As I am sure many of you know Australia just had a federal election under their variant of STV. As I am sure you all know STV is nonmonotonic. Does anybody know or even know how to calculate the fraction of voters who experienced a nonmonotonic effect? Specifically this is when a candidate would have been better served by having been ranked lower. In the last election this was 14%.
Although nonmonotonicity is probably the most ‘dramatic-seeming’ issue for STV and RCV/IRV, it is almost surely not the most pathological. Its primary pathologies are probably the double bind effect (whereby the voter is faced with the lack of any reasonable strategy, which it shares with approval).
Plus extreme vulnerability to easy-to-utilize over-exploitation of the bullet voting strategy (so far as I know, RCV/IRV are the only methods afflicted with significant bullet voting pathology). See (please DO CLICK the link to see the actual demonstration):
Below I have worked out an example of one of the major pathologies of “IRV” voting. As a mere example, it’s more than just a “vignette.” It is not explained in as much detail as I would prefer. And it’s very easy to make silly mistakes in such exercises, so feel free to point out any.
Ranked Choice Voting fails through “bullet voting”. Voters are numbered 1 through 28. Candidates are designated A through Z. Candidates within parentheses are the ones that get eliminated.
Well maybe those issues are more important than nonmonotonicity and maybe not. This is not what I am interested in at the moment. STV has a lot of issues.
I am interested in numbers for the magnitude of the effect specifically in the recent Australian election. If you have numbers for these other issues I will gladly take them.
This is part of the reason why I am interested in this. When communicating with Australians about why their system is bad it is easy to get them to agree that nonmonotonicity is bad. People often say it only effects a minority of voters. To counter this point I need to know the actual number. 14% is high enough for people to agree that it is a problem but I need an up to date number.
What’s the source for this figure?
I think I originally heard it from Warren but saw it again somewhere else. I have never seen a calculation or anything that would make me believe enough to cite it formally. This is sort of why I am wanting to get a number for this election. 14% is pretty damning if true.
If you think about it the number feels about right. Nonmonotonic effects are common in close races. This is the number of ballots who had any nonmonotonic effect at any round. The number would be much lower if we were counting the fraction of candidates who would have won under a different ranking or even those who would have done better.
The Australian election was not as close as the last one, so the number may be less. I would be interested in any measure of nonmonotonicity not just the one which I think was 14% in the last election.
I don’t think the raw ballot data is made publicly available, so whoever calculated the 14% figure must have modeled hypothetical preference flow rather than done an exact calculation.
Yes that is exactly the issue. I would doubt that it is an exactly calculated value but more of an estimate. I think they release the number of votes candidates have at each round. It seems a pretty good estimate could be derived from those. I would suspect that the person who calculated the 14% is well known. This is an interesting result.