Not a Job for Siri: Why Automated Redistricting Is the Wrong Fix for Gerrymandering
Every 10 years after the census, legislative bodies of all types – from the U.S. House of Representatives to state legislatures and city councils – must have their districts redrawn. The U.S. Constitution requires this redrawing of voting districts to ensure that districts of the same type have equal population. Unfortunately, in most states and in most communities, legislators draw their own maps and congressional boundaries. This is a clear conflict of interest that usually results in gerrymandering, the manipulation of districts for political advantage.
As gerrymandering driven by improved technology and big data has grown more sophisticated, it has resulted in elected officials creating districts that make it harder for voters to hold them accountable on Election Day. In response, Americans have fought back and sought creative public policy approaches to stop this threat to democracy. The results are the creation of independent citizen redistricting commissions, bipartisan politician commissions, new state constitutional standards, and other solutions that make redistricting fairer and more independent.
“Why can’t you just use make a computer draw all the redistricting maps?”
Improved technological tools have also generated suggestions for ways those tools can combat the problem. There is one such suggestion that we will address here: automated computer-generated districts. Although this appears to be a viable and unbiased solution on the surface and is one that we are asked about frequently, it is problematic and undemocratic for several reasons. The purpose of this blog post is to discuss why Common Cause is skeptical of automated redistricting.
“Doesn’t using computers to draw districts completely eliminate human bias?”
No computer program to draw districts will ever be free of human judgment and bias because human beings must input instructions for a computer program to follow. One or more human beings must make value judgments about whether to tell the program to prioritize competition between the major parties, keeping counties and or cities together, nesting state house districts into state senate districts, or one of the many measures of compactness, partisan symmetry, or responsiveness. Choosing which to consider or one over the other is a political and value judgment whether a person does it directly when drawing districts or programs a computer to do so.
“OK, but once humans have made all those decisions, can’t a computer automatically generate the best map?”
Ordering a computer to generate a map based on specific criteria does not result in one map. It results in an infinite number of maps.
In several court challenges to the gerrymandering of districts, Common Cause has hired University of Michigan political scientist Jowei Chen to testify as an expert witness. Professor Chen’s persuasive approach for identifying partisan gerrymanders is to program software to produce thousands of different maps using the criteria the defendant legislators purported to use, excluding only partisan advantage. He then shows that very few – or frequently none – of the computer-generated maps produce the partisan advantage for the majority party that the legislator-drawn maps produce. This demonstrates that the challenged map is such a statistical outlier that it must have been drawn with partisan intent and effect.
In his work, Professor Chen generally produces 1,000 maps per simulation. Another one of our expert witnesses produced more than 20,000. Other experts in the field describe how supercomputers can produce millions of maps. Unless a decision-maker chooses the first map a computer produces, an essentially arbitrary selection, political choices must be made. If the human mapmaker decides to choose between maps rather than make an arbitrary selection of the first one produced, value judgments are inevitable.
“I still think we should just have computers draw squares or rectangles and be done with it. What’s the problem with that?”
It’s illegal. Because of our country’s history (and current practice) of racial discrimination, Section 2 of the Voting Rights Act requires the drawing of districts in which racial and language minorities can elect their candidate of choice under Thornburg v. Gingles. The Gingles factors require a district to be drawn for a racial or language minority when that group (1) is sufficiently large and compact to constitute a majority in a district, (2) tends to vote similarly, and (3) is usually outvoted by the majority voting as a block.
As a result, a race-blind automated redistricting process that simply requires compactness and nothing else would violate federal law. Since a map-drawer is legally prohibited from drawing a perfectly compact grid without regard to communities of color, human input is needed to complete a more complicated map that begins with Voting Rights Act districts. Once human inputs are required to instruct the computer what to do, value judgments are needed. As discussed above, this belies the notion that this is an automated process.
Drawing squares also ignores one of the primary reasons for redistricting – to ensure that people living in the district can elect legislators who truly represent their interests. Arbitrary district boundaries ignore the fact that communities aren’t shaped like perfect squares and that people don’t live evenly spread out throughout a state. Communities are made up of people with common social, economic, and other interests. They are centered around schools, or playgrounds, or job-producing industries. Communities cluster around water or other resources. Without human input into drawing maps, these people-centric districts would be split up to the detriment of the people who live in them.
“Who should draw maps if you think automated redistricting is problematic?”
If even automated redistricting done by a computer requires human judgment, this raises the obvious question of how to minimize bias among the people making those judgments. As an organization that advocates for empowering We the People to actively engage in our democracy, Common Cause supports redistricting that gives the public a voice in their own political representation. This means creating a system that neutralizes partisanship, prevents individuals with a direct stake in the outcome of redistricting from drawing districts, and encourages robust public participation in a transparent process. Redistricting is fundamentally about fair representation, which cannot be discerned without hearing from those who will be represented.
Why we value voices of people from each community…
One of the reforms that Common Cause has supported across the country to end gerrymandering is the independent citizen redistricting commission. The most effective of these commissions meet the criteria described above. They neutralize partisanship by including an equal number of commissioners from the two major parties along with commissioners who are members of neither major party. They remove from the process individuals with a direct stake in the outcome of redistricting by outlining a list of conflicts of interest that will prevent a person from serving as a commissioner. These include categories such as being a current or recent elected official, candidate, lobbyist, political party official, or major donor. The best redistricting commissions encourage robust public participation in a transparent process by requiring all commission deliberations to happen in public and by mandating a minimum number of hearings across the jurisdiction with adequate public notice. If achieving all of these objectives is difficult, implementing one or more of them would still improve the process.
An example of this reform, the California Citizens Redistricting Commission, is now being emulated in other states. In the last redistricting cycle, the California Commission traveled throughout the state to hear from 20,000 Californians who had stories to tell about their communities. Their stories included descriptions of shared business districts and local environmental concerns, agricultural issues, language communities, religious backgrounds, shopping locations, and many others that census data alone does not show. Commissioners were screened for direct political conflicts of interest, so their primary concern was balancing the interests of communities as fairly as possible. The commission provided an opportunity to a wave of people from all walks of life to tell their stories to a decision-making body that, for the first time in the state’s history, was actually listening.
At the heart of Common Cause’s philosophy about the best path forward is an understanding that there is no such thing as a perfect map. It cannot be constructed with a program, in a lab, by elected officials, or by a redistricting commission. There are always trade-offs. The trade-off in an automated system is getting a nice-looking map while completely ignoring input from the people who will vote in those districts. That is not an acceptable trade-off to us. As a result, Common Cause does not seek a perfect map. We instead seek a transparent process that empowers people to shape their own representation.