Which is why we need instant-runoff voting in the US, so we can vote for third party candidates but still choose between the top two candidates if our preferred candidate loses.
There is no doubt that IRV would be superior to what we use today, plurality voting.
However, if we achieve the momentum necessary to reform voting in the United States, it would be a shame to squander that momentum to receive IRV rather than an easier and better model such as Score voting or Approval voting.
instant-runoff is sold on false premises. Advocates suggest that you can honestly vote and not see spoiler effects. But instant-runoff eliminates a candidate in each round, and when your 2nd choice gets eliminated before your 1st choice, your 2nd choice never gets counted. This results in a wide range of messed up results that spoil elections and can fall back to strategic voting.
The best answer is Star voting (Score Then Auto-Runoff), which just does one instant runoff between top scoring candidates, thus counting everyone's scores equally. See http://www.equal.vote for details
No, STAR is not the best answer; it has the same problem as all scoring (instead of ranking) based methods—there's no clear mapping between actual subjective preferences (which can be consistently reduced to ordinal, but not interval, measures) and ballot markings, with the result that the same preferences by different voters produce different ballot markings. This is especially problematic in multicultural jurisdictions, because score ratings aren't distributed among the space of markings consistent with a set of preferences consistently, but instead show clear cultural differences. (This is a well-documented, though rarely-addressed, problem with star ratings in general; one example of a difference on this axis is that with the same degree of satisfaction measured vy other means, Americans of European ancestry give significantly higher ratings than Asian-Americans on fixed-rate scales.
This problem (the inconsistency in general, not the cultural variation though that underlined the problem) makes STAR and other score-based methods utterly unsuitable for most public elections.
If you want to fix the problems caused by loser elimination in IRV, just drop loser elimination and accept as winner the candidate who gets past the bare-majority threshold first as you check first all first preference votes, then total of first and second preference votes, and so on.
> no clear mapping between actual subjective preferences (which can be consistently reduced to ordinal, but not interval, measures) and ballot markings, with the result that the same preferences by different voters produce different ballot markings
No, you completely misunderstand. There's no universal reference for what scores "mean". You can give candidates from, say 0-5 stars in terms of how much support you want to give them in the race. That's it. 5 = most support, 0 = no support. It's ONLY relative to the actual pool of candidates in the election. People can just learn to understand this clear fact, and ballots can be marked as such as "least support" to "full support". And educating people about this is easier than educating them about IRV and all it's weirdness (and more honest than making false claims about IRV). There are NO cultural issues or biases or problems here in terms of what a score "means".
> If you want to fix the problems caused by loser elimination in IRV…
I don't know what you mean here, but it isn't IRV, it's some other rank system. Tons have been proposed, none are perfect, all are far harder to calculate and understand than Star voting. Ranked Pairs is actually pretty good, probably the best way to handle ranked ballots. I'm not sure what you're proposing.
