Join 3,433 readers in helping fund MetaFilter (Hide)

A jigsaw of disunity
October 7, 2012 1:10 PM   Subscribe

The League of Dangerous Mapmakers. The byzantine trade of redistricting was long dominated by brainy eccentrics like Hofeller and his Democratic counterparts. But that began to change in the 1990s, when the availability of mapping software and block-by-block census data for the whole country opened up the field to a waiting world of political geeks. The democratization of redistricting is a lovely thing, perhaps. But as one redistricting veteran told me, “There’s an old saying: Give a child a hammer, and the world becomes a nail. Give the chairman of a state redistricting committee a powerful enough computer and block-level census data, so that he suddenly discovers he can draw really weird and aggressive districts—and he will.”
posted by Sebmojo (20 comments total) 14 users marked this as a favorite

Is it me or has the Atlantic been racking up a higher than usual number of decent FPPs recently?
posted by cacofonie at 1:25 PM on October 7, 2012

Is it me, or do politics provide more jobs for craven scumbags than any other profession?
posted by Ickster at 2:43 PM on October 7, 2012 [4 favorites]

Most people in politics have at least some kernel of belief in their heart that keeps them from bolting to the private sector. If they were truly craven, after all, they wouldn't pick a side.

The financial sector, on the other hand...
posted by Hollywood Upstairs Medical College at 2:54 PM on October 7, 2012

I just went through this after the 2010 census, it was really interesting! I'm on a non-partisan school board, so it wasn't partisan high-stakes, but the whole process was fascinating.

First, the state gets the census data from the feds and then the state releases it to sub-state bodies that may have to redistrict -- in my case, the city and the school district (and the county, but that's not really important to this story) had to do so. We were able to work together with the city and the county to all use the same set of computer data (I think it was Maptitude) and the same redistricting experts. Then, each body has to vote on the redistricting map (or maps). Then all of the info has to go out to voters and to candidates in time for them to file for election. This isn't such a big deal for Congressional races, which are on a two-year schedule, but lots of municipal elections come up every year or in odd-numbered years, so the longer it takes the feds and the states to send the data on down the pipeline, the more screwed municipal bodies are in terms of timing. You're also competing with the state for good redistricting experts to do your redistricting work, and they're working very intensely in a very short period of time.

By the time we got our map for the school district, we had to vote on it the night it was presented (rather than allow a couple weeks for public study and comment) or else people would have had to file to run for election without knowing which district they were running in because of the deadlines for municipal races. The city postponed their redistricting a month to bicker about the maps (largely because the city races are partisan), which cost taxpayers maybe as much as $40,000 in the cost of re-sending voter registration cards (once with the old data to meet the required voter registration card mailing deadline, then again with the new actually correct data after they decided what city council districts we would be in). (It was, in fact, confusing to me as a voter when I got one card, and then immediately got a new card, even though I was one of three dozen people in the city following this issue closely.)

So, the city and school district of Peoria are both under a federal consent decree dating to 1987 to address inadequate minority representation on the school board and city council. Complying with the decree requires really exhaustive analysis of the census data, which the computer programs have made much, much faster, but the guys doing it still have to fiddle the edges by hand to comply with the consent decree's requirements.

In all other school districts in Illinois, 7 school board members are elected "at large" from the entire school district. In Peoria, the school district is sub-divided into three districts (1, 2, and 3) and board members are elected from those. One member is elected from district 1, which is (by decree) the "southernmost" of the three districts and must be a majority-minority district. 2 and 3 each elect 3 members, for a total of 7. District 1 has approximately 11,000 voters; districts 2 and 3 each have approximately 33,000 voters. The districts must be contiguous and compact, follow "rational" boundaries as much as possible (that means major roads) rather than splitting neighborhoods, and respect the status of District 1. Strangely, board members represent not their sub-district's constituents, but the entire district -- that is, once they are elected, they are treated as if they are at large. Here's a short summary of the data (PDF) and the resulting map (PDF), where lines moved very very little from the prior map ... they mostly straightened out along a couple major roads to result in less bumpiness.

