Join 3,427 readers in helping fund MetaFilter (Hide)


Solve for (D)emocracy
June 3, 2014 10:56 AM   Subscribe

This programmer thinks he's solved the gerrymandering problem. Gerrymandering has been discussed on the blue many times. But with very little eye towards solving the problem. A programmer named Brian Olsen has come up with the idea of mapping districts using compactness. It's fun! Check your state!.

Of course it would require rethinking major parts of the voting rights act,. But it's not like that hasn't been happening already.
posted by lumpenprole (71 comments total) 12 users marked this as a favorite

 
A programmer, you say?
posted by 2bucksplus at 11:03 AM on June 3 [13 favorites]


Wrong solution for the wrong problem. Gerrymandering isn't an engineering problem, it's a political problem and as long as there are parties who profit from unfair redistricting and with the lack of integrity to profit from this (aka Republicans) gerrymandering will continue.
posted by MartinWisse at 11:05 AM on June 3 [42 favorites]


While he covers it in the about section, it's definitely worth noting that these maps don't take into account geographic features that make natural boundaries (mountains, rivers, etc), or make use of existing boundaries like county/city/neighborhood lines. All which do have a very valid place in the process of drawing districts.

That said, I'm not convinced those faults would make these any worse in the average case than the joke that is the district maps of many states today.
posted by evilangela at 11:06 AM on June 3 [9 favorites]


Take the quiz to see if you can spot the gerrymandered districts.
posted by the man of twists and turns at 11:07 AM on June 3 [8 favorites]


Why don't you measure travel time? It might be the right kind of thing to measure, but it would take too long...The large amount of map data and extra computer time to calculate all those travel times would slow the process down horribly

"My solution might be wrong, but it was easier to compute than the right solution."

Well then.
posted by cjelli at 11:07 AM on June 3 [20 favorites]


This is fine as a programming exercise, but the programmer hasn't thought through the impact of such a scheme. Compactness isn't a criterion that is neutral between the parties. Instead, compact districts tend to overrepresent Republicans since Democrats tend to live in areas that are highly concentrated with Democrats, while Republicans tend to live in areas that are moderately concentrated with Republicans. If you draw district lines based on compactness, you are going to wind up with a few districts that are nearly 100% Democratic, and a large number of districts that are mostly Republican.

http://themonkeycage.org/2012/11/15/not-gerrymandering-but-districting-more-evidence-on-how-democrats-won-the-popular-vote-but-lost-the-congress/

If you're going to write an algorithm to draw district lines, the fairest one would be the one that maximizes the number of competitive districts, since that would come closest to reflecting a state's voters' partisan preferences in the makeup of the legislature. The programmer's dismissal of this approach is pretty unconvincing.
posted by burden at 11:14 AM on June 3 [15 favorites]


the man of twists and turns: "Take the quiz to see if you can spot the gerrymandered districts."

Oh dear, WaPo... Your first major claim already befuddles the issue by calling redistricting "gerrymandering" as if all redistricting is gerrymandering. Redistricting is an important setup when you have a house of commons as we do in the US. A population (rather than state) based voting method of selecting representatives. Calling the general term redistricting "gerrymandering" is like calling all electronic music "Techno" or "Dubstep" (these days, apparently).
posted by symbioid at 11:15 AM on June 3 [7 favorites]


This guy is suffering from Engineer's Disease.
posted by desjardins at 11:16 AM on June 3 [16 favorites]


I think he took the Vermont Senate map and turned it into a House of Representatives. We have both. That's why the Senate map looks weirdly disproportionate at a population level.
posted by maryr at 11:17 AM on June 3


Dems profit from gerrymandering too, MartinWisse. There are currently fewer states gerrymandered towards the Democrats, but that's mostly because of the successful Republican efforts to control state governments during 2010 redistricting.

I also don't think any program can solve the geographic self-sorting issues that cluster ideologically similar people.

And of course this doesn't address at all the idea of multiple-member districts, which can fix some of the issues with minority representation as well as the issues burden mentions above.
posted by nat at 11:19 AM on June 3 [4 favorites]


This guy is suffering from Engineer's Disease.

This programmer knows how to use a hammer. You won't believe what he thinks is a nail.
posted by griphus at 11:22 AM on June 3 [46 favorites]


Your first major claim already befuddles the issue by calling redistricting "gerrymandering" as if all redistricting is gerrymandering.

I think you are reading their first sentence as equating the terms as in: "redistricting OR 'gerrymandering'" rather than what I think they were trying to say: "redistricting or what people are accusing in our several attached hyperlinks links of being 'gerrymandering'"
posted by The 10th Regiment of Foot at 11:24 AM on June 3


Dems profit from gerrymandering too, MartinWisse.

Sure, but on the whole the modern Democratic party is far less likely to gerrymander rather than redistrict.

