STV for party candidate lists?

[Mike Ositoff]

How about this?:

When your party's voters vote their pre-election rankings of 
your party's candidates, and you don't know how many seats that
party will win, then why conduct a count for _each_ possible number
of seats that your party might win.

I don't know how big your party is, but if it might just win one
seat, then I'd do a count by Condoret(EM) or Schulze to choose
who gets the seat if the party just wins one seat.

Then, using those same rankings, of course, do an STV count
for N = 2, for the possibility that your party might win 2 seats.
The winners of that STV count are the 2 candidates who will get
your party's 2 seats in the event that your party wins 2 seats.

Then, do an STV count for N = 3, and, likewise, record the 3 
winners, for the possibility that your party wins 3 seats.

(Of course you'd use a computer program that would do all this
very quickly)

So you'd do this for values of N from one, up to the total number
of seats in the house (or the total number of your party's candidates,
whichever is greater).

Then, however many seats your party wins, you know whom to give
those seats to.

***

But, it's just occurred to me that your election rules might
require just having an ordered _list_, rather than a set of
camdidates for each possible value of N, and that you won't
be able, therefore, to give the seats as I've described.

Ok, no problem. When Condorcet(EM) or Schulze picks the the
winner for when your party wins exactly 1 seat, the candidate that
it picks gets the top position in the list.

Then, as before, you do STV for successively larger values of
N. 

When the program does an STV count for N = 2, and if your 
number 1 candidate isn't one of them, he still keeps his 
number 1 position, and the number 2 position is given to one
of the 2 STV winners (For instance, you could give it to the
one with the most 1st place votes).

Then, when the program does its STV count for N = 3, the 
number 3 position goes to whichever of its 3 winners has the
most 1st place votes.

And so on.

***

You might say it's crude to use 1st place vote totals. 
Ok, then, use Condorcet(EM) or Schulze instead, amnng the
N winners, in the N-winner STV count, to choose which of
its N winnes gets the Nth seat.

Obviously, if the N-1 winners that you already have are among
the current N winners, then you don't even have to do the 
Condorcet(EM) or Schulze count.

***

I believe that this way of choosing & ordering your list
can't be beat.

***

I hope that it doesn't seem like I'm just beating the drum for
my favorite issue, single-winner methods. Choosing the candidate
for the Nth place in the list, from among the N winners that
STV chooses really is best done by a good single-winner method.

You could use Plurality, just giving it to the candidate with
the most 1st choice votes, but I'd recommend, instead, using
Condorcet(EM), Smith//Condorcet(EM), or Schulze, if you really
want to do it in the ideal best way.

***

Mike Ossipoff