A Workable Standard
Statistical Outlier Analysis and the Solution to Gerrymandering
The Consent of the Governed Part 3 // Listen to this as a podcast.
Part I: Authoritarian States
Thomas Hofeller, “the master of the modern gerrymander,” died in August 2018, leaving behind a trove of secret files, emails, studies, spreadsheets, and other documents relating to Republican gerrymanders and the Trump Administration’s 2019 attempt to add a citizenship question to the 2020 Census. Described in the New York Times as the “Michelangelo of gerrymandering,” Hofeller was a top Republican political strategist, who had played a large role in Republican gerrymanders and political maps across the country — in Arizona, Florida, Maryland, Mississippi, Missouri, North Carolina, Ohio, Tennessee, Texas, and Virginia — in the past decade, and was an advocate for adding a citizenship question to the Census in order to dilute Democratic representation by under-counting immigrants.
The files were discovered by Thomas Hofeller’s estranged daughter, Stephanie Hofeller, who learned of her father’s death in September 2018 when she randomly searched for her father’s name online and found an obituary posted a month prior. She drove to her parents’ home in Raleigh, North Carolina and visited Kathleen Hofeller, her mother. There Stephanie found a plastic bag with 18 USB thumb drives and 4 external hard drives, which contained about 75,000 files, including family pictures and the documents pertaining to gerrymandering and the Census question. According to an NPR article: “Before Stephanie arrived at her parents’ apartment, her father’s business partner, Dale Oldham, had removed a laptop and a desktop computer with Hofeller’s work files, Stephanie said her mother told her. ‘Dale got all the good stuff,’ Stephanie told attorneys.”
Stephanie Hofeller got in contact with Common Cause, a nonprofit government watchdog group, to search for a lawyer to represent her mother, and mentioned the files in passing. At the same time, Common Cause had recently filed suit in North Carolina state court against the Republican-gerrymandered congressional map, and the 75,000-file stash that insluded records of Thomas Hofeller’s construction of that same congressional map piqued their interest. The law firm representing Common Cause in the suit, Arnold & Porter (which, coincidentally, was also representing private plaintiffs pro bono in Federal District Court in Manhattan in the suit against the potential citizenship question), subpoenaed the Hofeller files.
Geographic Strategies, Thomas Hofeller’s company, attempted to prevent the publication of the Hofeller files, which Stephanie made available online in January 2020, because Geographic Strategies argues that the files contain “trade secrets.” These challenges, as well as the fear of the files being destroyed, prompted Stephanie to send copies of the files to the New York Times, the New Yorker, and other news organizations, and many other publications have since reported on them.
One irony revealed in the Hofeller files is that Hofeller emphasized discretion and email security in a presentation for legislators and congressional map–drawers, saying “emails are the tool of the devil” and “treat every statement and document as if it was going to appear on the FRONT PAGE of your local newspaper.” Of course, Hofeller’s files have now appeared in the New York Times and the New Yorker — hardly discrete platforms for Republican gerrymandering secrets.
In 2015, the files reveal, Hofeller was hired by the Washington Free Beacon, a conservative publication, to conduct a study of the potential impact of drawing legislative and congressional maps based on voting-age population, rather than total population.
As written by David Daley in the New Yorker: “Mr. Hofeller’s exhaustive analysis of Texas state legislative districts concluded that such maps ‘would be advantageous to Republicans and non-Hispanic whites,’ and would dilute the political power of the state’s Hispanics. The reason, he wrote, was that the maps would exclude traditionally Democratic Hispanics and their children from the population count. That would force Democratic districts to expand to meet the Constitution’s one person, one vote requirement. In turn, that would translate into fewer districts in traditionally Democratic areas, and a new opportunity for Republican mapmakers to create even stronger gerrymanders. The strategy carried a fatal flaw, however: the detailed citizenship data that was needed to draw the maps did not exist.”
The solution, Hofeller argued, was a citizenship question in the 2020 Census. In 2016, Hofeller and Dale Oldham, his business partner who took Hofeller’s laptop and desktop after his death, got in contact with then-president-elect Trump’s transition team and Mark Neuman, who was managing the Census transition and became an advisor to Commerce Secretary Wilbur Ross.
