# An Exercise in Line-Drawing: Deriving and Measuring Fairness in Redistricting

**Authors:** Jeanne C. Fromer
**Citation:** "An Exercise in Line-Drawing: Deriving and Measuring Fairness in Redistricting," 93 *Georgetown Law Journal* 1547 (2005)
**Source:** http://papers.ssrn.com/sol3/papers.cfm?abstract_id=599404

## INTRODUCTION

*p. 3*
Drawing a jurisdiction's district lines is a central and important task in a district-based election system-the principal American form-because the nature of districts often dictates the electability of candidates, if not the actual results of elections. 2 The people who draw these lines therefore have a great deal of power over who will represent each district. It is thus no surprise that one of the biggest debates in election law, both within the academy and the courts, is about policing redistricting to ensure that it is fair.

*p. 3*
It is imperative that a measure of fairness be elaborated in the redistricting context for at least two reasons. First, polities concerned with the impact of redistricting on legislative representation-and thus ultimately, substantive rights-require a measure of fairness to guide them in constructing districts at each redistricting opportunity. Second, if courts are to police states' redistricting schemes, they need manageable and preferably objective standards to adjudge the fairness of these schemes. A desirable model of fairness, then, is both constructive and descriptive. A constructive model of fairness instructs a polity on how to construct fair districting schemes, while a descriptive model of fairness allows courts and others to determine whether a particular districting scheme is fair.

*p. 3*
In Vieth v. Jubelirer, 3 the Supreme Court directly confronted the shortcomings of previous scholarly work on the construction and measurement of fair districting schemes in the context of partisan gerrymandering. 4 At issue in the case was when, if at all, partisan redistricting-that with a dominant purpose of preserving the power of the reigning political party, while concomitantly dimin-ishing that of partisan competitors-would be unconstitutional under the Equal Protection Clause. 5 In an outcome as difficult to parse as many redistricting schemes themselves, four of the Justices concluded that there was no way to measure fairness in districting without improperly stepping on politicians' toes, 6 one thought that there might be a way to measure fairness but did not yet know what it was, 7 and the remaining four Justices, in total, suggested three different ways to test the fairness of districts, each based on measuring the substantive qualities of the challenged district lines. 8 Nonetheless, a majority of the Vieth Court expressed its dissatisfaction with the existing measures of fairness and rejected them as a test for evaluating an unconstitutional partisan gerrymander. 9 In fact, some of the Justices, in their respective opinions, beseeched readers of Vieth to derive such a measure. 10 This Article addresses the bedeviling and pressing question of what constitutes a fair districting scheme, not only in the context of partisan gerrymandering, which divided the Justices in Vieth, but in general, while avoiding the pitfalls of previously proffered proposals. Prior academic literature primarily framed the debate over fairness in redistricting as a search for a substantive measure or standard of fairness. This Article argues that such a search is doomed to fail for a number of reasons, fundamental among them that no one can agree on what is "fair," 11 and that the proposed measures typically attempt to mask or subvert the political effects of redistricting and elections, despite the inherent nature of politics in redistricting. This explains why the Supreme Court-in Vieth and in various other voting rights cases-and the academy have struggled and failed to locate a substantive baseline by which to measure the fairness of a districting scheme.

*p. 4*
This Article seeks to reorient the discussion toward procedural standards and 5. See Vieth, 541 U.S. at 272 (plurality opinion). 6. See id. at 271-306. 7. See id. at 306-17 (Kennedy, J., concurring). 8. See id. at 317-42 (Stevens, J., dissenting); id. at 343-55 (Souter, J., dissenting); id. at 355-68 (Breyer, J., dissenting). 9. Id. at 281 (plurality opinion) (" [N]o judicially discernible and manageable standards for adjudicating political gerrymandering claims have emerged. Lacking them, we must conclude that political gerrymandering claims are nonjusticiable . . . ."); id. at 308 (Kennedy, J., concurring) ("[T]he parties have not shown us, and I have not been able to discover, helpful discussions on the principles of fair districting discussed in the annals of parliamentary or legislative bodies. Our attention has not been drawn to statements of principled, well-accepted rules of fairness that should govern districting, or to helpful formulations of the legislator's duty in drawing district lines."); id. at 345 (Souter, J., dissenting) ("Since this Court has created the problem no one else has been able to solve, it is up to us to make a fresh start."); cf. id. at 368 (Breyer, J., dissenting) ("[D]issenters might . . . believe that the more thorough, specific reasoning that accompanies separate statements will stimulate further discussion. And that discussion could lead to change in the law, where, as here, one member of the majority, disagreeing with the plurality as to justiciability, remains in search of appropriate standards.").

*p. 4*
10. This is particularly true of Justice Kennedy's concurrence and Justice Breyer's dissent. See supra note 9.

*p. 4*
11. Cf. Issacharoff, supra note 2, at 596 ("The difficulty . . . is in the development of a benchmark against which to measure the distortive effects of improper manipulation of the political process.").

*p. 4*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL measures of fairness. It puts forth a procedural model for constructing a fair redistricting scheme based on compromise, collective public participation, and dynamism. In this model, the potentially numerous public participants are aware of all of the possible districting configurations. Each participant ranks or assigns value to its configuration preferences, either by ranking each configuration individually or by establishing a policy-a ranked listing of criteria-by which it would rank the multitude of configurations. The participants' goals and preferences, encoded in their respective policies or rankings, are then blended together to derive a unique districting configuration. The derived configuration is the fairest configuration in terms of procedural fairness; effectively, the public participants have collectively compromised to arrive at the fairest redistricting scheme.

*p. 5*
There are a variety of ways to derive the redistricting scheme based on the participants' concerns; this Article provides some illustrations of such procedures. For example, in one version of the procedural model, the configuration with the most preferred overall ranking or value is deemed to be the fairest districting scheme. In another version, the participants alternate turns; at each turn, a participant eliminates his or her least-preferred remaining districting configuration. The turns continue until only one districting configuration remains. There are numerous other forms that this model can take, but what is most important about these illustrations is that they demonstrate the shift to thinking about fairness in terms of procedure rather than in terms of substance.

*p. 5*
Not only can this model be used to construct fair districting schemes, it can also be used descriptively and dynamically by courts (and others) to judge the fairness of existing districting schemes in varying contexts. One way to use the model to adjudge fairness is effectively to reconstruct the districting process by having a judge compare how close the actual districting scheme came to being crowned the fairest scheme under the model. The further it is from victory-in that either it was generally not preferred by the participants or in that some participant found that scheme to be particularly unfair-the less fair it is procedurally. This measure does not look at the substance of any districting scheme, but does require some difficult post hoc approximations of how the procedure should have worked. Another version of the model does not require the same post hoc reenactments and looks instead to the specific procedure used to construct a districting scheme and how far removed it is from a fair procedure.

*p. 5*
The procedural model is useful beyond judging the fairness of a particular districting scheme; it also offers a dynamic means of comparing districting processes across jurisdictions and over time. Democratic experimentalists can examine the various local debates about and rankings of districting values and then compare these disaggregated data points to derive insight about redistricting goals.

*p. 5*
The benefits of the procedural model are numerous. First, underlying the model are well-grounded notions of compromise as fairness. Second, the model 2005] 1551 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING places the political decisions in the hands of politically motivated public participants-not just the ones in political power-and minimizes the need for judicial interference. Because each participant decides individually how to rank districting configurations, participants can consider their particular interests in shaping their preferences. For that reason, the model does not require any choice among conflicting substantive measures and standards of fairness. The model incorporates all of these varied interests to yield a unique and fair solution to the redistricting problem. Relatedly, it captures the spectrum of what is fair by allowing multiple participants to input their conditions for fair districts and derive as output a single districting scheme that blends the participants' voices. Moreover, the model recognizes that it is impossible to ignore the political effects of redistricting and thus enables varied interests to incorporate their respective political goals to derive a fair solution. Finally, the model can generate a more fruitful public conversation about the structures of electoral politics, both through direct public participation in the redistricting process and through the ability to analyze and compare the various local districting values being articulated during the redistricting process.

*p. 6*
This Article proceeds as follows. Part I discusses the redistricting that took place in Texas in the past decade to demonstrate both the variance in how fairness can be defined by those involved in redistricting and the implications of mostly unconstrained redistricting undertaken by the political party in power. Part II describes the history of the federal government's actions to assure fairness in the field of voting rights. Part III considers and critiques scholars' attempts to define fairness in redistricting, both constructively and descriptively. Part IV, in response to the federal government's actions and the assessment of previous definitions of fairness, introduces and justifies the procedural model, which is grounded in compromise, public participation, and dynamism; details a number of illustrations of how the model could work in practice; and discusses how to implement the model. Part V concludes by addressing further avenues of research, including the determination of who gets to participate in redistricting, strategic gaming, and opportunity costs.

## I. REDISTRICTING AT WORK

*p. 6*
Every ten years, after states redistrict in accordance with the latest United States census results, 12 there are (unsurprisingly) cries from individuals and groups across the political and ideological spectrum-whose interests do not align with those of the redistricters-that the results of redistricting are unfair. 13 12. Texas and Colorado recently broke with this tradition and redistricted mid-decade. Toobin, supra note 2, at 63-65, 80. Adam Cox argues that increased frequency of redistricting can lead to at least two unfortunate consequences: first, it prevents the effects of partisan gerrymandering from eroding over the course of a decade, and second, it raises the likelihood that redistricting will be affected by partisan influences. Cox, supra note 4, at 754-55. 13. See, e.g., Dan Harrie, Leavitt To Endorse Redistricting Maps, SALT LAKE TRIB., Oct. 6, 2001, at A1; Daniel Ruth, Not Quite Tallahassee's Answer to the Round Table, TAMPA TRIB., Apr. 5, 2002, at 2 1552 [Vol. 93:1547 THE GEORGETOWN LAW JOURNAL They are typically disappointed with the amount of representation they are likely to receive in state and federal legislatures, and they are of the belief that a redistricting scheme should recognize their interests. To them fairness means having the ability to contribute to or participate in both the redistricting process and the new districting scheme.

*p. 7*
Even so, each interested group or individual has a different and often self-serving definition of a fair redistricting scheme based on their particular interests. For instance, in the 2001 redistricting process in Texas, the parties tried to dress up their motivations of protecting and gaining seats for their party in the garb of "fairness." The Chairman of the state House of Representatives Redistricting Committee, a Democrat, purportedly said, "People keep asking me if we will have a fair plan. It's going to be fair. I'm going to draw [a top Democrat]'s district first. Then mine. And then my good friends." 14 Tongue in cheek or not, he apparently is equating fairness with protection of presumably Democratic incumbent seats. By contrast, the leader of the Texas Democratic Party thought fairness was "respect[ing] the rights and choices of voters, preserv[ing] minority districts, defend[ing] minority voting strength and protect-[ing] our state's interests in Washington." 15 The Texas Republicans have interpreted both of these statements to mean that the Texas Democrats want to protect their incumbents against the dramatic Republican growth in Texas over the past ten years. 16 They instead associate fairness with "compact districts that respect counties, communities and areas of common interest" and a "plan that puts the interests of the people of Texas ahead of protecting incumbent politicians." 17 To Texas Democrats, the Republicans' statements really signify that they would like to increase their numbers in the state and federal legislatures. 18 It should come as no surprise that the Republican-controlled state Senate and the Democrat-controlled state House of Representatives would not agree on a redistricting scheme; therefore, a state legislative redistricting board drew one, which a federal court later approved, essentially preserving the partisan distribution of power in existence at the time. 19 In 2002, when the Republicans gained control of the state legislature, Tom DeLay, the U.S. House of Representatives Majority Leader and a Republican ("If the redistricting process was so fair, so open, so accessible, how did Temple Terrace, a modest little burg with only 21,000 residents wind up gerrymandered with three U.S. Congress members representing the city?"); cf. Toobin, supra note 2, at 63, 66, 78 (describing how a political party's control of state legislatures in Texas and Pennsylvania led to significant increases in congressional seats for that party even though the party received fewer overall votes statewide).

*p. 7*
1553 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING from Texas, encouraged the now-controlling party to re-redistrict then on the expressed basis that the then-existing scheme did not accurately reflect the state's political leanings. 20 Fifty-one Democratic state representatives fled to Oklahoma to deprive the state House of Representatives of the quorum it needed to pass such a re-redistricting bill, and eleven Democratic state senators did the same to the state Senate by fleeing to New Mexico. 21 Once the legislators returned, the Republican state legislators succeeded in enacting their desired redistricting scheme. 22 The U.S. Supreme Court has so far refused to impede it. 23 According to one commentator, the resulting redistricting scheme "may add as many as seven new Republicans to the G.O.P. majority in the House of Representatives and shift the state's delegation to 22-10 in favor of the Republicans." 24 One Democratic representative, unsurprisingly, did not appreciate the re-redistricting. According to him, "[The Republicans] did everything they could to bust up my political base . . . . They drew my farm and where I grew up into the Amarillo district, and they drew Abilene, where I live now, into the Lubbock district." 25 If both parties' assessments of the other major party are to be believed, the two major political parties in Texas each want to maximize their number of legislative seats and protect their incumbents. If that is right, then the parties deem their own behavior to be "fair" merely because it benefits them. It is readily apparent that both parties are equal-opportunity offenders of redistricting with the goal of protecting their own interests, sometimes in the guise of "fairness." 26 Although academic scholars and the Supreme Court typically do not have as blatantly politicized notions of fairness in redistricting as the Texas political parties seem to have, they have diverse, often conflicting notions of fairness. For instance, some scholars and judges think that fair districts are competitive, while others are convinced that fair districts are those constructed by nonpartisan commissions. Still others maintain that fair districts are those determined in a political free-for-all. 27 This Article now turns to the government's and scholars' varying notions of fairness in redistricting. 20. See Toobin, supra note 2, at 63. This mid-decade redistricting without new census data has been termed a "perrymander," after Texas Governor Rick Perry. See Note, A New Map: Partisan Gerrymandering as a Federalism Injury, 117 HARV. L. REV. 1196, 1196 & n.3 (2004) Voting is justifiably considered "a fundamental political right, because preservative of all rights." 28 It is primarily for this reason that many individuals and groups feel aggrieved at being excluded from voting or from having as effective a vote as others similarly situated. These individuals and groups have sometimes sought protection from the federal government to preserve these rights when they have thought them to be under attack. The areas in which they have sought protection include the guarantee of equally weighted votes for each individual, protection and promotion of minority group representation, protection of less dominant political interests in the redistricting process from more dominant political interests, and prevention of minority groups from too heavily influencing redistricting schemes. In each of these areas, the courts and Congress have struggled to determine an appropriate baseline by which to measure whether a voting or redistricting scheme is fair. In some instances, the federal government has conceded defeat and has announced that it cannot find a workable baseline; in those instances, the government has avoided measuring fairness. In other areas, by contrast, the government has fashioned such baselines. This Part provides an overview of the federal government's work in devising baselines by which to measure the fairness of voting and districting schemes.

## A. THE SUPREME COURT STEPS IN

*p. 9*
Until 1962, the courts would not interfere with legislative districting efforts, regardless of the circumstances of a particular scheme; the Supreme Court had determined redistricting "to be of a peculiarly political nature and therefore not meet for judicial determination." 29 In Baker v. Carr, 30 however, the Court decided, under the auspices of the Fourteenth Amendment's Equal Protection Clause, that it would wade into what Justice Frankfurter had memorably deemed the "political thicket," 31 at least to resolve challenges of unconstitutional state action in redistricting. 32 28. Reynolds v. Sims, 377 U.S. 533, 562 (1964) 34 Alabama's districts were apportioned based on out-ofdate 1901 census numbers. 35 In the intervening sixty years, Alabama's total population had increased from fewer than two million individuals to over three million. 36 Most of these gains had occurred in urban counties, while the rural counties had significantly decreased in population. 37 Because the state's districting scheme was based on outdated population numbers, districts containing rural counties had many fewer individuals than did districts containing urban counties. Therefore, an individual in a rural district had significantly more power (at least in the abstract) to elect a state legislator than did an individual in an urban district. 38 In holding this disparate scheme of apportionment to be unconstitutional, the Supreme Court reasoned that "fair and effective representation for all citizens is concededly the basic aim of legislative apportionment." 39 What motivated the Court to step into the "political thicket" was, in the words of one commentator, "a negative vision of an identifiable evil: uncontrolled partisan manipulation of apportionment and the accompanying distortion of true majoritarian preferences." 40 After Reynolds, at least in the area of ensuring one vote for every person, the fairness of legislative districts was now in judicial play. 41 Over time, the Court has come to rely increasingly on numerical equality of First, numerical standards could be drawn from unassailable empirical data, specifically the decennial census, and thus could provide an objective basis for measuring political equality. Second, the one-person, one-vote rule, based on strict population equality, could be readily managed by the courts, thus allowing a justiciable standard for judicial immersion into the political thicket of elected institutions. Third, the existence of objective measures would defeat attempts to gerrymander districting schemes and would implement the constitutional guarantee that each resident citizen has, as far as is possible, an equal voice in the selection of public officials. 43 Issacharoff's three enumerated goals are arguably admirable in the context of ensuring that each person, in the abstract, has equal power to elect a representative and to be represented by him or her. 44 Nonetheless, the Court has sometimes awkwardly relied upon Reynolds in more complex redistricting contexts and has at other times refused to act because it did not think that courts could achieve these three goals. 45 The following sections address these more complex redistricting contexts: minority vote dilution, racial gerrymandering, and partisan gerrymandering.

## B. MINORITY VOTE DILUTION AND RACIAL GERRYMANDERING

*p. 11*
"To be effective, a voter's ballot must stand a meaningful chance of effective aggregation with those of like-minded voters to claim a just share of electoral 42. See, e.g., Karcher v. Daggett, 462 U.S. 725, 730 (1983). "[A]s most political scientists recognize, population equality guarantees almost no form of fairness beyond the numerical equality of population." Andrew Gelman & Gary King, Enhancing Democracy Through Legislative Districting, 88 AM. POL. SCI. REV. 541, 553 (1994).

*p. 11*
43. Issacharoff, supra note 40, at 1648 (citations and internal quotation marks omitted). 44. See id. at 1669 ("One can understand the role of the one-person, one-vote rule as that of an externally imposed constraint prompted by the failure of legislative bodies to bind themselves to a meaningful precommitment strategy for apportionment. Given the magnitude of the temptation of incumbents to preserve their power bases, and given the manipulability of majoritarian processes, the political system was unable to generate an internal check on partisan distortions of the electoral process-with the most extreme case being simply the refusal to reapportion at all when reapportionment threatened vested power bases. The Supreme Court provided the functional equivalent of a first-order precommitment strategy by creating a presumption of unconstitutionality every ten years should the legislature fail to reapportion after the decennial census, and by forcing the legislatures to channel their reapportionment procedures through the threshold requirements of the equipopulation principle."). But see id. at 1650 ("[T]he instrumental aspect of the equipopulation rule lost its vitality as the one-person, one-vote rule became increasingly reified as the functional definition of what it meant for an electoral process to be politically fair."); McConnell, supra note 32, at 103 ("In order to bring legislative districts as close to 'precise mathematical equality' as possible, states must disregard preexisting political boundaries such as cities, townships, and counties . . . . Once freed from these traditional constraints . . . , legislative line-drawers were able to draw maps to produce the results they desired, rendering elections less a reflection of popular opinion than of legislative craftsmanship.").

