There was some discussion on my earlier thread about Arnold's propositions about rules for drawing district lines that would eliminate the problem of gerrymandering. Wacky Hermit proposed a rule limiting the ratio between the perimeter and area of a district, and Ben Bateman proposed creating a computer model to optimize over a wider range of constraints.
My own thought is that attempts to group people that "fit" together into a district based on demographics is inherently anti-democratic. I think districts should be based on city and county lines, and I would propose two simple rules for two types of districts.
1. Type A districts can contain multiple cities. Any city in a Type A district must be contained entirely within that one district. More than one city can be in a Type A district, but none of those cities can cross the district line.
2. Type B districts can contain part of only a single city. Any city too large to fit in a Type A district can be broken into multiple Type B districts, but no two cities can share a Type B district.
These two rules would ensure that district lines are drawn that represent the local government structure. State and federal representatives should work with the same lines as local representatives; it'll make everyone more accountable, and reduce confusion and waste.
Update:
And then there's bioregional democracy:
Bioregional democracy (or the Bioregional State) is a set of electoral reforms designed to force the political process in a democracy to better represent concerns about the economy, the body, and environmental concerns (e.g., water quality), toward developmental paths that are locally prioritized and tailored to different areas for their own specific interests of sustainability and durability.
Interesting concept, and the article has some examples of how it's put into practice. I'm not an environmental fanatic by any means, but I like the idea of political structures built with environmental interest geographical boundaries in mind -- that way, people in a given environment can treat it as well or as poorly as they want, with minimal effect on other groups.
Update 2:
Richard Tallent emailed with an algorithm he devised to allow voters to draw district lines (with some pixelization).
There is only one fair method to creating districts: as much as possible, let the voters choose for themselves with whom they want to vote. Create a number of similarly-populated geographic "blocks" using a semi-random computer algorithm that is simply taught to avoid splitting up cites or counties where possible. Let each voter pick up to "X" other blocks along with his own, where "X" is the size of the represented region divided by the average number of people in each block. Require that voters can only choose regions either neighboring their own block or sharing some border with at least two other blocks they have also selected. Aggregate these to determine the wishes of each block, and use these numbers to shape the borders with a deterministic best-fit algorithm.
Interesting and possibly effective, but voters won't trust a system that's too complicated for them to understand -- just remember the recent EU Constitution debacle. Why not just take the next step and untie representation from geography entirely?
Type-A would likely help Democrats. A city like Houston has enough conservatives for a Republican congressman or two, but a city-wide election with slates would go entirely D. I can't think of a counterweight that would help Republicans.
Type-B would require thousands and thousands of congressmen as there are many towns with populations in the hundreds.
RA: Maybe I wasn't clear. The system would have both types of districts at the same time, as appropriate for a given region. Type A would be preferred, and could be as geographically large as necessary to hold lots of little cities. Type B would be created when a city was too large for a Type A.
Perhaps the underlying problem is that people don't agree on how districts should work. One view, generally favored by liberals, is that the legislature should mirror the populace as closely as possible in terms of political views and proxies for political views, such as race. So if a state's citizenry is 10% black, then 10% of the legislators should also be black. If 5% of the citizens are communists, then so should 5% of the legislators, and so on.
This liberal view sees all political disagreements being resolved in the legislature, with essentially none of them being resolved in the selection of representatives.
The conservative view says that it's a good thing to resolve some of the political questions in the selection of representatives. On this view, if 5% of the citizens are communist, then none of the legislators should be communist, because it isn't a view that the public accepts. The purpose of districting, on this view, is to narrow and focus the political debate towards the center, to achieve a more precise version of what the majority wants.
Without any agreement, even in theory, on how to draw districts, the process may forever be dominated by majorities seeking to increase their control.
One paradox of the liberal view is that it runs counter to the way you draw lines to entrench a political majority. Gerrymandering done right means that you concentrate your opponents' voters into near-100% districts, and create your own districts with maybe 60/40 security. But if your goal is to ensure that the legislature contains kook from your side, then you have to concentrate your own voters into extreme districts, where those votes would have served you better elsewhere.
BB: That's a pretty good description of the two positions, I agree.
One question that's rarely addressed is, why should representation be tied to geography at all? Mainly because of tradition and ease, I suppose.
The water quality comment really perked my intrest and it relates quite well when considered. One of the key problems with the current issues of water quality is that for the most part it hasn't been done graphically (now there are those out there pushing that management method, but the regional boards are less than accepting of the practice, hell the southern california board rejected a model study because it was scaled on the Reynolds number!), but as a cut and dry this is the rule for everyone so deal with it.
When lines are drawn too arbitrally you don't deal with the issues that affect an area, to a degree you get the Senate, which the House isn't supposed to be. A example that comes to mind in water quality is ventura county where they high levels of naturally occuring selinium in their water, but they are require to meet TMDLs for the reciving water lower that the levels that are in there, its an impossible feat unless they dilute the reciving water enormously. We got nailed on a project in Kern for heavy metal (arsenic) polution in groundwater but its naturally occuring. Simply because the regulations aren't developed in a way to consider issues specific to that region. Much in the way arbitray districs with out some kind of common issue base in doesn't serve to fix any of the problems in the region. Plainly it just doesn't make sense to make a distirct 60% Palos Verdes and 40% Inglewood because the issues that affect those areas are radicaly different and in the process the ability to address to probems in inglewood are severly diminished.
Of course that being said every district should be atleast 60% republican.
When I mentioned grouping communities together based on various considerations, I was not thinking of race, such as creating an all- or majority-black district. I was thinking more along the lines of which communities have social and/or commercial ties with each other and thus have similar interests. For example, my parents used to live in a little rural town called Roosevelt. Roosevelt is about 45 minutes' drive from Vernal, a slightly larger small rural town. Because of some local politics that I won't get into, Vernal is the site of all the area's shopping (i.e. Wal-Mart) despite not being very much bigger than Roosevelt. It would make more sense to locals to lump Roosevelt into a district with Vernal than to lump it in with Heber which is more or less the same distance from it, because Roosevelt and Vernal have commercial and social ties. Nobody in Roosevelt drives into Heber to do their weekly shopping.