[EM] Juho: Seat% and vote%--What I mean by "unattainable".

[Juho Laatu]

This mail has again very little EM content. Feel free to skip.

Juho


On 16.7.2012, at 21.56, Michael Ossipoff wrote:

> Juho:
> 
> {referring to making seat% as close as possible to vote%]
> 
>> I note again that I have not set any such target.
> 
> [endquote]
> 
> Excuse me, I thought you had, because you repeatedly spoke of that as a goal.

Better quote my text directly so I can't complain about quoting me wrong.

> 
>>> You said that you agree that people have a right to equal
>>> representation for everyone (too the extent achievable).
>> 
>> Not really
> 
> [endquote]
> 
> Ok, now that's progress. If you don't agree that people have a right
> to equal representation, then that explains a lot.
> 
> You continued:
> 
> , but if we want to set equal representation as our ideal target, we
> can approximate that with different seat allocation algorithms.
> 
> [endquote]
> 
> With some approximating it more approximately than others :-)
> 
>> 
>>> 
>>> Equal representation for everyone means equal representation for each
>>> person. Equal representation for each person means an equal number of
>>> seats for each person. An equal number of seats for each person means
>>> equal s/q
>>> (where q is a unit of population or vote).
>>> 
>>> Therefore, you agree that people have a right to equal s/q, to the
>>> extent achievable.
>>> 
>>> If you disagree with one or more of the statements in the two
>>> paragraphs before this one, then don't hesitate to say which
>>> statement(s) you disagree with, and why.
>> 
>> I think that was quite ok with the addition that we have set equal representation as our target
> 
> Sure, and if you';re saying that you don't consider that a right, then
> all bets are off.
> 
> You said:
> 
> , and we are talking only about approximately equal representation.
> 
> [endquote]
> 
> With LR, we're talking about approximately equal representation. With
> SL, we're talking about maximally equal representation.
> 
> You said:
> 
> I'm not sure if those paragraphs intended to say that SL or s/q or
> seats per person is the only correct way to measure "to the extent
> achievable".
> 
> [endquote]
> 
> What I said wasn't intended as a way to measure "to the extent
> achievable". Those paragraphs didn't speak of how to measure "to the
> extent achievable". Stick to one question at a time. Those paragraphs
> were intended to say only what they estsblished. Maximally equal
> representation for all means maximally equal s/q.
> 
> You speak of how to measure it. Putting everyone's s/q as close as
> possible to that quantity's ideal equal value is a way. It's a good
> feasible way. There are other ways you could do. For instance you
> could minimize the maximum s/q pair difference. SL doesn't do that. So
> far as I'm aware, that goal would require a trial-and-error procedure.
> But what SL does is a pretty good substitute. Besides, by putting
> _everyonn's_ s/q as close as possible to the ideal, you're doing
> something for everyone.
> 
> By the way, I've found out that Webster is about 1.9% large-biased (in
> U.S. apportionment). In general, it can be a few percent large-biased.
> 
> I don't know anything, really, about LR's bias. 1) Maybe it's unbiased
> like Webster (up to a few percent of bias). 2) Maybe it's unbiased
> like BF (completely unbiased if a uniform distribution is assumed).
> 3)Or maybe it's completely unbiased no matter what the distribution.
> 
> I doubt that it's #3, because the distribution's slope varies with q.
> I don't know of a demonstration that it's #2. So it can only be
> assumed that it's #1. But maybe it's #2. It could be either.
> 
> You certainly can't believe what you hear. Some say that Webster is
> completely unbiased. Some say that Webster is small-biased. But
> Webster is a few percent large-biased.
> So what we hear doesn't tell us much about LR's bias. Not unless
> someone shows why it's as they say.
> 
> If it can be shown that, for LR, it's #2, then LR is more unbiased
> than SL, and that definitely would count for something.
> 
> Mike Ossipoff

You make a lot of assumptions on what I think about different topics. That's of course interestng, at least to me :-). But for fluent discussion, maybe sticking to the technical topics could work better. The methods under discussion are an interesting discussion topic also themselves, without taking position for or against every detail.

Juho