Brilliant teardown of the Pettigrew-Radley study. The framing of a 99.53% vs 99.96% acceptance rate as "10x worse" really exposes how quantitative researchers can weaponize statistics without context. I ran into similar stuff when reviewingML model evaluation papers where a 0.5% accuracy diffrence gets ballooned into "50% relative improvement." The 32x ratio of meaningful votes to errors is the real story here.
Yes, it is truly disappointing to see such sloppy "research" like this. Some are asking: is this just incompetence in the ivory tower, or a purposeful, even coordinated, attack by some academics against RCV?
i am surprised by claim only 17 percent improvement of effective votes under PR versus FPTP,.
under PR (whether list PR or STV) is usually about 80 percent,
and under FPTP is likely about only 50 percent.
with single non-transferable voting, single X voting in multi-member contest, effective votes are usually about same as FPTP (round 50 percent) but due to range of members and parties elected, higher rate of satisfied voters.
even if their own vote not used to elect anyone, voters is satisfied if someone of party they support is elected or someone of the like-minded spectrum _ left, right, environmentalist, men and women , varieties of races and ethnic backgrounds, etc.
but under FPTP often majority -- and always several minorities -- not represented.
with 30 percent improvement over FPTP's rate of effective votes, the .3 or 2-5? percent spoiled/rejected votes are but a smidgeon.
Tom, in the DSOS article, the reference to RCV is for its use in single-winner contests, not multi-seat proportional representation. Of course you are right, if we were talking about a PR method, the number of effective votes would be even higher compared to plurality/FPTP. Not to mention that, judging by the experience of other PR countries around the world, the voter turnout would be higher since voters would have more choice and multi-party democracy.
so here under RCV we are talking of difference from 33 percent to about 50 percent,
or more likely an improvement from 50 percent (or 48 percent?) to 65 percent, on average.
RCV ensures just minimum of about 50 percent in a district. But of course some are elected with 60 or 70 percent in some cases.
whether it is 33 to 50, or 48 to 65, the difference seems less when you see that under RCV, in many districts, only under optimum conditions is there guarantee that even half the voters see their vote used to elect the winner.
I would like to see how they came up with that 17 percent difference.
is it difference from about a third casting an effective vote to about half casting an effective vote?
meanwhile, so much more improvement is out there if you merely open to multi-member wards or city-wide districting.
say you look at Wellington city election three-member wards (STV)
Successful candidates Range from least- to most-popular
2812 to 4242
Ratio 1 to 1.5
While even under RCV due to differences from ward to ward (population, voter turnout rate, rate of exhausted votes), you see range in successful candidate's vote tallies in each city each time. that range would allow a mere minority of voters to elect a majority of city councillors and wield power and rule. And minority rule is what we don't want in elections.
Again, the failure to look at Australia is sufficient, in itself, to discredit this research. There have literally tens of thousands of RCV elections conducted here, with a variety of rules regarding requirements for a valid vote (most usual is a complete ranking, but we've also done "optional preferential" where you number as many candidates as you want. Voting is compulsory so there's no selection effect - . The percentage of invalid votes is small (a few per cent), and many of these are blank ballots (effectively, I'm only voting because I have to).
Yes John, could not agree more. Yet another "hole" in the research of US political scientists and their shaky political science. Here's another oddity abolut this: I have had US political scientists tell me that because Australia has compulsory voting, that somehow invalidates Australia as a research source, since it makes it hard to tease out if ranked ballots in the US are discouraging "certain" voters, i.e. racial minorities, from voting. If you study race in Australia, that effect won't show up, so the thinking goes, because racial minorities are required to vote. Another version of the "American exceptionalism" narrative that has never had much validity, in my view. Thanks for your thoughts.
I suppose it;s a fair point if you are specifically studying turnout. But that's a very limited question, and one that's relatively easy to test from exit polls and similar.
