Reading: Introduction to the Redistricting Problem

Law of Gerrymandering

Douglas M. Spencer & Guy-Uriel Charles. 2020. “The Law of Gerrymandering,” Political Geometry, (ed. Moon Duchin et al.) Boston, MA: Birkhauser Science

Summary:

• Partisan vs. Racial Gerrymandering: Voting Rights Act (VRA) of 1965 opens the courts for legal challenges of racial gerrymandering. Supreme Court has articulated judicially manageable standards to enforce Equal Protection Clause of the 14th Amendment of the U.S. Constitution (plaintiffs are required to provide evidence that race as a "predominant factor" in drawing district lines - Miller v. Johnson (1995)) and the VRA. Shape of districts can be used as evidence of a racial gerrymander, but not always. Discrimination need not be intentional, but show how the "totality of circumstances... interact with social and historical conditions to cause an inequality in the opportunities enjoyed by black and white voters to elect their preferred representatives" (p. 6)
  • Predominant factor standard and totality of circumstances test have been judicially manageable for cases on racial gerrymandering, and these standards may extend to partisan gerrymandering.
• Legal standards for evaluating partisan gerrymandering: (1.) What is a legal standard? (2.) What does it mean for a standard to be "judicially discoverable and manageable?" [Do federal courts have the competency to adjudicate a particular category of cases? (a.) Courts as an institution are able to address the central conflict of the case. (b.) Courts can look at U.S. Constitution for a theory of harm and provide a remedy from a constitutional framework (political theory), public policy, and precedent/empirical data.]
  • Guarantee Clause: Denial of individuals from a republican form of government, draws from Article IV of U.S. Constitution. Status: Not justiciable.
  • Equal Protection Clause: An apportionment scheme might deprive votes of equal protection of the laws. Status: Prevailing argument in gerrymandering court cases since 1962. There is an accepted condition that political consequences do come from state legislative redistricting activity, but a threshold needs to be established to define when partisan motives become illegitimate. In Vieth v. Jubelirer (2004), two Justices argued that partisan motivation becomes illegitimate when "the minority hold on power is purely the result of partisan manipulation and not other factors" (p. 10). Another consideration offers that a redistricting plan can be "extreme" when it "lacks any rational relationship to representative government", with a focus on real and predicted effects/outcomes from the redistricting plan.
  • First Amendment: Protects individuals from viewpoint discrimination by the government and can be adopted to protect the "associational rights of political organizations". Status: Plaintiffs unsuccessfully attempted a First Amendment claim in Gill v. Whitford (2018), but Court highlighted that there was no clearly articulated set of standards to consider. Lower courts had previously detailed that in a 1A claim, "plaintiffs must provide evidence that (1) the district was intentionally drawn to burden voters based on how they voted, (2) that the burden resulted in a “tangible and concrete adverse effect,” and finally that (2) was caused by (1)" (p. 13).
  • Elections Clause: Article 1§4 of the Constitution reads, “The times, places and manner of holding elections for Senators and Representatives, shall be prescribed in each state by the legislature thereof; but the Congress may at any time by law make or alter such regulations" (p. 14). Thus, partisan gerrymandering is a limited, enumerated power, and supports claims that partisan gerrymandering is a justiciable issue. Limitations include that elections clause is only applicable to federal gerrymandering, excluding state legislative districts. Status: Attempted in North Carolina (2016 map - No opinion on claim) and Pennsylvania (2011 map - State Supreme Court rejected claim).
  • State Constitutions: Most state constitutions include clauses protecting the right to vote and for free and fair elections, courts can draw from state constitutions to provide a constitutional standard. Status: Standard upheld in League of Women Voters of Pennsylvania v. Commonwealth of Pennsylvania (2019)
• Alternative Approaches
  • Race as Party: Focus on racial effects of partisan gerrymandering, since race and party gerrymander are often correlated.
  • State and Federal Statutes: Congress and state legislatures can enact statutes to serve similar purposes as constitutional protections.

Landmark Court Cases:

• Baker v. Carr (1962) : Landmark case that engaged the Supreme Court in redistricting cases. Baker v. Carr established connection to the Equal Protection Clause. Individual equality ("one person one vote") was a set standard but a standard for evaluating group harms was more difficult to clarify. First case to ask what is "judicially discoverable and manageable", distinguishing between cases that were political in nature and cases where judicial intervention was appropriate.
• Gaffney v. Cummings (1973): Applied the Constitution to partisan gerrymandering claims. Gaffney v. Cummings specifically was a bipartisan gerrymandering case; Court upheld gerrymandering plan and said it would not attempt "the impossible task of extirpating politics from what are the essentially political process of the sovereign States" (p. 2) Last time the Court agreed that partisan gerrymandering claims were justiciable.
• Davis v. Bandemer (1986): Partisan gerrymandering case from Democrats in Indiana -- plaintiffs claimed that Republicans violated individual constitutional rights by redistricting to dilute the Democratic Party's political power. Justices disagreed on justiciability, with three Justices stating that federal courts regulating partisan gerrymandering was "flawed from its inception". Court majority decided that partisan gerrymandering claims were justiciable. There was attempt to articulate a judicially manageable standard set for the lower courts: "The most important of these factors are the shapes of voting districts and adherence to established political subdivisions. Other relevant considerations include the nature of the legislative procedures by which the apportionment law was adopted and legislative history reflecting contemporaneous legislative goals. To make out a case of unconstitutional partisan gerrymandering, the plaintiff should be required to offer proof concerning these factors, which bear directly on the fairness of a redistricting plan, as well as evidence concerning population disparities and statistics tending to show vote dilution. No one factor should be dispositive,” 478 U.S. at 173 (p. 3).
• Vieth v. Jubilerer (2004): Justices split on judicability, looking back to Bandemer and attempting to draw a standard for adjudicating partisan gerrymandering. Lower courts had struggled since Bandemer to agree around a workable standard. Justices disagreed on (a.) whether courts could intervene on gerrymandering cases, and (b.) what a workable standard could be in identifying partisan gerrymandering. Justice Kennedy cast a split vote, stating that partisan gerrymandering is justiciable but that they had yet to agree upon a single judicially manageable standard.