(Regarding lack of public study of the maps before public comment: The available maps that met the federal consent decree's rules were all basically the same, with a total of three blocks moved here or there, and only one of those blocks had a school-aged child in it, and nobody from those three blocks had any questions or complaints, so it wouldn't have made much difference, but the principle of the thing was really aggravating; if redistricting is going to be a political process, people should get to comment on it, even if the comments won't make much difference. The map presented to us had the least "wobbles" along major roads and was the closest to the required racial breakdown of the districts, and therefore best met the federal criteria, by three blocks of compactness and like 1/10 of a percentage point in racial representation.)

This post caught my eye because "brainy eccentric" definitely described the lawyer who did our map, and who's done all of them since 1987. He was an aging hippie in lawyer clothes, long white ponytail, who was extremely soft-spoken, but argued reflexively with LITERALLY EVERYONE who spoke to him. He had an encyclopedic knowledge of redistricting federally and redistricting specifically in Illinois, which is a very weirdly narrow area of practice. Whenever he does municipal maps, usually the entire city council or county board or school board has turned over since the last time he did them, so he gives a little history of redistricting at the top of his presentation, and the dude was totally incapable of staying on point because he had so many little digressions and filigrees about 40 years of Illinois redistricting. It truly was fascinating, but looooooooong-winded in the way only lawyers can be long-winded. Our board attorney warned us, when we started the process, that if we happened to run into the redistricting guy or talk to him for any reason, NOT to even give him hints about where we live, because the redistricting guy takes it very seriously and gets VERY upset when he is "corrupted" by knowing where elected officials whom he might redistrict out of their districts live, and wants to be able to say, when presenting contentious maps, "I had no idea where any of you lived," if accusations start flying. When a couple voters at the meeting were commenting and began, "I'm in district 2 --" he'd interrupt and say, very irritably, "I don't care where you live!"

Peoria's City Council has 10 members (+ mayor); 5 of them are elected by district, and 5 are elected at large by fractional cumulative voting. That is, I as a voter have 5 votes and I can assign those votes to any number of candidates up to five: I can give all 5 to one guy, or I can vote for 3 guys and they each get 1 2/3 votes, or I can vote for 5 guys and they all get 1 vote. It's the only electoral body in the entire United States that uses this system, which is supposed to increase minority representation on the City Council (but in fact, no longer does, if it was ever effective; today it serves to overrepresent the wealthier, whiter areas of the city in the at-large slots). So their redistricting was a lot more contentious because they're under this odd system, they had five districts to draw and all the council members were jockeying to keep their slots, and trying to calculate whether they could be in a geographical district or could get elected at-large, and it's a partisan body so there was political party involvement. Their redistricting got a short mention on NPR (All Things Considered, maybe?) because of the uniqueness of the system.

Anyway, great article in the Atlantic, very interesting to read, and it was a very interesting process to go through at the municipal level.
posted by Eyebrows McGee at 3:27 PM on October 7, 2012 [24 favorites]

The Top Ten Most Gerrymandered Congressional Districts in the United States
posted by RobotVoodooPower at 3:39 PM on October 7, 2012

FYI in Missouri at the state level the Republicans have rather huge advantage in both the state Senate and the state House of Representatives due to the way districting works.

Just for example, in 2008, a slim majority of voters in the state senate districts up for election that year voted democrat--50.5% Democrat to 49.5% Republican. (See the Sec of State election results here.)

Not exactly a resounding victory, but a victory nevertheless. Right?

So guess what the proportion of Democratic to Republican senatorial winners was?

4 to 13.

Yup, Republicans picking up 49.5% of the popular vote resulted in winning 76.5% of the senate seats. Losing the popular vote, the Republicans ended up picking up two more seats in the Senate, increasing their already substantial majority to end up with a veto-proof majority.

Situation in the Missouri House is similar, though the smaller districts mean the situation is a little less lopsided. In 2010 the Republicans did pick up majority of voters and, of course, had an even huger super-majority of senators win seats.

How do they do do that?

