[EM] primary election thoughts

[Daniel Bishop]

Ted Stern wrote:

>What do group members think of the following primary election proposal:
>
>- Ballots allow a voter to rank 1st, 2nd or 3rd choice candidates.
>
>- Unlike IRV, more than one candidate can be chosen for any rank.
>  
>
If you aren't using IRV, then there's not a huge implementation 
advantage to restricting the number of ranks.  So why have the restriction?

>- Any 1st/2nd/3rd choice vote is considered an "approval" vote.
>
>- Use Condorcet to tally.
>
>- If a Condorcet winner exists, with more than 50% of the voters
>  approving, then that candidate wins immediately and the seat doesn't have
>  to be decided in the general election.
>
>- Otherwise, eliminate candidates with less than 1% approval.
>
>- On the general election ballot for that seat, candidates will be listed
>  with the Condorcet winner (if any) at the top, with remaining candidates
>  listed below in order of approval.
>
>This would be an alternative to either Louisiana-style top two runoff or
>closed party primary.
>
>I'm curious what advantages of full Condorcet might be lost by reducing the
>options to only 3 ranks.
>  
>
Suppose that the votes are:

499: A>C>B
492: B>C>A
  9: C>A>B

Then C, a Condorcet winner with 100% approval, wins immediately.

Suppose that candidates A1, A2, and A3 entered the race, and the voters' 
preferences were

499: A>A1>A2>A3>C>B
492: B>C>A>A1>A2>A3
  9: C>A>A1>A2>A3>B

If they voted sincerely, their ballots would be:

499: A>A1>A2
492: B>C>A
  9: C>A>A1
 
There is now no longer a Condorcet winner.  Candidate A3, with 0% 
approval, is eliminated, and the remaining candidates move on the 
general election with the ballot order A>A1>C>A2>B.

If the general election uses a Condorcet method, C still has a good 
chance of winning.  But suppose that now B1, B2, and B3 enter the race, 
and the voters' preferences are:

499: A>A1>A2>A3>C>B>B1>B2>B3
492: B>B1>B2>B3>C>A>A1>A2>A3
  9: C>A>A1>A2>A3>B>B1>B2>B3

The ballots now become:

499: A>A1>A2
492: B>B1>B2
  9: C>A>A1

A is now a Condorcet winner with 50.8% approval, and so wins immediately.

Thus, your proposed method suffers from a perverse failure of 
Independence of Clones: It's possible for one candidate (A) to benefit 
from clones, while another candidate (B) is hurt by them, in the same 
election!