votes. STV is well-known and has been used in real 
elections for awhile. GB 2.3 would also qualify, theoretically.

I've been wondering if the PR aspect of GB 2.3 could be applied to 
other voting methods besides Generalized Bucklin. Something along 
these lines:
1. collect the ballots.
2. count the ballots, calculate the Droop Quota.
3. Use (insert your favorite single-winner method here) to pick a 
winning candidate.
4. Do any seats remain to be filled? If not, you are finished.
5. identify the DQ strongest supporters of that candidate. Delete 
their ballots. (If your criteria for measuring strength of support 
gives you a set of equally-strong supporters larger than the DQ, then 
multiply all such ballots by the fraction (#supporting-ballots minus 
DQ)/DQ. )
6. go to 2. (repeating 2 is necessary because of the possibility that 
some ballots may be exhausted, so the DQ will change.)

Sometime I may get around to the chore of trying this with Ranked 
Pairs. But I suspect that the procedure above is flawed by itself, 
regardless of what single-winner method is used in it. Working 
through examples with GB 2.2, which would be an example of the above 
PR method, I found that it sometimes gave the "wrong" results. It 
seems to be better if a PR method picks all, or at least several, 
winners simultaneously.

Has anyone tried checking whether Ranked Pairs could do satisfactory 
PR by filling N seats with the N highest-ranked candidates? I suspect 
it would not; too much bias toward the center.

-- 
----------------------------------
John B. Hodges, jbhodges@  @usit.net
Do Justice, Love Mercy, and Be Irreverent.