[EM] Portland specifics, and ovals for approval cutoff

Tue Jul 16 06:00:53 PDT 2024

 So it seems that this is because PAV is an elect-all-at-once method rather than sequential, as opposed to there being something special about PAV per se. Of course it might be computationally too expensive to run PAV optimally in a given election anyway. There is also Phragmen, which is more proportional than PAV anyway, I would say. PAV fails what I would consider to be the basic PR criterion for an approval method - passing perfect representation when the number of elected candidates is the same as the number of voters. Phragmen doesn't have as good monotonic properties as PAV though. It passes, but in a weaker sense.
Toby
    On Monday 15 July 2024 at 23:20:51 BST, Closed Limelike Curves <closed.limelike.curves at gmail.com> wrote:  

 Richard, I'd suggest taking a look at the newest version of Wikipedia's PSC article (and the citations in there); nonpartisan PR algorithms other than STV are a very new field of research, and we're just barely starting to study what algorithms satisfy what proportionality axioms.
STV does try to put together a coalition, but this involves a very basic greedy search for cohesive/solid coalitions that support a single group of candidates. e.g. say a Hispanic Republican's preferences are something like "vote for a Hispanic Republican, then non-Hispanic Republicans, then a Hispanic Democrat"; whereas others Hispanic Republicans care more about ethnicity than about party. In this situation, there's no solid coalition for Hispanics or Republicans, so it's very prone to underrepresent one of these groups. IIRC the EVC's proposal for a proportional STAR algorithm works on basically the same principle, as does anything that assigns seats sequentially/greedily, which is why I tend to dislike them; whereas PAV is optimization-based, so it tends to find better apportionments across cross-cutting coalitions. I think Schulze STV also satisfies similar properties (but only locally, because global optimization across committees requires voters to give candidates ratings).
On Sun, Jul 14, 2024 at 10:22 AM Richard, the VoteFair guy <electionmethods at votefair.org> wrote:

On 7/13/2024 5:39 PM, Closed Limelike Curves wrote:
 > ...  my guess is the effect of STV is probably going to be limited
 > compared to other methods, because STV is only proportional for solid
 > coalitions, i.e. voting blocs moving in perfect lockstep with each
 > other, and only if they exceed a full quota. ...

When using STV, coalitions don't need to be "solid."

STV calculations automatically identify de-facto "coalitions."  Each 
candidate can be thought of as a de-facto coalition.

One way to think of this concept is that STV gives any coalition a 
"second try" when filling the second seat, and a "third try" when 
filling the third seat.

In contrast, IRV provides only "one try" for a coalition to elect their 
candidate.  So for IRV, a "solid" coalition voting with the same tactic 
does have an advantage compared to a loose (non-solid) coalition.

I'm not saying STV is better than PAV.  I'm saying this specific 
criticism of STV -- thinking that a coalition must be "solid" -- is not 
a valid reason to dismiss STV.