In November 2019, the House Oversight and Reform Committee released text messages and emails between Neuman, Hofeller, and Oldham revealing that the three were designing the language of a Census citizenship question with the goal of undercounting immigrants and diluting Democratic voting power in August 2017. Included in the released documents was a letter from the Justice Department to the Census Bureau that stated the question was necessary to guarantee “compliance with the requirements of the Voting Rights Act and its application in legislative redistricting.” As the addition of a citizenship question would likely undercount minority groups, this reasoning is, at best, malicious.
In another email to Hofeller, Neuman dismisses the idea of using the American Community Service (ACS) data as a source for voting-age population statistics. The ACS is a standard annual Census survey; I used it for voting-age population in my hypothetical election simulations to calculate efficiency gaps last episode and found it completely suitable — but I digress.
The Hofeller files were used in the suit against the Census citizenship question as proof of partisan intent behind the question and evidence that there is no other substantive argument to add the question other than helping Republicans at the ballot box.
In May 2019, a Justice Department spokesman stated that the 2015 Hofeller study “played no role in the Department’s December 2017 request to reinstate a citizenship question to the 2020 decennial census.” That “reinstating” refers to the last time a census citizenship question was considered — in 1950. Additionally, the Justice Department’s argument that Hofeller had no role in the formulation of the census question is dead wrong. From the New York Times: “In their court filings… lawyers for the plaintiffs said that ‘many striking similarities’ between Mr. Hofeller’s study and the department’s request for a citizenship question indicated that the study was an important source document for the Justice Department’s request.” To top off the blatant partisanship in the drive to add the citizenship question, the New York Times, again, wrote: “The filing also says flatly that [assistant attorney general for civil rights] Mr. Gore and Mr. Neuman ‘falsely testified’ under oath about the Justice Department’s actions on the citizenship question.”
The Supreme Court, in Department of Commerce v. New York, did not buy the Commerce Department’s argument that the question was necessary to fulfill requirements established in the Voting Rights Act and remanded the case to the district court. The Trump Administration later dropped the question.
The Hofeller files also revealed Thomas Hofeller’s role in the 2016 North Carolina Republican gerrymander. The gerrymander was one of the worst in the nation, if not in American history (the map was struck down last year as a violation of the state constitution). The 2016 map was a partisan gerrymander introduced by the Republican-controlled legislature as a replacement for the 2011 Republican-drawn map, as that map had been struck down as an unconstitutional racial gerrymander. The 2016 map was, to put it mildly, extreme: the congressional delegation in 2018 was comprised of 3 Democrats and 10 Republicans, even though Democrats received a majority of votes cast in House elections.
In order to draw the incredibly effective gerrymander, Thomas Hofeller compiled a database of voter registration, gender, race, likely partisan lean, and addresses (down to the residence hall) of 23,100 North Carolina college students, many attending historically black colleges and universities like North Carolina A&T State University. According to the New Yorker: “Some spreadsheets have more than fifty different fields with precise racial, gender, and geographic details on thousands of college voters.”
This intricate data collection paired well with North Carolina’s 2013 voter-I.D. law, which reduced the number of eligible voter-I.D.s, especially hindering ballot access for minority groups and students, limited early voting, and ended same-day voter registration. Hofeller was involved in the defense of the voter-I.D. law when the North Carolina N.A.A.C.P. challenged it in court. The result of this was described, again, in the New Yorker: “Perhaps one of the clearest and ugliest gerrymanders in North Carolina — or in the entire nation — is the congressional-district line that cuts in half the nation’s largest historically black college, North Carolina A&T State University, in Greensboro. The district line divided this majority minority campus — and the city — so precisely that it all but guarantees it will be represented in Congress by two Republicans for years to come.” All of Hofeller’s gerrymandering and Census work was bankrolled by the Republican Party with the explicit intent to disenfranchise Democratic voters.
The problem of rampant partisan gerrymandering in North Carolina is furthered by other authoritarian tendencies in the state’s government. In 2016, Democrat Roy Cooper edged out incumbent Republican Pat McCrory in the North Carolina gubernatorial election, defeating McCrory by about 0.2% of the statewide vote. However, before Cooper took office, the North Carolina legislature (which was — and is — Republican-controlled and Republican-gerrymandered) convened for a special legislative session to strip powers from the governor before he took office. According to Ratf**ked by David Daley, the legislature, among other changes to the state constitution, “…reworked those county election boards so that both parties shared control. Except they wouldn’t actually share power. Democrats would govern them in odd years, Republicans in even ones. Statewide elections, of course, are only held in even years,” (David Daley, RatF**ked, 228).