Brilliant teardown of the Pettigrew-Radley study. The framing of a 99.53% vs 99.96% acceptance rate as "10x worse" really exposes how quantitative researchers can weaponize statistics without context. I ran into similar stuff when reviewingML model evaluation papers where a 0.5% accuracy diffrence gets ballooned into "50% relative improvement." The 32x ratio of meaningful votes to errors is the real story here.
Yes, it is truly disappointing to see such sloppy "research" like this. Some are asking: is this just incompetence in the ivory tower, or a purposeful, even coordinated, attack by some academics against RCV?
i am surprised by claim only 17 percent improvement of effective votes under PR versus FPTP,.
under PR (whether list PR or STV) is usually about 80 percent,
and under FPTP is likely about only 50 percent.
with single non-transferable voting, single X voting in multi-member contest, effective votes are usually about same as FPTP (round 50 percent) but due to range of members and parties elected, higher rate of satisfied voters.
even if their own vote not used to elect anyone, voters is satisfied if someone of party they support is elected or someone of the like-minded spectrum _ left, right, environmentalist, men and women , varieties of races and ethnic backgrounds, etc.
but under FPTP often majority -- and always several minorities -- not represented.
with 30 percent improvement over FPTP's rate of effective votes, the .3 or 2-5? percent spoiled/rejected votes are but a smidgeon.
Tom, in the DSOS article, the reference to RCV is for its use in single-winner contests, not multi-seat proportional representation. Of course you are right, if we were talking about a PR method, the number of effective votes would be even higher compared to plurality/FPTP. Not to mention that, judging by the experience of other PR countries around the world, the voter turnout would be higher since voters would have more choice and multi-party democracy.
Sorry, Steven I missed that aspect
so here under RCV we are talking of difference from 33 percent to about 50 percent,
or more likely an improvement from 50 percent (or 48 percent?) to 65 percent, on average.
RCV ensures just minimum of about 50 percent in a district. But of course some are elected with 60 or 70 percent in some cases.
whether it is 33 to 50, or 48 to 65, the difference seems less when you see that under RCV, in many districts, only under optimum conditions is there guarantee that even half the voters see their vote used to elect the winner.
I would like to see how they came up with that 17 percent difference.
is it difference from about a third casting an effective vote to about half casting an effective vote?
meanwhile, so much more improvement is out there if you merely open to multi-member wards or city-wide districting.
say you look at Wellington city election three-member wards (STV)
https://en.wikipedia.org/wiki/2022_Wellington_City_Council_election#Council
Successful candidates Range from least- to most-popular
2812 to 4242
Ratio 1 to 1.5
While even under RCV due to differences from ward to ward (population, voter turnout rate, rate of exhausted votes), you see range in successful candidate's vote tallies in each city each time. that range would allow a mere minority of voters to elect a majority of city councillors and wield power and rule. And minority rule is what we don't want in elections.
More on this in my Montopedia blogsite
https://montopedia.wixsite.com/montopedia/post/only-pr-ensures-majority-power-rests-on-majority-of-voters-edmonton-fptp-elections-versus-welling
Again, the failure to look at Australia is sufficient, in itself, to discredit this research. There have literally tens of thousands of RCV elections conducted here, with a variety of rules regarding requirements for a valid vote (most usual is a complete ranking, but we've also done "optional preferential" where you number as many candidates as you want. Voting is compulsory so there's no selection effect - . The percentage of invalid votes is small (a few per cent), and many of these are blank ballots (effectively, I'm only voting because I have to).
Yes John, could not agree more. Yet another "hole" in the research of US political scientists and their shaky political science. Here's another oddity abolut this: I have had US political scientists tell me that because Australia has compulsory voting, that somehow invalidates Australia as a research source, since it makes it hard to tease out if ranked ballots in the US are discouraging "certain" voters, i.e. racial minorities, from voting. If you study race in Australia, that effect won't show up, so the thinking goes, because racial minorities are required to vote. Another version of the "American exceptionalism" narrative that has never had much validity, in my view. Thanks for your thoughts.
I suppose it;s a fair point if you are specifically studying turnout. But that's a very limited question, and one that's relatively easy to test from exit polls and similar.
Yes, of course