*p. 11*
2005] 1557 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING results." 46 There are two popular strategies that a redistricter might employ to diminish or accentuate the voting power of a particular group of individuals, including minority groups and political parties. One strategy "cracks" or fragments a group's votes by dispersing the group over multiple districts so that it constitutes a minority group in each one and therefore cannot effectively consolidate its power to elect its candidate of choice. 47 Another "packs" together or overconcentrates a supermajority of the members of a group into as few districts as possible so that it will waste votes that might have been effectively spent in another district. 48 Although the Supreme Court held in Whitcomb v. Chavis 49 whether minority groups may be cracked or packed in districts in a way that undermines their political power to be a political question, 50 the Court declared this issue to be justiciable just two years later in White v. Regester. 51 White involved an apportionment scheme in Texas with some multimember districts, that is, larger than normal districts that elected multiple members of the legislature. 52 African-American and Mexican-American groups challenged these districts as "being used invidiously to cancel out or minimize the voting strength of racial groups" by effectively diluting their power through consolidation of multiple individualmember districts into a multimember one, in violation of the Equal Protection Clause. 53 The White Court ruled in their favor, 54 showing the Court's "willingness to evaluate the outputs of the election process as measured by the candidates elected and the policies developed" for "traditionally excluded minorities." 55 The Supreme Court subsequently pronounced that it would measure the fairness of legislative apportionment schemes to minorities under the Constitution against an (unspecified) ideal standard of apportionment. 56 The Court has not needed to rely only on the U.S. Constitution to determine whether legislative districting schemes undermine the voting rights of racial minorities; it also has the power to adjudicate claims brought under the Voting Rights Act (VRA). 57

*p. 12*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL Fifteenth Amendment's pledge to eradicate slavery by ensuring that African-Americans could exercise their right to vote-and thus their freedom-by, among other measures, suspending literacy tests and other devices often used to prevent the franchisement of African-Americans. 59 Further, Congress, under section 5 of the VRA, subjected any change in voting procedures, including redistricting, to a determination that the scheme "does not have the purpose and will not have the effect of denying or abridging the right to vote on account of race or color." 60 This determination would be made through preclearance by either the Department of Justice or a federal court. 61 The Supreme Court has held, in the context of redistricting, that only plans that aggravate "the position of racial minorities with respect to their effective exercise of the electoral franchise can . . . have the 'effect' of diluting or abridging the right to vote on account of race within § 5." 62 As to intent, the Court has held that section 5 forbids a retrogressive intent, rather than a more generalized intent to discriminate. 63 In 2003, the Court elaborated, in Georgia v. Ashcroft, 64 on the factors that influence the aggravation (or enhancement) of "the position in racial minorities with respect to their effective exercise of the electoral franchise." 65 At issue in Ashcroft was whether a redistricting plan formulated in response to the 2000 U.S. Census results creating thirteen districts with a majority African-American voting age population, thirteen additional districts with an African-American voting age population of between 30% and 50%, and four other districts with an African-American voting age population of between 25% and 30% violated section 5 merely because the previous benchmark apportionment scheme contained eleven districts with a majority African-American population, ten additional districts had a majority African-American voting age population, and eight districts with an African-American voting age population between 30% and 50%. 66 Interestingly enough, all but two of the state's African-American legislators voted in favor of the new redistricting plan. 67 In its attempt to obtain 58. Voting Rights Act of 1965, Pub. L. No. 89-110, 79 Stat. 437 (1965) S. 130, 141 (1976). Nonetheless, section 5 does not permit jurisdictions to rely on it as authority to maximize the number of majority-minority districts. See Miller v. Johnson, 515 U.S. 900, 925 (1995). 63. See Reno v. Bossier Parish School Bd., 528 U.S. 320, 328 (1999). 64. 539 U.S. 461 (2003). 65. Id. at 477 (quoting Miller v. Johnson, 515 U.S. 900, 926 (1995) preclearance, Georgia argued that the new plan was designed in part to increase African-American voting strength by giving them a combination of majorityminority districts and "influence" districts, 68 in which they could have at least a significant impact on election results. 69 The Ashcroft majority agreed with Georgia's approach and held that "any assessment of the retrogression of a minority group's effective exercise of the electoral franchise depends on an examination of all the relevant circumstances, such as the ability of minority voters to elect their candidate of choice, the extent of the minority group's opportunity to participate in the political process, and the feasibility of creating a nonretrogressive plan." 70 It thought that a state could pass muster under section 5 even if it decreased the number of majority-minority districts for their redistricting scheme:

*p. 14*
In order to maximize the electoral success of a minority group, a State may choose to create a certain number of "safe" districts, in which it is highly likely that minority voters will be able to elect the candidate of their choice. Alternatively, a State may choose to create a greater number of districts in which it is likely-although perhaps not quite as likely as under the benchmark plan-that minority voters will be able to elect candidates of their choice. 71 In addition, the Court emphasized that the changes in a minority group's ability to participate in the political process-such as the changes in the number of "influence" districts, the influence and power of representatives of majorityminority districts, and the support of representatives of the majority-minority districts for the new plan-are also relevant to a section 5 inquiry. 72 The Court thus announced its willingness to wade deeper into the political thicket than it had previously done by looking not just at the composition of districts but also at the internal workings of a legislature to determine whether minority groups had lost electoral power from one districting cycle to the next. The measurement of fairness in section 5 retrogression cases has therefore become messy and unclear, requiring comparisons not just between apples and oranges, but also between the outside of one apple and the inner core of another. 73 The acceptability of such open-ended comparisons makes it easy for a jurisdiction to satisfy its burden of proving nonretrogression, which effectively eviscerates any probing analysis of fairness in redistricting under section 5, leaving the assess-68. A majority-minority district is one in which a minority group constitutes a majority of the individuals in a district. An influence district is one in which a minority group does not constitute the majority but rather can have significant influence in elections.

*p. 14*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL ment of such fairness primarily in the hands of the legislators. 74 In summary, the Court has decided to measure, pursuant to section 5, the fairness to minorities of a new legislative districting scheme using the previous legislative apportionment as a baseline; a fair redistricting scheme is merely one that is no worse-whatever that means-than the previous one.

*p. 15*
The VRA and the Supreme Court jurisprudence described thus far serve to effectuate the Fifteenth Amendment. In 1982, Congress awakened the mostly dormant section 2 of the VRA by amending it to prohibit, among other things, any practice that results in certain minority groups "hav[ing] less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice." 75 Congress made this amendment in response to a Supreme Court ruling that a plaintiff needed to show both discriminatory intent and effect to show that minority voting power had been diluted under the Fifteenth Amendment. 76 The amendment thus makes section 2 broader in its scope of prohibiting voting schemes than the Fifteenth Amendment itself.

*p. 15*
Because section 2 does not contain as ready a baseline as section 5 does, the Supreme Court has had to fashion one. In Thornburg v. Gingles, 77 the Court measured whether a redistricting scheme absolutely dilutes minority voting power in violation of section 2 by measuring both whether the minority group is sufficiently large that it could compose a majority in a compact district and whether there is racially polarized voting (that is, both whether the minority group is politically cohesive and whether the white majority votes as a bloc to defeat the minority group's preferred candidate). 78 This standard has been difficult for lower courts to apply, in determining both whether a minority can create a compact district 79 and whether voting is racially polarized. 80 74. Cf. Nathaniel Persily, Forty Years in the Political Thicket: Evaluating Judicial Oversight of Redistricting Since Baker v. Carr, in PARTY LINES: COMPETITION, PARTISANSHIP, AND CONGRESSIONAL REDISTRICTING (Thomas Mann & Bruce E. Cain eds., forthcoming 2006) (manuscript at 22, on file with author) ("Georgia v. Ashcroft defangs Section Five considerably, however, and it is too early to tell whether the effect will be the intended flexibility in dealing with apportionment plans that sacrifice control for genuine influence or rather the unintended effect of looking the other way when state legislatures break up minority districts into substantial but ultimately uninfluential 20 percent or 30 percent districts."). There may, however, be no other choice than that adopted by the Ashcroft majority if courts are required under section 5 to measure retrogression accurately. S. 55, 62-64, 66-69 (1980). 77. 478 U.S. 30 (1986). 78. Id. at 50-51. 79. See, e.g., Dillard v. Baldwin County Bd. of Educ., 686 F. Supp. 1459, 1466 (M.D. Ala. 1988) (determining that "a district is sufficiently geographically compact if it allows for effective representation"). 80. See, e.g., Nixon v. Kent County, 76 F.3d 1381, 1386-87 (6th Cir. 1996) (en banc) (holding that different minority groups cannot band together to bring section 2 actions); League of United Latin Am. Citizens v. Clements, 999 F.2d 831, 850 (5th Cir. 1994) (en banc) (holding that one has to take factors

*p. 15*
1561 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING In 1994, in Johnson v. De Grandy, 81 the Supreme Court widened the inquiry into minority vote dilution by ruling that courts must look to the totality of the circumstances rather than just to the Gingles factors to determine whether there has been a violation of section 2. 82 The Court focused on proportionality, in the sense of "equality of opportunity," rather than proportional representation, as a baseline for measuring vote dilution. 83 Despite this complexity, some commentators praise Gingles and its progeny for providing a practicable measure of fairness by looking to discrete voting events rather than to the functioning of the political process. 84 Nevertheless, the Ashcroft Court's readiness to investigate the functioning of the political process may signal that the federal courts will look there as well in the context of the Court's section 2 jurisprudence-albeit a section with different purposes than section 5.

*p. 16*
Although the Court has mapped out a baseline in the vote-dilution cases summarized above, it has categorically rejected other vote-dilution claims for an inability to locate a baseline against which to measure the fairness of a voting or redistricting scheme. In Holder v. Hall, 85 for example, the Court held that the size of a governing body was not subject to a challenge under section 2 because it could not locate a satisfactory baseline. 86 The Court reasoned that when "there is no objective and workable standard for choosing a reasonable benchmark by which to evaluate a challenged voting practice, it follows that the voting practice cannot be challenged as dilutive under section 2." 87 It thought that there was "no principled reason why one size should be picked over another as the benchmark for comparison." 88 It is uncertain whether this holding is still valid in light of the Ashcroft Court's willingness in the context of section 5 to look at factors relating to legislature composition in analyzing retrogression. 89 In recent years, the Court has begun questioning the VRA's constitutionality in light of the Equal Protection Clause. In Shaw v. Reno, 90 the Court rejected as unconstitutional, unless it satisfied strict scrutiny, a redistricting scheme that other than race, such as partisan affiliation, into account to determine whether those factors explain racially polarized voting). But see Clements, 999 F.2d at 907-09 (King, J., dissenting) (arguing that it is impossible in nonexperimental research to determine why people vote the way they do, so courts should look just at the statistical correlations

*p. 16*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL enhanced African-American voting power-as encouraged by the VRA-by carving out two bizarrely shaped majority-minority districts in North Carolina. 91 The Shaw Court reasoned that:

*p. 17*
[A] plaintiff challenging a reapportionment statute under the Equal Protection Clause may state a claim by alleging that the legislation, though race-neutral on its face, rationally cannot be understood as anything other than an effort to separate voters into different districts on the basis of race, and that the separation lacks sufficient justification. 92 In light of Shaw, it became questionable whether any racial gerrymanderingthat is, districting to favor racial minorities-would be upheld as constitutional. However, the current Court has upheld a redistricting scheme in which race may have been used as one factor among many in redistricting. 93 In so doing, Justices Breyer, Ginsburg, Souter, and Stevens endorsed a vision of racial plurality. They have consistently reasoned that racial minorities are essentially political groups and are thus deserving of political power. 94 Justice O'Connor joined these Justices but approved, by contrast, of racial opacity, in that "so long as [States] do not subordinate traditional districting criteria to the use of race for its own sake or as a proxy, States may intentionally create majority-minority districts, and may otherwise take race into consideration, without coming under strict scrutiny." 95

## C. PARTISAN GERRYMANDERING

*p. 17*
The Court has also adjudicated the fairness of partisan gerrymanderingthat is, whether redistricting in favor of or to the disadvantage of particular political parties can be constitutionally unfair and improper. 96 In Gaffney v.

*p. 17*
91. See id. at 644-49. 92. Id. at 649. 93. See Easley v. Cromartie, 532 U.S. 234, 243-58 (2001); see also Bush v. Vera, 517 U.S. 952, 990-95 (1996) (O'Connor, J., concurring).

*p. 17*
94. See Shaw v. Hunt, 517 U.S. 899, 918 (1996) (Stevens, J., dissenting) ("A majority's attempt to enable the minority to participate more effectively in the process of democratic government should not be viewed with the same hostility that is appropriate for oppressive and exclusionary abuses of political power."); Bush, 517 U.S. at 1053 (Souter, J., dissenting) ("[M]alapportionment measurably reduces the influence of voters in more populous districts, and vote dilution predestines members of a racial minority to perpetual frustration as political losers . . . .").

*p. 17*
95. Bush, 517 U.S. at 993 (O'Connor, J., concurring). 96. Scholars have set forth various theories why partisan gerrymandering is harmful to the populace in the first instance. With one exception, these theories-even though they use different terminologyare grounded in notions of political fairness and electoral competitiveness. See, e.g., Hirsch, supra note 47, at 204 ("Districting plans that combine extreme partisan bias with record levels of incumbency protection undermine political competition not only at the level of individual districts, where elections become foregone conclusions, but also at the institutional level, by effectively barring one political party from taking control of the legislative body even if it repeatedly garners a majority of the popular vote."); Issacharoff, supra note 2, at 600 ("[T]he harm in gerrymandering . . . . is the insult to the competitiveness of the process resulting from the ability of insiders to lessen competitive pressures.");

*p. 17*
1563 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING Cummings, 97 the Court held that partisan redistricting is not exempt from judicial scrutiny for unconstitutionality, although it is "plausible to assume that those who redistrict and reapportion work with both political and census data." 98 The Court held that recognition of political strength is an acceptable and expected part of the redistricting process:

*p. 18*
[N]either we nor the district courts have a constitutional warrant to invalidate a state plan, otherwise within tolerable population limits, because it undertakes, not to minimize or eliminate the political strength of any group or party, but to recognize it and, through districting, provide a rough sort of proportional representation in the legislative halls of the State. 99 Moreover, Gaffney accepted as politically fair district lines "drawn with the conscious intent to create a districting plan that would achieve a rough approximation of the statewide political strengths of the Democratic and Republican Parties, the only two parties in the State large enough to elect legislators from discernible geographic areas." 100 A decade later, in Karcher v. Daggett, 101 Justice Stevens, in a concurring opinion, spoke out in favor of disallowing as unconstitutional certain partisan gerrymanders. 102 Justice Stevens proposed that a politically salient and cohesive group (be it a political party, a racial minority, or other group) deserves a roughly proportional share of power relative to its voting strength, and when there are deviations from population equality or other objective standards there should be a rebuttable presumption of discrimination against the group. 103 Three years later, in Davis v. Bandemer, 104 a plurality of four Justices adopted Justice Stevens's conclusion that partisan gerrymandering might be unconstitutional under the Equal Protection Clause, albeit applying a different Samuel Issacharoff & Richard H. Pildes, Politics as Markets: Partisan Lockups of the Democratic Process, 50 STAN. L. REV. 643, 681 (1998) (maintaining that when a political group intentionally "locks up" the electoral results by rigging districting boundaries in its favor, there is a political monopoly and thus no "optimal partisan political competition"); Michael J. Klarman, Majoritarian Judicial Review: The Entrenchment Problem, 85 GEO. L.J. 491, 509 (1997) (opining that partisan gerrymandering "enables a temporarily preponderant political party to extend its hegemony beyond subsequent elections that should have returned it to minority status"); A New Map, supra note 20, at 1198 (arguing that partisan gerrymandering is a "federalism injury that state legislatures inflict when they interfere with the ability of the 'People of the several states' to elect their national representatives"). The Note's attempt to reconceptualize the harm caused by partisan gerrymandering as a federalism injury at best makes sense when state legislatures are mapping districts for their congressional representatives, but not when they are drawing district boundaries for themselves.

*p. 18*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL approach. In Bandemer, following the 1980 U.S. census, the Republicancontrolled Indiana legislature redistricted, which outraged the Democratic Party because of the skewed results of the post-redistricting elections. 105 For example, in a 1982 election, the Republican Party received 48.1% of the popular vote but won 57% of the legislative seats. 106 A majority of the Court held that Indiana's redistricting scheme was not an unconstitutional partisan gerrymander; 107 the plurality reasoned that there was no "threshold showing of discriminatory vote dilution" to make out a prima facie case of an equal protection violation. 108 Although no majority of the Court established a standard to evaluate the constitutionality of a partisan gerrymander, the plurality did establish that a partisan gerrymander is unconstitutional when there is "[a] claim . . . that each political group in a State should have the same chance to elect representatives of its choice as any other political group" that satisfies two conditions. 109 First, according to the plurality, plaintiffs must demonstrate intentional efforts by a political party to undermine the strength of another party. 110 Second, plaintiffs must show consistent degradation of the political process as a whole; the results of a single election are insufficient to prove such degradation. 111 Three Justices concurred in the judgment, arguing that partisan gerrymandering claims should be held to be nonjusticiable political questions because "[t]he opportunity to control the drawing of electoral boundaries through the legislative process of apportionment is a critical and traditional part of politics in the United States, and one that plays no small role in fostering active participation in the political parties at every level." 112 Justices Powell and Stevens thought that partisan gerrymandering claims were actionable but would look to a broader range of factors than the plurality would. 113 This confusing result culminated in an equally complicated ruling on the issue eighteen years later in Vieth v. Jubelirer. 114 In Vieth, the Pennsylvania Republican Party, which controlled a majority of both state houses and the governor's office, redrew the districting boundaries for the Commonwealth of Pennsylvania after the 2000 census with intent to hurt the Democratic Party. 115 The new districting plan was challenged as, among other offenses, unconstitutional under the Equal Protection Clause for its partisan intent and effect. 116 Five Justices, a plurality of four and Justice Kennedy, held that the plaintiffs could not establish the alleged constitutional violation. 117 Justice Scalia, speaking for the plurality, thought that such claims were nonjusticiable because no standard had been found in the years since Bandemer was decided by which to measure the constitutionality of a partisan gerrymander. 118 The plurality also rejected the proposed measures put forth by the plaintiffs and by the dissenters as vague and requiring courts to make political judgments, a task for which they thought the judicial branch ill-suited. 119 Most relevantly, the plurality reasoned: 117. See id. at 305-06; id. at 307-08 (Kennedy, J., concurring). 118. See id. at 306 (plurality opinion) ("Eighteen years of essentially pointless litigation have persuaded us that Bandemer is incapable of principled application. We would therefore overrule that case, and decline to adjudicate these political gerrymandering claims."). The plurality thought it "significant that the Framers provided a remedy for such practices in the Constitution. Article 1, § 4, while leaving in state legislatures the initial power to draw districts for federal elections, permitted Congress to 'make or alter' those districts if it wished." Id. at 275-76.

*p. 20*
119. The plaintiffs had proposed to measure an unconstitutional partisan gerrymander by showing (a) "that the mapmakers acted with a predominant intent to achieve partisan advantage, which can be shown by direct evidence or by circumstantial evidence that other neutral and legitimate redistricting criteria were subordinated to the goal of achieving partisan advantage," id. at 284 (internal quotation marks omitted), and (b) a "requisite effect when (1) the plaintiffs show that the districts systematically 'pack' and 'crack' the rival party's voters, and (2) the court's examination of the 'totality of circumstances' confirms that the map can thwart the plaintiffs' ability to translate a majority of votes into a majority of seats," id. at 286-87 (some internal quotation marks omitted). The Court rejected this test, which it thought was borrowed from the Court's jurisprudence on racial gerrymandering, because "a person's politics is rarely as readily discernible-and never as permanently discernible-as a person's race. Political affiliation is not an immutable characteristic, but may shift from one election to the next; and even within a given election, not all voters follow the party line." Id. at 287. Relatedly, Michael McConnell suggests that racial and partisan gerrymandering are different beasts, as the former is designed to increase minority voting strength so that it better approaches what it would be in a system of proportional representation and the latter has the purpose of entrenching the power of certain political factions. See McConnell, supra note 32, at 116. The Court also rejected its test as relying on proportional representation, something the U.S. Constitution does not guarantee. See Vieth, 541 U.S. at 288.

