I read
Kos' post in favor of prop 77 (the California redistricting initiative) and then this
response from calpolitic. Ended up tracking down the
analysis [pdf] referred to in the Mother Jones article referred to in the response. I just skimmed it, but it's interesting.
It appears that the problem with politics-blind redistricting (as proposed in prop 77) is that it puts Dems at a disadvantage relative to Republicans because of the higher skew in Democratic areas. The two classic gerrymandering strategies are "pack" and "crack" -- pack your opponents' voters into a few districts to limit the damage, and crack the rest across multiple districts controlled by your majority to dilute them. It appears that following natural geographical borders may result in an inherent "packing" of Democrats.
At the top of p.19 of the PDF doc there's a chart profiling the skew in districts won by Dems vs. Republicans in '02 that bears this out --Dems won 21 districts by 80%+ where Repubs won 0 by that margin. Democrats are both "packed and "cracked."
This is unfortunate, to say the least. We didn't really need "inherent geographic disadvantage" added to our list of problems.
I still support 77, sorry (I note that the bias documented in the paper is in existing gerrymandered districts -- I don't see anything indicating it will get worse in politics-neutral redistricting) but I wonder if there's some other reasonable standard that could be applied in redisctricting that would address this inherent bias. I doubt it, frankly, which is why I still support 77, but maybe somebody has a suggestion.