The 2016 North Carolina legislative coup foreshadowed similar events in Wisconsin and Michigan in 2018, when Democrats Tony Evers and Gretchen Whitmer, respectively, defeated incumbent Republican Governors. In both states, the Republican legislature convened for a special session and again took power from the governors’ offices before the elected Democrats assumed power. Two of those states — North Carolina and Michigan — have an efficiency gap over the proposed 2-congressional seat threshold of unconstitutionality in favor of Republicans — 3.52 and 2.29, respectively — and the third state — Wisconsin — has an efficiency gap of 1.42 in favor of Republicans, high enough, in my opinion, to still be considered a gerrymander.
On top of the extreme partisan gerrymandering and legislative coup-ing that has taken place in North Carolina in the past five years, there have also been accounts of election fraud, the most notable being in the 2018 congressional election in North Carolina’s 9th district. There, according to an NPR article, Republican political operative Leslie McCrae Dowless was “…the alleged ringleader in a scheme instructing his co-conspirators to sign certifications that falsely stated they had seen a voter vote by absentee ballot, and improperly mailing in absentee ballots for someone who had not mailed it themselves.”
All of these gross violations of North Carolinians voting rights and authoritarian undermining of election process point to the fact that, as the Electoral Integrity Project found, North Carolina is “no longer considered to be a fully functioning democracy,” (Carol Anderson, One Person, No Vote, 97). The Electoral Integrity Project is a Harvard University research project studying the integrity and responsiveness of governments around the world. University of North Carolina professor Andrew Reynolds used the Electoral Integrity Project’s criteria to measure how democratic and representative the North Carolina government was. According to RatF**ked, North Carolina “earned a failing grade with an overall electoral-integrity score of 58 out of 100 for 2016. That score, he wrote, ‘places us alongside authoritarian states and pseudo-democracies like Cuba, Indonesia, and Sierra Leone.’ […] When the professor measured the integrity of the district boundaries, he found something he almost could not believe: North Carolina earned a 7 out of 100. That’s not only the worst rigged-district ranking for any state in the country, but the ‘worst entity in the world ever analyzed by the Electoral Integrity Project,’” (Daley, 226).
So why did I just spend 1,980 words discussing the Hofeller files and democracy — or the lack thereof — in North Carolina? Well, first, gerrymandering is part of a much larger crisis of democracy in America; to improve our election system, we need to eliminate gerrymandering as well as make the ballot box more accessible in a number of ways including expanding mail-in voting, ending voter-I.D. laws, and implementing automatic voter registration, among others. Second, partisan gerrymandering harms people. For voters in North Carolina, there is little incentive to cast a ballot, as the district lines predetermine election victors. That mis-representation harms voters by not reflecting their actual policy preferences. And third, the problem of partisan gerrymandering is really bad, North Carolina probably being the most authoritarian, anti-democratic state in the union.
Having described the dire extent of the gerrymandering problem in North Carolina, I want to turn away from that state for the time being, and look at another potential solution for the Supreme Court’s request for a “workable standard” that has played a large role in recent judicial deliberations on the North Carolina congressional map — statistical outlier analysis.
Part II: Random Walks
In part 2, I discussed the efficiency gap — how it functions, its benefits and drawbacks, and its role in gerrymandering litigation going forward. In Gill v. Whitford, the efficiency gap failed to compel the court to rule on the constitutionality of partisan gerrymandering, and the measure was relegated to a secondary role in the future, only acting as a small piece of evidence in a larger set of measures. Statistical outlier analysis took the efficiency gap’s place in Rucho v. Common Cause and other state constitution suits. The method, in my opinion, is a far better measure of gerrymandering, and has greater potential to act as a “workable standard” in court.
Statistical outlier analysis is superior to the efficiency gap in its complexity; while the efficiency gap advertises itself as a “single tidy number,” statistical outlier analysis recognizes the fundamental intricacy of voting and, by extension, democracy (Nicholas Stephanopoulos & Eric McGhee, Partisan Gerrymandering and the Efficiency Gap, 831). Furthermore, the one-dimensionality of the efficiency gap is its undoing in valuing competitiveness, reflecting the long-term stability of gerrymandered plans, disregarding gerrymandering in small-congressional district states, conflating winning and losing wasted votes, and idealizing 75–25 district and state vote splits.