*p. 20*
Many scholars have long agreed with the plurality's sentiment that partisan gerrymandering is essentially a political question best undisturbed by courts. See, e.g., Peter H. Schuck, The Thickest Thicket: Partisan Gerrymandering and Judicial Regulation of Politics, 87 COLUM. L. REV. 1325, 1330 (1987) ("[T]he Court should have held partisan gerrymandering claims to be nonjusticiable. When the Court considers whether and how to regulate politics, it should resist reforms requiring blunt remedial tools that threaten to transform, rather than refine, the American political system."); cf. John Hart Ely, Gerrymanders: The Good, the Bad, and the Ugly, 50 STAN. L. REV. 607, 616-18 (1998) (agreeing with Justice O'Connor's Bandemer concurrence and highlighting approvingly her limitation of the reasoning of her concurrence to major political parties); Issacharoff, supra note 40, at 1660 ("[T]he Court introduced a specific partisan gerrymandering cause of action that contemplates judging the fairness of the electoral process by evaluating its outcome, thereby giving constitutional force to the manner in which federal judges posit the results of a properly functional electoral process. By taking this step beyond ordering the electoral procedures to guarantee access, the Court treads dangerously close to the divide between the justiciable and the truly political."). And some similarly maintain that courts should judge partisan gerrymandering cases only if a manageable principle can be found to guide the courts. See, e.g., Issacharoff, supra note 40, at 1664 ("The key question . . . must be whether the claim of partisan manipulation of the redistricting process is amenable to a limiting and readily manageable principle, as found in the one-person, one-vote rule of Reynolds. Absent such a guiding principle, the risk of a judicial free-fall into the realm of naked politics . . . must be seriously reassessed.").

*p. 20*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL "Fairness" does not seem to us a judicially manageable standard. Fairness is compatible with noncontiguous districts, it is compatible with districts that straddle political subdivisions, and it is compatible with a party's not winning the number of seats that mirrors the proportion of its vote. Some criterion more solid and more demonstrably met than that seems to us necessary to enable the state legislatures to discern the limits of their districting discretion, to meaningfully constrain the discretion of the courts, and to win public acceptance for the courts' intrusion into a process that is the very foundation of democratic decisionmaking. 120 Justice Kennedy concluded that, while there was currently no standard by which to measure whether a partisan gerrymander is unconstitutional, there may be one put forth in time and that he would therefore not yet go so far as to hold such claims to be nonjusticiable. 121 The four dissenters each thought that the plaintiffs should at least have their day in court to establish a constitutional violation under the so-called "pageant of [substantive] standards" 122 put forth by their respective dissents. 123 120. Vieth, 541 U.S. at 291 (plurality opinion). 121. See id. at 306-13 (Kennedy, J., concurring). Justice Kennedy reasoned that:

*p. 21*
[C]ourts confront two obstacles. First is the lack of comprehensive and neutral principles for drawing electoral boundaries. No substantive definition of fairness in districting seems to command general assent. Second is the absence of rules to limit and confine judicial intervention. With uncertain limits, intervening courts-even when proceeding with best intentions-would risk assuming political, not legal, responsibility for a process that often produces ill will and distrust.

*p. 21*
Id. at 306-07. He hoped that "in another case a standard might emerge that suitably demonstrates how an apportionment's de facto incorporation of partisan classifications burdens rights of fair and effective representation (and so establishes the classification is unrelated to the aims of apportionment and thus is used in an impermissible fashion)." Id. at 312. He thought that such a standard might emerge from discussions by parliamentary or legislative bodies of fair redistricting principles, id. at 308, or improved computer technology, id. at 312-13. Justice Kennedy also wondered whether claims of partisan gerrymandering should be evaluated under the First Amendment rather than under the Equal Protection Clause. See id. at 315.

*p. 21*
One scholar thinks that Justice Kennedy's hope that a standard for measuring impermissible partisan gerrymandering would emerge from legislative discussions, technology development, or under the rubric of the First Amendment is unwarranted. See Richard L. Hasen, Looking for Standards (in All the Wrong Places): Partisan Gerrymandering Claims After Vieth, 3 ELECTION L. J. 626 (2004). As to legislative discussions on redistricting, he reasons that they are at best unhelpful and are at worst full of examples of drawing district boundaries for partisan advantage. See id. at 634. With respect to computer technology, Hasen thinks that it is useless without first specifying the harms and quanta of harms that one is looking to prevent. See id. at 635. And as to existing First Amendment jurisprudence, notably in the Supreme Court's series of patronage cases, Hasen is convinced that the jurisprudence does not translate properly in the context of partisan gerrymandering. See id. Hasen suggests instead that the Supreme Court should not adjudicate partisan gerrymandering claims "until a societal near-consensus emerges regarding the permissible range of using voters' party affiliation in redistricting." Id. at 628. Until then, he thinks that redistricting done for partisan reasons should be left to the mercy of the political process. See id.

*p. 21*
122. Hasen, supra note 121, at 630. 123. Justice Stevens thought that "when partisanship is the legislature's sole motivation-when any pretense of neutrality is forsaken unabashedly and all traditional districting criteria are subverted for

## 2005] 1567 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING D. THE SUPREME COURT'S MEASURES OF FAIRNESS AND BASELINES

*p. 22*
As detailed in the previous three sections, the Supreme Court has moved from staying out of the democratic process to seeking to preserve fairness to individuals and groups in the democratic process, although its frustration with the lack of judicially appropriate baselines may be driving the Court back out of the "political thicket." The Court has not yet fully defined what it means by fairness, but it has come closest in assessing the baselines against which to judge whether plaintiffs seeking a change in the legislative districting process have been wronged.

*p. 22*
In the one-person, one-vote context, the Court has established a simple partisan advantage-the governing body cannot be said to have acted impartially." Vieth, 541 U.S. at 318 (Stevens, J., dissenting). He would draw on elements of the Court's racial gerrymandering analysis to adjudge the constitutionality of an alleged partisan gerrymander:

*p. 22*
Gerrymandering always involves the drawing of district boundaries to maximize the voting strength of the dominant political faction and to minimize the strength of one or more groups of opponents. In seeking the desired result, legislators necessarily make judgments about the probability that the members of identifiable groups-whether economic, religious, ethnic, or racial-will vote in a certain way. The overriding purpose of those predictions is political. It follows that the standards that enable courts to identify and redress a racial gerrymander could also perform the same function for other species of gerrymanders.

*p. 22*
Id. at 335-36. The plurality thought his standard was ill-advised on the ground that "setting out to segregate voters by race is unlawful and hence rare, and setting out to segregate them by political affiliation is (so long as one doesn't go too far) lawful and hence ordinary." Id. at 293 (plurality opinion). Justices Souter and Ginsburg thought that "[i]nstead of coming up with a verbal formula for too much, then, the Court's job must be to identify clues, as objective as we can make them, indicating that partisan competition has reached an extremity of unfairness." Id. at 334 (Souter, J., dissenting). They wanted to "make a fresh start," id. at 335, and proposed "calling for a plaintiff to satisfy elements of a prima facie cause of action, at which point the State would have the opportunity not only to rebut the evidence supporting the plaintiff's case, but to offer an affirmative justification for the districting choices, even assuming the proof of the plaintiff's allegations," id. at 346. To make out a prima facie case, they suggested examining (1) whether plaintiff belongs to a cohesive political group, (2) whether the plaintiffs' district did not pay enough attention to traditional districting principles, (3) whether there are specific correlations between the deviation from these principles and the population distribution of the group, (4) whether there was a hypothetical district which would accomplish the plaintiff's goals but deviated less from traditional principles than the actual district, and (5) whether the defendants intentionally acted to manipulate the shape of the district to diminish the power of the political group. See id. at 347-51. They saw no need to develop "a full-blown theory of fairness" to fashion a remedy for this sort of vote dilution. See id. at 354. The plurality rejected their proposed test on the basis that parts of it "require[] a quantifying judgment that is unguided and ill suited to the development of judicial standards." Id. at 295 (plurality opinion).

*p. 22*
Justice Breyer, in his dissent, thought that there was "at least one circumstance where use of purely political boundary-drawing factors can amount to a serious, and remediable, abuse, namely, the unjustified use of political factors to entrench a minority in power." Id. at 360 (Breyer, J., dissenting). He thought that:

*p. 22*
[w]here a State has improperly gerrymandered legislative or congressional districts to the majority's disadvantage, the majority should be able to elect officials in statewide racesparticularly the Governor-who may help to undo the harm that districting has caused the majority's party, in the next round of districting if not sooner.

*p. 22*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL rule-rough numerical equality-to define fairness. This measure of fairness does not account for electoral outcomes or the redistricting process itself; the sole concern is the numbers.

*p. 23*
In the minority vote dilution context, to measure fairness, the Court looks to the VRA section 5 baseline of minority voting power and electoral representation under a previous legislative districting scheme. Under the Constitution or VRA section 2, it looks to an ideal baseline that it has struggled to define in cases like Gingles and De Grandy. Yet the current Court thinks it is unfair to redistrict by relying too much on racial factors, as evidenced by the Shaw line of cases. In the context of minority vote dilution and racial gerrymandering, the Court is concerned both with how previous electoral outcomes will reflect on future electoral outcomes and with which people are in which districts; the Court examines the redistricting process only to the extent it reveals discriminatory purpose.

*p. 23*
As for partisan gerrymandering claims, it seems that until a majority of the Supreme Court is satisfied that there is an adequate standard of fairness by which to measure the constitutionality of partisan gerrymanders, the Court will not entertain such constitutional challenges. Even before Vieth, the Court had a very strict threshold for what constitutes fairness-as long as there was no intentional discrimination or consistent degradation of a political party's power, political parties could throw about their weight in redistricting schemes. 124 These differing measures of fairness and motivation make it difficult for redistricters to create a single redistricting scheme that satisfies them all. 125 To address this problem, a simpler, unified, and constructive measure of fairness that doubles as a useful descriptive measure for the courts is introduced in Part IV 126 after a description and analysis in the next Part of the strengths and shortcomings of previous proposals for fairness in redistricting.

## III. PREVIOUS PROPOSALS FOR FAIRNESS IN REDISTRICTING

*p. 23*
Previous proposals for fairness in redistricting tend to fall into two main categories: constructive models and descriptive models. Constructive models provide a method for constructing a fair (or the most fair) legislative redistricting scheme, while descriptive models provide a standard to check whether a 124. The Court's fairness standard under Bandemer looked to both the composition of districts and a pattern of future electoral outcomes, and the Court examined the redistricting process only to the extent that it revealed intentional discrimination. Samuel Issacharoff argues that "[t]he closer the Court comes to assessing pure fairness matters in the outcomes of elections, the closer the judiciary is brought to a straightforward declaration of its political preferences." Issacharoff, supra note 40, at 1677. 125. Cf. Issacharoff, supra note 2, at 630-31 ("One of the perverse consequences of the absence of any real constitutional vigilance over partisan gerrymandering is that litigants must squeeze all claims of improper manipulation of redistricting into the suffocating category of race.").

*p. 23*
126. But cf. Persily, supra note 74 (manuscript at 2) (asserting that it is a mistake to "inappropriately group[] a wide array of cases under the heading of redistricting and thereby obscur[e] the important differences between litigation under the Voting Rights Act, different provisions of the U.S. Constitution, and state laws").

*p. 23*
1569 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING particular legislative redistricting scheme is fair. 127 This Part details and critiques previous proposals for each in turn.

## Computer Algorithms

*p. 24*
Since the advent of computing technology, many scholars have turned to the computer as the arbiter of fairness in redistricting. To these scholars, a computer software program would examine details about a jurisdiction's population and would rely on a set of algorithms to draw districting lines for the jurisdiction, and they cite a number of advantages of automated redistricting for both courts and legislatures. First, legislatures would avoid political deadlock by limiting their discretion, 128 and courts would never have to enter the "political thicket" by developing redistricting plans. 129 Second, it avoids human discretion, as compared with redistricting by legislators or other interested parties, meaning, the argument goes, legislators cannot easily manipulate redistricting plans. 130 Third, one can encode desirable social goals into these algorithms to produce socially beneficial districts. 131 Fourth, to automate redistricting, one needs to encode principles; thus, the automation of redistricting encourages political debate over redistricting goals, rather than over particular redistricting configura-127. These categories are not as distinct as they may at first seem. With some additional information, any constructive model can be converted into a descriptive model and any descriptive model can be transformed into a constructive model. To convert a constructive model to a descriptive model, one merely needs to construct a fair districting scheme (or a set of fair districting schemes) using the constructive model and then compare the resulting districting scheme(s) with any other districting scheme to determine if they are a match. To transform a descriptive model to a constructive model, one must generate every possible redistricting scheme, testing each one for fairness under the descriptive model. One can either consider any of the redistricting schemes in the set labeled "fair" to be constructively fair or determine which of the schemes in the fair set is the most fair via an additional test.

*p. 24*
There are numerous other plausible and useful categorizations of proposed redistricting models. For instance, Adam Cox divides the measures into those that are process-based, those that are outcomebased, and those that are institution-selecting. See Cox, supra note 4, at 756. Nearly all of the measures that Cox places in one of the preceding three categories, other than that of a temporal floor on redistricting, fall within the category of what this Article labels as constructive or descriptive substantive models. The classification of models as constructive or descriptive in this Article is particularly helpful in the context in which it occurs, namely a discussion about how to construct fair districting schemes and how to measure whether a particular districting scheme is fair.

*p. 24*
128. Cf. id. at 760-61 (reasoning that the current redistricting rules "change the effect of legislative deadlock by authorizing courts to refashion district lines when the political branches fail to do so").

*p. 24*
129. See James B. Weaver & Sidney W. Hess, A Procedure for Nonpartisan Districting: Development of Computer Techniques, 72 YALE L.J. 288, 290 (1963) ("To avoid th[e] 'political thicket,' a court may desire to limit its own discretion in creating new legislative districts. One means of accomplishing this end could be to adopt a mechanical formula which makes the actual drafting of district lines non-discretionary once general principles of representation have been determined." (citations omitted) (quoting Colegrove v. Green, 328 U.S. 549, 556 (1946) 132 Fifth, courts, legislatures, and the public can explicitly review the goals of the automated process by reviewing its encoding. 133 However, the neutrality and lack of bias promised by automated redistricting programs is a mere illusion: "[R]edistricting standards and processes embody political values and . . . automation of this process hides the fundamental conflict over values." 134 Legislators must still select the goals of an automated redistricting program and encode them, at which point any claim to neutrality disintegrates. 135 The best that can be said of automated redistricting is that "by virtue of the explicitness required in computer programming, the computer can be used as a tool to make previously articulated theories of representation practicable." 136 Another problem in reality with automated redistricting proposals is that the transparency promised by automated redistricting is at best an opacity: the technical algorithms and descriptions of the automated processes tend to confuse and obscure underlying political goals. 137 Sneaky (or perhaps clever) legislators "could attempt to manipulate automated redistricting by selecting 132. See id. at 84-85; Michelle H. Browdy, Note, Computer Models and Post-Bandemer Redistricting, 99 YALE L.J. 1379, 1381 (1990).

*p. 25*
133. See Altman, supra note 130, at 85 & n.16; Issacharoff, supra note 40, at 1697 ("[T]he mandatory use of computer technology can force the manifold political considerations behind any given redistricting plan into the open where the individual steps in ordering the political process may be subject to both the sanitizing effect of public scrutiny and the possibility of meaningful judicial review."); Browdy, supra note 132, at 1381, 1385 n.36 ("Previously, only the resulting map and perhaps the legislative process used to approve the map would be available for judicial review. With automation, then, rather than having to glean the subjective intention of legislators by analyzing redistricting maps, the Court will be able to analyze the objective 'intent' of the redistricting procedure as preserved in the record of the computer program." (citation omitted)).

*p. 25*
134. Altman, supra note 130, at 86; see also Robert G. Dixon, Jr., Fair Criteria and Procedures for Establishing Legislative Districts, in REPRESENTATION AND REDISTRICTING ISSUES 7, 7-8 (Bernard Grofman et al. eds., 1982); Issacharoff, supra note 40, at 1696 ("The attempt to remove political considerations from computer-assisted redistricting was stillborn.").

*p. 25*
135. See Altman, supra note 130, at 85 & n.17; see also Gordon E. Baker, Gerrymandering: Privileged Sanctuary or Next Judicial Target?, in REAPPORTIONMENT IN THE 1970S 121, 139 (Nelson W. Polsby ed., 1971) ("While the computer is neutral, those who program it are not."); Robert G. Dixon, Jr., The Court, the People, and "One Man, One Vote", in REAPPORTIONMENT IN THE 1970S, supra, at 8, 30. "Since there is no general consensus over what objectively neutral goals are, or whether they exist at all, no amount of automation can make the redistricting process objectively neutral." Altman, supra note 130, at 86; see also Brian Hayes, Machine Politics, 84 AM. SCIENTIST 522, 522 (1996) ("This superabundance of solutions is an invitation to political (or meta-political) mischief."); Browdy, supra note 132, at 1384 ("[M]ere advocacy of the use of objective standards accomplishes nothing in the abstract. What is needed is a proposed set of objective standards, or at least a mechanism which ensures the development, implementation, and review of a workable set of standards.").

*p. 25*
136. Browdy, supra note 132, at 1385. 137. See Altman, supra note 130, at 134. Nonetheless, it is theoretically possible to devise relatively transparent and comprehensible redistricting automata. See infra Part IV.B-C (proposing some to implement the presented procedural model); cf. Christopher Avery et al., The Market for Federal Judicial Law Clerks, 68 U. CHI. L. REV. 793, 868-82 (2001) (describing the well-established and easy-to-understand medical-residency matching program in the context of pondering the best system for matching federal judges with law school graduates seeking a clerkship with such judges).

*p. 25*
1571 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING weights which generate politically desirable results." 138 Related are the concerns that a rigid automated redistricting process will not capture the goals valued by the public 139 and that it is difficult to encode many political values computationally. 140 A third disadvantage of automated redistricting is that proponents would typically like to carve into silicon stone an algorithm for all future redistricting occasions. 141 Many politicians, individuals, and groups would reasonably prefer to battle under the actual conditions of each redistricting occasion based on the political goals most important to them at the time. 142 All in all, automated redistricting programs are only as fair as the goals that underlie them. 143 Therefore, in and of itself, redistricting software does not provide any solution to the fairness problem. The only additional fairness that an automated process brings is the possibility of public records explaining each algorithmic choice. 144 Given the difficulty that courts and the general public have in understanding mathematical and computational technicalities, this measure of fairness generally provides little comfort indeed. 145 138. Browdy, supra note 132, at 1387; see also Issacharoff, supra note 40, at 1702. However, Michelle Browdy notes that under such a scheme, "legislators must discuss the criteria of fair redistricting, and the weights assigned to those criteria, explicitly. Thus, they are to some degree accountable to the voters and the courts in their decisions." Browdy, supra note 132, at 1387. There is still a strong concern, however, that sneaky legislators with computers will make gerrymanders "rather more sophisticated, since those in charge of designing constituencies can draw on a wealth of data that would have been unavailable or unmanageable before the electronic age." Baker, supra note 135, at 139.

*p. 26*
139. See Altman, supra note 130, at 82 ("[T]o use automated redistricting, a function must be written which is rigid enough for computer processing but subtle enough to meaningfully capture the social worth of districts.").

