Gerrymandering

Both Kevin Drum and Matthew Yglesias are talking about a subject that we can agree on--gerrymandering sucks!

Kevin's idea to fix it is a bit wacky, and Matthew's fix is a radical reform from the American district concept (For a look at what his ballot could look like in California see under the extended entry something I shamelessly stole from commenter "Bumperstickerist").

My concept is shared by some other people in a few of the comments. Take a fairly simple computer program and feed it only a very few variables. Make compactness an important variable or maybe general fit to county lines. Feed it only population data, no race data, no party affiliation data, no religious data, no economic data. Make the program simple enough that it can be replicated on any home computer. Use publically available census data. After the first decade draw the lines with a slight weighting to keeping the district close to its original boundary. That's it. No safe districts based on gerrymandering.

Based on the Hare system for proportional representation. Ballot for California.

Please Rank the Following candidates in order from 1-54. Where 1 is the candidate you most prefer and 54 the least. By least prefer we mean 'you don't like so much'. Please assign only one number per candidate.

___ Candidate A
___ Candidate B
___ Candidate C
___ Candidate D
___ Candidate E
___ Candidate F
___ Candidate G
___ Candidate H
___ Candidate I
___ Candidate J
___ Candidate K
___ Candidate L
___ Candidate M
___ Candidate N
___ Candidate O
___ Candidate P
___ Candidate Q
___ Candidate R
___ Candidate S
___ Candidate T
___ Candidate U
___ Candidate V
___ Candidate W
___ Candidate X
___ Candidate Y
___ Candidate Z
___ Candidate AA
___ Candidate BB
___ Candidate CC
___ Candidate DD
___ Candidate EE
___ Candidate FF
___ Candidate GG
___ Candidate HH
___ Candidate II
___ Candidate JJ
___ Candidate KK
___ Candidate LL
___ Candidate MM
___ Candidate NN
___ Candidate OO
___ Candidate PP
___ Candidate QQ
___ Candidate RR
___ Candidate SS
___ Candidate TT
___ Candidate UU
___ Candidate VV
___ Candidate WW
___ Candidate XX
___ Candidate YY
___ Candidate ZZ
___ Candidate A1
___ Candidate A2
___ Candidate A3
___ Candidate A4
___ Candidate A5
___ Candidate A6
___ Candidate A7
___ Candidate A8
___ Candidate A9
___ Candidate A10
___ Candidate A11
___ Candidate A12
___ Candidate A13
___ Candidate A14
___ Candidate A15
___ Candidate A16
___ Candidate A17
___ Candidate A18
___ Candidate A19
___ Candidate A20
___ Candidate A21
___ Candidate A22
___ Candidate A23
___ Candidate A24
___ Candidate A25
___ Candidate A26
___ Candidate B1
___ Candidate B2
___ Candidate B4
___ Candidate B5
___ Candidate B6
___ Candidate B7
___ Candidate B8
___ Candidate B9
___ Candidate B10
___ Candidate B11
___ Candidate B12
___ Candidate B13
___ Candidate B14
___ Candidate B15
___ Candidate B16
___ Candidate kind of dumb idea in practice, no?
___ Candidate B17
___ Candidate B18
___ Candidate B19
___ Candidate B20
___ Candidate B21
___ Candidate B22
___ Candidate B23
___ Candidate B24
___ Candidate B25
___ Candidate B26

Please check your ballot to make sure you have 54, and only 54 marks,and that no mark is repeated.

I have enough trouble with the CA ballot. I don't need this!

Comments

[sidereal]

"Take a fairly simple computer program ..."

Done.
Go here, click through to 200404, create an account (or BugMeNot), and get the source code for the article "SQL, Data Mining, & Genetic Programming" by Brian Conolly. Or just buy the back issue. It's a genetic algorithm for developing fair, compact districts. Non-deterministic, but Congress can specify the random seed for reproducability.

The problem here is not that no solutions are available, it's that there are way too many interested parties for any attempted solution to make it through bureaucracy unscathed.

[Fledermaus]

Gerrymandering is a real problem. I like your computer program idea, although it would still be vounerable to manipulation in its creation, at least it would be an argument about rules rather than specific districts.

Of course, as aptly pointed out by Publius the real problem is with legislative control.

[Sebastian Holsclaw]

Yes, as Publius says, the major problem isn't the availability of an easy solution--it is the fact that the solution would have to be be passed by Congressman who are elected under the current system. (Or maybe by referenda?)

I hate the referedum system, but this would be the classic case for it--dealing with a simple issue in a simple way that goes over the heads of the legislature because they are personally interested in the outcome.

