Redistricting: Should Computers Draw the Lines?

What’s more, the maps only would be as good as the formula that Tim or Mike Fortner or anyone else wrote. Fortner pointed out that some elements of redistricting are awfully difficult to code. How do you tell a computer to abide by the Voting Rights Act, which is a matter of judicial interpretation?

Fortner did say that a computer might be able to handle a relatively simple map, like Iowa’s congressional districts. Iowa is overwhelmingly white, so it doesn’t have to worry about Voting Rights Acts compliance in redistricting. It only has five congressional seats. And it has a rule that counties can’t be divided. Of course, the same factors that make it an easier place for a computer to draw a map also make it an easier place for humans to draw maps. “The computer is best as a tool to guide people,” Fortner told me, “and let them know how they’re doing.”

Just to make sure my idea was bad, I posed it to a couple of panelists discussing technology in redistricting at the National Conference of State Legislature’s annual meeting in Louisville (where I am this week). Kim Brace, president of Election Data Services Inc., said there actually has been quite a lot of academic research into using computers to redraw maps — with lousy results. “If you think legislative plans look crazy,” he said, “wait until you see what a computer does.”

Mark Stratton, who works on redistricting with Indiana’s Legislative Services Agency, made another point: A redistricting plan only does any good if legislators are willing to vote it into law. “The computer,” he said, “can’t vote for the plan.”

At first, that comment struck me as missing the point. Whether the maps are drawn by humans or computers, there are plenty of ways to structure the redistricting process so that lawmakers have no choice but to accept proposals that abide by objective criteria. Fortner’s constitutional amendment would have done just that in Illinois.

But, the more I think about it, Stratton was making the critical point. People are perfectly capable of drawing fair, reasonable congressional and legislative lines. When states end up with unfair, unreasonable lines, it isn’t because no one could figure out how to draw a better map. The reason is politics.

I’ve talked before about redistricting approaches on this blog before.

Briefly:

• Prop 11 is not perfect but I do support it as significantly better then the previous highly-partisan approach.
• To me, an ideal solution would be to use good multi winner winner election methods to select a body of representatives rather then using single winner districts. Ideally representatives would be chosen in a way that each person is equally represented by someone they (helped to) choose and whose political viewpoints closely resemble their own. This could be achieved through proportional representation or, even better, asset voting.
• Where single winner districts are needed I would prefer that they be selected in an automated fashion – without much human input.
• …but the voting rights acts (both federal and state) apparently requires the creation of minority-majority districts – which effectively requires gerrymandering
• Given these constraints, an independent redistricting commission approach seems to be the only viable option. Prop 11, although flawed creates such a commission.

So… should I apply for the commission? Do I have enough to offer? Is the potential time commitment worth it?

I support it. Please vote yes on 11.

I’ve written a fair amount about redistricting approaches so I’ll reference some of these posts in my explanation for why I support prop 11.

The problem

Currently the legislature is responsible for drawing districts in which those same legislatures are elected. This creates a clear conflict of interest. This conflict currently occurs in that:

• legislators want to protect their own seats
• legislators want to maximize the seats that their party gets

In CA I would describe the current status quo as a ‘bipartisan gerrymander’ in that the problem is more characterized by seats being protected then by a power grab where one party has a disproportionate number of seats and controls the redistricting system to maximize the number of seats controlled by that party. (I may be wrong in this regard; If you have evidence to the contrary please leave a comment.)

The result of this gerrymandered system is that rather than having voters selecting their representative we frequently have representatives choosing their voters via partisan or bi-partisan gerrymandering.

See also:

• An introduction to gerrymandering
• Gerrymandered districts yield a highly predictable outcome for who will be elected. Prop 11 claims that 99% of incumbent politicians were elected in recent elections
• An excellent tool for understanding how redistricting can be used to control who will be elected see the Redistricting Game

My ideal solution

In my view the problem of redistricting is best handled in two ways.

1. Use a multi-winner election method with larger (or state-wide) districts. Election methods that do this are known as proportional representation systems.

2. Where districts are used have the method of creating the districts be automated by software without any human input once the algorithm is determined. This is called algorithmic redistricting.

See also:

• single member districts vs proportional representation
• An algorithmic redistricting proposal called the shortest splitline method
• several posts about algorithmic redistricting

Prop 11 – an ‘independent commission’ solution

Prop 11 proposes having an independent commission which is responsible for doing the redistricting. This is a considerable improvement over having the state legislators do it. Much of the proposition details how this commission is selected.

What I like:

• The power to create and shape districts is taken out of the hands of people who have a direct stake in the outcome.
• Prop 11 only addresses districts for CA legislators. It does not address districts for US House seats. This is important since other states (Texas notoriously) are known to gerrymander in favor of having more Republican US House members and some people opposed previous CA redistricting reform since they considered this to be ‘disarming’ relative to TX behavior. It is my view that national redistricting reform is needed to address how US house districts are selected.