*p. 26*
140. See Richard L. Morrill, Ideal and Reality in Reapportionment, 63 ANNALS ASS'N AM. GEOGRA-PHERS 463, 470 (1973). Moreover, even the fastest computer may not be able to solve many of the computational problems inherent in automated redistricting. Altman, supra note 130, at 95, 99, 107-08. But see id. at 113 ("If the problem cannot be made tractable through restricting the values or data, a probabilistic method might solve the problem most of the time, or a deterministic algorithm might work well in most cases. Alternatively, a method that quickly finds an approximately-optimal solution may be developed.").

*p. 26*
141. See Browdy, supra note 132, at 1395. This concern dovetails with the issue of how often redistricting should or could occur-decennially or more frequently. See supra note 12.

*p. 26*
142. Another possible drawback for automated redistricting is that no rational legislature would trade away its advantages of redistricting under regular political conditions. See Hayes, supra note 135, at 522. However, this drawback is a problem for any redistricting model that purports to be fair; any political figure with an existing political advantage would rationally prefer not to play by rules of fairness. Ex ante preferences are another story entirely.

*p. 26*
143. As Bert Huang has aptly noted, it is not "computers that kill democracy, it's people with computers that kill democracy." E-mail from Bert Huang to Jeanne Fromer (Aug. 10, 2004) (on file with the author); cf. Toobin, supra note 2 (suggesting that use of computers to redistrict has produced less fair districting schemes than in the days of drawing boundaries on maps with markers).

*p. 26*
144. See Browdy, supra note 132, at 1389. 145. But see Issacharoff, supra note 40, at 1647 ("The goal of this exercise is not to take the politics out of reapportionment (an undoubtedly doomed enterprise) but to define a deliberative process that can be both vibrant and reviewable even in the world of clear political preferences.").

*p. 26*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL 2. Nonpartisan Commissions Some scholars think that fairness in redistricting can derive only from apolitical decisionmakers. They suggest forming nonpartisan commissionspossibly consisting of geographers, political scientists, mathematicians, or reasonable people generally-to draw the district lines typically delineated by legislative representatives. 146 Geographers can concern themselves with respecting communities of interest and traditional boundary lines, 147 political scientists can address good governance concerns, 148 mathematicians can ensure that districts are compact, contiguous, and of equal population, 149 and reasonable people can voice their representational concerns with "less opportunity for insider manipulation" of the redistricting process. 150 Twelve states currently entrust some aspect of their redistricting process to a nonpartisan or bipartisan commission. 151 Robert Dixon saliently criticizes the nonpartisan commission: "The idea of a so-called nonpartisan or neutral commission offers no certain path to representational value. Indeed, it is essentially a 'three monkeys' policy: speak no politics, see no politics, hear no politics, and hope that men are angels, after all." 152 Moreover, there is the well-documented problem that even if nonpartisan commission members are the angels that Dixon skeptically invokes, it is very possible-even likely, because the commission is not considering the political ramifications of its choices-that their redistricting plan will have unintended or undesirable political effects. 153 For example, Nathaniel Persily describes how 146. See Ely, supra note 119, at 634-35 ("Having a 'nonpartisan' commission rather than the legislature draw the district lines is a frequent suggestion, one that the Court's pattern of certiorari denials suggests may provide a safe harbor from federal court intervention. And indeed, when employed, such commissions do seem to represent something of an improvement insofar as they appear to be understandably less prone to protect the seats of incumbents than legislatures are."); BRUCE E. CAIN, THE REAPPORTIONMENT PUZZLE 2-3 (1984) he once helped draw a redistricting plan that unknowingly moved an uninhabited swamp from the district of one legislator, who had initiated environmental projects relating to the swamp, to that of another; the legislator who was to lose the swamp argued that the change would disrupt his projects. 154 It therefore seems preferable that a plan with political effect be adopted with some prescience rather than without.

## Bipartisan Commissions

*p. 28*
Numerous scholars acknowledge that every redistricting plan has some partisan intent or bias, and therefore seek fairness in minimizing the partisan advantage that a dominant political party can obtain over the other in the typical two-party state through a bipartisan redistricting commission. 155 In a statistical study, Gary King demonstrates that bipartisan redistricting does somewhat reduce partisan bias. 156 Robert Dixon argues that a bipartisan commission "permits a focus on the realities of political representation in all proposed plans, thus avoiding a process of shadowboxing with pseudostandards such as contiguity and compactness." 157 Proponents of bipartisan commissions typically neglect the fact that these commissions at best offer fairness only along one dimension: the partisan dimension, and typically for the Democratic and Republican parties at that. 158 subdivision boundaries and with the most regular shapes, it could very well result in a landslide election for a given political party."); Nathaniel Persily, In Defense of Foxes Guarding Henhouses: The Case for Judicial Acquiescence to Incumbent-Protecting Gerrymanders, 116 HARV. L. REV. 649, 678 (2002) ("With nonpartisan expertise, however, often comes detachment from the policy goals of the political branches. For example, it is quite typical for nonpartisan experts to attempt to make district lines as coterminous with political subdivision boundaries as possible. Pursuing such a goal, however, often conflicts with attention to communities of interest that straddle such boundaries and with a state's public policy goal of regionalism in uniting cities and suburbs. Such crossover districts can link the representational fates of richer suburbs with the poorer cities off of whose services suburbanites might be likely to free-ride. Also, the overzealous pursuit of neutral redistricting principles often leads to boundary changes that disrupt ongoing programs, such as environmental or infrastructure projects, instituted by an incumbent for his or her district."); see also Issacharoff, supra note 40, at 1695 ("[R]eapportionment, even under a purportedly neutral commission, necessarily confronts the redistricting question in a context where the outcomes of the decisionmaking are regularly ascertainable."); infra Part III.A.4 (discussing how every redistricting plan has partisan or political effect, even if not through partisan or political intent).

*p. 28*
154. See Persily, supra note 153, at 678 n.95. 155. See Dixon, supra note 134, at 10 (approving of a bipartisan commission with a tiebreaker as logical and preferable to other redistricting methods, but admonishing that it is necessary that the commission have access to political data); Dixon, supra note 135, at 30, 36; Jeffrey C. Kubin, Note, The Case for Redistricting Commissions, 75 TEX. L. REV. 837, 846 (1997).

*p. 28*
156. Gary King, Representation Through Legislative Redistricting: A Stochastic Model, 33 AM. J. POL. SCI. 787, 818-19 (1989).

*p. 28*
157. Dixon, supra note 135, at 36. For a discussion of contiguity and compactness, see infra sections B.3(a) and (b) of this Part.

*p. 28*
158. See Ely, supra note 119, at 635 ("[S]uch commissions have done nothing visible to open the door to the possibility of third-party representatives. Nor does turning the problem over to a commission carry any evident advantage in terms of . . . complying with a judicial requirement of race-blind districting without substantially reducing the (limited) minority representation that currently exists.").

*p. 28*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL Many interests are at work within any particular state: those of racial minorities, third parties, religious groups, geographic communities, business interests, and political organizations to name a few. To be sure, some of these interests tend to correlate with particular partisan interests-such as, stereotypically, big business or rural communities with the Republican Party and labor unions or urban communities with the Democratic Party. 159 But not all do. For instance, Catholic groups tend to be both pro-life, which is typically associated with the Republican Party, and against the death penalty, which tends to be associated with the Democratic Party. 160 To the extent this association is accurate, neither the Republican nor the Democratic Party can fully account for the Catholic groups' interests, at least in this regard. There also can be vast differences between party members' interests depending on their extra-partisan interests. For example, urban Democrats have been shown to be significantly more liberal with regard to economic and social welfare issues than are rural Democrats. 161 Thus, packing rural Democrats together will tend to yield more liberal Democratic representation. 162 To the extent that extra-partisan interests do not correlate exclusively with either of the political parties represented in a bipartisan commission, there is no guarantee that such a commission will consider these interests in redistricting.

*p. 29*
Robert Dixon, a strong proponent of bipartisan commissions, dispenses with this issue by noting that the bipartisan commission may take account of other interests in redistricting, such as those of racial minorities. See Dixon, supra note 135, at 36 ("[T]here could be some gain for ethnic minorities, for, although it might not be advisable to attempt to conduct all commission business in open session, the commission's existence would bring the process into the open and provide a responsible focal point for making known the interests of particular groups concerning representation."). However, there is no guarantee that bipartisan commissions will do anything other than conspire to protect the parties' interests (or those of the incumbents) at the expense of other interests. See Issacharoff, supra note 2, at 600-01 (arguing that bipartisan "political market manipulation threatens a core tenet of democratic legitimacy: accountability to shifting voter preferences," that "market division agreements should be inherently suspect, if not per se prohibited," and therefore that redistricting by incumbents should be deemed unconstitutional).

*p. 29*
159. E.g., Stephen Ansolabehere & James M. Snyder, Jr., Reapportionment and Party Realignment in the American States, 153 U. PA. L. REV. 433, 436-37 (2004) One can argue that the commissions need not consider these interests on the ground that the political parties in power, as representatives of these diverse interests, are responsible for representing their interests in redistricting just as they are in general legislation. However, this argument overlooks the fact that substantive legislation and representation differ qualitatively from redistricting: redistricting reestablishes the political foundation upon which representation is built; and in redefining the rules of the electoral game, these interests have more reason than with general legislation to have a direct say in the process. Furthermore, redistricting affects the political parties and incumbents directly by establishing how much representation they are likely to attain; they therefore have a distinct interest in ignoring or diminishing their constituents' interests for their own gain. 163 Finally, there are those that would argue that bipartisan commissions are unfair to the dominant political party in a state, if there is one. 164 Redistricting, on this argument, is a spoil belonging to the state's political victor and is not to be split between the parties. This argument sounds in fairness to the political victor.

## Partisanship

*p. 30*
Those who contend that bipartisan commissions are unfair to the political victor often support redistricting in the legislature under regular partisan conditions. The arguments in favor of automated redistricting, nonpartisan commissions, and bipartisan commissions, as well as the arguments discussed below in favor of proportional representation and particular redistricting standards argue against a freewheeling partisan redistricting extravaganza. The biggest concern with partisan redistricting is that "the majority party can forestall its opponents from effectively competing for dominance, perhaps even perpetuating itself in office." 165 That is, there is reason to worry that there ought to be a limit on how long a majority party can lock up power against challenge.

## Proportional Representation

*p. 30*
Peter Schuck presents an argument that a fair districting plan is one that leads to fair representation. 166 Schuck views proportional representation-the situation in which each group is represented in the legislature in numbers proportional to its percentage of the state population-as the fulfillment of fair 167 The hypothesis underlying the theory of proportional representation is "that groups are the effective motive forces in the legislative process, but also that the electoral rules that structure that process should be evaluated, at least for purposes of constitutional analysis, according to how well they facilitate group formation and maintenance." 168 According to this view, unrestrained redistricting allows the "majority party to impede and distort the processes by which minority parties form, grow and compete to become the majority and represent their members effectively." 169 Congress and the Supreme Court have explicitly rejected proportional representation as a benchmark for fairness. 170 Although a proportional-representation system does have appeal, one would have to eliminate the traditional districtbased system to implement it effectively. 171 Most states would not adopt proportional representation systems without a judicial or legislative mandate, which is highly unlikely given the Supreme Court's jurisprudence and Congress's stance. As to the merits of such a system, proportional representation is likely to lead to a multiparty system, which tends to lead to ineffective, fragmented legislatures. 172

## Symmetry

*p. 31*
A districting plan is considered to be symmetric when all groups "receive the same proportion of seats for a given vote proportion." 173 That is, group A gets x percent of the legislative seats for y percent of the population's votes whenever any group B gets x percent of the legislative seats for y percent of the population's votes. 174 Symmetry is considered by its advocates to be fair because all groups get the same number of legislative seats for the same number of votes.

*p. 31*
1577 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING Scholars typically discuss symmetry in the context of partisan symmetry, and the two major parties, at that. 176 As with bipartisan commissions or partisanship in general, partisan symmetry measures only the partisan dimension of fairness, typically only for the Democratic and Republican Parties.

*p. 32*
Gary King, John Bruce, and Andrew Gelman discuss the notion of symmetry in the racial context. 177 They write that the notion of partisan symmetry is not easily adapted to racial symmetry. Racial minorities are infrequently nominated, and nomination is necessary for election, whereas in the partisan situation, both Democrats and Republicans are typically on the ballot. 178 Therefore, they transform the partisan-symmetry standard so that "blacks would be allocated the proportion of seats in the legislature they would get if voters and candidates were color-blind, and if ideology were independent of race." 179 They recognize the extremism of this standard because it assumes that minorities have no "distinctive political positions." 180 As beautifully simple as symmetry sounds as a measure of fairness, there are a number of ugly complications. First, symmetry is very hard, if not impossible, to implement practically because it involves theoretical calculations that are difficult to translate into districting plans. 181 REV. 588, 622-23 (1993).

*p. 32*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL but the general idea is that a district be packed tightly to ensure effective representation and to ensure that those in the district share interests, by virtue of proximity. 185 By restricting the possible districting schemes available to redistricters, compactness is thought to reduce the opportunities to gerrymander. 186 The Supreme Court has hinted that compact districts are a prudential feature of good districting schemes, even if not constitutionally required. 187 The value of compactness has been questioned in recent years as technology and rapid transit have diminished the need for compact districts. 188 Moreover, compactness is often thought unnecessary to achieve good governance: Bruce Cain suggests that representatives in noncompact districts could have multiple district offices or that they could delegate to dispersed staff to compensate for the diffusiveness of the districts. 189 Furthermore, it is questionable whether compact districts will enclose communities of interest and whether noncompact districts may do so at least as well. 190 Finally, some scholars decry compact districting schemes for unfairly disadvantaging racial minorities and urban 185. See, e.g., CAIN, supra note 146, at 32 (noting that the press and public measure compactness by shape); MARK MONMONIER, BUSHMANDERS & BULLWINKLES: HOW POLITICIANS MANIPULATE ELECTRONIC MAPS AND CENSUS DATA TO WIN ELECTIONS 64-70 (2001) (noting that "compactness is a bit like pornography-although we know it when we see it, individual sensitivities and community standards vary widely," and describing one way to measure compactness by "dividing the area of the district by the area of the smallest circle that completely encloses the district"); Altman, supra note 130, at 83 n.6; Dixon, supra note 134, at 16; Harris, supra note 130, at 220-21; Henry F. Kaiser, An Objective Method for Establishing Legislative Districts, 10 MIDWEST J. POL. SCI. 200, 200 (1966) (defining compactness as "making the average, over districts, of the relative moments of inertia of the geographical configurations of the districts as small as possible"); Weaver & Hess, supra note 129, at 290-91, 296-307 . 301, 326-39 (1991).

*p. 33*
187. See, e.g., Shaw v. Reno, 509 U.S. 630, 647 (1993) (striking down a legislative districting scheme containing two noncompact minority-majority districts unless it satisfied strict scrutiny); Thornburg v. Gingles, 478 U.S. 30, 48-51 (1986) (holding that in determining whether there has been minority vote dilution, one must look to, among other factors, whether the minority group in question could form a compact district); cf. Aleinikoff & Issacharoff, supra note 184, at 623 ("[T]he Court in Shaw insisted that it was not seeking to make compactness an independent constitutional requirement."); Issacharoff, supra note 40, at 1692 ("The Supreme Court has, to date, given no indication of constitutionalizing independent rules of compactness for districting arrangements. Rather, the lack of compactness is generally taken as evidence of an ulterior purpose that may call into question the permissibility of state districting practices.").

*p. 33*
1579 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING Democrats, who are typically packed tightly in cities and therefore can be packed into few districts, and for unfairly advantaging suburban and rural Republicans, who are spread more loosely and can be distributed advantageously over many districts. 191 (b) Contiguity. A contiguity standard mandates that "every part of the district [be] reachable from every other part without crossing the district boundary." 192 This standard is a usually uncontested measurement of fairness, 193 but at least one scholar has disparaged contiguity as a baseless "pseudostandard." 194 (c) Equal Population. As elaborated in Part II.A, the standard of equal population is constitutionally mandated to ensure fair voting and/or fair representation. That Part discusses the benefits and disadvantages of such a requirement.

## Protection of Political Boundaries and Communities of Interest

*p. 34*
Another oft-cited criterion for redistricting is that neither political boundaries, such as cities or counties, nor communities of interest should be broken into multiple districts. 195 The justifications for this criterion overlap substantially with those reasons for having compact districts-namely, that those who share interests or values, typically those in the same city or county, achieve effective representation when they aggregate their votes. The Supreme Court has accepted this reason as a valid basis for redistricting choices. 196 One criticism of protection of political boundaries is that it unnecessarily perpetuates discriminatory and unhelpful boundaries. 197 Additionally, Richard Thompson Ford argues that automatic protection of political boundaries ignores important normative questions about democracy. 198 Ford claims that these boundaries are both arbitrary-because there are multiple other justifiable boundaries-and controversial-because there are groups excluded from a political boundary that want inclusion and groups included in a political boundary that prefer exclusion. 339, 346-48 (1960) (holding that a city's redrawing of its lines to keep African-Americans out of the city while retaining all the white residents was violative of the Fifteenth Amendment).

*p. 34*
198. See Richard Thompson Ford, Geography and Sovereignty: Jurisdictional Formation and Racial Segregation, 49 STAN. L. REV. 1365, 1368 (1997).

## Protection of Incumbents

*p. 35*
There is strong concern that incumbent legislators will attempt to protect their seats by redistricting favorably to themselves 200 and that a party in power will attempt to dislocate incumbents from other parties. 201 This concern has in fact stimulated much of the academic literature on how to redistrict fairly. 202 However, there are countervailing fairness concerns that incumbents be protected in the redistricting process to ensure that they need not compete against one another in the next round of elections and that they need not lose many of their constituents to a different district. 203 The concern for incumbent protection stems not from personal concern for continuing incumbents' employment but rather from concern for their constituents. 204 The argument goes that having rapid turnover of incumbents leads to inexperienced legislators and thus to poor representation. 205 Incumbents have the responsibility to learn the needs of their constituents and represent their constituents; changing constituent bases by redistricting thus also leads to ineffective governance. 206 This measure of fairness is very controversial: there are arguments that incumbents' election advantages-name recognition, the ability to raise funds, and a say in the redistricting process-are sufficient that they do not need any 200. E.g., Issacharoff, supra note 2, at 624-26; Weaver & Hess, supra note 129, at 295-96 ("[L]egislatures have generally districted on a partisan basis, often with the goal of re-electing as many incumbents as possible."). However, a competing concern motivating those involved in redistricting is protection of a partisan advantage; this concern competes with incumbent protection by making incumbents give up votes to spread out their party supporters among more districts, thus allowing their party to capture more legislative seats. See Gelman & King, supra note 42, at 541. "Since political party gain is the most predictable common ground for otherwise competing incumbents, party advantage will often take precedence over individual incumbents' advantage in the ultimate political compromise represented by a redistricting plan." Id. at 542. 201. See Issacharoff, supra note 40, at 1661-62. One technique to accomplish this dislocation involves "kidnapping" a particular political party's incumbent from one district and lumping him or her with another incumbent from that party so that that party loses a seat and the incumbents can battle to preserve their position. See, e.g., Toobin, supra note 2, at 65-66.

*p. 35*
202. Cf. Bruce Cain, Assessing the Partisan Effects of Redistricting, 79 AM. POL. SCI. REV. 320, 320 (1985) ("[R]edistricting effects are tied closely to incumbency effects.").

*p. 35*
203. See CAIN, supra note 146, at 12. 204. See id. ("[T]he incumbents have been duly elected by the voters in their old districts, and . . . a commission or computer program that removed a large number of them from their seats would in effect be depriving those voters of their elected representatives."). These arguments about incumbent protection in redistricting are analogous to those involved in the national debate over term limits.