Meanwhile the real solution is simple: don't use districts, use proper proportional representation.
posted by MartinWisse at 11:25 AM on June 3 [8 favorites]


Sure, but on the whole the modern Democratic party is far less likely to gerrymander rather than redistrict.

I'm a life long democrat raised by much more liberal democrats - and that made me laugh out loud.


Given the opportunity the democratic party would be just as craven as the repubs
posted by JPD at 11:27 AM on June 3 [10 favorites]


Can you imagine how much of a disaster a PR system would have been when states rights parties were all the rage? My God.
posted by JPD at 11:28 AM on June 3


Casting gerrymandering as a problem of "the maps are all squiggly" instead of "Democrats are concentrated in cities and drawing fair districts is very difficult" is another example of what I call the dodge.

You can tell when the dodge is in play because it takes a hard problem (how do you draw competitive districts) and turns it into an easy one (how do you draw convex shapes), and in doing so disproportionately helps the people (Republicans, usually) who are trying to get you to do something that is ultimately in their favor, not yours.

The other great modern play of the dodge is people advocating for a flat tax because they just did their taxes and it was so difficult and complicated! But what's difficult about taxes isn't that they're progressive, it's that it's hard to calculate your total 'income' as defined by the government. A flat tax simplifies the simplest step of taxes (looking up a value in a table), but doesn't solve anything related to how complicated it is to calculate your income. You still have to deal with calculating your actual capital gains, adding up your W-2s, actual withholdings, deductions, and so on. The only people who really benefit from a flat tax are the rich, but they want you to think that you'd be better off with it.

The dodge. Watch for it.
posted by 0xFCAF at 11:29 AM on June 3 [66 favorites]


Meanwhile the real solution is simple: don't use districts, use proper proportional representation.

QFT. Gerrymandering is a solved problem. Just count the damn votes.
posted by anonymisc at 11:30 AM on June 3 [3 favorites]


Compactness isn't a criterion that is neutral between the parties.

Consideration of party is how we get gerrymandered districts in the first place.

Population, and to a very slight extent geography, should be the only things allowable in the apportionment process.
posted by madajb at 11:30 AM on June 3 [2 favorites]


I'm going to float my solution to the district definition problem (at least for the US House) again: We now have the communications and transport technology necessary to go back to the original district sizes. Make district sizes about 35,000 and grow the membership of the House to about 10,000. There's absolutely no reason the reps need to go to D.C. and sit together in the same building. Everything can be handled remotely; maybe they can convene in person 2 or 3 times a year. If they must sit in the Capitol building, they can take turns, or do it by seniority, or whatever.
posted by mr_roboto at 11:30 AM on June 3 [6 favorites]


Population, and to a very slight extent geography, should be the only things allowable in the apportionment process.

Because of the high geographic concentration of Democratic voters, saying this is equivalent to saying "Republicans should be overrepresented in legislatures."
posted by burden at 11:34 AM on June 3 [2 favorites]


> Wrong solution for the wrong problem. Gerrymandering isn't an engineering problem, it's a political problem

There has been no solution forthcoming to the political problem, and one big reason behind this is that there isn't any agreed-upon rational criteria as to what "best districting" is.

Even if academics simply agree, "This is the best redistricting" and politicians ignore them, this is a big step forward. In every dispute involving gerrymandering, someone can hold up a map and say, "Science says that this is the best redistricting" - and by comparison with this, the obviously bogus redistrictings proposed by the machine will look extremely lame.
posted by lupus_yonderboy at 11:34 AM on June 3 [5 favorites]


Don't use districts at all. Allow multiple winners instead, like in single transferable vote.
posted by jeffburdges at 11:36 AM on June 3 [3 favorites]


The simplest solution I've heard of that seems like it would work is to put a limit on the ratio of a district's perimeter to its area. That still gives lawmakers flexibility to draw the districts to fit whatever criteria they want, but not so much flexibility that they can make very oddly shaped districts just to split the vote.
posted by burnmp3s at 11:39 AM on June 3 [4 favorites]


On the technical side of things, it's not true that there has been "very little eye towards solving the problem". Wikipedia has a whole subsection on proposed objective rules to reduce gerrymandering, namely: Of course, as Wikipedia discusses, all of them have considerable intrinsic limitations, even putting aside the political hurdle of getting them made into law. Mr. Olson's proposal is a fourth: minimize the average distance to the center of a district. He doesn't seem to realize it, but that's equivalent to applying the k-means algorithm to the population distribution. And the limitations of k-means are also well-known...

So no, I'm afraid there is no best redistricting, even from a purely academic perspective.
posted by narain at 11:40 AM on June 3 [6 favorites]


"My solution might be wrong, but it was easier to compute than the right solution."

Well then.


more like " my solution might be wrong, but it's still possible to do"
posted by Dr. Twist at 11:40 AM on June 3


Make district sizes about 35,000 and grow the membership of the House to about 10,000.