Indeed, no gerrymandering measure can perfectly capture the nuance of the democratic voting process. As mathematician Moon Duchin said in an interview with Quanta Magazine this year, “We fundamentally don’t know how to and should not try to turn the whole complicated picture of representative democracy and its ideals into an objective function… no objective function really captures the complexity of what we’re trying to do when we vote.” What this means for gerrymandering measurement is, as Jonathan Mattingly and Christy Vaughn wrote in a study entitled Redistricting and the Will of the People, “The ‘will of the people’ is not a single election outcome but rather a distribution of possible outcomes,” (Jonathan Mattingly and Christy Vaughn, Redistricting and the Will of the People, 1).
The essential question that statistical outlier analysis seeks to answer is: What would a congressional district map look like if partisanship had not had a role in the drawing of district lines? To this end, the method simulates tens of thousands, million, even billions of congressional maps, modeling for state-determined redistricting criteria and controlling for Supreme Court–mandated population equality, and produces a curve describing the probability of state maps to produce different partisan outcomes. This is then compared to the map under scrutiny to determine if it is a statistical outlier and is therefore a partisan gerrymander.
In state redistricting, the most basic unit to construct a map is the precinct, though this can be further broken down with more intricate, detailed data. In my home state of Colorado, there are currently 3,219 precincts. In the 2016 presidential election, there were 170,850 precincts in the entire United States. Precincts can then be divided into trillions and trillions of different possible maps; mathematician Moon Duchin wrote in an article for Scientific American, “By the time you get to a grid of nine-by-nine, there are more than 700 trillion solutions for equinumerous rook partitions, and even a high-performance computer needs a week to count them all.” Incredibly, mathematicians have yet to calculate the number of possible maps in a simple six-by-six grid split into two districts where the districts do not have to be equal-sized, as it would take a computer over a week to calculate the number of possible maps. When approaching the number of precincts in a state, Duchin said in an interview, “we’re probably looking at the google range, by which I mean 10 to the 100.” This far exceeds the number of atoms in the universe, which is between 1078 and 1082 (Mattingly & Vaughn, 11). To state the obvious, this far exceeds Justice Alito’s suspicion that, in drawing the lines, there’s “maybe dozens, maybe hundred, maybe even thousands of ways.”
There are so many possible plans that it would be near-impossible to calculate all the possible maps in a state, so political scientists and mathematicians use Markov Chain Monte Carlo, which is a method to take a random sampling of maps that is representative of the universe of possible maps — this universe of possible maps is called an ensemble of maps (Moon Duchin, Gerrymandering Metrics: How to Measure? What’s the Baseline?, 4). Markov Chain Monte Carlo is essentially a random walk through a map; imagine you’re standing at one point on the edge of a precinct and walk along the edge of the precinct until another precinct’s border intersects the edge you’re walking along. You then have a choice to continue along your current path to the next vertex or to turn and walk along the edge of the intersecting precinct. Markov Chain Monte Carlo is a mathematical way to simulate this decision-making on a much larger scale, finding a sampling of the ensemble of maps.
Rather than producing a completely random ensemble of maps, however, mathematicians input criteria to influence the probabilities of following different precinct edges and turning — or not turning — at intersections (Duchin, Gerrymandering Metrics, 4). There are a few federally mandated criteria for drawing district lines including population equality between districts (for example, you cannot have one district with a population of 100,000 and another with a population of 150,000), district contiguity, and there must be, to some extent, minority opportunity to be elected and represented (as mandated by the Voting Rights Act of 1965). Other redistricting criteria are set by state laws and constitutions, such as Iowa’s emphasis on keeping counties together and Arizona’s mandate to prioritize creating competitive districts, and these can also be worked into the Markov Chain Monte Carlo calculation (Duchin, Gerrymandering Metrics, 5).
Once political scientists and mathematicians are satisfied with the number of maps they have simulated (for example, in Rucho v. Common Cause, the plaintiffs simulated 24,518 maps), they can then lay them out on a curve revealing the probability of maps following state and federal redistricting criteria to produce different partisan outcomes using real-world votes (Rucho v. Common Cause, Kagan’s dissent, 20). The map under scrutiny is then compared to the curve of probable maps to determine if it is a statistical outlier. For example, if the average Democratic seats in an ensemble of maps for a state is 10, and in 90% of simulations, Democrats received between 7 and 12 seats, but under the current maps Democrats only receive 4 seats, the map is probably an outlier and, therein, a Republican gerrymander.