*p. 35*
205. See id.; Persily, supra note 153, at 672 ("[E]ffective policymaking requires experience. A system of representation intentionally biased against the cultivation of experience undermines the institution of the legislature and weakens it vis-a `-vis the other branches of government.").

*p. 35*
206. See CAIN, supra note 146, at 12; Persily, supra note 153, at 671 ("Voters develop relationships with their representatives. Long-term representatives have a chance to learn about and understand the unique problems of their districts and to pursue legislation that remedies those problems. To voters, 'entrenched' representatives are the ones who know the most about them and can do the most for them . . . . On the contrary, novice representatives are likely to be systematically inferior to 'entrenched' representatives when it comes to the effective representation of their constituents' views.").

*p. 35*
1581 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING special protection. 207 And incumbents who have enough of a say in the redistricting process might "lock up" an election for themselves regardless of their merit. 208

## Competitiveness

*p. 36*
The desire to make districts competitive contrasts starkly with the desire to protect incumbent legislators in redistricting, because making districts competitive means that incumbents do not have a significantly stronger chance than any other candidate to be elected. Supporters argue that making districts competitive ensures good government by "maximiz[ing] effective voter choice." 209 In addition, if voters of a district always elect a member of the same political party, the candidates need "answer only to primary voters, who represent the extreme partisan edge of both parties." 210 The downsides of using competitiveness to measure fairness are numerous. First, all the reasons for protecting incumbents speak against making districts competitive. Second, it is very difficult to measure whether a district is truly competitive until "after comprehensive consideration of both the general and 207. Cf. Issacharoff, supra note 40, at 1701-02 ("[D]isrupting the district alignments of incumbents is likely to be the greatest source of resistance to computer-generated redistricting, both from within the political process itself and from a federal judiciary that is unfortunately becoming accustomed to measuring political fairness by the impact of redistricting on incumbent power bases.").

*p. 36*
208. "Only four challengers knocked off incumbents in the November 2002 general election-a modern record low not only for a redistricting year, but for any election year." Hirsch, supra note 47, at 182. Hirsch attributed this result to the redistricting that occurred in response to the 2000 U.S. Census results, which typically had the purpose of protecting incumbents. See id. at 186-89. Nathaniel Persily takes issue with attributing incumbent reelection to gerrymandered districts. He first argues that, "to the extent incumbents have unfair and growing advantages over challengers, redistricting is not to blame. We know this because statewide elections unaffected by redistricting, such as electors for governor and U.S. Senate, have shown parallel growth in rates of incumbent reelection." Persily, supra note 153, at 650; see also id. at 659-66 (describing other causes for incumbency reelection, including the decision of high-quality challengers to avoid election battles with an incumbent, the rise of candidate-centered politics, and rising campaign costs). Second, in his view, "there has been steady and significant turnover both in Congress and in state legislatures . . . ." Id. at 654; see id. at 654-57, 654 fig. 1, 657 fig. 3 (stating that 10-20% of incumbents in the U.S. House of Representatives have lost reelection campaigns in the past thirty years; that in some states without term limits, more than 30% of state legislative incumbents have lost reelection campaigns; that partisan control shifts in about 13% of state legislative houses with each election; and that, during the 1990s, twenty-nine state legislative houses had a shift in partisan control). But see Persily, supra note 74 (stepping back from this conclusion in light of the 2002 election results on the basis that redistricting played a substantial role in making 90% of districts in the House of Representatives noncompetitive).

*p. 36*
209. Gordon E. Baker, The Unfinished Reapportionment Revolution, in POLITICAL GERRYMANDERING AND THE COURTS 11, 23 (Bernard Grofman ed., 1990); see also CAIN, supra note 146, at 68; Issacharoff, supra note 2, at 615 ("[T]he concept of 'fair' representation has no meaning outside an appropriately competitive electoral process. The electorate can only express a 'free and uncorrupted choice' if it has the ability to select among competing political prospects."). Persily counters that "there is no evidence . . . that safe districts have produced unresponsive representatives." Persily, supra note 153, at 660 (reasoning that election margins in previous elections are not good indicators of future election margins).

*p. 36*
210. Toobin, supra note 2, at 64 (explaining that extremists of each party tend to be the ones going to the polls to vote in primaries).

*p. 36*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL idiosyncratic factors that are likely to affect the campaign and the election." 211 Third, competitiveness is not a stable trait of districts, making it difficult to measure the fairness of a redistricting plan. 212 Fourth, "competitive districts may lead both to greater power for the ideologically centrist, median voter and to almost half of the voters[] being unhappy with their representative." 213

## Diversity

*p. 37*
There are numerous proponents of ensuring diversity in legislative bodies. It is typically said that ensuring diversity requires diverse districts, in the sense that some are either majority-minority districts or districts in which minority groups have influence. 214 Heather Gerken theorizes that that there are two possible orders of diversity that a districting scheme can provide. 215 A scheme that generates a number of majority-minority districts in relative proportion to a polity's population will tend to lead to the election of a legislature that is first-order diverse-it reflects the composition of the polity. 216 The scheme is also second-order diverse in the sense that there is interdistrict diversity, as some districts will be controlled by (perhaps different) minority groups and some will be controlled by the majority group. 217 Gerken argues that there is independent value to second-order diversity in districting, 218 because even though the representatives of majorityminority districts will constitute a minority group in the legislature, they can still effectively represent and reward their constituents via legislative committee seats, bill sponsorship, and provision of assistance to constituents. 219 Although the goal of diversity has an admirable basis, a number of details remain unsettled, which makes judging the diversity of districting schemes difficult. First, it is unsettled whether influence districts are preferable to minority-majority districts or vice versa. 220 218. See id. at 1126 ("Second-order diversity thus offers an alternative strategy for achieving fairness in a disaggregated democracy. Rather than giving electoral minorities 'influence' over each decision, it gives some electoral minorities the chance to be in the majority on a decisionmaking body and ensures that someone who is not in the middle of the political spectrum will have the chance to be the swing voter."); id. at 1144-45 ("Turning the tables promotes a healthier democratic process for reasons that sound in the language of process. It may help electoral minorities feel that they have gotten a 'fair shake' from the majority and thus feel more invested in the political process.").

*p. 37*
219. See id. at 1134-35. 220. Cf. supra text accompanying notes 64-74 (discussing the Supreme Court's thoughts on the topic in the context of section 5 of the VRA).

*p. 37*
1583 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING much representation or influence each minority group must achieve. Third, numerous potential minority groups exist in religious, political, ethnic, racial, and other spheres. It is indeterminate which set of these groups must be accounted for in a districting scheme for it to be "adequate."

## C. ADDITIONAL PROBLEMS WITH THESE STANDARDS AND METHODS

*p. 38*
In addition to the individual downsides discussed for each of the above standards and methods, most, if not all, share some common problems. First, aside from a handful of fairness criteria-such as partisanship and bipartisan commissions-these standards and methods ignore partisan effects in redistricting. That notion is problematic, because "[e]very districting method helps someone at least to the extent of hurting someone else." 221 That is, every redistricting choice is ultimately political in intent or political in effect by necessity. 222 Seen in this light, what is commonly thought of as gerrymandering is nearly identical in effect to "innocent" redistricting. 223 For instance, as noted above, compactness, a seemingly neutral criterion of fairness, tends to gerrymander by helping Republicans and hurting Democrats. 224 To attempt to remove political considerations from redistricting is to cast a blind eye on the inherent nature of redistricting. In building a model of fair redistricting, therefore, it is preferable to account for political considerations. 225 Second, although each of these suggested methods or standards for fairness in redistricting has flaws, each also has dependable arguments in its favor. It is difficult, if not impossible, to assemble neutrally a proper subset of these fairness standards or methods as preferable and exclude the others. 226 225. "To the extent that important political data is kept concealed, the districting process will take on the quality of Russian roulette." Dixon, supra note 135, at 30; cf. Hirsch, supra note 47, at 180 ("Rather than pretending to 'take politics out of the process' by creating supposedly apolitical redistricting commissions, States should place these same overtly political concerns-partisan fairness, responsiveness, and accountability-squarely on their list of redistricting criteria and should implement them systematically and empirically.").

*p. 38*
226. Cf. Persily, supra note 153, at 678 ("[C]oncerns about representation and governance are of equal weight to concerns about electoral competition, and there is no philosophically uncontestable reason why judges should force one set of values rather than another down the throat of state governments. More subtly, while the process of redistricting may be considered as determining the rules of politics, it should also be seen as a substantive area of public policy."); Richard H. Pildes, Avoiding Balancing: The Role of Exclusionary Reasons in Constitutional Law, 45 HASTINGS L.J. 711, 724 (1994) ("When values are diverse but important, the preservation of this tension between values-rather than 1584

*p. 38*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL more, many of these fairness goals conflict with one another. 227 For instance, to achieve competitive districts necessarily means one cannot provide incumbents much protection. Similarly, to respect political boundaries may mean undermining diversity interests when minority groups are spread across various and possibly distant cities. 228 Because of such conflicts, any established rules of fairness would require value judgment in picking among the standards and methods. Relatedly, each proposed standard or method captures some different fairness value along some subset of the possible dimensions of fairness. But no fairness method or standard above captures fairness along all possible dimensions.

*p. 39*
Third, these standards and methods put judges reviewing redistricting schemes in the unfortunate position of having to implement their own substantive vision of democracy by picking and choosing which standards to apply. 229 This picking and choosing disserves judicial restraint and minimalism. 230 Judicial decisions in which judges choose from the grab bag of available standards and methods, for which democratically accountable actors are unquestionably better suited, deprive these democratically accountable actors of a role in fashioning any alternatively fair democratic vision. 231 That is precisely the unfortunate the total triumph of one set of values-fosters the richness of a complex society with multiple aspirations.").

*p. 39*
227 229. Of course, legislatures might specify the precise standard by which courts should judge whether a redistricting scheme is fair, which would make judicial review of redistricting schemes more democratic. Nonetheless, given the numerous problems discussed in this Article and in the literature generally with having an unrestrained legislature dictate redistricting procedure, it would be troublesome for the legislature to set the fairness baseline for redistricting schemes.

*p. 39*
230. See, e.g., Cass R. Sunstein, The Supreme Court, 1995 Term-Foreword: Leaving Things Undecided, 110 HARV. L. REV. 4, 6-9 (1996).

*p. 39*
231. See Bruce E. Cain, Election Law as a Field: A Political Scientist's Perspective, 32 LOY. L.A. L. REV. 1105, 1110 (1999) ("By choosing to protect one theory of representation, the Court altered structures of representative government that had been in place since the eighteenth century. In so doing, they deprived the American people of an entire class of institutional mechanisms for compromise which could be used to solve collective action problems."); id. at 1111 ("By declaring that apportionment was justiciable, the Court intruded into the political design process. In a number of areas, the courts have entered into political issues which were previously thought to be more properly decided in arenas outside the courts."); Michael A. Fitts, The Hazards of Legal Fine Tuning: Confronting the Free Will Problem in Election Law Scholarship, 32 LOY. L.A. L. REV. 1121, 1135 (1999) ("[C]ourts . . . seeking to resolve the structure of our electoral system should not assume that all choices are capable of normative legal resolution. Some important subset of decisions needs to be left to the political process itself."); cf. Richard H. Pildes, The Theory of Political Competition, 85 VA. L. REV. 1605, 1611 (1999) ("The way to sustain the constitutional values of American democracy is often through the more indirect strategy of ensuring appropriately competitive interorganizational conditions. It is in this way that central democratic values, such as responsiveness of policy to citizen values and effective citizen voice and participation, are best realized in mass democracies."). But cf. Hirsch, supra note 47, at 180 ("When the normal state political processes become deadlocked and redistricting is thrown into the courts, judges should invite empirical evidence of likely electoral outcomes and should evaluate the competing remedial maps for partisan fairness, responsiveness, and accountability, rather than relying solely on formal criteria such as compactness and respect for county and municipal lines.").

*p. 39*
1585 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING result of Georgia v. Ashcroft, 232 in which the Court struggled to compare, in determining fairness, whether more influence districts are better than fewer majority-minority districts. 233 As Cass Sunstein suggests, "a minimalist path usually-not always, but usually-makes sense when the [courts are] dealing with an issue of high complexity about which many people feel deeply and on which the nation is in flux (moral or otherwise)." 234 Fourth, a number of these standards-including those adopted by the Supreme Court to analyze racial gerrymandering and by some members of the Court to measure partisan gerrymandering-do not look at the fairness of an entire jurisdiction's line-drawing of district boundaries as a whole but rather at the fairness of individual districts. Even though a particular district in a jurisdiction might seem unfair to a certain group, all of the other districts might be favorably drawn to that group's advantage. It is therefore not sensible to judge the fairness of a districting scheme by looking only at the internal workings of each individual district; that would be like examining just one data point rather than looking at the pattern that emerges from drawing lines connecting multiple data points. Instead, it is preferable to have standards and methods for fair districting that operate by examining all of the districts within an entire jurisdiction. 235 Finally, as the breadth of the above standards and methods shows, it is difficult to define one substantive standard of fairness. Commentators themselves are cognizant of this difficulty. 236 Choosing a single substantive standard or method for the public at large makes the courts seem paternalistic. It would be preferable to allow political actors-or better yet, the public itself-to choose these standards for themselves in a fair fashion. 237 The challenge, thus, is to address these deficits by formulating an alternative measure and model of fairness in redistricting, grounded in procedural compromise.

## IV. AN EXERCISE IN FAIR LINE-DRAWING

*p. 40*
After arguing that collective public compromise is fairness in the redistricting context, this Part introduces a procedural collective-compromise model for constructing fair districting schemes, which is designed to minimize, if

*p. 40*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL not to bypass, the problems inherent in substantive or assertedly apolitical models of redistricting. The model works by allowing any number of public participants to rank the possible districting configurations in a jurisdiction, and then deriving a single districting scheme based on the various participants' rankings. A variety of procedures can be used to map the rankings onto a resulting districting scheme, and this Part presents two proceduresconfiguration elimination and meta-election-as illustrations of the procedural model. After presenting a generalization of the procedural model, this Part discusses the implementation of this model on a computer, and then argues that the procedural model provides fairness in the redistricting context above and beyond existing models. This Part then details how the model can also be used descriptively to both measure the fairness of a particular redistricting scheme and compare redistricting processes over time and across jurisdictions by examining the local substantive redistricting values at play in the various redistricting processes.

## A. COMPROMISE AS FAIRNESS

*p. 41*
In a compromise, "the parties to the conflict give up their insistence on their original demands, and alter what they are willing to accept." 238 Deeming compromise to be fairness seems like a realistic, not an idealistic, notion of fairness, 239 and as set forth below, the notion of compromise is rooted in the concept of democracy. Many recognize the positive and negative views of compromise. While compromise facilitates problem solving and cooperation, in the realm of politics, compromise can be seen as improperly requiring parties to concede dearly held values. 240 However, just as many also recognize that compromise is to be favored in the political sphere, because it is "the only fair and rational way of reaching a reasonable agreement between . . . differing points of view." 241 238. BARRY JAY SELTSER, THE PRINCIPLES AND PRACTICE OF POLITICAL COMPROMISE: A CASE STUDY OF THE UNITED STATES SENATE 8 (1984). For compromise to occur, there must be a middle ground, see id. at 22, and in the typical case, the vast range of districting configurations, embodying varying goals and impacts, provides plenty of ground upon which to meet.

*p. 41*
239. See id. at 15 ("The call to compromise is based ultimately on the demand for realism, for accepting the world's options, for not being stubborn." ("As we want to say that the person of goodwill is a person of firm principle, we are often inclined to suppose that the willingness to compromise is a sad but sure sign of moral turpitude. Yet we also want to picture the morally decent person as a person who is striving to be concessive and accommodating with fellow humanity, and so we are sometimes moved to regard the willingness to compromise as a profound expression of moral goodwill."). The tradition of compromise as fairness, or justice, is recognized by philosophers as diverse as Socrates, Aristotle, John Locke, Jean-Jacques Rousseau, Immanuel Kant, Friedrich Nietzsche, and John Rawls. Socrates and Nietzsche both regard compromise agreements between self-interested people on relatively equal footing as justice. 242 Aristotle views principled compromise as necessary to achieve justice. 243 Locke, Rousseau, Kant, and Rawls offer a more benign interpretation of justice: an accommodation among mutually respecting people to live together via a social contract. 244 In fact, Edmund Burke stresses that democratic government is positively founded on compromise. 245 As one scholar puts it:

## FRED CHARLES IKLE

*p. 42*
[O]ne might go so far as to suggest that the primary purpose of truly democratic institutions is to inhibit the formation of policies that are not, in large measure, the product of accommodation on the widest possible scale.

*p. 42*
Here arises the suspicion that a true democrat is a person without any sense of the limits of legitimate compromise and that at the foundation of democratic theory is a morality of "pure compromise." 246 Although there are those that emphasize that only matters of morality should not be compromised, 247 it should be clear at this point in the discussion that there are few moral and many political concerns inherent in the range of identified several reasons for being cautious about appealing to principle in the political world. Such appeals err in overlooking conflicts of principles; they cut against the element of pragmatism and tolerance in the American political system . . . .").

*p. 42*
242. See Kuflik, supra note 240, at 56 (discussing Socrates and Nietzsche). 243. See David Resnick, Justice, Compromise, and Constitutional Rules in Aristotle's Politics, in COMPROMISE IN ETHICS, LAW, AND POLITICS, supra note 240, at 85 ("Justice for Aristotle is not simply an abstract ideal but a practical political problem rooted in the reality of class conflict and competing political claims. Justice must promote the common interest, but the common interest requires that each class be willing to accept the fact that its own interests as well as the true concept of distributive justice requires compromise . . . . For Aristotle, compromise is not incompatible with the principles of justice, but rather true justice requires a principled compromise.").

*p. 42*
244 248 Instead of having a dominant voice drown out softer voices in American pluralistic society, it is intuitively more fair to blend all of these voices. 249 Further, if the participating parties perceive the redistricting process as fair, by virtue of their participation, the winning districting configuration will have an air of legitimacy. 250 One can judge a compromise in two ways: either by measuring the substantive outcome of a compromise agreement or by looking to the method by which the compromise agreement was reached. 251 The first judgment tends to find a fixed set of compromise agreements to be fair, while the second judgment views a limited number of compromise processes as fair. 252 This Article does not assume a fixed set of fair districting configurations in the abstract for the reasons principally set out in Part III.C, but rather argues for procedural fairness, in that there are a limited number of ways to compromise and choose a districting configuration that is fair, regardless of the characteristics of the ultimate districting scheme. This procedural model grants each of the participants a degree of moral legitimacy by allowing each one to participate in the redistricting process. 253 Behind the veil of ignorance-before parties know how much political power they will wield-they would rationally want to participate in the procedural model to ensure that they have some voice.

## B. THE PROCEDURAL MODEL

*p. 43*
The proposed procedural model anticipates that there are n participants involved in redistricting. Who the participants can be is discussed in more detail in Part V.A. For now, imagine as typical participants various public entities, such as political parties, minority groups, majority groups, and other interest 248. "In a liberal democracy good compromises respect or enhance the personal autonomy and dignity of citizens and strengthen liberal and democratic institutions by satisfying legitimate aspirations, by empowering disenfranchised citizens, or by opening a closed system to greater public participation." J. PATRICK DOBEL, COMPROMISE AND POLITICAL ACTION: POLITICAL MORALITY IN LIBERAL AND DEMOCRATIC LIFE 167 (1990).

