Become a Patreon!
Excerpted From: Aviel Menter, Calculated Discrimination: Exposing Racial Gerrymandering Using Computational Methods, 22 Columbia Science and Technology Law Review 346 (2021) (260 Footnotes) (Full Document)
In 2004, a plurality of the Supreme Court held that challenges to partisan gerrymanders were nonjusticiable, finding in Vieth v. Jubelirer that there were no judicially manageable standards by which such claims could be adjudicated. However, the Court's decision did not completely shut the door on future partisan gerrymandering challenges. Justice Kennedy concurred in the judgment, providing the fifth vote to uphold the challenged map. But he left open the possibility that judicially manageable standards might one day be developed that would enable federal courts to distinguish between acceptable and unconstitutional partisan gerrymanders. If “new technologies” could be developed to more precisely measure “the burdens gerrymanders impose,” Justice Kennedy would find partisan gerrymandering claims justiciable.
Political scientists and data scientists quickly picked up Justice Kennedy's gauntlet, developing a flurry of techniques and metrics designed to objectively identify excessively partisan district maps. Simulated redistricting software, powered by advanced machine learning algorithms, promised to be one of the most effective techniques. These algorithms take as input the state legislature's neutral and non-partisan redistricting criteria--for example, compactness, contiguity, or the desired number of majority-minority districts-- and produce thousands of viable district maps satisfying those criteria. To show that a state legislature's district map was politically gerrymandered, a litigant compares the political bias of the legislature's map to the average political bias of the maps randomly generated by the simulated redistricting algorithms. In the event the legislature's map exhibits an unusually strong partisan lean compared to those produced by the algorithm, one might reasonably assume the legislature had drawn districts primarily for partisan advantage.
In the end, however, Justice Kennedy left these experts at the altar. One term after he retired, the Supreme Court in Rucho v. Common Cause rejected simulated redistricting tools as a source of judicially manageable standards for the detection of partisan gerrymanders. Instead, the Court held that no clear line could be drawn between an acceptable and unacceptable level of partisanship in redistricting--and that constitutional challenges to partisan gerrymanders therefore did not present a justiciable question.
But simulated redistricting tools may still have a promising future. Although federal courts may not use these algorithms (or, indeed, any other technique) to overturn partisan gerrymanders, these algorithms can still help to identify racial gerrymanders. Since at least the 1960s, the Supreme Court has held that racially motivated redistricting not only presents a justiciable legal question, but also violates the Constitution.
Unsurprisingly, detecting racial motive in the redistricting process presents much the same challenge as detecting political motive. State legislatures are usually guarded about the unconstitutional purpose behind their discriminatory legislation. Instead, they frequently obfuscate their intent by pointing to other, facially legitimate justifications for the district maps they produce. But simulated redistricting algorithms can control for these facially legitimate criteria. And just as a legislature's map can be compared to a set of automatically generated maps to uncover political bias, the same comparison can also be performed to identify unexplained racial bias.
Existing simulated redistricting algorithms generally require little alteration to identify racial gerrymanders. Many of the justifications that legislatures provide to obfuscate political motive are also used to justify racial gerrymanders. For example, the Voting Rights Act (VRA) generally prohibits state legislatures from drawing districts in such a way as to break up majority-minority districts--i.e., politically cohesive racial minority voting blocs numerous and geographically compact enough to comprise a majority in a single district. The VRA therefore opens the door for legislatures to argue that their consideration of racial data in the gerrymandering process is necessary to ensure that the resulting district map does not violate the Act. But simulated redistricting algorithms--designed to draw the required number of majority-minority districts--will produce a realistic sample of permissible maps that the legislature could pass. In the event the legislature's proposed map evinces measurably greater racial bias than those in the sample, the claimed VRA compliance motivation can be assumed to be pretextual.
Indeed, after Rucho, tools designed to detect hidden criteria in the redistricting process may be more necessary than ever. Now that the Supreme Court has declared partisan gerrymandering challenges nonjusticiable, it is easy to imagine that state legislatures will be particularly brazen about using their stated political motives to disguise their underlying unconstitutional goals. Indeed, the Supreme Court has already faced several such challenges. If partisan gerrymandering is permissible, but racial gerrymandering is not, courts adjudicating racial gerrymandering challenges must disentangle partisan and racial motives--a difficult task in jurisdictions in which race and politics are highly correlated. But to the extent that such distinctions can be detected, automated districting algorithms can help separate such motives. By holding constant the number of minority party districts, in the same way that they already hold constant the number of majority-minority districts, simulated redistricting algorithms can determine whether a proposed map's racial bias is really the result of purely political considerations, or whether it is unexplained by constitutional criteria alone.
