[EM] An unusual "absolving" PR method

[Kristofer Munsterhjelm]

Some methods are algorithmically simple by making the voters or 
candidates do their work instead. For instance, Approval gets some of 
its simplicity by telling the voters to figure out where to put their 
cutoff. Let's say that such a method "absolves itself" of work that 
other methods do (and the back-end complexity that arises as a consequence).

In that vein, here is one such PR method with parties: Voters rank party 
candidates. An order of finish is computed using a majoritarian method, 
say ranked pairs. Separately, each voter's ballot is counted as a vote 
for whatever party their favorite candidate belongs to, like in ordinary 
party list.

Then the method uses some well-known party list method (say Sainte-Laguë 
with Pukelsheim's method) to determine the number of winners per party 
per district. These winners get filled up with candidates in the order 
of the district's order of finish.

For instance, if district A's order of finish is 
R1>G1>B3>R2>G2>R3>G3>B2>B1 and the party list calculation gives the 
district two seats to R and G each and one seat to B, then the 
allocation is:

Red: R1, R2
Green: G1, G2
Blue: B3

Now, all else equal, you'd expect these seats to be filled up with 
centrists since we're using a majoritarian method and candidates closer 
to the median voter are preferred by it. But here's the "absolution": 
the parties are motivated to retain their distinct identities. Knowing 
that centrists will be elected, but that the party list component is 
based on favorites, they may simply not nominate candidates who are too 
centrist, because they're going to win the seat anyway and they don't 
want to dilute their ideological position.

However, there is a tension to this reasoning. If they nominate a bounch 
of extremists, then that will put voters off voting for them and they 
lose first preferences. The parties thus have to find a balance between 
not running too centrist candidates (who won't have an ideological base 
in the party itself) and candidates who are too far into the wings.

In practice, this might just turn into ordinary party list PR, because 
the tension that "if we run too extreme candidates, people won't vote 
for us" also applies to ordinary party list. What it would do is keep 
parties from fielding candidates beloved by the leadership but not the 
people, as the outcome ordering would force these candidates to the back 
and they wouldn't be elected -- unless everybody else on the list is worse.

-km