deluxev2
11hr ago
STV run with one seat and a 50% quota (under most methods of doing STV) is equivalent to IRV. I prefer the mental abstraction of STV, but your right that it isn't common parlance to use it in that way.
-
IRV is clone independent but still falls to the center squeeze (where several nearby candidates can choke out the center of the group). Also, every single Condorcet method is clone independent if there exists a Condorcet winner (which polling suggests is over 90% of elections). Many of these algorithms will choose something from the Smith set where it isn't even clear mathematically what you could do that is better, it just chooses a different one from the Smith set if there are clones. These are all ranked choice systems that can be computed in a single pass over the ballots. Why are we looking at IRV among the ranked choice methods?
-
Approval at least has the guarantee that you should never rate something you like less higher than something you like more. Approval and IRV have relevant tactical voting in the same situations: when your preference is close to losing to something more moderate. Both ask you to downrate the moderate if you think you can win, or uprate the moderate if you think you can't. IRV requires you to tell an outride falsehood to do this. In terms of the benefit you get from tactical voting, it is pretty similar across the two methods (both about 10-20% of what you get in FPTP).
-
IRV does not choose a Condorcet winner, and thus does not invoke the Black median voter theorem. There are many ranked choice algorithms that do, but not IRV.
I only bring up Condorcet as if we are going to be doing ranked choice ballots we have all these other better options. Options where counts can be aggregated from different polling centers and that provide better mathematical guarantees. I don't think later-no-harm is a good outcome. If a candidate B becomes more popular with a subgroup, but doesn't reach the threshold of being their first choice, candidate B should be more likely to win the election. I'd much rather have monotonicity, so that rating a candidate higher makes them more likely to win.
I'll grant that your threshold choice is inherently tactical, but I think it is a much better brand of tactical than other things that fall under the label. In IRV, tactical votes are misrepresenting your preferences to the voting system. You have to lie about who you actually like and dislike. In approval voting, you are compressing down the vote to provide as much information as possible to the voting system. It is a "true preference" that you like everyone you voted for more than everyone you didn't. It doesn't reward liars, it rewards those in touch enough to know roughly the bounds of possibility for this election.
But yeah, I guess there are some real selling points of IRV:
-a lot of the other ranked choice methods aren't clone independent which is definitely a problem that needs a solution
-you can just walk in and list your candidates in order and be voting "optimally for your desired outcomes" a good chunk of the time.
I still think ranked methods aren't worth the cost (really try it, with a group of even 5 people trying to decide what to do, ranked methods are demolished by approval voting in terms of implement-ability, then extend to the entire country). IRV among them is particularly bad for counting but is among the better set of election properties and especially explain-ability.
(Also, you've been a great conversation partner, kudos and gratitude)
More options
Context Copy link