There is no set threshold for outliers to be declared gerrymanders, but this is not as great a detractor for using the method as it may appear. Fundamentally, statistical outlier analysis is descriptive of maps relative to the most probable, nonpartisan map rather than it being prescriptive of maps being gerrymanders. No single measure can perfectly capture the complexities of voting or gerrymandering, but the statistical outlier method effectively indicates when partisan intent may have played a role in the drawing of a map. The method shows when there is something wrong with a map that cannot simply be explained away as a result of state political geography or attempting to succeed in state or federal criteria, as geography and criteria are both built into the Markov Chain Monte Carlo.
Statistical outlier analysis has been used in gerrymandering litigation in, most recently, League of Women Voters of Pennsylvania v. Commonwealth of Pennsylvania, which was the successful challenge to the Republican-gerrymandered Pennsylvania congressional map ruled illegal under the state constitution. Additionally, statistical outlier analysis was used by plaintiffs in Rucho v. Common Cause, the unsuccessful suit against the Republican-gerrymandered North Carolina congressional map, which was brought before the Supreme Court in 2018 — Rucho challenged the same map that Thomas Hofeller helped draw.
In League of Women Voters of Pennsylvania, the Republican legislature-drawn congressional map was thrown out by the state supreme court after a statistical outlier analysis by mathematician Jowei Chen. According to a New York Times article, “In the view of the majority of the Pennsylvania Supreme Court, ‘perhaps the most compelling evidence’ that Republicans sacrificed traditional redistricting criteria for partisan gain was a political scientist’s [Jowei Chen’s] simulation of 500 possible congressional maps.”
After the old map was discarded by the court, both the legislature and Democratic Governor Tom Wolf proposed replacement maps. Mathematician Moon Duchin ran a statistical outlier analysis of the new plans for the governor and found that “there is less than a 0.1% chance that the Turzai-Scarnati plan [the plan drawn by the Republican-controlled legislature] was drawn in a non-partisan way,” (Moon Duchin, Outlier analysis for Pennsylvania congressional redistricting, 1). Additionally, the report found that “the GOV [governor’s] plan does not meet even the looser standard for statistical significance, and in fact when it exhibits any partisan skew, it is not skewed in the Democratic-favoring direction.” (Duchin, Outlier analysis, 4).
The political geography of Pennsylvania implicitly favors Republicans because Democrats are highly concentrated in large urban centers in Philadelphia and Pittsburgh and many other Democrats live in small and medium-sized cities that are surrounded by heavily-Republican rural areas such as State College, Erie, York, and Lancaster. It is difficult for these cities to be represented by a Democrat, even if the district lines are drawn without partisan intent. Duchin’s statistical outlier analysis reflects this: “The full range of possibilities I encountered in trillions of trials against recent [state] Senate vote geography was 4 to 10 seats for Democrats [out of 18], but the 5-seat outcome is relatively rare and the 4-seat outcome is vanishingly rare.” (Duchin, Outlier analysis, 3).
In Duchin’s analysis, she used two different methods of Markov Chain Monte Carlo: a simple and a weighted random walk (Duchin, Outlier analysis, 2). The simple random walk took into account federally mandated redistricting requirements, such as contiguity and population equality, but ignored normative state and federal criteria, such as compactness and minimizing county splits (county splitting is when a county is split into multiple districts even though its population is large enough to minimize splitting). The weighted random walk, on the other hand, did take into account normative criteria and limited cross-district population deviation.
Duchin also made maintaining communities of interest a priority in her weighted random walk, treating her defined “geoclusters” (cities, neighborhoods, and geographic areas) similar to counties, in that the algorithm probabilistically prefers to take random walks which do not split communities of interest (Duchin, Outlier analysis, 10). By doing so, Duchin argues that minority opportunity is protected as minority-majority communities of interest are less likely to be split.
Finally, Duchin measured the resulting probability curves and was able to compare her produced maps to the proposed and current maps using two traditional gerrymandering measures: mean-median difference (which is a form of partisan bias measure) and the efficiency gap, which I discussed in episode #2. She found the legislature’s proposed replacement map to be a gerrymander, and the Pennsylvania Supreme Court threw out the Republican map again and drew their own map. The new map was not a gerrymander, and in the 2018 midterm elections, Democrats received 9 of the state’s 18 seats, roughly mirroring their 53.9% statewide vote share.