*p. 43*
249. For compromise to occur, the parties involved must typically relinquish their respective ideal districting configurations to arrive at some less-preferred configuration; by virtue of the plethora of fair districting criteria and methods, parties should recognize that there are many competing values involved in the districting process and should therefore be amenable to compromise. See SELTSER, supra note 238, at 23. 250. "When a procedure is perceived to be fair because it satisfies certain criteria of fairness, it is more likely to lead to outcomes that are viewed as legitimate by all the parties. These outcomes will be more durable because, when seen as the product of a fair process, they are less likely to incite disaffected parties to try to derail them." STEVEN J. BRAMS & ALAN D. TAYLOR, THE WIN-WIN SOLUTION: GUARANTEEING FAIR SHARES TO EVERYBODY 8-9 (1999).

*p. 43*
251. Golding, supra note 240, at 7. 252. See id. at 7-8. 253. See id. at 16 ("The compromise process is a conscious process in which there is a degree of moral acknowledgment of the other party. The other party is accorded some degree of moral legitimacy, and so are some of his interests.").

## 2005] 1589 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING

*p. 44*
groups. 254 There is a finite, but typically very large, number of possible districting configurations for each district. 255 Each participant must rank the possible districting configurations, either by ranking in order each individual districting configuration or, because the sheer number of districting configurations would in most realistic cases make such a ranking unmanageable, by providing a "policy"-a ranked listing of criteria (such as preferring a maximum probable number of representatives for the participant, followed by preferring competitive districts in districts not probably controlled by the participant) set by the participant-from which the ranking of districting configurations can be automated without the need to review each possible configuration. 256 The model then creates a collective compromise among the assorted public participants by integrating all of their rankings or policies to derive a unique districting scheme.

*p. 44*
In one illustration of the integration of the participants' rankings or policies, the n participants take turns eliminating their least-preferred districting configuration, from all possible districting configurations, until there is one remaining configuration, which is deemed to be the fairest districting scheme. In another illustration of this integration, the most preferred configuration is determined by comparing, for each configuration, the sum of the participants' rankings for it. Participants' rankings can all be evenly or unevenly spaced or can include both negative and positive rankings. 257 The most notable aspect of these illustrations is not that they use a particular method to achieve a unique districting scheme; rather, it is that they illustrate how participants' varying configuration preferences can be blended together to yield a unique districting scheme. Underlying this model is the hypothesis that a properly constrained compromise constitutes fairness in the redistricting context. 254. Obviously, in creating the procedural model, a substantive-and arguably political-choice must nonetheless be made about who gets to participate. The problems created by choosing the participants are discussed in Part V.A, infra.

*p. 44*
districting configurations for y districts with a total population of x people. Altman, supra note 130, at 97. In practice, there will be fewer configurations because configurations with highly unequal populations between districts are not constitutional.

*p. 44*
256. It might be the case that only policies composed of constitutional criteria would be constitutional.

*p. 44*
257. The idea of having participants choose between individual districting configurations rather than having them choose individual lines to draw is inspired by a series of computer science problems, particularly in the field of artificial intelligence. For this set of problems, the classical artificial intelligence approach is to examine the various potential configurations of the environment at issue and determine which configuration maximizes the solution to a particular problem. Take, for instance, the game of chess. Rather than have a computer chess software program examine individual chess pieces, the classical artificial-intelligence approach has the program evaluate the possible moves it can make at any particular time (that is, all of the potential chessboard configurations that can exist after one move), then all of the possible moves its opponent can choose in response, then all of the possible moves it can choose in response to that, and so forth for a finite number of moves. The program seeks to maximize its chessboard configuration at each turn, while minimizing the configuration of its opponent. See, e.g., STUART J. RUSSELL & PETER NORVIG, ARTIFICIAL INTELLIGENCE: A MODERN APPROACH 122-45 (1995).

## Configuration Elimination

*p. 45*
The method of harmonizing participants' preferences and goals via configuration elimination to derive a fair redistricting scheme is first presented by two examples and then a general model. Even though the illustrations use partisan and racial-minority participants, the participants could just as easily consist of an individual representing only himself or herself, an environmental cause group, a religious minority group, and a labor union, to name but a few other or additional potential participants. Ignoring the order of districts and any data other than political party membership, there are three possible configurations of equipopulous districts, as depicted in Table 1: one configuration with all of the Depublicans packed into one district, one configuration with a two-member majority Depublican district, and one configuration with one Depublican in each district. 259 How can the parties compromise to pick a fair districting scheme? After enumerating its goals and preferences, each party takes turns eliminating the configuration it dislikes the most until only one redistricting configuration is left. That configuration is the most preferred in a compromise and is thus the fairest districting scheme.

*p. 45*
Suppose the Depublicans want as much representation and influence as 259. In the real world, each configuration would likely consist of sets of individual people, rather than sets of labels. But for the sake of illustration, it is clearer to strip these individuals of the complex of attributes that make them each unique and flatten them into one or two dimensions. For that reason, the number of configurations is far smaller than those that would be derived upon considering each individual as a unique entity under the formula provided in note 255, supra. Pursuant to such individualized consideration, there would be 3,025 distinct districting configurations, not all of which would be constitutional due to population disparity. 260 They dislike Configuration #3 the most because they will not win any representation. They like Configuration #1 more because they will control one district but dislike having no influence in any other. They prefer Configuration #2 the most because they will have one representative and will have some influence in District #2.

*p. 46*
Suppose the Remocrats also want as much representation and influence as possible. They would rank the configurations, starting with the most disliked, as depicted in Table 1: Configuration #1, Configuration #2, Configuration #3. They dislike Configuration #1 the most because they only win two districts, and they have no influence in District #3. They like Configuration #2 more because they only win two districts but they have some influence in District #3. They prefer Configuration #3 because they win all three districts.

*p. 46*
Imagine that the Remocrats make the first move in the redistricting process, and they eliminate, according to their true preferences, Configuration #1. The Depublicans go next and eliminate Configuration #3. The only configuration left is Configuration #2, and that is deemed to be the fairest redistricting scheme. 261 In this example, there is no benefit in gaming strategy. If each party picks as expected, the Depublicans have no incentive to game the system; they have achieved their preferred districting scheme. If the Remocrats try to eliminate Configuration #2 to spite the Depublicans, the Depublicans will eliminate Configuration #3 as before, and the "fair" redistricting scheme will be deemed to be Configuration #1, the most disliked configuration for the Remocrats. 262 (b) Partisan and Racial Minority Participants. Consider a slightly more complicated example, modified from the one introduced above. Further complicate the model by introducing the following racial interests: there are three African-Americans in the state, two of whom are Remocrats and one of whom is a Depublican. There are thus four types of people in Small Island: white Remocrats (R w ), white Depublicans (D w ), African-American Remocrats (R a ), and 260. For the sake of this example's simplicity and clarity, the possibility that the Depublican incumbent representatives, party leadership, and party membership each have different goals at stake in ranking districting configurations is ignored. Cf. Issacharoff, supra note 2, at 600 ("[T]here should be greater constitutional concern and, correspondingly, greater warrant for judicial intervention when political parties have joined together to squeeze the competitive juices out of the process. In the case of an inequitable gerrymander, dissatisfied voters are forced to seek recourse in a distorted political process. The expression of their political will may be hampered, but they will at least have an ally and a willing institutional voice through the minority political party. No such mechanism is easily available in the context of the bipartisan gerrymander. In such cases, the only relevant market actors are fully complicit in the cartelized political market."). If the preferences of these three Depublican interests might in fact be distinct, one of many ways to accommodate these differences would be to let each interest participate separately, with its own rankings. See infra Part V.A. 261. In this example, the same result would obtain if the Depublicans were to make the first move. 262. This Article considers gaming issues for configuration elimination in Part IV.B.1.c.ii and for the procedural model generally in Part V.B.

*p. 46*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL African-American Depublicans (D a ). As depicted in Table 2, ignoring order of districts, there are sixteen possible equipopulous districting configurations based on Small Island's partisan and racial composition.

*p. 47*
In districting, the Remocrats and the Depublicans could have the same strategy as above, making them indifferent as to race. But if only for the sake of assuming a unique ranking of configurations, assume that after preferring a districting configuration that gives them as much representation and influence as possible, Remocrats prefer configurations with the most African-American people in non-Remocrat-controlled districts, followed by configurations with the most total African-American majority districts, then configurations with the most and strongest African-American influence (that is, presence in districts), followed by configurations with the most total African-American Depublicans in Remocratic districts, then configurations with the fewest majority white Remocratic districts, then configurations with the most cross-party coalition African-American districts, and then configurations with the least white Remocratic influence. These preferences stem from African-American people's likelihood to identify with Remocratic values, even if they are not Remocrats. Thus, the Remocrats rank the configurations as set out in Table 2.

*p. 47*
Similarly, after applying the criteria from the previous section, Depublicans prefer configurations with more total white people in non-Depublicancontrolled districts, followed by configurations with the most white-majority districts, then configurations with the most white influence, then configurations with the most white Remocrats in Depublican districts. Then, Depublicans prefer configurations with the fewest majority African-American Depublican districts, followed by those with the weakest African-American Depublican influence, then configurations with the most and strongest white Remocratwhite Depublican coalition districts, and then those with the fewest majority African-American Remocratic districts. These preferences are shaped by the likelihood that white people identify with the Depublican Party. Thus, the Depublicans rank the configurations as listed in Table 2.

*p. 47*
The African-Americans prefer a districting configuration that gives them as much representation and influence as possible, with tiebreaking criteria being the more Remocratic majorities in non-minority-majority districts followed by the more districts with Remocratic influence in non-minority-majority districts, the more Remocrats in an African-American district, the more African-American Remocrats in an African-American district, and the more African-American Remocrats in a Depublican district, because Remocrats identify with African-American needs. They then prefer configurations with the most total African-American Depublicans in a Depublican district, followed by those with the most and strongest majority African-American Remocrat-white Remocrat coalition districts, and then configurations with the most and strongest majority African-American Depublican-white Remocrat coalition districts. Thus, the 2005] 1593 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING African-Americans rank the configurations as set out in Table 2. 263 Assume the Remocrats play first, the Depublicans play second, and the African-Americans play third. The Remocrats begin by eliminating Configuration #2; the Depublicans eliminate Configuration #13; and the African-263. In this example, no white group gets to participate. Putting aside constitutional considerations of permitting whites-as opposed to a minority group-to participate as a distinct group, cf. United States v. Carolene Prods. Co., 304 U.S. 144, 152 n.4 (1938) (noting but not ruling on the possibility that "prejudice against discrete and insular minorities may be a special condition, which tends seriously to curtail the operation of those political processes ordinarily to be relied upon to protect minorities, and which may call for a correspondingly more searching judicial inquiry"), there is no reason that such a group cannot be accommodated by this model. This illustration omits them merely because including a fourth participant would render the description of this illustration even more unwieldy.

*p. 48*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL Americans eliminate the most undesirable configuration not already eliminated, Configuration #9. These rounds continue until one districting configuration is left: Configuration #7. This configuration assures the Remocrats two districts, the Depublicans one district, and the African-Americans one district.

*p. 49*
(c) Generalization. The generalized form of configuration elimination anticipates that there are n participants. Each participant must rank the districting configurations, which can be done by ranking each individual districting configuration in order or, because the sheer number of districting configurations would in most realistic cases make such a ranking unmanageable, by giving a policy 264 by which one can automate the ranking of districting configurations. 265 To sort out the participants' varying preferences requires z turns, where z is the total number of possible districts minus one. 266 By eliminating their least preferred districting configuration until they end up with one configuration, the participants arrive at a scheme, which is deemed to be the fairest districting scheme. Although it is useful to describe configuration elimination by invoking a group of participants (sitting around a large table, perhaps) taking turns eliminating configurations, for all practical purposes-given the considerable number of participants and configurations-the elimination would have to be at least partially automated. The participants' rankings would be encoded into a computer, which would "take turns" on their respective behalves eliminating their least-desired configurations until it derived a unique districting scheme.

*p. 49*
i. Turn-Taking Considerations. In configuration elimination, the more participants, the more advantaged is the participant that takes the last turn as compared with participants that choose earlier. The reason for this advantage is that participants who choose later have more chance that other participants, in their turns, have already eliminated their less-preferred redistricting configurations; thus, later-choosing participants have more opportunity to get closer to their most-preferred redistricting configuration. 267 To make the process fairer in the context of configuration elimination, it is necessary to minimize the last-participant advantage. A simple way is to have 264. This policy would be a ranked listing of criteria, such as preferring a maximum number of representatives for the participant, followed by preferring competitive districts in those districts not controlled by the participant.

*p. 49*
265. If a participant is indifferent as to a certain set of districting configurations, the participant can randomly order those for elimination.

*p. 49*
266. See supra note 255 (noting the equation for determining the total number of possible district configurations).

*p. 49*
267. In an alternating game to divide desired items fairly, the participant who chooses first has a distinct advantage because that participant always gets to pick its most desired item first, which other participants might also want and will fail to get. See BRAMS & TAYLOR, supra note 250, at 27. But because participants are eliminating their least-desired districting configuration with every turn, participants who choose last have a distinct advantage.

*p. 49*
1595 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING the parties alternate taking first turns. 268 If, for example, the Remocrats and the Depublicans are the only participants and the Remocrats are chosen to play first, the Remocrats would choose, followed by the Depublicans, followed by the Depublicans taking another turn, followed by the Remocrats, followed by the Remocrats taking another turn, and so forth. For more than two participants, the same principle applies by reversing the order after every participant has had a turn. 269 If, say, the Remocrats, the Depublicans, and the African-Americans are the only participants and the Remocrats are chosen to play first and the Depublicans are chosen to play second, the Remocrats would choose, followed by the Depublicans, followed by the African-Americans, followed by the African-Americans taking another turn, followed by the Depublicans, followed by the Remocrats, followed by the Remocrats taking another turn, and so forth. 270 As to number of turns per participant, participants can either take an equal number of turns (at least until the last round, in which case some participants may get an extra turn) or a varying number of turns. For example, one may think it fairer for participants to get to take turns in proportion to their strength in the state to comport with proportional representation. On this view, by giving participants a number of turns proportional to their presence in the state, they have a proportional opportunity to choose the fairest district. 271 Relatedly, one might contend that turns should be allocated based on current legislative power, in accordance with the argument that to the political victor go the spoils. 272 But adopting either of these positions would be a mistake. The size of the group represented by a participant is already accorded significance and proportionality by virtue of the redistricting configuration itself, with large groups achieving better representational results on average through these models. Granting partici-268. See id. at 39. This "alternation, at least at an intuitive level, is more equitable because the [earlier] chooser[s] ha[ve] more opportunities to pick" more less-preferred districting configurations "under balanced alternation than [they] do[] under strict alternation." Id. at 45. 269. See id. at 45-47. 270. Steven Brams and Alan Taylor suggest that one can sometimes increase the fairness of an alternating game through a process they call "balanced alternation." Even when alternating taking first turns in choosing desired items, the first participant may have an advantage if the participants rank their items in the same order or if there is "a bigger difference between their . . . most-preferred items than between their . . . least-preferred items." Id. at 39-40. To improve the fairness, the participants alternate alternations: that is, when there are two participants, x and y, and x is chosen to go first, x chooses an item first, followed by y, followed by y taking another turn, followed by x, followed by y, followed by x, followed by x taking another turn, followed by y, followed by y, followed by x, followed by x, and so forth. See id. at 41. This improvement makes less sense in the context of configuration elimination, because if all participants rank the districting configurations the same way, that is great; they will all end up with their most-preferred districting configuration. Similarly, it does not matter if there is a bigger difference between the least-preferred configurations than between the most-preferred configurations, because unlike an alternation game that divides items with varying values, the redistricting process chooses a single districting configuration, and the value of the process is in the ultimate districting configuration, not in the values of the eliminated districting configurations. pants representing larger groups more turns would thus make their successes exponentially disproportional as compared with their numbers in the relevant jurisdiction.

*p. 51*
For instance, in the examples above, the Remocrats dominate Small Island: even in the worst possible scenario, they control two of the state's three districts. Giving them at least two-thirds of the turns in configuration elimination instead of one-half of them would significantly increase their impact on the redistricting scheme, beyond the impact they would have based on their numbers. Conversely, giving the Depublicans but one-third of the turns would diminish their chance to contribute to the redistricting scheme beyond the minimal impact they would have based on their numbers. Thus, the exponential effect of proportional input is unfair, and the proportional effect of equal input is fairer. 273 Each participant is thus treated the same whether or not they are similarly situated in terms of size. Although it is generally accepted that equal treatment is ideal when everyone is similarly situated, 274 it may be ideal to ignore varying situations in determining fairness. This is particularly true in configuration elimination, in which differences among participants' strengths are reflected in the ultimate districting scheme even if participants get an equal number of turns. Participants with little political support in a state cannot do much to strengthen their positions under any possible redistricting scheme by virtue of limited backing. By contrast, a participant with abundant political support can reflect that support under a variety of districting schemes. 275 ii. Limitations and Variations. In configuration elimination, by knowing other participants' elimination preferences, a participant can reorder its ranking of redistricting configurations. For instance, if a participant wants to eliminate a redistricting configuration early in the game-because it views that configuration as putting it at a distinct representational disadvantage-and if the participant knows or suspects that other participants likewise want to eliminate that configuration, the participant may rank that configuration as one to remove later in the game rather than earlier. By this strategy, the participant hopes that another interested participant will pick off its disliked configuration and that it can use its earlier turns to pick off less disliked configurations that other participants are less interested in eliminating early.

*p. 51*
Such strategic gaming behavior is risky. First, a participant may not truly 273. The counterargument is that equal input creates a disincentive for larger mainstream groups to permit smaller (possibly fringe) groups or interests to participate. The more that dominant groups think these smaller groups will negatively impact their ultimate control of districts, the less likely they will be to permit their participation.

*p. 51*
274. YOUNG, supra note 271, at 79-80. 275. Dominant participants may not perceive the procedural model as fair because they believe that they deserve more input into the ultimate districting configuration by virtue of their power. These strong participants, however, if operating behind the veil of ignorance, would have viewed the model as fair, because they would not know whether they would be strong, weak, or average-strength participants.

*p. 51*
1597 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING know how another participant will behave; if the other participants do not behave as expected, the disliked configuration may end up becoming the selected districting scheme. Second, if all participants have a similar strategy, they will all rank many of their disliked configurations for elimination late in the game and may end up with one of those configurations as the fairest districting scheme. The stakes of redistricting are high, and participants may not want to assume this risk.

*p. 52*
In addition to gaming concerns, the structure of configuration elimination permits a single participant to "hijack" a districting configuration. Namely, a participant can choose, in accordance with either true rankings or gaming behavior, to eliminate a particular districting configuration when its turn comes. That configuration thus has no chance of being selected as the state's actual configuration-even if the configuration is the preferred one of every other participant. To minimize this particular concern, the rules of configuration elimination could be modified to allow the resurrection of an eliminated configuration if, say, during the round in which the configuration was hijacked, a majority of participants votes to "rescue" it.

## Meta-Elections

*p. 52*
Another illustration of the model would sift through participants' varied districting goals by holding meta-elections-that is, elections of districting schemes for election of representatives. 276 (a) Partisan Participants. Recall that the Depublicans in Small Island have ranked the configurations in Table 1, starting with the most disliked: Configuration #3, Configuration #1, Configuration #2. Assuming evenly spaced rankings, they have therefore assigned Configuration #3 a ranking of 1, Configuration #1 a ranking of 2, and Configuration #2 a ranking of 3. The Remocrats have ranked the configurations, starting with the most disliked: Configuration #1, Configuration #2, Configuration #3. They have therefore assigned Configuration #1 a ranking of 1, Configuration #2 a ranking of 2, and Configuration #3 a ranking of 3. Configuration #1 therefore has a total ranking of 3, Configuration #2 has 5, and Configuration #3 has 4. Configuration #2 is the most preferred ranking based on the participants' varied preferences and would be elected as the fairest districting scheme. 277 This ranking scheme does not account for the possibility that the difference between the Depublicans' preference for Configurations #3 and #1 might not be 276. The current standard method of redistricting is a meta-election in the sense that state legislatures vote on a districting scheme for the election of representatives. The election of the districting scheme in the instant proposal, however, falls more squarely within the category of what is typically perceived as an election in the sense that the public has the possibility to participate more directly than under the current standard method of redistricting. See infra Part V.A.