Well, most democratic districts are urban, so a nice 'rational' compact district automatically gives districts that are 80%-90% democratic--sometimes even more!

Meanwhile the typical republican district is a mixture of medium and small towns, rural areas and maybe some suburban areas. Those are typically 50%-65% Republican--but basically none are in the 80%-100% range where most Democratic-leaning districts are.

The result: Republicans get a lot more winning seats out of their voters.

And the solution? (From the Democratic point of view, or the urban/city point of view--obviously the Republicans are quite pleased with things the way they are.)

The problem is, there is not a good solution. Urban areas are by nature highly concentrated and the voters are far more uniformly Democratic voters. Any reasonably compact districts are going to include a very, very high percentage of Democratic voters by nature.

To fix the problem you would have to gerrymander like crazy--or maybe draw many long thin districts that included part of the urban core, part of the suburbs, and part of the surrounding countryside. An urban area and its surrounding countryside would have a district map looking something like a pie, with every district being a slice of that pie that takes in a bit of the urban core in the center and runs all the way out to the 'crust' in farm country.

And in Missouri, state senate districts must by law follow county lines. So it doesn't even take any nefariousness at all on the part of Republicans to end up with this un-earned super-majority. Just follow county lines and it happens all by itself.

So I don't know what the solution is, but I know that districts that have guaranteed winners are bad for the parties, bad for our democracy, and bad for America.

Districts that are competitive for both parties are better for everybody--except maybe the incumbent.

The problem of lopsided districts is far more serious than the problem of gerrymandering. We should draw districts to have as many competitive races and as few shoo-ins as possible, geography be damned.
posted by flug at 4:24 PM on October 7, 2012 [3 favorites]

Just a little update--in 2010 in the Missouri state senate races, Republicans won 67% of the popular vote and 76.5% of seats.

So not quite the lopsided effect we saw in 2008, where 49.5% of the vote brought 67% of the seats, but still a pretty decent magnifying effect--definitely a greater proportion of seats than popular vote.
posted by flug at 4:43 PM on October 7, 2012

The problem is, there is not a good solution.

The solution is simple, take the politicians out of the equation.
It would not be super difficult to automate the drawing of districts based solely on population (no rigging for party registration, minority status, age, or anything else).

Just a straight population count divided by the number of desired districts with districts conforming as near to a square shape as possible.

You could then create a random selected board of voters to review the results, and nudge the boundaries here and there to account for, say, a dividing natural feature or if you have a small neighborhood that clearly belongs to a city end up outside the district boundary.

Again, the board would only have access to population counts, no other demographic data provided.

Of course, it would not end up perfect, but it would be better than the situation we have now, where just this year I got gerrymandered from my city based district into one that includes the rural eastern half of my county.
This was, naturally, in order to cement the majority party's power.
posted by madajb at 4:46 PM on October 7, 2012

The Redistricting Game has been posted before, but I came across a Ward-level version for DC. (Unfortunately, the sidebar incorrectly reports unacceptable values as acceptable, so you'll have to keep an eye on the numbers.)
posted by ChurchHatesTucker at 4:48 PM on October 7, 2012

The solution is simple, take the politicians out of the equation.

I could not agree with this statement more.

However, I think the goal should be to make sure that, as much as possible, the people in each district have similar concerns so that those they elect to represent them don't have to make as many compromises in representing their district.

The thing that probably colors my thinking on this is that I used to live in Woodbury, MN. Despite the fact that the suburb of St. Paul is mostly upper-middle class, well educated families, it is part of district 6 and currently represented by Michele Bachman (and on behalf of all Minnesotans, I'm really sorry). Most of her votes come from the rural areas surrounding St. Cloud (a solid two-hour drive from our house in Woodbury). I think that the district is designed to take democratic votes away from some of the other urban and suburban districts. The issues that are important to the people in Woodbury are completely different than the people in St. Cloud and I think the districts should reflect that.
posted by VTX at 5:06 PM on October 7, 2012

Just a straight population count divided by the number of desired districts with districts conforming as near to a square shape as possible.