The new, court-drawn map, however, is very favorable to Democrats: under the new map, Democrats received more congressional seats than they would have under any of the 500 randomly drawn maps in Chen’s analysis used in League of Women Voters of Pennsylvania. That being said, the new map is not a statistical outlier, and satisfies traditional, nonpartisan redistricting criteria. According to an Upshot analysis: “The new Pennsylvania map… meets every standard nonpartisan criteria. It’s compact, minimizes county or municipal splits and preserves communities of interest. But it consistently makes subtle choices that suggest that partisan balance may have been an important consideration.” The decision to roughly match proportionality is significant in that it is a unique choice among states, but it is not a reflection on the statistical outlier method in itself, the method generally recommending a more favorable map for Republicans. Regardless, the new map still slightly favors Republicans relative to a proportional partisan distribution.
Statistical outlier analyses of North Carolina, similar to that of Pennsylvania, reveal extreme Republican gerrymanders. In 2014, mathematicians Jonathan Mattingly and Christy Vaughn published one such analysis, revealing that, in a sampling of 100 maps in their ensemble, in no map did Democrats receive 4 or fewer of North Carolina’s 13 congressional seats, and Democrats averaged 7.6 Democratic seats (Mattingly & Vaughn, 3). At the time, Democrats held only 3 congressional seats in the state.
It should be noted that North Carolina’s congressional map was changed in 2016 after it was ruled an unconstitutional racial gerrymander. However, Mattingly & Vaughn’s findings hold true with the redrawn map because, although the new map visually looked a lot better, it produced similar results and remained heavily Republican-gerrymandered.
In the study, two different methods were used to produce ensembles of maps: “Long Period” and “Short Period.” The Short Period method produced more maps, but was less restricted by the inputted state and federal redistricting criteria, while the Long Period was more precise but had a narrower ensemble (Mattingly & Vaughn, 4). The two methods had similar results, determining that a random nonpartisan map drawn to some extent within the confines of state and federal criteria would produce approximately 7 Democratic seats, with the Long Period method giving Democrats slightly more seats than the Short Period method (Mattingly & Vaughn, 5).
Part III: A Workable Standard
The greatest advantage of statistical outlier analysis over the efficiency gap, partisan bias, and other gerrymandering measures is that it clearly does not use proportionality as a baseline. Instead, rather than comparing a map’s current partisan distribution to a proportional seat allocation, statistical outlier analysis reveals what a map would look like had there been no partisan intent in the redistricting process.
In Rucho v. Common Cause in 2018, Jowei Chen, who had run a simulation for League of Women Voters of Pennsylvania, did a statistical outlier analysis of North Carolina’s congressional map, and that formed the core of the plaintiffs allegations that the Republican-controlled legislature had committed a partisan gerrymander. Justice Kagan remarked on the fact that statistical outlier analysis did not rely on proportionality in its determination of gerrymanders in oral arguments for Rucho: “What’s quite interesting about the statistical analysis in this case is that quite a lot of it does not run off a proportional representation benchmark. In other words, all the computer simulations, all the 25,000 maps… really do take the political geography of the state as a given. So… if Democrats are clustered and Republicans aren’t, that’s in the program. And all the other redistricting requirements or preferences, like contiguity, like following natural boundaries, that’s all in the program. So… the benchmark is not proportional representation. The benchmark is the natural political geography of the state, plus all the districting criteria, except for partisanship. And if you run those maps, right, what did you get? You got 24,000 maps and this — and 99 percent of them, 99 plus percent of them, were on one side of the map that was picked here,” (Rucho v. Common Cause oral arguments, 22:43).
Another advantage of statistical outlier analysis is that it gives legislatures some (limited) wiggle room in the redistricting process, allowing the drawing of a range of maps with a range of outcomes. This preserves the Constitutionally described role of the legislative branch and state governments in the drawing of district lines, while giving the judicial branch a much-needed role in overseeing the legislative branch in the redistricting process.
Additionally, statistical outlier analysis does not require any kind of hypothetical elections, a problem inherent to both partisan bias and the efficiency gap. All the data is sourced from real-world elections, and in the case of uncontested races, party registration and prior election data can be used in the affected precincts. Ultimately, the confluence of the benefits of statistical outlier analysis makes it, as Justice Breyer put it in Rucho oral arguments, “absolutely a workable standard,” (Rucho v. Common Cause oral arguments, 21:17).
