[EM] List PR

[Forest Simmons]

The d'Hondt count is the same rule that is used in sequential PAV:

In sequential Proportional Approval Voting, the approval winner gets a
seat. Then that winner is included in all pairs under consideration for
two seats. The two members of the winning pair are included in all triples
considered for three seats, etc.

In this context the rules of PAV simplify to the d'Hondt count, as
mentioned in a previous posting in a response to a question of Hugo Harth:

In the n_th stage, each ballot is weighed by a factor of the reciprocal of
one more than the number of already seated candidates approved on that
ballot. 

So Craig's system can be thought of as an ingenious hybrid of sequential
PAV and list PR (although d'Hondt came before sequential PAV, and Craig
wasn't thinking of this connection).

By the way, sequential PAV is computationally effective, so it can be
applied to any size multiwinner election, thus avoiding lists altogether.
The only problem is that it creates a natural pecking order among the
winners where it may not be desirable to have one.

When a pecking order is desirable, then sequential PAV is the method of
choice, in my opinion.

Suppose that seniority is important in certain important committee
assignments, etc. Then a pecking order for the representatives elected on
the same day might be desirable.

By the way, STV picks winners sequentially, too, so it is no better than
sequential PAV in this regard, though some would say that the randomness
it inherits from IRV would effectively remove the "order" from "pecking
order"  :-) 

If you believe strongly in IRV and STV, then you probably believe that STV
picks the strongest winners first.


Forest


On Fri, 15 Jun 2001, LAYTON Craig wrote:

> 
> I should have explained the d'Hondt count.  It is my preferred method for
> allocating seats in a list PR system.  Seats are awarded in "rounds".  In
> each round, divide each party's total number of votes by the number of seats
> they have already been awarded +1.  The party with the highest total gets a
> seat (tied parties all get a seat).  Continue until there are no seats left.
> In this scenario, P5 could not get a seat.
> 
> The d'Hondt count is often said to disadvantage minority candidates, but as
> long as there is no minimum quota, it is the fairest way to allocate seats.
> The party getting the final seat will end up with one candidate for every n
> voters, where n is always greater than the number of votes for any party not
> recieving any seats.
> 
> Craig
> 
>