What, you don't think we have enough problems with gridlock?
posted by desjardins at 11:42 AM on June 3


burnmp3s: The simplest solution I've heard of that seems like it would work is to put a limit on the ratio of a district's perimeter to its area.

Minimum isoperimetric quotient
Such rules would prevent incorporation of jagged natural boundaries, such as rivers or mountains. When such boundaries are required, such as at the edge of a state, certain districts may not be able to meet the required minima.
posted by narain at 11:43 AM on June 3 [2 favorites]


> The dodge. Watch for it.

"The algorithm-based districts make a lot of intuitive sense."
"... bypasses these issues completely in favor of straightforward geographic compactness."

The Dodge is often accompanied by truthiness. If someone has to say "It just feels right, doesn't it?", it's probably wrong.
posted by benito.strauss at 11:45 AM on June 3 [8 favorites]


My home Congressional district here. It's designed that way to have a Hispanic majority district. Why is that such a bad thing?
posted by Pararrayos at 11:48 AM on June 3 [1 favorite]


Make district sizes about 35,000 and grow the membership of the House to about 10,000.

You've got 237 districts in NYC alone, then, which makes even less sense than the current system.
posted by griphus at 11:54 AM on June 3


In our increasingly peripatetic and virtual age, why do we have nothing but geographic districts? What if you could opt out of your geographic district and into a superdistrict representing a primary concern of yours? My rep is in the education-of-special-needs-children superdistrict, or the agricultural-labor superdistrict, or more likely: the Internet-shut-in superdistrict. Geography seems like a good criteria if you want face time with your rep or a local pork barrel project, but otherwise not too useful.
posted by BrotherCaine at 11:56 AM on June 3 [4 favorites]


I suspect it's a hop, skip and jump to the Hunger Games after the People Who Believe What They Hear on Fox News Superdistrict is formed.
posted by griphus at 12:00 PM on June 3


Internet-shut-in superdistrict

That's the only one in which local issues might not be important, and even then, somebody's gotta run the cable and deliver your groceries. Best make sure the roads aren't impassible.
posted by asperity at 12:03 PM on June 3


Olsen ... Olsen ... wait, is this the same guy who got dropped on his head by the nanny and then crushed all the beetles?
posted by freecellwizard at 12:04 PM on June 3 [3 favorites]


Such rules would prevent incorporation of jagged natural boundaries, such as rivers or mountains. When such boundaries are required, such as at the edge of a state, certain districts may not be able to meet the required minima.

That doesn't sound like a completely unsolvable problem. You could easily have rules around the jaggedness of state borders not being counted in the formula. The point is mainly that the ratio between perimeter and area is a decent metric for how nonstandard the district boundaries are, so even if the rule was that future district maps cannot have an average ratio higher than the current configuration it still would make it harder for gerrymandering to happen in the future.
posted by burnmp3s at 12:05 PM on June 3


I seem to recall one of the first Decode DC episodes talked about the House of Representatives, and if we still had to go by the original number of constituents instead of the hard limit imposed a while back, the number of reps would be at least 1500 by now.

Trying to think what the world would look like when every major city had several representatives. I imagine things would look a bit different.
posted by fifteen schnitzengruben is my limit at 12:07 PM on June 3 [1 favorite]


The suggestion of 35,000 US reps is an interesting one. It prompted me to look into the number of state-level legislators. If my cribbing from wikipedia is right, there are only 7577 legislators, 5518 in "lower" houses and 2059 in "upper" (Nebraska's unicameral is treated as an "Upper" house, while DC's Council is called "lower", for whatever reasoning). Not sure whether or how this bears on the idea to change the US reps, but there you go..
posted by jepler at 12:20 PM on June 3


Screw geography. Divide the population randomly into virtual districts, assigned when their address changes. Geography doesn't mean shit anymore since we live wherever we can afford a house the size we want, don't know our neighbors, our front doors would be webbed in with cobwebs if we don't sweep because we never leave except via the garage and think nothing of driving across a third of a state to go to the Wal-Mart anyway. Home is just the first word in Home Theater System.
posted by George_Spiggott at 12:20 PM on June 3 [2 favorites]


hamburger
posted by George_Spiggott at 12:23 PM on June 3 [1 favorite]


Just because a given optimisation problem for redistricting is computationally infeasible to solve exactly, doesn't mean that approach has to be rejected outright. It seems to me that a competitive approach could work: announce a period during which anyone can submit a district map, and at the close of the period, accept whichever one scores the best.

Presumably the Dems and the Reps would each exert lots of computing power on the task, lest they let the other party walk away with it. That should yield a reasonable map if the objective function isn't badly behaved.

Compare combinatorial auctions: determining the best allocation all by yourself is NP-hard, but taking bids works well in practice.
posted by finka at 12:25 PM on June 3 [4 favorites]


...somebody's gotta run the cable and deliver your groceries. Best make sure the roads aren't impassible.

