[EM] List PR

[LAYTON Craig]

Forest,

>I think the simplicity of your original suggestion out weighs any possible
>benefits from the tweaks that I suggested, especially in the case of
>ordinary public elections.

I do like the idea of using PAV or sequential PAV to decide between the
candiates of each list.  With all the ballots being computer counted, it
wouldn't really make much difference to counting the ballots, or
understanding the basic principles of the method, and would make no
practical difference at all to actually vote using the method.

>One question: What if the d'Hondt quota for a party's list exceeds the
>number of candidates on the list?  Are all of the "extra" votes wasted? Or
>are they redistributed proportionally when the list is exhausted?

This is a bit tricky.  For simplicity, I would lean towards wasting the
vote.  This may also counter-balance the incentive for parties to limit the
size of their lists (which negatively effects voter choice).

I guess the alternative would involve some kind of fractional transfer of
voters who have spread their votes over more than one list.  The process
could go something like; continue awarding seats to parties even when the
number of seats exceeds the number of candidates.  These seats are "surplus
seats".  When the number of seats left to be filled is equal to the number
of surplus seats (or maybe the number of surplus seats plus one),
redistribute as follows (I apologise in advance for the imprecision of my
mathematical explainations);

1) Calculate the average of votes per seat for each party with surplus seats
(total votes received/total seats awarded).  Call this value "a".  

2) The surplus now needs to be reduced so that these voters don't get an
unfair advantage over other voters (who don't get their surplus
redistributed).  Off the top of my head, I would using the following formula
(n-1)*a +1 where n is the number of surplus seats (so if there are 3 surplus
seats, the redistribution value will be equivalent to 2 seats plus one
vote).

3) Divide this number by the total number of votes for that party (to get a
basic fractional distribution value).  

4) Multiply this fractional distribution value by the fraction of the
person's vote that went to that party (ie by .5 if the voter voted for one
candidate each of two parties).

5) Distribute the value obtained at (4) between all parties not getting
surplus seats in proportion to the voter's support for those parties.

I've tried to simplify this process by only requiring redistribution once
during the count, but it remains overly complex.  I also think there are
slight errors in steps 4 and 5 of the process.  I definately think the votes
should be wasted.

Craig