*p. 52*
277. This method of election is known in the literature as a Borda count. See, e.g., Grant M. Hayden, Note, Some Implications of Arrow's Theorem for Voting Rights, 47 STAN. L. REV. 295, 302 (1995).

*p. 52*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL the same as the difference between their preference for Configurations #1 and #2. The same goes for the Remocrats. To account for these variations, these participants could assign a number of points to each configuration from a total of, say, 100 points. For instance, the Depublicans might assign 85 points to Configuration #2, 10 points to Configuration #1, and 5 points to Configuration #3, while the Remocrats might assign 55 points to Configuration #3, 30 points to Configuration #2, and 15 points to Configuration #1. There is therefore a cumulative assignment of 25 points to Configuration #1, 115 points to Configuration #2, and 60 points to Configuration #3. Configuration #2 is, by this measure, the most preferred scheme. 278 This scheme could easily be adjusted to accommodate negative rankings to express dislike of a scheme in addition to positive rankings, but the results would be indistinguishable under either an evenly or unevenly spaced ranking scheme. Incorporating negative rankings would merely shift the range of ranking numbers below zero, but would nonetheless mark the same differentiation between potential configurations. The only value of negative rankings in this context would be expressive, that is, to note active disapproval of particular configurations. But this structure would have no effect on the determination of the winning configuration.

*p. 53*
(b) Partisan and Racial Minority Participants. In our slightly more complicated situation, there are now white Remocrats (R w ), white Depublicans (D w ), African-American Remocrats (R a ), and African-American Depublicans (D a ), with possible districting configurations and configuration rankings as set out above in Table 2. Assuming evenly spaced rankings, for the Remocrats, Configuration #2 has a value of 1, Configuration #1 has a value of 2, and so forth, with Configuration #15 having a value of 16. For the Depublicans, Configuration #13 is assigned a value of 1 and so forth until Configuration #7 is assigned a value of 16. For the African-Americans, Configuration #2 is assigned a value of 1 and so forth until Configuration #7 is assigned a value of 16. Applying this data to the model yields Configuration #7 as the most highly valued and therefore the chosen districting configuration. 279 As set out previously, the participants can choose a districting configuration in a cumulative voting scheme. Because of the proliferation of configurations, it suffices to note here that analogous principles would work to generate the configuration with the highest cumulative score.

*p. 53*
278. Although Configuration #2 has been selected, and is thus the fairest districting scheme, in configuration elimination and both versions of meta-election, there is no a priori reason that the same result be reached in these variations. For elaboration on why it is acceptable that there can be different districting schemes emerging from varying implementations of the model, see infra Part IV.B. In an evenly spaced ranking scheme, the least preferred districting configuration of a participant is assigned a value of 1, the second least preferred configuration a value of 2, and the most preferred configuration a value of at most the total number of districts. 280 The cumulative ranking of a districting configuration is the sum of all of the participants' rankings of that configuration. The participants' varying preferences and goals are blended together to yield a unique districting configuration with the highest cumulative ranking, which is thus deemed to be the fairest one. 281 In an unevenly spaced ranking scheme, each participant assigns a value to each districting configuration, such that the sum of each participant's total assigned value to all of the districting configurations is a specified value m. 282 The cumulative ranking of a districting configuration is defined as above, and again, the configuration with the highest cumulative ranking becomes the state's districting scheme.

*p. 54*
There are no turn-taking considerations in meta-elections. In fact, metaelections favorably avoid turn-taking altogether, in that all of the participants effectively choose at the same time. There does exist, however, the related concern whether participants with larger followings should be allocated more votes than those with smaller followings. For the reasons discussed above, 283 280. The equation for the total number of districts is given supra note 255. 281. Cf. KENNETH J. ARROW, SOCIAL CHOICE AND INDIVIDUAL VALUES 4 (2d ed. 1963) ("If we admit meaning to interpersonal comparisons of utility, then presumably we could order social states according to the sum of the utilities of individuals under each . . . .").

*p. 54*
282. "Under cumulative voting, which is a procedure designed to give some control to minority shareholders, each shareholder gets a block of votes equal to the number of shares he owns multiplied by the number of directors to be elected. The shareholder may then cast his entire block for one candidate or may distribute his votes among any number of candidates in whatever proportion he desires." Edward S. Adams & Christian J. Lane, Constructing a Jury That Is Both Impartial and Representative: Utilizing Cumulative Voting in Jury Selection, 73 N.Y.U. L. REV. 703, 733-34 (1998).

*p. 54*
In the context of jury selection, Edward Adams and Christian Lane propose that a cumulative voting system replace peremptory challenges in jury selection. See id. at 714. Just as parties involved in redistricting want their state's delegation to be the most sympathetic to them, opposing counsel want to secure juries composed of individuals sympathetic to their respective goals. See id. at 708. Adams and Lane are convinced that the current system, with its peremptory challenges, has the effect of excluding a disproportionate amount of racial minorities from juries and therefore undermines the legitimacy of the justice system. See id. at 709-10 (reasoning that the justice system is undermined both when society mistrusts the system because some groups of people are regularly excluded from juries and when juries are not impartial because potential jurors are struck on the basis of race). They describe a system in which the prosecution and defense would assign from its block of votes varying numbers of positive or negative votes to each potential juror, with the jurors receiving the highest cumulative votes being seated. See id. at 745-46. If the sides cancel out each other's votes, according to Adams and Lane, juries would instead be randomly selected, which they think would avoid gaming the system and would promote expression of true preferences for jurors. See id. at 736, 753 ("The effectiveness of the system depends primarily on the strength of a group's interest in choosing a particular representative.").

*p. 54*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL each participant should be allocated an equal number of votes.

*p. 55*
There are nonetheless strong concerns for strategic gaming behavior in conducting meta-elections. Notably, participants have every incentive to rank low what they presume will be other participants' preferred districting configurations to decrease the chance that those configurations will be selected. This incentive is most exaggerated when there are unevenly spaced rankings because a participant can assign all of its points to its preferred districting configuration and zero points to every other one. One possible response is that a participant's choice to so game the meta-election system-and eliminate others' preferred districts-is as legitimate a theory by which to rank as is, say, protection of incumbents or symmetry. 284 Another response would be to minimize such strategies, at least in the context of unevenly spaced rankings, by threatening the participants with a randomly selected districting configuration in the event that they each game the system. This threat would work only if all participants had the same strategy-assigning points only to their most preferred districting arrangement-which would result in a tie between the various districting configurations, assuming that each participant prefers a different configuration. 285

## The Bare-Bones Procedural Model

*p. 55*
The above illustrations-configuration elimination and meta-elections-are merely two of numerous forms that the bare-bones procedural model can take. The illustrations may or may not be the most appropriate forms of the procedural model. They are, however, useful primarily as a starting point for elaborating on and analyzing the engineering of procedures that satisfy the following four criteria: (1) participants rank the potential districting configurations in their state, either by ranking each configuration individually or by establishing and applying a policy to rank the multitude of configurations; (2) these rankings are mapped onto a unique districting configuration; (3) the mapping incorporates each participant's rankings (and thus its goals and preferences) to some degree, thereby sorting through the varied and possibly conflicting goals of participants; and (4) the unique districting configuration is a member of the set of the potential districting configurations.

*p. 55*
The bare-bones model does not claim that there is a particular districting configuration that is the fairest, and in so doing, sidesteps any political choices about districting criteria. Rather, the model claims that the ultimate districting configuration derived is fair because the procedure used to map the participants' 284. This response has less force in the context of hijacking concerns for configuration elimination because one participant can remove a configuration preferred by all others. At least in this context, the participants' rankings are blended together to derive a districting scheme.

*p. 55*
285. That is precisely what Edward Adams and Christian Lane threaten in the context of jury selection, in which there are few (typically two) players. See Adams & Lane, supra note 282, at 750. The threat of randomization is obviously more threatening in a context with fewer players, like jury selection, because the possibility of a tie is more likely than in the context of redistricting with many players.

*p. 55*
1601 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING preferences to the configuration is fair, in that it blends the participants' differing preferences and goals in a collective compromise. 286 At this juncture, the model also does not strongly prefer any one mapping procedure over another. There may be good reason to adopt one method over another or some other variation of the model entirely.

*p. 56*
It may be worth noting that various potential mapping or selection procedures might yield differing districting schemes even when the procedures are applied to identical participant rankings of potential districting configurations. That, however, is beside the point. First, the procedural model is unconcerned with the substantive merits of the ultimate districting scheme and is instead concerned only that the procedure applied to derive that districting scheme is fair. Second, there may exist a set of procedures that are sufficiently fair, even if there are differences among them and each can produce different results. Third, there may well be a mapping or selection procedure that more fairly blends the participants' varying goals than the rest; deriving the best procedure and proving its worth is beyond the scope of this Article. This Article instead attempts to shift the debate away from determination of the best substantive measure or standard of fairness in districting toward discussion of the most appropriate procedural standards or measures of fairness in districting.

## Automation

*p. 56*
Given the vast number of districting configurations under consideration in the procedural model, computers would prove useful, if not essential, in implementing procedural fairness. 287 In fact, given the vast number of districting configurations and the likely high number of participants, 288 the most practical way to execute the procedural model would be to automate it using computer software. The bare-bones process described above in Part IV.B.3 can easily be modeled on a computer. The automated model would then accept as input each participant's redistricting rankings, the possible districting configurations, and the mapping procedure, from which it could execute the redistricting process in a straightforward manner to derive a unique and fair redistricting scheme.

*p. 56*
An automated version of the procedural model would avoid many of the problems that are described above as seemingly inherent in algorithmic implementations of redistricting on a computer, while retaining their discussed advantages. 289 First, having each participant encode a policy or ranking of districting configurations would tend to encourage political debate on districting goals generally. Second, anyone-including courts, the media, academics, and inter-

*p. 56*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL ested individuals-could review these encoded policies and rankings. Third, the concern that redistricting software is just as dangerous as old-fashioned redistricting is absent in the automation of the procedural model. Recall that the worry is that redistricting software is only as fair as the goals that underlie it, and redistricting goals-even those to be encoded in computer software-are often nefarious, harmful, misguided, or underinclusive. The procedural model posited is fair because it incorporates multiple and often conflicting goals expressed by varied participants to yield a compromise solution; therefore, its encoding in software is fair to the same extent. Fourth, the procedural model has no pretense to neutrality, the illusion of which is troublesome for other redistricting software. Rather, it brings together varying political goals of different participants by applying a neutral procedure to derive a unique districting scheme. Fifth, despite an encoding of the procedural model in software, participants can battle out redistricting each cycle under the goals important to them at the time by ranking the districting configurations decennially. Finally, the concern that political goals become opaque when encoded in technical form in software has less force when encoding the procedural model in software. Principally, the procedural model is a shell materially devoid of political choices. 290 Moreover, transparency is not a major concern for a software version of the procedural model, as it might be with other redistricting software, because different participants will check each other's interests. In sum, it is not only practicable to encode the procedural model as computer software, it is also desirable to do so.

## Enumeration of Districting Configurations

*p. 57*
Enumerating all of the possible districting configurations for a polity-which is necessary to instantiate the procedural model-is likely to be intractable, that is, "solvable in principle, . . . but requir[ing] so much time or space that [it] can't be used in practice." 291 The enumeration of all districting configurations is equivalent to the problem of dividing a set into numerous partitions. 292 Partitioning sets in this fashion is suspected to be intractable because no algorithm that uses a limited amount of time or space has been found to do it; however, no one has yet proved this problem to be intractable. 293 Even if the districting-configuration-enumeration problem is intractable, one can satisfactorily approximate the enumeration in a number of ways. First, one could use a probabilistic method, in which one would repeatedly "guess" a districting configuration at random until a sufficient number of unique configura-290. Nonetheless, by translating participants' preferences with regard to districting configurations to absolute rankings, the purposes underlying the varying preferences will be lost. 294 Alternatively, one could use a deterministic, but approximate, method to enumerate a sufficient number of districting configurations. 295 That is, a specified approximation algorithm would be capable of using established rules to enumerate a certain number of districting configurations, but not all of them, in a reasonable amount of time.

## MICHAEL SIPSER, INTRODUCTION

*p. 58*
One might think that an approximation is improper and undermines any fairness inherent in the procedural model. Of course, it would be preferable to enumerate all districting configurations with which to instantiate the redistricting process. But an approximate or probabilistic method, as described above, typically operates very similarly to the actual, desired method. 296 Therefore, approximate or probabilistic solutions are characteristically functionally equivalent to the possibly intractable ideal method. 297 D. WHY THE PROCEDURAL MODEL IS FAIR This model guarantees procedural fairness, in that all participantsessentially members of the public-get to contribute to the districting outcome. And all participants get to choose which districts they prefer based not on some possibly paternalistic, ostensibly neutral standard but based on the districting criteria important to that participant. Because all districting criteria have some political intent or effect, all participants will have their own peculiar preferences for particular districting criteria. It is inherently fair to allow participants to contribute to the redistricting process with their own voices, so long as the proper set of participants is involved. 298

## Traditional Models of Compromise

*p. 58*
Steven Brams and Alan Taylor present four measures by which to judge the fairness of models of compromise: "(1) proportionality, (2) envy-freeness, (3) equitability, and (4) efficiency." 299 Proportionality means that each party, among 294. See Altman, supra note 130, at 113 (suggesting the applicability of probabilistic methods generally).

*p. 58*
295. See id. (suggesting the applicability of approximate methods generally). 296. For a technical discussion of approximation algorithms, see SIPSER, supra note 291, at 333-35, and for a technical discussion of probabilistic algorithms, see id. at 335-47.

*p. 58*
297. Another possible way to enumerate districting configurations is to have each participant "nominate" a limited number of possible districting configurations. The slate of nominees would thus be a proper subset of all the possible districting configurations. Each participant would then rank the slate of nominees, and the mapping procedure would operate on the set of nominated configurations and rankings of them, rather than on rankings of all possible configurations. The downside of this approach is that each participant's nominated configurations will be somewhat biased in its own favor from the get-go, and the configuration most acceptable to the participants overall might thus not be located. The upside of nomination instead of enumeration is that participants have some incentive to nominate configurations that they think have a plausible chance of being accepted by the other participants, that is, of being ranked highly by them, and they might not proffer extremely biased configurations.

*p. 58*
298. x parties to a compromise, thinks it is getting 1/x of the total value involved. 300 Envy-freeness means that no party to a compromise is willing to surrender what it has received in exchange for any other party's share. 301 Envy-free compromises are always proportional, but not all proportional compromises are envyfree. 302 Equitability means that each party thinks it receives the same fraction of the total value. 303 Equitability is different than envy-freeness, in that parties to an equitable compromise may be envious of what the other parties have received because of different valuations for items at play in the compromise. An efficient compromise is one in which there is no other allocation that is subjectively better for some party without being worse for another party. 304 Three of the most prominent methods of compromise that can embody at least some of these characteristics are alternation, cakecutting techniques, and adjustedwinner systems. 305 (a) Alternation. Alternation involves parties taking turns picking items; it assumes that parties rank the involved items honestly. 306 The basic case of alternation involves parties picking items they would like to keep during their turns. Configuration elimination is a variation of alternation. A party can have the strategy of guessing or knowing the other parties' preferences and not choosing items early in which the other parties have little interest. 307 However, this strategy is a gamble because one side may not be sure of the other sides' preferences and the other sides may game in return out of spite if they guess that the one side is gaming. 308 All in all, it seems intuitive to assume that rational participants "will . . . [n]ever choose their least-preferred item [and will n]ot 'waste' choices on a desired item that they know will remain available and, hence, can be chosen later." 309 As for measuring the fairness of alternation, Brams and Taylor note that it can cause envy because of both the item-by-item choices and the difficulty of judging equitability from rankings. But it is efficient when the participants are sincere in their choices. 310 Yet it is hard to apply envy-freeness and equitability to the configuration-elimination version of the redistricting model because in each round, participants do not pick items to keep; rather, they pick districting configurations to eliminate and the ultimate goal is to devise an ultimate districting configuration. One can, however, judge the efficiency of the redistricting process; just as the basic case of alternation is efficient if parties rank sincerely, so too is the ultimate redistricting configuration efficient if the parties rank sincerely. It is efficient because there is no other redistricting configuration that would make someone better off without making anyone worse off. Efficiency implies the satisfaction of "classical utilitarianism, which asserts that goods should be distributed so as to maximize the total welfare of the claimants." 311 Among the strongest benefits of alternation are that it is simple to comprehend and that it does not require much from participants in the way of preparing their preferences. 312 A weakness of alternation, as discussed above in the context of configuration elimination, is that there is a strong first-chooser advantage; the order of turn-taking by the parties can matter significantly. 313 Another weakness, at least in the context of redistricting, is that a participant can eliminate a configuration with which all other participants would be satisfied. Part IV.B.1.c above discusses how to minimize these disadvantages to make alternation more robust in redistricting.

*p. 60*
(b) Cakecutting. Cakecutting procedures involve a generalized version of divide-and-choose. When there are two participants, one participant divides a "cake" and the other participant chooses the preferred piece. There are many sophisticated algorithms to cut cakes among more than two participants fairly. 314 Nonetheless, cakecutting procedures are not suitable to the redistricting problem because no participant gets to "keep" a district. It seems nonsensical to have participants each choose one district or districting configuration, especially when there is no necessary one-to-one mapping between the number of participants and the number of districts, meaning not everyone could easily get a turn. 315 (c) Adjusted-Winner Systems. In an adjusted-winner system, each participant distributes a set number of points, say 100, between all the items at play to indicate how much that participant values each item. 316 Each participant is 311. YOUNG, supra note 271, at 9. 312. See BRAMS & TAYLOR, supra note 250, at 34. 313. See id. 314. See, e.g., id. at 53-65; BRAMS & TAYLOR, supra note 305, at 57-62. 315. One could create a one-to-one mapping by allowing participants to choose multiple districts or portions of districts depending on their number in relation to the number of districts, but ultimately this solution ignores the fact that the ideal goal of redistricting is to maximize the fairness of districting configurations, and thus, representation, not to give each participant a district, or subdistrict, with which to play.

*p. 60*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL initially assigned the items that they ranked higher than the other participants. The one with the lower total is given the items on which the parties tied. Then, one transfers items (or fractions thereof) among the parties until their point totals are equal. Items are picked for transfer based on the ratio of the points of the initially assigned winner to the points of the initially assigned loser-those items with lower ratios are transferred. If items tip the balance too much, they must be split. When there are more than two parties, it is mathematically provable that one can satisfy at most two of envy-freeness, equitability, and efficiency. 317 This procedure is also not suitable for the redistricting problem. Only one districting configuration can ultimately emerge victorious, and it makes little sense to have participants assign point values to possible individual districts instead, because that dissects the problem into unnatural pieces that when fit together will far from ensure fairness. 318 A variation that is sensible in the redistricting context is the cumulative-voting scheme described above as a meta-election.