But how much pull does your rep actually have to address local infrastructure issues if they lack seniority or are on the wrong committee? You could just as easily have this addressed by watchdog groups who use legal action to make sure infrastructure funds get fairly distributed. Plus a lot of that
is usually at the county/city level.
posted by BrotherCaine at 12:29 PM on June 3 [1 favorite]


Compactness is a virtue in redistricting, but not the only one. In particular, if followed blindly it can lead to the under representation of minorities. Also, it's usually a good idea to pay at least some attention to internal political boundaries, keeping counties within a single district when possible rather than chopping them up, etc.

Also, it's entirely possible to have 'compact' districts that are still gerrymandered. One of the favorite Republican tactics of the 2010 redistricting was taking a city that used to be a more or less single unit and chopping it into bits to attach to the surrounding, much more Republican districts. A pie-cut breakdown of a city and suburbs may be quite 'compact' yet still still effectively disenfranchise urban inhabitants.
posted by tavella at 12:37 PM on June 3 [3 favorites]


Well, Ds do re-district as well as Rs, but there is such a thing as taking it too far. In Texas, we have had real issues with re-districting. Historically, there was a re-districting after every census, with the party in power coming out on top. However, in the early 2000s, the Republicans started a cycle of redistrict to get more Rs in, then immediately after the elections, re-district again to give themselves even more power. This lead to, in 2003, all the Ds literally leaving the state and hiding out in OK, and the Speaker sending the Texas Rangers to try to arrest them all and drag them back so that there would be a quorum.

This has been done before, but it hasn't happened in a while -- I don't think there are enough Ds left in the state house to stop the Rs from doing pretty much anything they want. Witness the Wendy Davis fillibuster, which was defeated by simply calling another session. (They had to pass the bill during a special session because in the regular session they would have needed a 2/3rds majority, not a simple plurality.)
posted by pbrim at 12:41 PM on June 3 [3 favorites]


My perfect world solution is to get rid of districts, and let each state party publish a list of nominees for House seats, in ranked order. Voters vote for parties, which they already do, and the number of seats each party receives is proportional to the number of votes they received. This solution makes everyone's vote in a state equally likely to change the results, plus keeps the party representation in the House much closer to the actual popular vote.
posted by gsteff at 1:02 PM on June 3 [1 favorite]


There has been no solution forthcoming to the political problem, and one big reason behind this is that there isn't any agreed-upon rational criteria as to what "best districting" is.

Even if academics simply agree, "This is the best redistricting" and politicians ignore them, this is a big step forward.
Ah, the engineer's standard fallacy, the idea that a new standard/solution for something will sweep aside all others, that something as complex as this can be agreed upon, that you can ignore the political and sociological reasons for the problem in the first place.

See also every internet standards war ever.
posted by MartinWisse at 1:48 PM on June 3 [3 favorites]


"Don't use districts at all. Allow multiple winners instead, like in single transferable vote."

1) STV still uses districts.
2) At-large elections generally give terrible results, e.g. Detroit's city council.

Increasing the size of the legislature (though probably not to 35,000) is a good idea, and would do a lot to mitigate the lack of actual representation in the legislature.

Pretty much all other proposals here are solutions in search of a problem, or betray a fairly sizable unfamiliarity with the U.S. Constitution. It would be easier to increase the size of the legislature — which would still take an amendment and be pretty wildly unpopular, given the approval ratings of Congress in general — than to do things like end first-past-the-post elections or institute proportional representation. If you want proportional representation, you're going to have to institute it in a state before the country, and even there, you're probably going to have to bring over a couple of counties or cities before the state would go for it.
posted by klangklangston at 3:02 PM on June 3 [1 favorite]


In every dispute involving gerrymandering, someone can hold up a map and say, "Science says that this is the best redistricting"

Yeah, this is kind of the textbook definition of Engineer's Disease.

Take an extremely complex, nuanced problem, grind away all political, sociological, and other considerations until it is computable, come up with an optimal solution for your new, simpler situation, and then declare that SCIENCE shows you are right.

I cringed as soon as I read "This programmer thinks he's solved". If the thing he's solving isn't a programming problem, red alert. The problem is programmers seem to think everything is a programming problem.

Population, and to a very slight extent geography, should be the only things allowable in the apportionment process.

This being MeFi, I'm surprised no one has brought up the issue of race. This sort of simple population/geography analysis fails for various reasons that others have already pointed out (failing to account for rural/urban populations, for example) but it also fails in terms of possible racial discrimination. It has long been a problem that minorities are either concentrated or scattered so that their political influence is diminished. This has been a huge problem with drawing districts that avoid disenfranchising people. The article actually does discuss this problem, but them dismiss them, saying:

Brian Olson's algorithm bypasses these issues completely in favor of straightforward geographic compactness. It would be a huge lift to actually put programmatic redistricting into practice — among other things, we'd need to retool major portions of the Voting Rights Act. But given what's at stake — the very idea of representative democracy — it's worth considering.

