Objective measure of gerrymandering?

[lpetrich]

If The Supreme Court Reads This Study, It Could End Partisan Gerrymandering Forever | ThinkProgress

For the last decade, the Supreme Court of the United States has openly refused to police partisan gerrymandering even in egregious cases where the state legislature or its congressional delegation bears little resemblance to the will of the people. A new study out of Duke University, however, casts serious doubts on the reasoning of the justices who have thus-far refused to strike down unconstitutional gerrymanders.

In 2012, Democratic U.S. House candidates in North Carolina received 81,190 more votes that Republicans. Republicans received just under half of the votes earned by the two parties. And yet, the GOP walked away with 9 of the state’s 13 congressional districts. So, despite the fact that they earned just over 49 percent of the two-party vote, Republicans won nearly 70 percent of the state’s congressional seats

Jonathan C. Mattingly and Christy Vaughn considered this question, and they wrote a program that composes lots of random districts that are constrained to be contiguous. Here is the paper that they wrote:
[1410.8796] Redistricting and the Will of the People

They generated districts randomly, while imposing some constraints on them. They must be contiguous, they must be approximately even splits of the population, and they must be compact. The importance of the first one was absolute, while the relative importances of the second and third ones could be adjusted with a weighting parameter.

They used a division of the state into Voting Tabulation Districts (VTD's) and they found each one's neighbors. They used a random walk to generate the districting. Each step in the walk was to look for a VTD with a neighboring VTD in another electoral district and have that other one annex the first VTD. The step would always be accepted if it improved the constraint satisfaction, and part of the time if it did not. They ran the redistricting first in a "hot" phase, making it wander over the space of possibilities, then in a "cold" phase, to make it home in on a nearby good possibility. After settling on that possibility, they made the system "hot" again to look for a new one. They ran each phase for 20,000 or 50,000 steps, their short and long phases.

They ran their simulation for 100 hot-cold alternations, getting a districting for each cold phase. They then found out for each of them how many Democrats the districts elect. They found that over half of them elect 7 or 8 Democrats and the rest of them elect 6 or 9 Democrats. They hardly ever found a districting that elects only 4 Democrats. That's pretty much what one would expect from the proportion of voters who voted Democratic.

It would be interesting to extend this work to other states, to see how gerrymandered they are.

When a gerrymandering case came before the Supreme Court, the conservative justices refused to get involved on the ground that there seemed to be no good objective way of recognizing gerrymandering. Anthony Kennedy seemed more optimistic than Antonin Scalia that such a way could emerge in the future. This work looks like it could be the beginning of work that Justice Kennedy, at least, might find satisfactory.

 

[Bronzeage]

If one is not able to define obscenity, but knows it when one sees it, the same should apply to gerrymandering.

 

[NobleSavage]

This study is just North Carolina. And keep in mind that Section 2 of the Voting Rights Act requires that districts be gerrymandered such that demographic groups which are a _minority_ of the population make up a _majority_ of the voters in those districts. When states fail to gerrymander for democrats ^H^H^H^H^H^H^H^H^H^H black people, the federal government intervenes and forces gerrymandered districts.

Let's look at Maryland's 3rd congressional district:

http://www.newrepublic.com/article/...cas-most-gerrymandered-congressional-district

 

[fromderinside]

Here's what we should do. Every county that touches either ocean or navigable body of water (wet) should be considered a democratic county and every county that does not (dry) should be considered a republican county. So Democrats should gerrymander by minimizing the number of republican voters from dry counties and republicans should gerrymander by minimizing persons from wet counties in voting districts. Each party should start with an equal number of districts.

My favorite would be to make pie slices originating at the center of cities with greater than 300,000 residents outward to rural areas thus assuring democrats control all elections.

 

[Togo]

What would happen if you randomly assign each voter to a district in their state based on nothnig more than chance, and keep the allocations secret until the votes are cast?

