[EM] How to make a summable version of STV

[Kristofer Munsterhjelm]

Jameson Quinn wrote:
> This method leaves open the possibility for "candidate hijacking". 
> Consider: Sojourner Truth has 1000 votes who vote Truth > Douglass > 
> Everyone. Robert E. Lee has 1100 strategic voters who vote Truth > Lee > 
> Everyone > Douglass. They dominate the honest Truth voters, so Lee is 
> considered to be Truth's second choice. The Droop quota is 1900. Truth 
> has no quota and is eliminated; her 2100 votes pass to Lee; and Lee is 
> elected. That kind of pathology could cause a civil war.
> 
> Part of this is the LNH property of STV biting us. With a system that 
> can see past the top choice, you'd have 2100 votes which were 
> essentially half-Douglass-half-Lee. In systems like Schulze STV, that 
> would at least be helping Douglass, so the Lee voters would have to 
> think twice about the tactic. And in systems like RRV, that wouldn't be 
> any more likely to elect Lee than Lee's 1100 honest votes, so the tactic 
> would be useless.  

Couldn't a more complex variant work in Range as well? Say you have 
honest voters who prefer A > B > C, and some dishonest voters who prefer 
  D to win. They could either vote D first and play honestly, or they 
could vote A > D > ...

By voting A > D, their tactic is to get the second seat and also to 
downweight the A > * ballot in doing so. They know that they can get D 
anyway, but if they poison the A > * ballot with D, they can both have 
their cake (get D) and eat it too (make the honest voters' latter 
choices count less).

If there are only two seats, this has no point. If there are many, it 
might pay for a substantial fraction of the D-voters to vote A > D so as 
to weaken the competitions for later seats.

Of course, the tactical voters would have to decide whether they get 
most bang for the buck by voting D, A>D, or a combination of the two. It 
might turn out that the downweighting makes the strategy suboptimal in 
any case.