[EM] Once and for all--Bias vs transfer properties

[MIKE OSSIPOFF]

  With the Bias-Free method, because, in every cycle, there is no net dS/q, 
that means that, any two states, in any two cycles, even at opposite ends of 
the population range, will never have any difference in their 
seats-per-quota expectation. Bias-Free is genuinely entirely unbiased.

  That means that Webster is very nearly unbiased, because Webster's 
round-off points are nearly the same as those of Bias-Free. And it means 
that Huntington-Hill gives significantly more seats per person to smaller 
states, because Huntington-Hill's round-off points are considerably lower 
than those of Bias-Free. Huntington-Hill is significantly biased in favor of 
smaller states.

What about Hill's transfer property? Webster and Hill both are as 
proportional as possible, according to their respective measures of 
proportionality-difference. Bias-Free doesn't, strictly speaking, have a 
transfer property. Let me point out two differences between bias and 
transfer properties:

  1) Because Bias-Free is unbiased, any increased unproportionality, in 
comparison to Hill or Webster, must  be random and un-directed. Bias, on the 
other hand, is a directed, systematic seats-per-quota difference against a 
specific known end of the population spectrum. That's worse than a little 
random, undirected unproportionality. Random, undirected unproportionality 
isn't unfair. Bias is unfair.
  Bias-Free completely eliminates bias.

  2) Hill's transfer property is about a _matter of degree_. Hill, by its 
proportionality standard, has less unproportionalilty than any other 
allocation. Less doesn't mean none. Less doesn't even mean a lot less. Any 
method has some unproportionality. It's unavoidable when we have to award 
whole numbers of seats. But Bias-Free _entierly_ elilminates bias.

  Of course, not only is Webster simpler than BF, and precedented, but it 
also compromises, having a transfer property in addition to nearly no bias. 
Still, advocacy and adoption of BF would show an
   intention to genuinely entirely eliminate bias.

  And please don't let anyone tell you that bias is difficult to define. 
Bias is a systematic disparity in
  seats-per-quota expectation for states of different population. Bias-Free 
has no bias.

Mike Ossipoff

_________________________________________________________________
View Athlete’s Collections with Live Search 
http://sportmaps.live.com/index.html?source=hmemailtaglinenov06&FORM=MGAC01