[EM] How unfair is distribution-caused bias?

[MIKE OSSIPOFF]

I prefer that an apportionment result have little or no correlation between 
q and s/q. And CW & AR achieve that. (Well, AR might need a little 
improvement, but it's potentially the best method).

BF is unbiased if the state-size frequency distribution is uniform. With the 
existing distribuiton, when BF is used, the greater frequency of smaller 
states will mean that if you're in a small state, it's more lilkely than not 
to be in the lower part of a cycle. So BF will show a very little large-bias 
for that reason.

Now, I don't say it's the state's fault that they're small states--even if 
they actively discourage people from moving there :-)  And it's not their 
fault if the small states are more numerous than a uniform distrlibution 
would have it. But that isn't the method's fault either. So I suggest that 
maybe it's worthwhile to distinguish between method-cause bias and 
distribution-caused bias. Maybe the latter isn't as unfair. Maybe BF isn't 
doing anything wrong when s/q is affected by the distribution.

Mike Ossipoff

_________________________________________________________________
Get FREE Web site and company branded e-mail from Microsoft Office Live 
http://clk.atdmt.com/MRT/go/mcrssaub0050001411mrt/direct/01/