- What they are referring to as RCV is called Instant Runoff Voting (IRV). IRV is a RCV, but RCV isn't IRV. Think "Rock and Roll is The Beatles" vs "The Beatles are a Rock and Roll group."
- IRV isn't that great when it comes to Voter Satisfaction Efficiency (VSE)[0]. There's even other RCV methods like Ranked Pair[1] that are exactly the same for the voter. Essentially IRV is a 6% improvement upon plurality (what we currently use) but RP is a 15% increase from plurality. So I ask, why go to IRV when we can do twice as good? (or better! Keep reading)
- Fair Vote claims that IRV has a high resistance to spoiler candidates (a candidate that takes away votes from another similar candidate). This is actually why the distinction from RCV and IRV is important and why Fair Vote is misleading. While RCV is better than plurality in this respect, some RCVs are better than others. IRV doesn't adequately solve this issue (but there are other RCVs that do!). IRV fails when two candidates are similar in positions and similar in popularity. [2] This is specifically the kind of spoiling that we are trying to prevent. We're trying to prevent a candidate like Sanders from spoiling Biden (or vise versa) not Jill Stein spoiling Biden. IRV prevents the latter, but not the former. I am specifically upset with Fair Vote because they are misleading in this context. Because they claim IRV := RCV, they take all the things different RCV methods solve and claim IRV solves them. This is simply not true (and why I make the first bullet).
- There are just better methods besides ranking (also called "ordinal" methods). There's a class called Cardinal[3], that is actually simpler. The simplest is called approval. An example of approval is Netflix's rating system (binary). It does pretty well. Better yet is range/score voting. This would be the old Netflix system where you rated out of 5 stars. The difference in ranked vs cardinal is that you rate each candidate or policy independent of the others. This allows you to be more precise/honest (you may actually like two candidates equally!). It also reduces complexity because a common accident in ranking is rating two candidates the same. In ordinal systems this is a bug, in cardinal this is a feature.
- I want to specifically mention STAR voting[4]. While STAR isn't perfect, it much better handles various types of spoiler candidates, has a high VSE (similar to that of RP), and handles strategic voting well. The last is the major reason you should choose STAR over RP (and there's no reason you should choose IRV over RP, because RP is strictly better). While all three systems encourage honest voting (in a single agent scenario) strategic voting is always an issue. Under no circumstance is STAR worse than IRV because STAR handles strategic voting. Specifically the most important kind is the "1-sided strategy"[5] which is arguably very common.
- Lastly, I want to appeal to authority. Kennith Arrow (who you may be familiar with from Arrow's Impossibility theorem and Gibbard's extension) is a Nobel Laureate and said cardinal systems are "probably right."[6] And I think many of us know science speak into the confidence that conveys. That link is a full interview with him from Election Science. It is a good read/listen.
TLDR: We want a good voting system, use STAR, not IRV.
[0] https://electionscience.github.io/vse-sim/vse.html
[1] https://en.wikipedia.org/wiki/Ranked_pairs
[2] https://www.electionscience.org/library/the-spoiler-effect/
[3] https://en.wikipedia.org/wiki/Cardinal_voting
[4] https://en.wikipedia.org/wiki/STAR_voting
[5] https://electionscience.github.io/vse-sim/VSE/
[6] https://www.electionscience.org/commentary-analysis/voting-t...