[Anarch]

I hate the referedum system, but this would be the classic case for it--dealing with a simple issue in a simple way that goes over the heads of the legislature because they are personally interested in the outcome.

But it would have to be a multiplicity of referenda simultaneously enacted across all the major states in order to work. That's the catch: it's either multilateral disarmament or nothing at all.

[marguerite]

maybe if every other state had a referendum for a law that said "we'll do it if Texas does it"? (Or, all the liberal states say "we'll do it if Texas does it," all the conservative states say "we'll do it if California does it", and California says "we'll do if Texas does it.)

completely harebrained, I know, but I don't know of a better way.

[CharleyCarp]

I'm not sure the "disarmament" effect ought to count for much. Assuming that fo any possible congressional district centrists outnumber extremists of either party -- and I think this IS a fair assumption -- the majority of the people of each district win even if only their district is constitued on a non-partisan basis, without regard to any other district much less an other state.

A national movement of state referenda. Sort of like term limits in the early 90s.

[John Kelsey]

Maybe what's needed is some kind of delay between when the law is passed and when it's implemented. Pass a law changing the formula for redistricting in a way that is algorithmic and known in advance, but don't have it apply for, say, ten years.

Also (others have probably thought through more of this), it might make it more palatable if it were possible to generate, say, three maps, from different seeds, and let some human agency choose which of the three was best. This would prevent some kinds of worst case situations where the algorithm somehow gave a silly result, at the cost of allowing a certain amount of ruling-party influence on redistricting, but no gecko-shaped districts.

--John

[st]

John Kelsey has the seed of an interesting idea. If California could write the law so it takes effect in 10 years, but could make its effect conditional: i.e., we will institute this new pattern if and only if similar laws have been passed in some supermajority of states, or in some other key bellweathers like NY and TX. Make it an election off-year, so states that get held up by budgets, or other hangups have time to have their floor fights.

That way, political pressure can be put on that date by the above-referenced centrists in other states. The date itself could become a symbol for making the case: "Fair representation by 2015" or something.

[st]

er, I mean "bellwethers," sorry.

[Tom]

Y'know, there's bound to be a way to exploit the old saying of getting two boys to share a pie by having one cut and the other choose.

[Tom]

On another topic, there's the potential for multi-member constituencies elected by single-transferable vote (sometimes called proportional representation).

I'll note here that proportional representation was one of the policies favored by Lani Gunier, Clinton's first nominee for Deputy AG for Civil Rights, who was a critic of districts gerrymandered for racial representation. Her opposition to such gerrymandering didn't stop the WSJ opinion page from sliming her as supporting gerrymandering, and so her nomination was dropped. One more reason to read the FT instead of the WSJ.

[Tom]

"Please Rank the Following candidates in order from 1-54. Where 1 is the candidate you most prefer and 54 the least. By least prefer we mean 'you don't like so much'. Please assign only one number per candidate.

[long list of candidates]"

Sebastian, on a technical point as a single-transferable vote aficionado (one who's counted a few STV votes): you don't have to assign preferences for all in the list, even on a 54-member constituency. Your vote would probably get counted even if you only voted for 6-10 candidates. The only way your vote would not get counted is if all the candidates you voted for were eliminated as not being able to get over the quota, which would be unlikely. (More likely is a portion of your vote might not get transferred after one of your candidates were elected if you didn't have sufficient candidates numbered down-ballot).

Example:

I vote as follows:

1. Jody McStravick, Radical American Moderate Party
2. Nanci Pelosi, Democratic Party
3. Thingy Noname, Democratic Party
4. Anna Eshoo, Democratic Party
5. Somebody Alioto, Democratic Party

Let's say the quota for election is 0.25 million votes.

First count: Jody McStravick is eliminated. My vote gets transferred to Pelosi. Let's say Pelosi has 1.25 million first preferences. She's over the quota, and so my vote then gets transferred to Thingy Noname at a value of 0.8 of a vote.

Thingy Noname later gets eliminated. My vote then gets transferred to Eshoo at 0.8 of a vote. Eshoo gets elected in this count with 0.30 million votes. My vote then gets transferred to Alioto at a value of 0.8 x 0.166 = 0.1333 of a vote. Let's say Alioto later gets elected with 0.27 million votes. The rest of my vote would get spoiled, but that would be about 0.01 of a vote.

So you don't have to list a huge number of preferences to have most of your vote counted in STV.

[Anarch]

Y'know, there's bound to be a way to exploit the old saying of getting two boys to share a pie by having one cut and the other choose.

Strangely enough, I spent an hour or two surfing various online expositions of "the cake-cutting problem"...