The decision in Vieth v. Jubelirer, a Pennsylvania political gerrymandering case.
Since an equal protection argument under the 14th amendment seems unavailing, I would be strongly in favor of a Constitutional amendment requiring that "All legislative districts shall be drawn according to an algorithm that takes no notice of political affiliation." That still leaves authority with the states and smaller units to choose a method of drawing legislative boundaries, but removes the kind of political gerrymandering we've seen in Colorado, Texas, Pennsylvania, and elsewhere in recent years. Justice Stevens quotes a North Carolina state senator as saying "We are in the business of fixing elections", with regards to drawing district boundaries.
When it came to choice of algorithm, as a first guess, I would probably go for one that used high-population precincts as seeds and drew boundaries to equalize population while minimizing the ratio of district perimeter to district area. An appropriate weighting function for a least-squares minimization ought to be pretty easy to define. An alternative, one I know to be favored by Ginger, acts to preserve historically-related communities within single districts; it would be reasonably easy to design a weighting function that took both our concerns into account.
Regardless of the exact choices, redistricting for partisan benefit clearly violates fairness in a fundamental way. If the courts do not find enough guidance to act to preserve fairness, we should act as a people to ensure that voting for our legislative leaders is a fair and dynamic process.
I take the opposite tack. I think we should gerrymander the crap out of our districts to weight towards every one having an even 50%-50% split of Democratic and Republican voters. Then districts will be won by whoever motivates their base and mobilizes their voters the best. We'd see more turnout, less apathy, and fewer guarantee outcomes.
But then, I'm crazy.
The minor problem with that scenario is that it denies third parties. Our system is already skewed toward remaining two-party, and inertia makes it difficult for an insurgent party to replace one of the existing parties. I'd still prefer not to build in additional hurdles.
The major problem is that it only works in states with a 50-50 split. Suppose the state is 55-45; either all of the districts will also be 55-45, or all but one of them will be 50-50 and the remaining district will be 80-20. In the former case, you've got instant state-wide gerrymandering via minority party dilution; in the latter case, you've got an artificial distribution whose final district's composition and likely geographic appearance are gigantically kluged. Ugly.
Perhaps Andy was implying that we should kill off the excess until we have a true 50:50 split. He *is* crazy, you know.
The problem with basically letting a computer generate districts through geometric constraints is that it completely ignores the human element.
One human element with near universal support is that racial minorities deserve representation. It would be hard have a computer program ensure such representation when the program takes into consideration perimeters and square miles.
And once you've considered that human element, the camel has its nose under the tent.
One could easily adapt the weighting function to include a term representing racial composition, depending on what you wanted to do with it. E.g., the weighting function could prefer districts that had the same racial distribution, or it could prefer districts that had maximally different racial distribution, or it could prefer districts that had racial distribution most similar to their seed precincts, etc.
The more you put into the weighting function, the less stable it is.
However, I think that the primary impetus for minority-majority districts has been the historical fact that back in the day, districts were drawn to eliminate minority representation. A race-neutral algorithm won't do that.
I'd argue, in fact, that if we continue to cast politics as black v. white, we're perpetuating racial injustice instead of ameliorating it.
A race-neutral algorithm would probably satisfy equal protection under the 14th, and hence pass the VRA.
I think you have the concept of precincts and redistricting backwards. Precincts are drawn *after* redistricting is completed. If you didn't draw precincts after redistricting is complete, it would be near impossible to simulataneously create equal-population districts for Congress, city council, state rep, state senate, SBOE, etc. with a single precinct scheme. Therefore, it would be difficult to identify "seed" precincts prior to redistricting.
Your idea of "seed" precincts is entirely unworkable on a practical basis. Precinct lines themselves are a creation of politicians. What is to prevent a county official from creating several large conservative white suburban precincts with the intent that they will be identified as "seed" precincts?
Your treatment of the VRA is problematic, too. Court opinions have basically said that performing minority districts cannot be dismantled. Courts will continue to apply the Gingles test to districts even if your plan is adopted. (The Gingles test does not need evidence of intent to declare a violation of the VRA.) A computer that destroys a performing minority district would be just as distressing to a court as if politicians had done it. Your suggestions appear to me to repeal the VRA as it has been interpreted for 40 years. I am not willing to go that far.
I support the VRA and am confident that even a GOP Congress will extend its life beyond 2007 when it is set to expire.
Fine; if not precincts, then "population cells"--basically the smallest geographic unit of population. I'd doubt we'd want a unit smaller than a single building, but if the data exists, one can design the algorithm to use it.
Of course, the more cells, the more computing time necessary to optimize the boundaries. Even if we're dealing with an n-squared problem, we're probably not bumping up against any practical limits.
Remember that a properly-optimized scheme shouldn't depend on how the districts were seeded, so politicization of the seed cells ought not to matter. Depending on how the problem is defined, you might not need seed cells at all.
It may be that I'm misunderstanding how precincts are organized, but I have never noticed that my local precinct offered any split ballots, which means that it is necessarily within a single district for all possible district-based elections. (Of course, a properly-designed interface for an electronic voting scheme would hide that.)
As to the VRA, it's not going to expire in 2007. Certain elements of it, primarily pre-clearance of districts in the South due to the South's history of fucking over blacks, are going to expire, but the fundamental provision that, hey, you gotta let everybody vote, that's permanent.
It's been forty years. Almost two generations of people have grown up under pre-clearance. It seems likely that they've learned their lesson, and it behooves us to stop assuming that they're racist assholes.
I think that a non-partisan, race-blind approach to drawing districts is a significant enough departure from the process heretofore that it can be distinguished from the existing jurisprudence. Not to mention, if it's implemented by way of a constitutional amendment, it'll automatically trigger new independent new jurisprudence.