[EM] Hare (aka IRV) versus STAR

Tue Apr 9 17:31:10 PDT 2024

(This is a re-send with a few typos and omissions fixed. I was tired 
when I sent it the first time.)

It has been recently claimed here that STAR is better than Hare, and 
that the promoters of STAR  have been much
more honest than the US promoters of Hare (aka IRV or RCV).

I strongly disagree.

Score is Approval with a  "I wish to weaken the effect of my vote for 
the sake of being more sincere/expressive" box/button.
So it is strategically equivalent to Approval while being more 
complicated and less fair.

And Approval has a quite good reputation here because it meets Favorite 
Betrayal  (aka FBC) and compared with FPP the winner
will strongly tend to have higher social utility and  be much more 
likely  a sincere Condorcet winner.  Also, and not unrelatedly,
it has a bias toward centrists that some people think is wonderful.

But some people seem to think that adding a Top-Two Runoff (automated in 
the case of STAR) to Score (to make STAR) is just
a harmless little gimmick that just makes the method "a bit more 
accurate", brings it into compliance with Condorcet Loser
and so must make it more "Condorcet efficient".   ("Sky-high" according 
to CLC here).

But actually it makes the method profoundly different and very bad. It 
seems to me that the inventors of STAR must have been
motivated by three priorities:

(1) the method isn't  Hare,

(2) the method, in a purely technical and completely useless way, 
apparently meets Mono-raise (aka Monotonicity).

(3) subject to being saleable to and understood by  not-so-deep 
thinkers, the method be as bad as possible.

 From the "equal-vote" website: https://www.equal.vote/

> Ranked Choice Voting, where voters rank candidates in order of 
> preference has been lauded as a solution, but in elections where the 
> third candidate is actually competitive,vote-splitting remains a 
> serious issue <https://www.youtube.com/watch?v=yhO6jfHPFQU&t=169s>and 
> RCV only offers a marginal improvement compared to a primary and  
> general election with Choose-One Plurality voting.

> Luckily, many voting methods are can effectively prevent 
> vote-splitting. As it turns out, when voters can weigh in on each 
> candidate individually, when all ballot data is counted, and when 
> voters are able to show equal preference, vote-splitting can be 
> eliminated. All voting methods which do this pass theEqual Vote 
> Criterion <https://electowiki.org/wiki/Equal_Vote_Criterion>, 
> including STAR Voting <https://www.starvoting.us/star>,...

The "Equal Vote Criterion" is just  propaganda nonsense: 
https://electowiki.org/wiki/Equal_Vote_Criterion
> The Equal Vote Criterion orEquality Criterion 
> <https://www.equal.vote/theequalvote>is avoting method criterion 
> <https://electowiki.org/wiki/Voting_system_criterion>which requires 
> that a voting method ensure that every voter may cast a vote which is 
> as powerful as a vote cast by any other voter. Voting methods which 
> pass the Equal Vote Criterion do not exhibitvote-splitting 
> <https://electowiki.org/wiki/Vote-splitting>or the "Spoiler Effect," 
> ensuring that every vote can cast anequally weighted vote 
> <https://electowiki.org/wiki/Equally_Weighted_Vote>.
> Choose-One Plurality Voting (First Past the Post) and Instant Runoff 
> Voting (often referred to as Ranked Choice Voting) do not satisfy the 
> Equal Vote Criterion.

This is just dishonest blather. If anything meets this very vague and 
confused "criterion" IRV (aka Hare) certainly does.

The classic scenario that motivated some people get negative about Hare 
(and also methods like Min-Max Margins):

49 Bush
24 Gore
27 Nader>Gore

Gore>Bush 51-49,   Bush>Nader 49-27, Nader>Gore 27-24.

Hare eliminates Gore and elects Bush, so the Nader voters whose Gore> 
Bush preference was strong had incentive to use the Compromise
strategy and vote Gore>Nader ("betraying" their sincere favourite).  If 
the method was Approval they could have approved both Nader and
Gore, preventing the election of Bush without having to vote their 
sincere favorite below equal-top.

But in this type of scenario STAR does no better than Hare. The Nader 
voters would have incentive to give Nader zero points.

"Traditionally" Hare's  vulnerability to Push-over strategy has said to 
be a result of it's failure of Mono-raise.  But STAR is much more vulnerable
to Push-over.

Say you are sure that your favourite will make the final two. In that 
case then you have incentive to give every candidate that you are sure your
favourite can beat 4 or 5 stars.  If 5 stars then you are relying on you 
favourite winning the runoff without your help, but if 4 stars then you 
might
fail to get one of the predicted sure-loser turkeys into the final.

In a Hare Push-over strategy scenario, the strategists rely on their 
favourite winning the runoff against their own votes, i.e. with their 
votes supporting
the turkey against their favourite. This makes it much more risky (more 
likely to backfire) and difficult to coordinate than is the case with STAR.

The equal-vote site has a link to a quite ok video on the Favorite 
Betrayal Criterion.  I find that weird and misleading, because STAR 
badly fails FBC.
https://www.youtube.com/watch?v=JtKAScORevQ

 From https://www.starvoting.org/

>
>     Why STAR Voting?
>
> Voting reform is the keystone. A single cause with the potential to 
> empower us to be more effective on every other issue we care about.
>
>  *
>
>     Honesty is the best strategy. Strategic voting is not
>     incentivized. <https://www.starvoting.org/strategic_voting>
>
>  *
>
>     Even if your favorite can’t win, your vote helps prevent your
>     worst case scenario. <https://www.starvoting.org/how_to_vote>
>
>  *
>
>     Highly accurate, no matter how many candidates/parties are in the
>     race. <https://www.starvoting.org/accuracy>
>

I'm not sure exactly what "accurate" is supposed to mean, but I refute 
the suggestion that these claims are more true of STAR than they are of 
Hare.

In the poll I will vote STAR below Hare and Approval and all the 
Condorcet methods.

Chris

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240410/f512bed8/attachment-0001.htm>