What I dislike:

• The independent commission that is created is designed to have a precise balance of Republicans, Democrats, and independents on it. I find it extremely distasteful to enshrine into law an assumption that there are two major parties and that the parties have an approximately equal power balance. This does not necessarily reflect future reality even if it closely resembles the current reality. (And it does not even represent the current reality in CA. The Pew Center reports that currently 39% of CA voters identify themselves as Democrats compared to 28% as republicans. (Pew Research citation) That is a 3:2 ratio not a 1:1 ratio.
  See also:
  • Do independent or bi-partisan commissions solve gerrymandering?
  • All my posts about gerrymandering

Conclusion

Prop 11 is not perfect. It’s not how I would design a solution. But, on balance, it is a clear improvement over the current perverse system. I recommend voting yes on prop 11.

Prisons Warp the Vote

…apportionment plans must be based on population and where voters reside should not affect the power of their vote. As a result, an industry or other economic interest can only exert political influence in proportion to the number of people that support it. The Reynolds case intended to permanently the settle the question of whether other interests besides population can be the basis of representative democracy.

Except it didn’t.

The Census Bureau counts people in prison where their bodies are located on census day, not where they come from and where they will return, on average, 34 months later. Forty-eight states bar prisoners from voting, and most states have constitutional clauses or election law statutes which explicitly declare that prisoner remains legal residents of their home addresses. When states draw districts based on the Census Bureau’s padded counts of prison locales, they give those districts extra representation just because the prison industry has a facility there.

…
Of the 12,594 people in the district containing the New Hampshire State Prison, 11 percent are incarcerated. As a result, every group of 89 residents in this prison district has the same political power as 100 residents elsewhere in the state.

The article references an advocacy site for this issue called Prisoners of the Census. That site has some good information about the issue including historical background for why the Census counts prisons in this way and arguments for why it is appropriate to change the way that prisoners are counted but not to change the way that college students are counted.

You can read more about how this hazard seems to occur in practice.

Quoting Eli Rosenbaum from the WASHINGTON POST Saturday, 29 October, 2005:

So why don’t independent commissions work? One big reason is that, as the current commissions demonstrate, equal representation on the panel for both parties tends to favor the status quo. The commission is indeed bipartisan, not nonpartisan, and each party’s delegates on the panel are closely connected to their state parties and politicians. To avoid gridlock and approve a plan, commissioners must draw a map that is pleasing to both sides, and of course nobody on either side really wants a competitive district. Political scientists even have a name for this type of redistricting scheme: bipartisan gerrymandering.

On the other hand, non partisan redistricting seems to have worked well in some places.

Take Canada, where gerrymandering is almost non-existent.

The Canadians have an “independent non-partisan agency” to draw districts (“Elections Canada”). Jean-Pierre Kingsley, the Chief Electoral Officer, appointed 1990 to serve until age 65 or resignation, is Canada’s best paid civil servant. This post is appointed directly by parliament and reports to them. The C.E.O. can be removed from office only for cause, by the Governor General after a joint request following a majority vote by the House of Commons and Senate.

(Lazyweb: I can use some images to back the claim that Canada is not very gerrymandered.)

Iowa, which has used a non-partisan redistricting service since 1980 is also rather nice.

The Rose report points out that these algorithms could be unconstitutional and seems to consider algorithmic redistricting approaches to be politically naive.

A recent letter-to-the-editor in The Appeal-Democrat suggested we just draw district lines according to latitude across the state to create the areas our legislators represent. Many people say such things, not unreasonably, because they are ignorant of the fact that such districts would inevitably be unconstitutional.
…
The idea behind such proposals is simple: people want to take the “politics” out of redistricting politics in a complete and total manner. Of course, in practice, trying to take the politics out of politics is as nonsensical and impossible as the verbal formulation. Further, such reforms might cause more problems than they would ever solve. Such exercises can be valuable as thought experiments, but realistic reforms strive for something much more moderate and, for lack of a better word, political.

I disagree with the assertions made by the Rose report that algorithmic approaches to redistricting are forever unrealistic reforms.

The Rose report directly and indirectly references a number of other algorithmic redistricting proposals that have been made. Not surprisingly the shortest-splitline algorithm is not the first such proposal.

A list of references (caveat: I have only read a few of these):

• The Rose Report post which includes excerpts from an interview with Barry Keene:
  

  HICKE: Ah, yes, there comes the Elections Committee again. [Laughter]

  KEENE: Yes, it’s redistricting time, so you get to do redistricting.
  Well, the politics of redistricting was fascinating, and it was coming up, but it was not really my cup of tea. Having power over the survival of lots of legislators by drawing districts is a source of great power, but it s also not where I wanted to be.
  …
  HICKE: Did you propose redistricting by computer, or was that just something that happened?