Again, textbook Engineer's Disease, ignoring the nuance and complexity of the problem in favor of a simple, "elegant" solution.
posted by Sangermaine at 3:11 PM on June 3 [3 favorites]


No empirical polsci proof that gerrymandering exists in a distorting form currently in the US. In fact it has been strongly rejected over the past few decades. It is a problem that exists in the news and internet, but not in reality to a meaningful degree.
posted by jjmoney at 4:23 PM on June 3


"First, we assume a spherical district..."
Whether a particular group (Republicans, minorities, etc.) is over- or under-represented depends on your own agenda.
posted by Tool of the Conspiracy at 4:32 PM on June 3 [2 favorites]


That's not correct at all, and seems like you've reacted to a misconception in one direction (that gerrymandering reduces competition and increases polarization) by substituting it in another, that it has no distorting effects.

Gerrymandering isn't the bogeyman that many people assume, especially to the extent of justifying the cost of redistricting on some impersonal algorithm, but there are multitudes of empirical proofs showing specific gerrymandering effects — California's redistricting is the source of a ton of literature, as is Texas's.

To dismiss it like you did, that it is not a problem in reality, would also fail the common sense test of recognizing that it's something that political parties fight over bitterly, which they'd be unlikely to do if it had no actual effect on outcomes.
posted by klangklangston at 4:37 PM on June 3 [3 favorites]


It's all fun and games to use computers to create new district maps until an off-by-one error goes and gives Long Island statehood.
posted by mccarty.tim at 6:58 PM on June 3


US needs what we have an Australia - a fully funded, national, deeply independent electoral commission that is protected from the vicissitudes of venal politicians.

Our districts get redrawn pretty much every election - they a based around population moreso than size, and it's all done by the AEC. If anyone, especially, a politician was caught trying to influence the process, they would not only be hounded out of town, but quite likely into jail as well. It would be viewed here akin to vote fraud on a massive scale.

*wipes tear* I love our AEC.
posted by smoke at 8:20 PM on June 3 [5 favorites]


There does seem to be some value in a defined algorithm for at least creating proposals for humans to look at. But that algorithm would have to take into a lot of things (some noted in this thread: natural boundaries, county lines, streets, racial makeup, etc.) And even with an algorithm, we'd probably still need some kind of human process to make districts that people think are basically fair (there will always be people who don't think the result is good enough of course). But the benefit of an algorithm is not that it's some kind of magic computer thing that is objective, but that it would encourage more open processes.
posted by R343L at 8:32 PM on June 3


If you're going to write an algorithm to draw district lines, the fairest one would be the one that maximizes the number of competitive districts

This is the same as a very strong gerrymander in favor of the local minority party. It can't be not that.
posted by ROU_Xenophobe at 8:38 PM on June 3 [2 favorites]


burden: "If you're going to write an algorithm to draw district lines, the fairest one would be the one that maximizes the number of competitive districts"

That is a terrible idea:

* How are you going to make Wyoming's single congressional district competitive? There's just one of it.

* How will you make districts involving cities even vaguely competitive? Have every single one of New York state's districts comprise 1/27th of the city, and a long long stretch of upstate? I don't think anybody would look at that and consider it reasonable.

* If you do manage to get every district to be 50% Republican and 50% Democratic, when there's a 1% nationwide swing to the Republicans, they'll win almost every district. A 1% swing the other way, and the Democrats win every single one. That's not really conducive to democracy.

* What do you do when people in some district are very happy with their representative, and vote for her in large numbers? Continually change the district around to include people who wouldn't like her, so that it's competitive?

* What do you do about third parties? Try to put together a weird snaky district that combines all the Green voters in California, so they have a chance at winning a district?

There are realistic alternatives, that have actually been tested in other countries and found to work: proportional representation, and neutral electoral commissions.
posted by vasi at 12:45 AM on June 4 [5 favorites]


This is one thread where I don't want to read all the comments because I'm already confused enough after reading the title - Solve for Democracy. It triggered a Solve What? moment. I have already done plenty of reading about Gerrymandering to know that for some it is a tactical political tool but for others it actually is more fundamentally related to how they want to be represented.

Should divisions favour homogeneity or balance? Maybe Representation in general needs to be re-affrimed. Are the people being represented to their satisfaction? Hahahaha :) Try and gerrymander a referendum on that question. Well it's easier than I thought :/
posted by vicx at 1:55 AM on June 4


As is typical of the programmer caste, he clearly didn't bother to look and see if there were any prior projects or research that might inform his work. If he had, he might have discovered that Iowa has been algorithmically determining districts since 1981. The program spits out a map, and the legislature can vote it up or down. If they reject it, the program will generate another map, and that one is binding. It's a brutally simple system that legislators in other states are terrified their constituents might find out about.
posted by ulotrichous at 5:43 AM on June 4 [4 favorites]


"First, we assume a spherical district..."

