Another problem with RCV is that it’s difficult to understand, at least when you get into the runoffs. That makes it harder for people to understand and trust the election system, which is also very important.
This is one big issue for me with IRV vs Cardinal systems. Both handle _weak_ spoilers but IRV doesn't handle ''strong,, spoilers. Understanding that these are two different types of spoilers is really important. It is "Bernie vs Hillary" and "Cruz vs Trump." Even with IRV we'd still see the parties trying to prevent strong primary candidates from running independently because they could spoil the election. With cardinal methods (approval, score, star) this is far less of a concern and thus I think more democratic of an election. I honestly think we should be much more concerned with strong spoilers than weak spoilers (weak spoilers don't affect elections much anyways).
STAR: 5, Score 4, approval 4, IRV 1, FPTP 1
It would be ranked in this order too, but there's more expressivity here than an ordinal system can provide.
FWIW, this already happens. That will also happen under any system. I'm not sure why you think IRV encourages moderates. It actually encourages extremist candidates which is why I'm particularly against it.
IRV/Approval/Borda/Condorcet voting systems all improve those factors. That they have some edge conditions that could be theoretically exploited belies that fact that plurality voting (ie, FPTP / one person one vote) is a complete train wreck for spoilers and favoring extreme voting.
No voting system is perfect (Arrow's impossibility theorem) but some are harder to exploit than others. Staying with a broken system because the proposed replacement isn't perfect is a poor justification.
The non-monotonicity shown there for instant runoff voting seems to be more undesirable than any downsides of approval voting.
IMO the point of elected representation is to be representative, not to elect specific people, so the possibility of a lesser known candidate winning over a major party's favorite candidate isn't a problem.
Arrow's impossibility theorem only says that no _ranked_ voting system is perfect, so it doesn't apply to cardinal methods such as Approval or Score.
> That they have some edge conditions that could be theoretically exploited belies that fact that plurality voting (ie, FPTP / one person one vote) is a complete train wreck for spoilers and favoring extreme voting.
I agree, but I think the nature of the exploitation should still be considered. For IRV, we would be introducing a dependence on unauditable voting machines to tabulate votes, which would make it harder to verify the integrity of future elections.
I tend to not get into this fight because 1) it extends to cardinal systems with Gibbard's and 2) Arrow's is more about social choice than voting.
I think the problem here is that people get caught up in VSE. I get it, that's what I was caught up in when I first started learning about this stuff. Then I had to really dig through strategic voting and spoilage. The nuance of the systems actually do make a very big difference in the outcomes.
Because for many municipalities, it's not whether they move to Approval vs. RCV or IRV or Star, etc - it's moving away from the dumpster fire of plurality voting.
I'm curious why you think this? I think exactly the opposite. RCV/IRV has major problems, especially for strong spoilers. They are also more difficult to tabulate and with the recent events of the last election the importance of ease to tabulate went up in my priority list (making me even question my preference of score over approval). In fact, IRV also favors extremists, which is something I really don't want.