Arrow's impossibility theorem does not apply to STAR voting. STAR voting is strictly more expressive than Ordinal Voting systems (i.e. RCV), while simultaneously being simpler to implement (and explain/understand). STAR voting also provides other benefits over RCV, such as being district summarizable.
I don’t have to know how STAR voting works to know that Gibbard’s 1978 theorem applies.
The only game-forms which are straightforward (meaning, one’s best strategy in what option to choose not depending on how one expects others to choose, or on others’ preferences) are the probability mixtures of “do whatever voter X said to”, “between option A and option B, determine how many voters expressed a preference for A over B and how many for B over A, and if the proportion preferring A is more than p%, pick A, otherwise B”, and “pick outcome C”, with the mixes being any mix of those over any combinations of A,B,C, and X.
(So for example, you could flip a coin, and if heads, select a random pair of candidates, and the winner is whichever of the two got more pair wise votes, if tails, select a random voter and go with who they said was their top pick, and if the coin lands on its edge pick Neil deGras Tyson. This would count as “straightforward” because in choosing how you want to vote, it doesn’t matter at all how anyone else is voting.)
(Another example is if there are just 2 candidates period. This is also “straightforward”)
Outside of such probability mixtures, there can always be situations where your estimate of how other people will vote would influence the most effective way for you to vote in order to best improve your expected outcomes according to your preferences.
Outside of such “straightforward” game-forms, there can’t be a function from (your preferences (like, vNM utility, not just a ranking) between the different outcomes) to (a dominant strategy for you).
No, it doesn't, but optimal voting under STAR is less dependent on other voters than in both Plurality and RCV systems. Under STAR voting, it would be far, far more practical to vote your conscience irrespective of how others are voting. This would encourage the middle majority to actually vote.
When further combined with other reforms, such as multi-member districts, this could completely transform the political landscape.
STAR is not easier to implement than ordinal voting methods in general, it may be easier to implement than IRV [0] in particular. Likewise, while IRV is not district summarizable, some ordinal methods are.
And outside of specialized circumstances that don’t apply to usual candidate elections, the additional information on STAR ballots compared to unforced preference ordinal ballots is noise of no consistent meaning.
Also, Arrows impossibility theorem applies to any balloting system that expresses ranked preferences and produces a ranked result, whether or not it provides additional information or compresses the available rankings for either the inputs or the result (typically, it applies to single winner election results which compress the results the same way FPTP ballots compress ballots—one first place and everyone else tied for not-first-place.)
[0] “RCV” is a name used by advocates to conflate IRV with ordinal methods generally, and accepted by some opponents of ordinal methods to conceal that their arguments apply only to IRV and not the broader class.
> STAR is not easier to implement than ordinal voting methods in general, it may be easier to implement than IRV [0] in particular. Likewise, while IRV is not district summarizable, some ordinal methods are.
Could you provide an example? Thanks in advance.
> Arrows impossibility theorem applies to any balloting system that expresses ranked preferences and produces a ranked result
My understanding is that Arrow's impossibility theorem does not apply to STAR or any other Cardinal Voting system. Is that incorrect?
As far as I know, STAR voting is a form of RCV. However, the Fairvote organization has advertised RCV with instant runoff voting as the counting method as being RCV.
Technically speaking, STAR voting is a form of cardinal voting, while ranked-choice encompasses the various ordinal methods.
The distinction here is that in STAR voting, I could rank two candidates equally to show that I have no preference between them. Also, I don't have to rank the candidates in order: I could give my first-place a score of 5, my second place a score of 2, and my third place a score of 0. If I scored my second place candidate as 3, this ballot could result in a different outcome.