If we could only assume an educated and rational population there wouldn't really be any issue.
posted by koolkat at 6:20 AM on June 4 [1 favorite]


I should have defined my terms more carefully. I would define a competitive district as one in which the number of Democratic voters in the most recent Presidential election is close to the number of Republican voters (within 2%, 1%, or 0.5% or whatever). If there's a popular Republican incumbent that represents such a district for 40 years, that's still a competitive district.

In states that are split roughly 50/50 between Democrats and Repblicans, you might be able to draw all districts as competitive districts. In states that lean to one party over the other, the process of creating as many competitive districts as possible will inevitably lead to the creation of some districts that are safe for the majority party. So California will still elect more Democrats than Republicans since there will be some safe Democratic seats in addition to the competitive seats.

Obviously you can't do anything about single-member states by changing how district lines are drawn, since there aren't any.

This proposal doesn't do anything to help third party candidates get elected, but at least in competitive districts, candidates would have an incentive to take the concerns of third-party supporters seriously, instead of blowing them off as safe incumbents mostly do now.

I guess I'd like you to explain why a district that incorporates a bit of NYC and a strip running upstate is unreasonable. It looks "funny" but surely the objection to such a district isn't only aesthetic.

Clearly it is not possible to draw all 435 Congressional districts as competitive districts, so I don't think it's likely that a 1% shift will lead to total turnover in the House. But even that would be better than our current system, in which shifts in the electorate are unable to change the control of the House of Representatives.

I'd strongly prefer proportional representation to any system relying on districts. But I think neutral commissions that don't take party and competitiveness into account are no good since they naturally produce lines that favor Republicans.
posted by burden at 8:01 AM on June 4


I should have defined my terms more carefully. I would define a competitive district as one in which the number of Democratic voters in the most recent Presidential election is close to the number of Republican voters

Again, this is exactly the mechanism you would use to gerrymander in favor of whichever party is the statewide minority. This is how you get the NY Senate.
posted by ROU_Xenophobe at 9:51 AM on June 4 [1 favorite]


I don't think that's accurate, ROU. I don't think that a self-interested minority party with the ability to redistrict as it sees fit would create competitive districts. Why wouldn't such a party try to engineer a legislative majority for itself by creating a bunch of districts (though less than a majority) packed with the other party's voters, thus creating a majority of districts leaning towards the minority party?

I don't know much about the NY State Senate but is it really accurate to say that its district lines are drawn to maximize the number of districts that are nearly balanced in terms of voter partisanship? If so, can you provide some additional information about that? I would be deeply surprised if that were the case.
posted by burden at 10:33 AM on June 4


burden: "In states that lean to one party over the other, the process of creating as many competitive districts as possible will inevitably lead to the creation of some districts that are safe for the majority party."

Isn't that unfair to the people in "safe districts", who now can't really affect the balance in Congress? Who decides which people end up in a safe district?


"at least in competitive districts, candidates would have an incentive to take the concerns of third-party supporters seriously, instead of blowing them off as safe incumbents mostly do now."

This doesn't just "not help" third parties, but actively hurts them. In close districts, third parties have trouble getting votes because the wrong lizard might win.

Anyhow, my point wasn't about whether you're helping or hurting third parties, but about the practicalities of redistricting for "competitiveness" in the presence of third parties. Let's say the Reform or Green or Libertarian party does well, reaching 20% of the vote—not so far-fetched, Perot got 19% in 1992.

How do you redistrict now? Draw districts that are all 40/40/20? It's much harder to draw districts like that than to just balance two parties. Plus, it's unfair to the third party, who might have areas of strength where they could win a district. Maybe you should draw some districts that are 34/33/33, and then others that are 50/50/0, and then a few safe districts that are 100/0/0? What if a fourth party arises? You see that it's a lot more complicated than just "oh, make them all competitive".


"I guess I'd like you to explain why a district that incorporates a bit of NYC and a strip running upstate is unreasonable. It looks "funny" but surely the objection to such a district isn't only aesthetic."

Sure. The US is a representative democracy. Note the representative part. Your congressperson isn't just a proxy for a party, but is supposed to represent you, a voter in their district. If you have a concern, you ought to be able to go to your congressperson and tell them. If there are issues your district particularly cares about, your congressperson should probably care about them too.

Local concerns do matter. If there's a gross chemical factory that's polluting your neighbourhood, we'd like your congressperson to really care that they could lose a large chunk of votes. We don't want the neighbourhood split thirty ways, so thirty different congresspeople all say "oh well, it's just a dozen votes."