Of course, the Court rejected simulated redistricting as a viable tool to identify partisan gerrymanders in Rucho. But this does not suggest that it would do the same in the context of racial gerrymanders. Indeed, the Court in Rucho took great pains to distinguish partisan gerrymandering and racial gerrymandering, clarifying that its holding did not touch on the latter. The line-drawing concerns that caused the Court to reject simulated redistricting in the partisan context have already been resolved with respect to racial gerrymanders. Accordingly, this Note argues that simulated redistricting algorithms present an empirically helpful and legally viable tool to identify hidden racial motive in the redistricting process.
Part I of this Note explains the purpose and function of simulated redistricting algorithms. These algorithms were originally designed to solve the computationally difficult problem of creating a representative sample of all possible district maps. Data scientists and political scientists have since developed a number of techniques to address this problem. And although the Supreme Court has rejected the use of simulated redistricting algorithms in partisan gerrymandering challenges, that decision does not reflect any technical deficiency in the algorithms, but rather the Court's reluctance to wade into the political thicket.
Part II of this Note summarizes the law of racial gerrymandering, explaining that a district map is presumptively unconstitutional if racial motive predominated in the district drawing process. This Part also discusses common defenses to racial gerrymandering challenges, including that a district map was actually motivated by a partisan purpose, or that consideration of racial data was necessary to comply with the Voting Rights Act.
Finally, Part III of this Note argues that simulated redistricting algorithms are particularly well suited to identify racial gerrymanders. The ability to generate a representative sample of district maps can help identify any number of hidden districting criteria, not just partisan motive. And because simulated redistricting algorithms can control for the number of partisan or majority-minority districts, they can help detect whether a particular district map is justified by Ruchosanctioned partisan goals or Voting Rights Act compliance.
State and federal legislatures have access to a number of means by which they can combat the problem of gerrymandering. They can rely exclusively on neutral criteria when drawing districts, defer to districts drawn by computer, delegate the redistricting process to an independent commission, or even adopt a different election system not susceptible to gerrymandering. Courts, however, have a relatively limited number of tools to address this problem--especially after Rucho. So long as the redistricting process remains in the hands of self-interested state legislatures, plaintiffs will seek empirical methods by which they can separate a legislature's facially legitimate motive from its hidden unconstitutional one. Even after Rucho, simulated redistricting provides such a method.
[. . .]
Pretextual redistricting has become more common since state legislatures have realized that they can abuse the VRA to harm minority voters. And it may become even more common yet, now that the Supreme Court has given its imprimatur to partisan gerrymanders, often nearly indistinguishable from racial gerrymanders. Racial gerrymandering cases present an abundance of thorny legal questions regarding the extent to which legislatures can consider race when drawing election district boundaries. But when intentional racial gerrymanders arise--not out of a good faith effort to comply with the VRA or enhance minority representation, but from an attempt to suppress the minority vote--courts face a practical problem as well: legislatures are wary of admitting their invidiously discriminatory motive. Accordingly, courts need a way to distinguish necessary and lawful consideration of race (or race-correlated factors) from unlawful discrimination.
Simulated redistricting algorithms offer help with this narrow but crucial task. Although originally developed to provide a more objective measurement of a map's partisan bias, simulated redistricting provides a generalizable technique for isolating and identifying the considerations that entered into a legislature's redistricting process. A comparison between a legislature's map and maps generated algorithmically pursuant to neutral criteria can therefore provide powerful evidence that the legislature considered a factor--such as race--that it claims to have ignored.
Of course, the theoretical appeal of simulated redistricting does not guarantee the practicality of its application. More empirical research will be necessary to verify that simulated redistricting algorithms are actually capable of isolating racial motive on real-life district maps. And even if these algorithms can fulfill their purpose in practice, courts may still reject them. Rucho may not have foreclosed simulated redistricting in racial gerrymandering cases, but--even before Rucho--the judiciary has sometimes been hostile to or confused by statistical argument. Misunderstanding the algorithms, courts may still dismiss empirical evidence produced by simulated redistricting as “sociological gobbledygook.”
However, even if simulated redistricting algorithms are put to good use, the problem of intentional gerrymandering will be far from solved. When legislatures draw district maps, there are necessarily winners and losers. And though courts may be able to prohibit legislatures from intentionally picking losers on the basis of race, the judiciary ultimately cannot prohibit the consideration of every criterion on which a legislature might gerrymander. To truly prevent representatives from choosing their voters, the redistricting process may ultimately need to be removed from legislative hands, or made irrelevant by the adoption of an election system in which redistricting is either difficult to manipulate or does not take place at all. But, as long as legislatures remain mapmakers--and as long as they continue to draw maps on the basis of race--courts will need empirical tools to help them to identify intentional racial gerrymanders.
J.D. 2021, Columbia Law School.
Become a Patreon!