Unintentional Gerrymandering

Chen, J., & Rodden, J. (2013). Unintentional gerrymandering: Political geography and electoral bias in legislatures. Quarterly Journal of Political Science, 8(3), 239-269.

• Summary:

Contrary to common perceptions that intentional gerrymandering is the main cause of partisan bias in electoral maps, the authors of this study demonstrate that patterns of human geography contribute to significant levels of bias, which they refer to as "unintentional gerrymandering." Using automated districting simulations the authors show that the electorally inefficient distribution of Democratic voters in urban population centers produces an electoral bias, which favors Republican voters.

• “In majoritarian political systems like the United States, the extent to which electoral support for a party translates into legislative representation is driven by the geographical distribution of votes across districts” (p. 240)
• Different considerations in the districting process (e.g. preserving county boundaries; compactness) can have unintended consequences sometimes with racial or partisan implications even when maps are drawn without nefarious intentions.

Redistricting Algorithms

Becker, A., & Solomon, J. (2020). Redistricting Algorithms. arXiv preprint arXiv:2011.09504.

Summary

Redistricting algorithms can be categorized into two groups: (1) generative - design new plans and (2) assessment - determine whether a plan is evident of gerrymandering

Sample Algorithms: (see Table 1.1 for more details)

1. Enumeration
2. Random Unit Assignment
3. Flood Fill
4. Iterative Merging
5. Flip Step Walk
6. Recombination Walk
7. Power Diagrams

cut edges - object function used by practitioners to quantify compactness; cut edges counts how many pairs of neighboring units are assigned to different districts in a given plan (goal: minimize)

• local search methods: hill climbing, simulated annealing, tabu search, evolutionary algorithms

Discussion

Why not automate the process of the drawing districts?

• Need to consider the importance of human factor when there are critical tradeoffs
• There should be a procedural requirement to have human factor that makes it more democratic than an automated process
• Difficult to define: What is optimal? How to find the best optimal plan?
• Pro of automation: unbiased [but it may not be possible to fully remove subjectivity - Flood fill having some element of unseen bias]
• Con of automation: intractable due to complexity of features/combination/weighting of features

Redistricting algos (e.g. enumeration) is a Hard NP problem to generate all possible districting maps. Additionally, sample uniformly across space is also Hard NP problem.

The Ensemble Approach to Political Redistricting

Clelland, J., DeFord, D., & Duchin, M. (2020). Aftermath: the ensemble approach to political redistricting. Math Horizons, 27(3), 34-35.

Summary:

Gerrymanderring is defined as drawing districts to provide biased results to one political party. As an attempt to solve this issue scientists have proposed geometric techniques to detect potential gerrymanders. Even with the application of computational methods, resulting plans still favor one partisan. Districting plans are composed of graph partitions that satisfy a collection of rules determined by each state. As a result, there is no one-size-fits-all approach for taking on gerrymandering. Mathematicians around the country are using variants of Markov chain Monte Carlo to sample for a large collection of redistricting plans called an "ensemble". These plans are in turn used to generate a non gerrymandered baseline for comparisons. A certain district in the challenged plan is an outlier with respect to the ensemble if its partisan makeup is far different than the comparators. As a result, federal and state judges have adopted certain enacted maps unconstitutional as partisan gerrymanders. The ensemble method offers a breakthrough tool for detecting gerrymanders, for supporting reform efforts, and guiding mapmakers and commissions as they attempt the difficult task of drawing fair lines.

Specific Items of Note:

How do we decide how to compare the distribution plans?

• Certain districts are outliers if the partisan makeup is far different than the comparators.
• Two different sampling methods identify the same set of outlying districts.
• The sampling method allows us to evaluate plans, and rules for constructing plans, in advance of adoption.
• The plan can be generated from a neutral distribution (partisan distribution)
• The enacted should not fall within one standard deviation from the mean.

General thoughts (Next Steps):

• How to get the data together to solve the problem that can work with the computational tools
• Need some geometries that expand the entire state to satisfy contiguous constraint
• Shapefiles datafile -> geopandas -> geodataframes (operate as a data frame, column variable geometry that’s organized as polygons, what’s the boundary of this projection) * Population (descriptives) * Paristan (vote totals) * VRA compliant (racial demographics from Census)