This isn't just academic, either. In Saskatchewan, there were no city districts. Each city was split between four districts, which each also had a giant rural area attached. There was a small vote shift towards the rural-favoured party, so that party won all those districts. Now voters in the cities of Saskatchewan have no real representation, nobody in Parliament who cares about their city. (Thankfully, in Canada there's a neutral electoral commission that is in the process of fixing this.)

Think of how much worse this would get in NY. It's not just "a bit of NYC" in each district. To get vaguely competitive, contiguous districts, each one would have to have 1/30th of Long Island, a long but just 400 foot wide strip of Manhattan, and then a giant swath of upstate. What does a thin strip of Manhattan have to make it a community? What does it have in common with the Finger Lakes, so that we want them to have the same representative? What if the city candidate wins the vote in every district, and all of upstate NY has no representation in Congress? It's quite likely that no minority candidate will win, because we've split up every community—is that not a problem?

And yes, obviously there are downsides of single-seat representative democracy. It's not very fair to widely distributed communities, for one thing.


"Clearly it is not possible to draw all 435 Congressional districts as competitive districts, so I don't think it's likely that a 1% shift will lead to total turnover in the House. "

Ok, maybe not every seat. But I bet you could get 75% of seats to be competitive this way, if competitiveness is your primary concern. Would you be happy if Republicans won 52% of the vote this fall, and got a veto-proof majority of Congress?


"But even that would be better than our current system, in which shifts in the electorate are unable to change the control of the House of Representatives."

I'm sympathetic to your concerns that the current system plurality reversal, where the overall-seat-winner is not the overall-vote-winner. But remember that the Democrats held the House for 46 consecutive years. Because of a single plurality reversal in 2012, you're advocating a drastic change to an untested system with obvious flaws.

I know you're trying to help the Democrats, but this is populist Republican thinking at work: "There's a problem! Something has to be done! We don't believe in studying what solutions work, so here's something we pulled out of our ass...at least it's something!" You can do better.
posted by vasi at 2:30 PM on June 4 [1 favorite]


I don't think that's accurate, ROU. I don't think that a self-interested minority party with the ability to redistrict as it sees fit would create competitive districts.

Sorry, but you're just wrong about this. You maximize your seat share by maximizing the number of districts where you have a slight lead.

I don't know much about the NY State Senate but is it really accurate to say that its district lines are drawn to maximize the number of districts that are nearly balanced in terms of voter partisanship?

It's close to that; nobody anywhere plays completely hardball. The easiest way to see it is in the party registration. You'll see an absurd number of districts with a very small R edge in registration.
posted by ROU_Xenophobe at 3:28 PM on June 4 [1 favorite]


Yeah, that's weirdly one of the biggest misunderstandings of gerrymandering — the goal isn't to produce 100 percent safe districts, it's to spread your supporters as thin as possible while maintaining a majority.
posted by klangklangston at 4:03 PM on June 4 [1 favorite]


Vasi, most of us are in safe districts now. In any system that relies on districts, some people are going to live in safe districts. Wouldn't the competitive-district approach lead to fewer people living in safe districts than any other district-based approach? If you think that district-based systems are unfair to people living in safe districts, then shouldn't you support the competitive-district approach over, for example, districts drawn by commissions?

You're right that the competitive-district approach would not preserve communities of interest within district lines, though I would note that proportional representation has exactly the same problem. If you support proportional representation despite this problem, why do you raise it as an objection to the competitive-district plan? The lack of minority representations is a real problem with this approach, but this could be mitigated somewhat by drawing the "safe" districts as majority-minority districts (this would pretty naturally happen anyway in states like NY, CA, IL).

I also don't see the problem with a situation where a party that gets 52% of the national vote gets the ability to set policy. I mean, isn't this basically what happens in parliamentary systems? If you get the majority, you get to implement your program. I think that's good for democratic accountability, since it allows/forces parties to show voters that their platforms work if implemented. I think it's a heck of a lot more legitimate than having a party that received a minority of votes elected to the majority of a legislature.

I don't understand your point about the 46-year Democratic majority. Did Republicans ever receive more votes than Democrats for House seats during that period (I honestly don't know). Besides, my goal isn't "to help the Democrats" but to create a system where voters have their preferences reflected better in the legislature. PR would work best for this, but Americans seem to like having districts, so maybe this plan is a next-best approach.

ROU, you got me intrigued about the NY senate, so I went back and took a quick and dirty look at the 2012 election results. Of the 63 districts, only 8 had the top two candidates within 10,000 votes of each other (most districts had like 100,000-120,000 votes cast so 10,000 votes is like an 8-10% margin). And there appeared to be numerous super-safe seats on both sides, which is not what you'd expect if the districts were drawn so as to maximize the number of competitive seats. To me, it looks like the NY state senate districts are designed with incumbent protection as a higher priority than competitive districting. If you have better data to look at, I'd be happy to do so.