 

[Jayjay]

What would happen if you randomly assign each voter to a district in their state based on nothnig more than chance, and keep the allocations secret until the votes are cast?

How would they know whom to vote, if the districts are secret? Or should they vote for candidates in all the districts, akin to bloc voting?

 

[Bronzeage]

What would happen if you randomly assign each voter to a district in their state based on nothnig more than chance, and keep the allocations secret until the votes are cast?

It would be an impossible bit of electoral calculus. It would probably work out to be a de facto "at large" election, where a slate of candidates, say 10 qualify for 3 offices. Each voter votes for three. It means that if a particular majority dominates, they elect their three candidates. This does not lead to representative government.

The only solution is to responsibly draw districts which try to balance the identified demographic groups, based on race and economics. It's a difficult task, but it can be done. If gerrymandering were not a real thing, the national GOP would not have spent so much time and money to take control of state legislatures, with the purpose of redrawing Congressional districts.

 

[Bomb#20]

What would happen if you randomly assign each voter to a district in their state based on nothnig more than chance, and keep the allocations secret until the votes are cast?

It would be an impossible bit of electoral calculus. It would probably work out to be a de facto "at large" election, where a slate of candidates, say 10 qualify for 3 offices. Each voter votes for three. It means that if a particular majority dominates, they elect their three candidates. This does not lead to representative government.

That's an artifact of the arbitrary rule that each voter has to vote for three different people. "At large" elections don't produce that result if each voter only gets one vote, or if each voter is allowed to cast all three of his votes for the same person. Either of those is an effective way to get representative government, provided minority groups are able to get the word out to their members about voting strategy.

The only solution is to responsibly draw districts which try to balance the identified demographic groups, based on race and economics. It's a difficult task, but it can be done.

That's a second-worst solution to a completely artificial problem. Why the heck should race and economics be the only groupings by which voters are allowed to make common cause with one another? Democracy is supposed to be voters choosing their rulers. Geographic districts, either gerrymandered or even "responsibly" drawn, are rulers choosing their voters. The better solution is to prevent the problem from arising in the first place. Either abolish districts altogether -- see above -- or else abolish the demand that they be geographic and let each voter choose which abstract district she'll be a member of for herself.

How the heck can anyone imagine that government could even in principle ever possibly be representative, if there's a rule that you and your neighbor -- the one who disagrees with you about everything -- both have to be "represented" by the same person?

 

[Blahface]

If you allow humans to draw the districts, you are always going to have a high risk of gerrymandering. The best solution is to use the shortest split line method to draw districts.

Also, non-partisan primaries with approval voting would help curb the effects of badly drawn districts. It would ensure everyone has a say regardless of which party dominates the district. In a competitive environment without vote splitting, even small minorities can have a big influence in elections so they can't be completely ignored.

 

[lpetrich]

If you allow humans to draw the districts, you are always going to have a high risk of gerrymandering. The best solution is to use the shortest split line method to draw districts.

Or else the algorithm like what I'd reported on in my OP -- create some random initial districts, alter them, accept or reject the alteration in simulated-annealing fashion, then repeat LOTS of times. It uses a penalty function that gives penalties for departures from population equality and compactness. One can easily add other penalty-function terms, like ones that make a certain number of districts be majority-minority ones.

 

[Jayjay]

If you allow humans to draw the districts, you are always going to have a high risk of gerrymandering. The best solution is to use the shortest split line method to draw districts.

Or else the algorithm like what I'd reported on in my OP -- create some random initial districts, alter them, accept or reject the alteration in simulated-annealing fashion, then repeat LOTS of times. It uses a penalty function that gives penalties for departures from population equality and compactness. One can easily add other penalty-function terms, like ones that make a certain number of districts be majority-minority ones.

The random method is fine for research purposes where you want a statistical sample, but what would stop whoever is in charge of the districts to keep rolling the dice until he gets a districting that he likes most?