I don't think this "works" quite as well as you think. The way flug describes it (and I think this is more or less consistent with American demography), because liberal voters are more highly concentrated in very liberal (generally urban) areas, following ideologically-neutral guidelines such as "compactness" might lead to strange results.

To make an extremely stylized example, take a state with 3 million people, 1 million of whom live in the city and are 70% Democratic, and 2 million of whom live in the rural areas and are 40% Democratic, and you are selecting 6 representatives. Half of the voting population is Democratic (.7*1 + .4*2 = 1.5 million), but if you follow natural boundaries, such as a city line, then only 1/3 of the representatives (2 of the 6) will be Democrats. This effect would then be amplified by trying to ensure minority representation, which result in a few very Democratic districts surrounded by vaguely Republican districts, with the result that the House is probably more Republican than it would be otherwise.

Now, I'd prefer nonpartisan districting using these criteria anyway, since I think the benefits clearly outweigh the costs. But, I don't think there's any escaping the inherent tension between geographic representation (in America, single-representative districts) and national politics.
posted by dsfan at 5:14 PM on October 7, 2012 [1 favorite]

So: in terms of national popular vote margin, the Congressional 2012 new redistricting advantage is R+1.2+/-0.1%.

posted by ersatz at 5:33 PM on October 7, 2012

Here's how you solve the problem:

let ANYONE submit a redistricting map after the census, so long as variation in population between districts is sufficiently low.

Take each submitted map, and measure the total length of all the inter district boundaries.

The one with the lowest total gets used.
posted by ocschwar at 5:38 PM on October 7, 2012 [1 favorite]

The solution is simple, take the politicians out of the equation.

Then you get the districts that the people drawing the districts happen to like, or districts that the people creating the automated system might like.

Just a straight population count divided by the number of desired districts with districts conforming as near to a square shape as possible.

Variations in population density will make that impossible in most states.

Again, the board would only have access to population counts, no other demographic data provided.

You don't need to be provided with any external data to know that these here blocks in urban Whereverville are mostly Irish Catholic Democrats, or no notice that they're rural farmland, and so on.

I think that [MN6] is designed to take democratic votes away from some of the other urban and suburban districts.

Other way. Heavily Republican districts are a sign of a Democratic gerrymander (intentional or not), and heavily Democratic districts imply a Republican gerrymander.
posted by ROU_Xenophobe at 5:40 PM on October 7, 2012 [2 favorites]

represented by Michele Bachman (and on behalf of all Minnesotans, I'm really sorry)

Considering Bachmann won her district quite handily, including many precincts in Woodbury and eastern MN 6, you apparently don't speak for all Minnesotans.

Despite the fact that the suburb of St. Paul is mostly upper-middle class, well educated families, it is part of district 6

Portions of Stillwater, Lake Elmo, Woodbury and Afton were recently drawn into the very safely DFL MN 4.
posted by lstanley at 6:07 PM on October 7, 2012

When you use the computer redistricting program, you can tell it to observe a variety of rules, and to prioritize those rules. For example, you could say the MOST important thing is no districts are more than +/- 2% of the average district in terms of population. And then next, computer, I want you to keep political units together -- defined neighborhoods most important, cities and towns next, counties if you can manage it but don't worry too much.

You could give it information on race and ethnicity and request it create a certain number of minority-majority districts, or to try to racially balance every district. You could identify suburban, urban, and rural zones and tell it to try to keep like areas together even if you make a suburban donut. You can tell it you want the most compact possible districts, or that it can vary from the most compact by 10% but no more. You can use interstates or commuter rail lines as "magnets" that try to center districts along themselves (on the theory that all the people on commuter rail line X have interests in common).