You and klang are totally correct that gerrymandering isn't about producing 100% safe districts, it's about producing the largest number of districts with a reasonably reliable edge for your side. But there has to be a lower bound on this, doesn't there? A self-interested party wouldn't draw districts where it has a one-voter advantage over the other party; instead, I'd think it would try to draw as many districts as possible with like a 3-5% advantage over the other party, even if it could draw fewer of those than single-voter-advantage districts. The goal with the competitive district plan is to set the percentage that defines a competitive district lower than the percentage that makes gerrymanderers comfortable.
posted by burden at 5:40 PM on June 4


You're confusing personally safe incumbents with safe districts, which is why I suggested looking at party registrations.
posted by ROU_Xenophobe at 5:51 PM on June 4


Is that information available online somewhere? Or a summary of it? Would love to see some support for your position.
posted by burden at 5:56 PM on June 4


NY latfor. Too lazy to look it up myself.
posted by ROU_Xenophobe at 6:32 PM on June 4


Found it!

http://www.elections.ny.gov/NYSBOE/enrollment/senate/senate_apr14.pdf

I count 14 districts in which Republicans have a registration advantage, out of 63 districts. I went ahead and calculated the percentage of the two-party voter registration that is held by the Republicans in each of these.

- The smallest registration advantage held by Republicans is 2.2%, in the 59th District.
- The largest registration advantage held by Republicans is 16.8%, in the 43rd District.
- The average registration advantage in these 14 districts is 9.4%
- There are four districts in which the Republican registration advantage is less than or equal to 5%.
-There are six districts in which the Republican registration advantage is greater than or equal to 10%.

Frankly, I don't see this data as supporting your position. It looks to me like the district mappers did not try to draw competitive districts, or districts maximizing the number of districts with small Republican majorities. These look like incumbent protection districts to me. In any case, it seems to me like the districts would look very different if the mappers were required to draw as many districts as possible with, say, under 2% discrepancies between Republicans and Democrats. If you have a different take on it, I'd be interested.
posted by burden at 6:40 PM on June 4


If you want to look at why his plan is bad, look at his California map and compare it to the current congressional districts. He whacks far northern California in two, as does the current map... except that the current map does it while following the county lines to a very high degree. He has the far east of California broken into quite random chunks even where you could follow county lines between Mono et al. Central California is just a bunch of blobs.

And it's simply for little gain. His districts have population differences in the thousands, while the current one differs by no more than *2* between any district. The improvement of distance for average voter to the center of the district is all of six miles. The actual districts are generally quite compact -- the only really potentially dodgy one I can see is 21 pulls in a bit of Kern but doesn't do it the obvious way instead wiggling around to pick up the southern suburbs. When they do have to break a low population county between districts, they appear to only split them once. For the high population counties, there isn't any crazy Austin style snaking -- 51 pulls a bit from San Diego is about as close as you get, and it seems likely that was just to even up populations since San Diego already had to be broken up into multiple districts and Imperial doesn't have enough on its own.
posted by tavella at 9:12 PM on June 4


burden: "most of us are in safe districts now. In any system that relies on districts, some people are going to live in safe districts."

Yes, most of us are in safe districts now. But in places with good commissions, like Canada where I live now, if we're in a safe district it's because a certain party or representative is organically popular in the area where we live. If the party leadership or representative does something bad, or a new popular candidate emerges, we could easily become a competitive district.

Under your system, if you're in a safe district, it's because the commission arbitrarily chose to put you there. If the commissions remain partisan, they'll choose for partisan reasons—perhaps put their party's least-committed voters in the safe districts. I think there's a big difference between being denied influence because of organic factors, and being denied influence intentionally, because of your political leanings.

And under your system, what if the dominant party in your state does something you dislike, and your safe district actually votes against them? Well, if the commission would rather your census block be in a safe district, you'll go right back into another safe district soon. There may be no escape.

In short: if you intend to intentionally deny some people influence for the good of the many, you better find a damn good rule to decide whom.


"You're right that the competitive-district approach would not preserve communities of interest within district lines, though I would note that proportional representation has exactly the same problem."

Most forms of proportional representation retain a regional component, or even districts: Mixed-Member Proportional, Single Transferable Vote, and even List-PR with multiple districts. Only in a very few countries are there no districts at all.


"The lack of minority representations is a real problem with this approach, but this could be mitigated somewhat by drawing the "safe" districts as majority-minority districts (this would pretty naturally happen anyway in states like NY, CA, IL)."

Almost all southern states currently vote Republican in presidential elections. Therefore, your system would divide these states into districts that are either 50% R/50% D, or safe Republican. There would be no safe Democratic seats. Your scheme might very well completely eliminate African-American representatives in the South.
posted by vasi at 12:57 AM on June 5


« Older There are lots of dialects of English. Which one d...  |  Urban Dictionary Spelling Bee:... Newer »


This thread has been archived and is closed to new comments