Richard Fobes

On 7/13/2024 5:39 PM, Closed Limelike Curves wrote:
> Ahh, yeah, any PR method will outperform winner-take-all on this metric. 
> If cities are using STV for their city councils, that might imply 
> single-winner IRV actively reduces minority representation.
> 
> That said, my guess is the effect of STV is probably going to be limited 
> compared to other methods, because STV is only proportional for solid 
> coalitions, i.e. voting blocs moving in perfect lockstep with each 
> other, and only if they exceed a full quota. So, for example, if ethnic 
> minorities tend to vote for a mix of white and minority candidates, or 
> if they're split across party lines (e.g. Hispanics only go about 60-40 
> for Democrats), this will tend to break up and dilute their interests. 
> It's not enough to have a quota of Hispanics who rate Hispanic 
> Republicans and Hispanic Democrats highly; you need to have a full quota 
> rank /either/ a Hispanic Republican or a Hispanic Democrat at the top of 
> their ballots. The effect is the same as center-squeeze, where your vote 
> gets "stuck" with a weak candidate who nevertheless has enough first 
> preferences to stay in the race for several rounds. By the time this 
> candidate is eliminated, the more-electable candidates with fewer first 
> preferences have been defeated.
> 
> But from what I can tell, this is much trickier to resolve than 
> center-squeeze; proportional Condorcet methods don't seem to have any 
> guarantees on how out-of-whack a coalition can get, although I think 
> Schulze STV has a nice local stability property.
> 
> Most ethnic minorities aren’t solid coalitions, although 
> African-Americans happen to be so overwhelmingly Democratic they might 
> act like one. Something like PAV that approximately satisfies the core 
> property should do better at giving minority voters more representation.
> 
> On Thu, Jul 11, 2024 at 8:40 PM Richard, the VoteFair guy 
> <electionmethods at votefair.org <mailto:electionmethods at votefair.org>> wrote:
> 
>     On 7/11/2024 10:47 AM, Closed Limelike Curves wrote:
>       > ... given there's no theoretical basis to think IRV would affect
>       > city council and mayoral seats differently. ...
> 
>     In the new Portland elections, "ranked choice voting" for city-council
>     seats is STV (the Single Transferable Vote) with three seats per
>     district, whereas "ranked choice voting" for Portland mayor is
>     single-winner IRV (instant-runoff voting).
> 
>     Of course Portland's three-seats-per-district STV city-council
>     elections
>     are going to yield more gender and racial diversity compared to
>     single-winner IRV for electing Portland's mayor.
> 
>     Richard Fobes
> 
> 
>     On 7/11/2024 10:47 AM, Closed Limelike Curves wrote:
>      > I'm going to go ahead and say I'm skeptical either of these
>     results will
>      > replicate, given there's no theoretical basis to think IRV
>     would affect
>      > city council and mayoral seats differently. My guess is both results
>      > would disappear if you used a hierarchical/partial pooling model to
>      > reduce the noise in the estimates.
>      >
>      > On Thu, Jul 11, 2024 at 10:45 AM Michael Garman
>      > <michael.garman at rankthevote.us
>     <mailto:michael.garman at rankthevote.us>
>     <mailto:michael.garman at rankthevote.us
>     <mailto:michael.garman at rankthevote.us>>>
>      > wrote:
>      >
>      >     And that’s bad! But it doesn’t negate the point I made. In
>     council
>      >     races it helps. And that’s good!
>      >     On Thu, Jul 11, 2024 at 1:44 PM Closed Limelike Curves
>      >     <closed.limelike.curves at gmail.com
>     <mailto:closed.limelike.curves at gmail.com>
>      >     <mailto:closed.limelike.curves at gmail.com
>     <mailto:closed.limelike.curves at gmail.com>>> wrote:
>      >
>      >          > In mayoral elections, RCV seems to *decrease* gender and
>      >         racial/ethnic diversity.
>      >         Sounds substantial!
>      >
>      >         On Thu, Jul 11, 2024 at 10:28 AM Michael Garman
>      >         <michael.garman at rankthevote.us
>     <mailto:michael.garman at rankthevote.us>
>      >         <mailto:michael.garman at rankthevote.us
>     <mailto:michael.garman at rankthevote.us>>> wrote:
>      >
>      >              > In city council elections, RCV has a small but
>     positive
>      >             effect on racial/ethnic diversity
>      >             Sounds substantial!
>      >
>      >             On Thu, Jul 11, 2024 at 1:26 PM Closed Limelike Curves
>      >             <closed.limelike.curves at gmail.com
>     <mailto:closed.limelike.curves at gmail.com>
>      >             <mailto:closed.limelike.curves at gmail.com
>     <mailto:closed.limelike.curves at gmail.com>>> wrote:
>      >
>      >                 Empirically, IRV adoption has no substantial
>     effects on
>      >                 diversity
>      >               
>       <https://www.newamerica.org/political-reform/reports/rcv-impact-on-candidate-entry-and-representation/ <https://www.newamerica.org/political-reform/reports/rcv-impact-on-candidate-entry-and-representation/>>. In theory, I'd expect a small decrease in representation for minorities, because people of color tend to have higher rates of spoiled ballots, so IRV should hit them the hardest.
>      >
>      >                 In general, the theoretical advantages of IRV
>     over FPP
>      >                 are outweighed by its practical costs (spoiled
>     ballots,
>      >                 lower trust in elections, cost, etc.).
>      >                 ----
>      >                 Election-Methods mailing list - see
>      > https://electorama.com/em <https://electorama.com/em>
>     <https://electorama.com/em <https://electorama.com/em>>
>      >                 for list info
>      >
>     ----
>     Election-Methods mailing list - see https://electorama.com/em
>     <https://electorama.com/em> for list info
> 
----
Election-Methods mailing list - see https://electorama.com/em for list info

----
Election-Methods mailing list - see https://electorama.com/em for list info

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240716/94c75516/attachment.htm>