[EM] Top 6, Top 2, Head to Head Primary

[Kristofer Munsterhjelm]

On 12/08/2012 05:42 AM, Don Hoffard wrote:
> *Top 6, Top 2, Head to Head Primary*
>
> *Nominations:*
>
> 1. In order for a candidate to get into the primary they must get
> registered voters to sign nominating cards for them.
>
> 2. Each candidate must get at least one quarter of 1% of the registered
> voters in their district/state to sign a nominating card for them.
>
> 3. Only the top 6 candidates with the most signed nominating cards will
> be included on the primary ballot (see exception below).
>
> 4. A registered voter may sign a card for more than one candidate, but
> only one for each candidate.
>
> 5. Only registered voters from the same district/state as the candidate
> may sign a nominating card for that candidate.
>
> 6. No more than two candidates from each political party will be
> included on the primary ballot.

That would blunt the clone (teaming) tendency of using Approval for 
nominations, but it could still be gamed. Perhaps it's unrealistic, but 
I'll still mention it:

- Party X is at the center and knows its candidate will be elected. 
However, to appeal to as many voters as possible, they've been 
deliberately vague and know their candidates may be hurt in the primary.

- Thus party X makes decoy parties and tells supporters of X to vote for 
all candidates in all the decoy parties.

- By doing so, party X gets their near-top count for their candidate 
replicated across numerous parties.

- These clone candidates then take up most of the spots on the primary 
ballot, crowding out many of the serious competitors and making party 
X's task easier.

> 7. Only registered voters, from the same political party as the
> candidate, may sign a nominating card for that candidate. However, a
> political party may allow other registered voters to sign their
> candidates nominating card.

Does this mean only registered Democrats can support a Democrat, or that 
non-Democrats can also support the Democrat? The first seems to say that 
only registered Democrats can do so, while the second seems to say that 
if the party wants (say) Republicans to also be allowed, they can do so.

> 8. Any registered voter may sign a nominating card for any
> non-affiliated candidate, but only one for each candidate.
>
> 9. In non-partisan position elections 6-7 above does not apply.

How about each party using an internal method to determine the two who 
will go on to the primary? If a party wants to use, say Schulze, then 
they'll be free to do so. This shouldn't cause much disenfranchisement 
because in the context of the better primary and general election 
system, if the leadership twists the nomination method to their own 
ends, those who disagree can form their own party.

(This actually happened in New York under STV, where some democrats 
split to an "insurgent Democrat" group.)

I suppose that would make the clone attack much easier, though.

> *Primary election:*
>
> 1. The primary will allow voters to rank each of the candidates from 1
> to 5. Number 1 being their top pick and 5 their 2^nd lowest (their
> lowest being their one non-ranked candidate).
>
> 2. A registered voter is allowed to rank all candidates regardless of
> political party.
>
> 3. The two candidates that receive higher ranking than any of the other
> candidate will move on to the general election. Each candidate is
> compared head to head with each of the other candidates. [If A>B, A>C,
> A>D, A>E, and A>F, (> meaning more preference votes) then candidate A
> moves on to the general election and if B>C, B>D, B>E, and B>F then
> candidate B moves on to the general election.]
>
> 4. If no candidate meet the head to head criteria in 3 above then the
> candidate with the lowest preference vote is eliminated first, then the
> next lowest eliminated next, until only 2 candidate remain and they will
> move on to the general election. [i.e. the Instant Runoff System]

I'd suggest that 3 and 4 be interleaved, i.g. first you try to find the 
pairwise winner, but if there isn't one, you eliminate the one that's 
last, then you look for the pairwise winner among the ones left, and so on.

Doing so ensures the system doesn't radically change its character from 
a Condorcet method into a non-Condorcet method just because someone 
manages to engineer a cycle (or just because a cycle appears benignly). 
It's thus more robust, retaining the logic of both methods despite 
perturbations.

Furthermore, the changed method would pass Smith. If there are three 
serious candidates and three (extremist) not-serious candidates, and 
pairwise, all the serious candidates are preferred to the not-serious 
candidates but there's disagreement (a cycle) about which serious 
candidate is the best, then as soon as all but two serious candidate 
have been eliminated, those win.

One might make it even more rigorous by treating it like a single-winner 
method and then electing first and second place from the resulting ordering.

Unfortunately, neither of the fixes above would remove the tendency for 
this method to elect similar candidates. All majoritarian methods would 
do so. Say a bare majority (50%+1) puts A and B ahead of everybody else. 
Then A and B gets elected, even if the other 49% voted C first; and it 
could be the case that if A and C's platforms were investigated in 
further detail, enough people would shift from A to C to make C win.

Resolving that - i.e. getting more variety into each round - would 
probably have to involve using a proportional method in both stages: 
something like PAV for the nomination stage and something like a 
proportional ordering for the primary->general stage. This would make 
the system a lot more complex, though.

> 5. If one candidate meets the head to head criteria then that candidate
> moves on to the general election and if a second does not meet the head
> to head criteria then only the remaining 5 candidates would be subject
> to the elimination method described in 4 above.
>
> *General Election:*
>
> 1. No write-in votes will be allowed in the general election.
> (exceptions: one candidate dies or with-draws)
>
> 2. It is possible that the two candidates would be from the same
> political party.
>
> 3. The winner is elected to the office.

This stage seems reasonable.