  KEENE: Yes, I did. I ve always felt that redistricting more than anything else poisoned the well in the legislature, particularly the assembly, where Republicans felt that they would never be permitted to become a majority, and decided that they would throw bombs within the institution and bring the institution to a halt. They were probably correct in their judgment on the first part, that under redistricting by Democrats they would never be permitted to become a majority.
  …
  People didn’t like commissions, because they always figure that no matter how much you try to balance commissions, there’s always going to be that odd vote that s going to be the deciding factor
  …
  So I said, “What else can we do?” And maybe borrowing on my-why I borrowed on it, I’ll never know-on my coin-tossing experience as a way of producing staff reduction, [laughter] we were going to create redistricting by computer, and use something called a random number generator to resolve differences. So after all of the criteria had been applied, the criteria of equal numbers of districts, the criteria of minority representation-there’s some federal criteria on that-after all of the attempts to consolidate areas of interest and not cross too many jurisdictional boundaries, you would still be faced with the capacity to create an infinite number of districts. It would just be a smaller infinity than you had before.

  So how do you resolve that? You do it with a random number generator. Well, nobody understood this business. I shouldn’t say nobody. Most legislators said, “Well, we can’t turn redistricting over to a computer, its not politically good for us, and we don’t understand what it is that you’re proposing.” John Garamendi, I believe, in one of his moments of appealing to the television spotlight, got up and said, “You know the old saying, Barry: it’s garbage in, garbage out.”

  Well, it died a fairly uneventful death.

• Letter to the editor of The Appeal-Democrat proposing just using latitude lines
• Brian Olson’s “fair redistricting” algorithm
  
• George L Clark’s redistricting algorithm
• PDF with redistricting proposal by Roland Fryer and Richard Holden
• political science dissertation by Micah Altman on redistricting formulations ( I have not read this but I understand that he describes desirable properties of redistricting and then demonstrates that the finding an optimal solution is computationally complex (NP hard) )
• A pair of posts by the Geomblog on algorithmic redistricting. post 1, post 2

The Ballot Access News reports:

According to this story in the Riverside, California Press-Enterprise, wealthy Congressman Darrell Issa of San Diego County has agreed to pay to get an initiative on the California June 2008 ballot. That initiative would provide that each U.S. House district elect its own elector. The initiative already collected 100,000 signatures in August, then had been abandoned for lack of funding.

Here is Jack Santucci’s take on this development (Jack was an analyst at Fairvote.org and I’ve disagreed with him frequently on other matters. Here I am nearly full agreement with him.):

CA Congressman Darrell Issa (R-49) will help bankroll the effort to split California’s Electoral College votes by congressional district (CD allocation). And he’s defending it as a move to “proportional representation.”

“This is about making people’s votes count,” he said. “It’s about proportional representation.” […]

Issa insists that he has not endorsed a candidate for president and said the effort is not motivated by politics, but by a desire to increase voter turnout in the state.

“If Florida had proportional representation [in 2000], Al Gore would be president today,” he said.

In another post I highlight some problems with CD allocation. The biggest (in my opinion) is that doing so would drastically raise the stakes of redistricting wherever the system were implemented. Bluntly, gerrymandering would affect presidential elections.

The folks at Why Tuesday have made a 5 minute video segment where they visit Chris Swain, director of the University of Southern California’s Electronic Arts Game Innovation Lab to discuss The Redistricting Game.

View the segment.

While you are at it, you can visit the Redistricting Game blog

This certainly occurs with gerrymandered single winner districts and is possibly inherit to single winner districts that use the conventional plurality voting system (vote for one, whoever gets the most votes wins).

Only a few factors can lead to a very solid prediction of who will win. Incumbency is one such factor. Incumbents almost never lose.

Take the 2006 midterm elections as an example. This is viewed as ‘a great change election’. In that election over 94% of incumbents remained in power. And this, indeed, was a shake-up. Normally an incumbent will win an election 98% of the time.

To see just how predictable elections are, consider the projections that FairVote makes in it’s Monopoly politics report. They make their predictions 2 years ahead of the elections and have had an accuracy rate of 99.8% in projecting winners in the 1,613 races the called between 1996-2004 (for 552 races they did
not make predictions but instead labeled these races as competitive or vulnerable). The only data that they use in predicting the winner for a district is the results from recent federal elections in the district and the incumbent’s party and seniority.

So incumbents win 98% of the time and 75% of US legislature elections are predictable with 99.8% accuracy more than one year ahead of the elections. In other words, we don’t really have a two party dominated political system. Instead each voter is effectively dominated by one party.

An open question:

The ‘one party’ article references the ‘political oddsmaker’ site by Ron Faucheux. However, the link to it is broken. Is this still online? What is the link?

The “issue” that gerrymandering to make majority-minority districts is required by law… sounds interesting. I encourage you to actually look at the law and inform us what the heck you are talking about.

Alas, I never did do that research. I had originally suspected that this might be so from the redistricting game.

It looks like Warren has now heard the same message from CA Secretary of State Debra Bowen while attending the Electronic Voting technology Workshop in Boston.

Here is what Warren has said on the matter:

CA secretary of state and heroine D. Bowen was there listening to the talks. I sat next to her at lunch and tried my best to give her the goods. But she’s very knowledgeable. I’m impressed. She said shortest splitline would be illegal under the voting rights act. Basically her argument was that gerrymandering is mandated by the VRA to create “majority minority” districts and splitline would prevent that hence is illegal, QED.

I suppose that I (or someone) should do the legwork to read the voting rights act to see what it has to say on the matter.