Anyway, with the computer programs, you have a lot of options for trying to make "rational" districts that will help the maximum number of people feel represented. You don't have to "just draw squares" to get fair, non-partisan maps; you can tell it to respect counties and cities and towns, to try to create "urban" and "rural" districts; etc. Or, as is sadly more often the case, you can tell it "We need a lot of my political party in the legislature, make it so, computer!" The U.K.'s Boundary Commission, for example, uses these criteria:
"One of these rules requires them to take account of local government boundaries. A second rule requires that each constituency should have an electorate within 5% of the average electorate of all constituencies. A further rule allows a Commission to take into account “special geographical considerations”. The rules also require the Commissions, if altering constituencies, to take account of inconveniences this will cause and local ties that would be broken.
They're required to ignore voting patterns in drawing their districts. I'm sure our UK compatriots can tell us in what ways the system works and in what ways it's broken!

Anyway, part of the problem is U.S. congressional districts are just so big. The average population represented by each US Rep is about 650,000 people. The largest is Montana (all of it) at a bit over 900,000; the smallest is Wyoming at just under 500,000. (Geographically, the largest is all of Alaska at 572,000 square miles; the smallest is New York 15 (upper Manhattan) at 10 square miles.) UK constituencies are closer to 70,000 people; Canadian ridings are right around 100,000; Australia about 130,000.
posted by Eyebrows McGee at 6:25 PM on October 7, 2012

Take each submitted map, and measure the total length of all the inter district boundaries.

This is a silly criterion. Natural boundaries such as rivers are preferable to artificially straight lines, and rivers are exponentially longer thanks to all the kinks and turns. This scheme might work well in Iowa, with their square counties and more even population density, but would totally collapse in the original 13 colonies.

The problem is, there is not a good solution. Urban areas are by nature highly concentrated and the voters are far more uniformly Democratic voters. Any reasonably compact districts are going to include a very, very high percentage of Democratic voters by nature.

^^This. The median blue precinct is far more partisan than the median red precinct (and there are a larger raw number of red precincts, which is how we end up even), a large part of this is due to race. Black voters are a monolithic Democratic bloc, and black majority precincts can vote upwards of 80% Democratic.

It's an under-reported problem, especially in Southern states, that more unscrupulous black legislators forge unholy alliances with white Republicans to keep black voters in black districts (and therefore out of neighboring districts). Breaking up some of the majority-black districts would absolutely turn some of the surrounding suburban districts into Democratic seats, leading to a more liberal overall outcome for the legislature. Surely a black voter who votes for their black Democrat wants a Democratic majority in the legislature, but the incentives do not always align between the voter and the entrenched incumbent!

Because of these racial constraints, it is a fair bit easier for a Republican legislature to pack Democrats than it is the other way around. Democrats are hamstrung when they try to unpack these 70, 80% Democratic seats that are huge holding pens for black voters.
posted by Hollywood Upstairs Medical College at 6:34 PM on October 7, 2012

There have been efforts to develop objective measures of 'bizarreness' of the shape of districts intended to prevent gerrymandering (see A Measure of Bizarreness in the Quarterly Journal of Political Science).
posted by James Scott-Brown at 3:41 AM on October 8, 2012

RobotVoodooPower: The Top Ten Most Gerrymandered Congressional Districts in the United States

South Florida takes a lot of heat for gerrymandering, but the maps can be misleading. Population density outside of those spaghetti strings on the Southeast coast mean that everything West of the Monroe county line has to be one district to have enough people in it to be a district, and that includes everybody in the Keys. We're mostly talking about the Everglades, an alligator-infested swamp with people living on the eastern edge.

The area around Jacksonville and the St. James River has the same issue, mostly minus alligators.

You can see this clearly on Google.
posted by halfbuckaroo at 4:23 AM on October 8, 2012

The real solution is to get rid of districts entirely and switch to proportional voting by party. If the needs of a given geographical area (e.g. rural voters) are so particular, then those voters can form a party to represent them. Tying representation to geography is a throwback to the days before fast travel, mass media, and computer networks, back when a representative would spend a significant amount of time just getting around his district (and at that time they were all male).

It would really be better for everyone except the two major political parties and the incumbents, which means it will never happen.
posted by jedicus at 3:18 PM on October 8, 2012 [1 favorite]

« Older In the wake of the venerable Boston Phoenix changi...  |  The Ohio State University Marc... Newer »

This thread has been archived and is closed to new comments