[EM] FW: IRV Challenge - Press Announcement

[Raph Frank]

So, this is a legal test of PR-STV not IRV?

In any case, the two arguments are

- Loss of vote strength due to exhausted ballots

This seems fair enough, would you consider the following a reasonable
implementation that avoids the issue?

You could exclude the exhausted ballots from consideration for transfer.

If 'Doug' received first choices

6000: Doug > Meg
2000: Doug > Sue
5000: Doug
Total: 13000
Surplus: 3000

with a quota of 10000

The 4000 ballots which voted for Doug only would remain with Doug.

The 3000 surplus would be transferred in proportion to the 2nd choices

2000/(6000+2000) * 3000 =  750 to Sue
6000/(6000+2000) * 3000 = 2250 to Meg

So the voters would have their votes split:

Doug>Sue ballots:
Doug: 1250
Sue: 750
Total: 2000 (unchanged)

Doug>Meg ballots
Doug: 3750
Meg: 2250
Total: 6000 (unchanged)

Doug only ballots
Doug: 5000 (unchanged)

Total
Doug: 10000 (quota)
Meg: +2250
Sue: +750

Each ballot has been split into parts, but the sum remained at a single vote.

The could cause problems if Doug received more than a quotas worth of
"Doug only" ballots.  In this case, he would remain at more than a
quota worth of votes.  For example,

11000: Doug
1000: Doug>Meg
500: Doug>Sue
Surplus: 2500

I think the fairest solution here would be

Doug: 11000 (still has 1000 surplus)
Meg: +1000
Sue: +500

Otherwise, the Doug>Meg and the Doug>Sue voters would effectively get
more than 1 vote (though technically, they would get negative votes).
In effect, Doug keeps more than a quota.

It might also be worth recalculating the quota on the fly, but that
just adds more complexity.

I am not sure what your concern is about transfers occurring first?

The next issue is rounding problems.  This can be solved by using
rational numbers.  However, if the rounding was to many decimal
places, like say 10, then it is only a problem in theory.