## Fairness Above and Beyond the Previous Proposals

*p. 61*
The procedural model presented in this Part provides fairness more consistently than the standards and methods discussed in Parts II and III. 319 First, it acknowledges the role that politics intrinsically plays in redistricting, while ensuring that no political participant can dominate the redistricting process beyond the participant's actual representation in a jurisdiction. The model allows participants to rely on the political-or apolitical-criteria that matter to them through its ranking of districting configurations. Unlike "objective" substantive criteria with strong, possibly unintended political effects, such as compactness, competitiveness, and symmetry, the procedural model allows participants to choose their own political or "apolitical" criteria for redistricting. The model then allows each participant input into the redistricting process through invocation of the procedure mapping the participants' rankings onto a unique districting scheme. 320 Second, the model is participatory in a way that no other approach-either the Supreme Court's or scholars'-is because it enables the public-effectively, 317. See id. at 84. 318. See supra note 235 and accompanying text. 319. It is admittedly tautological to define fairness as procedural and then to argue that the procedural scheme presented herein is fair. Nonetheless, I am using the label of "fairness" to represent the more complex notions, argued for above, that a good redistricting scheme and process cannot fail to account for political criteria, should not ignore significant political interests, should incorporate the interests of various public groups and individuals, but at the same time should not put judges in the position of having to rely consequentially on politics to judge the goodness of redistricting schemes.

*p. 61*
320. With regard to the specific models of configuration elimination and meta-elections, the game does not allow strong participants to bully weaker ones as could happen under unconstrained partisan redistricting or through bipartisan commissions, because each participant gets an equal number of input or turns.

*p. 61*
1607 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING multifarious groups, interests, and individuals-to partake in the redistricting process. By virtue of participation, the public can impact the way redistricting happens. Because every districting configuration affects the characteristics of representation, these participants can influence the rules of governance, thus enabling true democracy via collective public compromise. Moreover, the model is participatory in the sense that it can generate a public conversation about the structure of electoral politics, most specifically in redistricting criteria.

*p. 62*
Third, it is important to have a single method or standard that integrates the various interests at play in redistricting. Even though race, partisanship, and other interests are qualitatively different issues, 321 it is unfeasible to have one redistricting method that produces fair results as far as race is concerned, another method that produces fair results as for partisanship, and yet another method that produces fair results for other interest groups or incumbents. Each state ultimately needs a single redistricting plan. Without a single integrative method or standard, it is unclear how to merge the results of the varying fairness methods for different interests into a single redistricting configuration.

*p. 62*
One might object-albeit conceding that compromise is a good measure of fairness in the redistricting context-that these participants instead should sit down and hammer out a substantive compromise through interaction. This interaction would be preferred, on this argument, because the participants would create value through creative solutions, instead of just trying to capture it. 322 Ideally, a substantive compromise instead of a procedural one might be preferable. But in the realistic redistricting context, it offers few guarantees of fairness. With a proliferation of public participants, structuring a substantive compromise in the traditional sense collaboratively becomes impossible. Second, there is little assurance that the interests of all parties to the compromise would be considered. 323 Ultimately, the model's promises of procedural fairness provide a more realistic assurance of fairness in redistricting.

*p. 62*
In any event, there is a real sense in which a substantive compromise is achieved by employing the procedural model. When each participant ranks the possible districting configurations and a procedure is used to map these rankings onto a unique redistricting scheme, the participants have in effect specified an 321. See Issacharoff, supra note 40, at 1677 ("Partisan affiliation is not an immutable characteristic nor is such an affiliation irreversible once chosen . . . .").

*p. 62*
322. Creating value and capturing value are two important elements of successful negotiations. See generally ROBERT H. MNOOKIN ET AL., BEYOND WINNING: NEGOTIATING TO CREATE VALUE IN DEALS AND DISPUTES (2000).

*p. 62*
323. The process of redistricting, as envisioned in this Article, is thus distinct from jury deliberations. "Jurors do not merely vote individually on a verdict. The verdict is a collective democratic act. Jurors try to reach agreement through discussion, something that can involve hours of face-to-face interaction . . . . [The jury] engages in the process of compromise . . . ." Gerken, supra note 214, at 1152-53 (footnotes omitted). One might question whether substantive compromise, were it practical, is ideal in the context of redistricting. Power relations and implicit biases can skew substantive bargaining. These problems are avoided when the participants "meta-redistrict" by creating a redistricting scheme via a procedure that is one step removed from direct participant interaction.

*p. 62*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL overlapping terrain between the positions and interests of all the participants from which the redistricting scheme is chosen. 324 What the participants have done by accepting and applying the procedural model is achieve what Cass Sunstein labels an "incompletely theorized agreement," by which the participants converge-and thus agree-on a redistricting result despite not necessarily having concurred on a uniting fundamental principle for the resulting scheme. 325 Sunstein argues that incompletely theorized agreements are characteristic of governance decisions in a pluralistic society. 326 It might even be said that the procedural model for redistricting is an archetype of these agreements in the sense that the model is driven, in substantial part, by incorporation of the various participants' differing visions of representation and fairness into a redistricting scheme.

## E. A DESCRIPTIVE PROCEDURAL MEASURE OF FAIRNESS

*p. 63*
1. A Judicial Measure Justice Stevens, concurring in Karcher v. Daggett,327 suggested that a procedural descriptive measure might be fruitful for adjudicating partisan gerrymandering cases:

*p. 63*
A procedural standard . . . may also be enlightening. If the process for formulating and adopting a plan excluded divergent viewpoints, openly reflected the use of partisan criteria, and provided no explanation of the reasons for selecting one plan over another, it would seem appropriate to conclude that an adversely affected plaintiff group is entitled to have the majority explain its action. On the other hand, if neutral decisionmakers developed the plan on the basis of neutral criteria, if there was an adequate opportunity for the presentation and consideration of differing points of view, and if the guidelines used in selecting a plan were explained, a strong presumption of validity should attach to whatever plan such a process produced. 328 Justice Powell subsequently agreed in Davis v. Bandemer 329 that an inquiry as to the process by which redistricting occurs should inform whether unlawful 324. The model of configuration elimination is an intuitive example of this, because the participants take turns eliminating their least-preferred districting schemes until the commonly least objectionable districting scheme remains.

*p. 63*
325. See, e.g., Cass R. Sunstein, Incompletely Theorized Agreements, 108 HARV. L. REV. 1733, 1735-36 (1995) 330 As Justices Stevens and Powell thought it useful with regard to partisan gerrymandering claims, the procedural model can serve also as a standard against which to evaluate how fair an actual districting scheme is in varied contexts. A descriptive procedural measure is of obvious use to both courts and scholars. There are two ways to use the model as a standard.

*p. 64*
Once a mapping procedure has been chosen to convert the participants' rankings into a unique districting scheme, one can determine how fair a districting scheme is by recreating the instantiation of the procedural model for that jurisdiction with the relevant participants. 331 In the case of configuration elimination, one keeps track of the stage at which each districting configuration is eliminated. Upon completion of the recreation, one looks at what stage the actual districting scheme was eliminated. If it was eliminated too far away from the fair districting scheme, the actual districting scheme is unfair. 332 For metaelection, one looks at the distance of the actual districting scheme from success in the vote. The further it is from success, the more unfair it is. Requiring a recreation of the mapping suffers from the problem that it is post hoc and therefore difficult to recreate accurately. It is difficult, if not impossible, for the participants to submit their rankings as they would have chosen them at an earlier time. In the most innocent scenario, the rankings might no longer accurately reflect what the participants would have chosen at the time of redistricting. Viewed through a more cynical lens, the participants might try to change their rankings to manipulate the mapping's outcome to affect a court's judgment of fairness. 333 A preferable standard would instead measure only the procedure actually used against an ideal procedure. That is, the greater the deviation of the procedure actually used for redistricting from the ideal, the less likely it is that redistricting was fair. The distance between procedures can be measured in a 330. See id. at 173, 175 (Powell, J., concurring in part and dissenting in part) (maintaining that "the nature of the legislative procedures by which [an] apportionment law was adopted" is a primary factor to examine when adjudicating partisan gerrymandering claims, and finding it troublesome that "[t]he legislative process [at issue] consisted of nothing more than the majority party's private application of computer technology to mapmaking").

*p. 64*
331. Choosing the relevant participants in a recreated redistricting process imparts a judgment call to the recreator. One can imagine ex ante rules to determine who should have participated. For instance, in a suit to determine if there was improper partisan gerrymandering, possibly only political parties should get to participate in the recreation, while in a suit to determine if there was racial gerrymandering, possibly only political parties and minority groups should get to participate. Alternatively, all relevant participants might get to play in the recreation regardless of the evaluative context.

*p. 64*
332. Like choosing who gets to play in the recreated districting game, choosing how far away is too far calls for the proper exercise of judgment under the circumstances. Cf. Vieth v. Jubelirer, 541 U.S. 267, 344 (2004) (Souter, J., dissenting) ("[T]he issue is one of how much is too much, and we can be no more exact in stating a verbal test for too much partisanship than we can be in defining too much race consciousness when some is inevitable and legitimate.").

*p. 64*
333. Although the participants may have ulterior motives in their post hoc rankings, they might not be able to derive their preferred districting outcome because of the complexity involved in the mapping process.

*p. 64*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL number of dimensions, including which participants were involved and how participants' interests were accommodated into the ultimate districting configuration. 334 Both of these suggested standards do not look at the actual districting scheme in place; rather, they inquire whether the procedure used to generate the scheme was just. The use of a procedural standard to measure fairness in the judicial context provides an improvement over the fairness methods and standards presented in Part III, because it allows judges to rule minimalistically by avoiding implementation of any substantive models of democracy. 335 A procedural standard leaves to the participants in redistricting-more democratically accountable actors-the crucial political decisions required to generate a redistricting scheme.

*p. 65*
These descriptive measures provide a useful way for courts to evaluate claims of partisan gerrymandering, racial gerrymandering, minority vote dilution, and all sorts of other districting manipulation. 336 For instance, in judging a claim of minority vote dilution, a court might measure whether the relevant minority groups had the requisite opportunity to participate in redistricting, which would arise only if the minority group qualified for participation in the redistricting process, as outlined below. If a court would, by contrast, choose to measure minority vote dilution by recreating the instantiation of the procedural model, it could judge whether the actual districting scheme under scrutiny would have come remotely close to being deemed the fairest districting scheme. To be fair under this measure, the actual districting scheme would not have to match the procedural model's outcome; that would be statistically improbable to satisfy. Rather, the actual districting scheme would have to have fared reasonably well in the redistricting process, in that it would have been eliminated fairly late in configuration elimination, would have ranked fairly highly in a meta-election, or would have ranked highly in other procedural mappings. In evaluating a claim for minority vote dilution, a court would need to choose, for example, whether to include all theoretical participants in a recreation or rather to include some limited subset of them, which would include the minority group at issue.

*p. 65*
The proposed descriptive measures are facially constitutional. In recent years, the Court has questioned the constitutionality of the use of racial information to determine districts. 337 The Court has struck down redistricting schemes that 334. To the extent that participants are ranking districting configurations by establishing a policy of various ranking criteria, courts might want to measure the constitutionality of the ranking criteria themselves. That question is not as fraught with politics as is the Supreme Court's current measurements of substantive fairness.

*p. 65*
335. See supra text accompanying notes 230-31. 336. Were these descriptive measures to be endorsed by courts, the question would nonetheless remain whether courts would want to adjudicate all gerrymandering claims. With the descriptive measures described herein, courts would have the tools to make judgments on nearly any claim, such as whether left-handed people were treated unfairly in the last redistricting, but it might be imprudent or unjustified to allow all claims of gerrymandering on any basis to be justiciable.

*p. 65*
2005] 1611 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING have relied excessively on race, and has upheld schemes that did not rely too much on race. This model, which does not, would be facially constitutional under the Shaw line of cases. Even if a particular mapping scheme took race into account-either by allowing racial groups to participate in redistricting or by allowing any participant to account for race in its rankings-it would be only one of many factors influencing the outcome of the process. 338 A particular redistricting process, if it takes only race or too much racial information into account, however, may be unconstitutional under Shaw and its progeny. 339 Not only are these descriptive measures facially constitutional, but they also can claim constitutional support. First, Article IV, Section 4 requires that "[t]he United States shall guarantee to every State in this Union a Republican Form of Government." 340 By definition, a republican form of government is one that governs through representatives. 341 It would seem that an effective and manageable way-if not the most effective and manageable way-to ensure that states maintain a republican form of government would be to secure to the various interests in a state a voice in their representation, and hence their districting scheme that affects their representational capacity. 342 Moreover, the Fourteenth Amendment guarantees individuals "equal protection of the laws." 343 In the context of voting rights-the rights preservative of all other rights-it seems constitutionally problematic when the voting strength of certain groups in a state is minimized or is extinguished; if these groups' interests are not incorporated in some fashion into the state's districting scheme, their voting strength 338. Cf. Issacharoff, supra note 40, at 1700 ("In light of Shaw, advocates of enhanced minority representation may actually find protection in computer-mandated redistricting. Shaw seems to find the exclusive reliance on race to the detriment of all other factors to be constitutionally offensive. One reading of this case would allow for minority-representation concerns to be operative so long as they are not the exclusive focus of the redistricting process. Thus, if the minority representation concerns were coupled with other state objectives in a global redistricting strategy, the minority representation concerns could emerge constitutionally intact.").

*p. 66*
339. Similarly, if a participant used unconstitutional criteria to rank districts, it would be unsurprising if a federal court held the particular redistricting process in which those criteria were used to be unconstitutional. 340. U.S. CONST. art. IV, § 4. 341. See WEBSTER'S NINTH NEW COLLEGIATE DICTIONARY 1001 (1998) (defining a "republic," inter alia, as "a government in which supreme power resides in a body of citizens entitled to vote and is exercised by elected officers and representatives responsible to them and governing according to law"); cf. Robert G. Natelson, A Republic, Not a Democracy? Initiative, Referendum, and the Constitution's Guarantee Clause, 80 TEX. L. REV. 807, 822-23 (2002) (positing that there is general agreement that a republic is based on majority rule, absence of a monarch, and the rule of law).

*p. 66*
342. "A districting scheme so malapportioned that a minority faction is in complete control, without regard to democratic sentiment, violates the basic norms of republican government . . . . Constitutional standards under the Republican Form of Government Clause are ill-developed, but surely a government is not 'republican' if a minority faction maintains control, and the majority has no means of overturning it." McConnell, supra note 32, at 105-06. Nonetheless, the Supreme Court has refused to rely on this constitutional provision in adjudicating voting rights claims because it considers it to lack "judicially manageable standards." Baker v. Carr, 369 U. S. 186, 223 (1962).

*p. 66*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL will effectively be undermined. 344 In sum, the procedural model presented in this Article seems suitable for the Court to use as a baseline for measuring a redistricting scheme's fairness. Since the Supreme Court declared voting rights cases to be justiciable, the Court has struggled to devise a standard of fairness by which to judge districting configurations, particularly an automated, relatively objective, and manageable standard. In fact, in its most recent decision on voting rights in Vieth v. Jubilerer, a majority of the Supreme Court could not presently locate any standard of fairness by which to adjudge partisan gerrymandering. 345 The procedural model provides a baseline for the Court to advance its jurisprudence without getting mired in political questions unsuitable for courts. 346 The Court need only judge the redistricting process, not offer its own political criteria for redistricting, to determine the procedural fairness of a districting scheme.

## An Experimental Dynamic Measure

*p. 67*
There is another way the procedural model can be used descriptively. Using it as a dynamic institutional structure that offers a means of comparing districting processes over time and across jurisdictions, democratic experimentalists can collect data points on who participated in particular instantiations of the procedural model and which substantive values mattered to the various participants. 347 Use of the model thus provides a vast amount of information about local electoral values, which can then be analyzed and compared across jurisdictions and over time, both by governmental bodies and by the public. This information can then ultimately be used to generate more robust redistricting procedures, based on classifications of the conditions under which various substantive districting criteria might be most appropriately employed, determinations of which public bodies and individuals should be included as participants in the procedural model, and measurements of other aspects of the various instantiations of the procedural model. The public would effectively be able to judge the success of their respective redistricting procedures through access to these metrics, which could serve to keep various jurisdictions accountable to 344. Cf. White v. Regester, 412 U.S. 755, 765-66 (1973) (noting that the Court will "entertain[] claims that multimember districts are being used invidiously to cancel out or minimize the voting strength of racial groups" under the Equal Protection Clause when "the political processes leading to nomination and election were not equally open to participation by the group in question-that its members had less opportunity than did other residents in the district to participate in the political processes and to elect legislators of their choice").

*p. 67*
345. See supra notes 114-23 and accompanying text. 346. At the very least, it would make it possible for the Supreme Court to identify arguably extreme cases of redistricting without any "politicized" investigation into the fairness of the resulting redistricting schemes.

*p. 67*
347. See Michael C. Dorf & Charles F. Sabel, A Constitution of Democratic Experimentalism, 98 COLUM. L. REV. 267, 283-89, 314-24 (1998) (calling for the collection and pooling of political experimental information between the units of the U.S. government-principally, the federal branches and the states-to strengthen American democratic practices).

*p. 67*
1613 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING their electorate. 348 All in all, the procedural model could serve as more than a one-shot generator of districting schemes; the debates and data the model would generate could also be used for collective learning and for making electoral structures more robust.

## V. FURTHER AVENUES OF RESEARCH

*p. 68*
This Part addresses in turn a number of further avenues of research raised by the introduction of the model, namely who gets to participate in redistricting, possibilities of gaming the model, the opportunity costs created by the model and whether they are justified, and the relationship between procedural fairness and substantive fairness in this context.

## A. WHO GETS TO PARTICIPATE

*p. 68*
The generalized bare-bones model allows any number of participants. 349 There are no theoretical limits on who can play-political parties can, of course, participate as can other interest groups, such as racial, ethnic, professional, and environmental groups. In addition to groups, individuals can also participate. One can imagine each person living in a jurisdiction as a participant. Clearly, most state denizens do not care or know enough to provide detailed rankings of districting configurations, but total participation is conceivable under the procedural model, the benefit being that each individual could rank the districts with which he or she is interested in associating.

*p. 68*
The flexibility of who can participate is also useful when there are splits among members of a group. For instance, political party leaders may have different, possibly opposing, views as to how to district from the incumbents of their party. And both of them might have different views than the party membership. As discussed above, incumbents typically care about keeping their districts intact, packed with party members and supporters, to benefit from name recognition and fundraising. Conversely, political parties are usually interested in maximizing the number of districts they can win, which requires spreading out party members and supporters. 350 Under the procedural model, the incumbents, the party leadership, and the party members can participate separately to protect their respective interests. 351 Similarly, pro-gun Democrats and anti-gun Democrats might participate separately, as might pro-life Democrats and prochoice Democrats. There is no theoretical limit on how many interest groups 348. See id. at 288-89. 349. For configuration elimination, it would not make sense to have more participants than districting configurations, because some participants would not get a turn to participate in selection of a districting scheme.

*p. 68*
350. See supra Part III.B.5. 351. Who gets to speak for specific groups is a decision that takes place outside of the model by which groups will self-define and somehow choose a representative. To the extent that ostensible members of a group are unhappy with their representatives, they can seek separate participation in the procedural model.

*p. 68*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL can participate. There is also no material negative impact on the application of the procedural model to have more participants rather than fewer, or vice versa.

*p. 69*
Despite the flexibility of the model and although the model favorably avoids making any overarching political determinations as to districting criteria, it forces a substantive determination of who gets to participate in redistricting. Admittedly, determining the participants is a political assessment, and deciding who gets to participate is a critical question with political effect. Because the procedural model must offer some method or standard by which to determine who gets to participate, it obviously cannot remain entirely agnostic with regard to political choices. 352 In that sense, there is some resemblance between the complications inherent in deciding who gets to participate and the complications afflicting the existing standards of redistricting, discussed above-political decisions remain unavoidable, even if they are pushed to a different place in the procedural model. Nonetheless, embedding substantive judgments in the procedural model hardly disqualifies the utility of the process in addressing redistricting questions.

*p. 69*
Although the ultimate resolution of this difficult question of fairly determining who gets to participate is beyond the scope of this Article, one can nevertheless have objective standards in deciding who gets to participate. The same way that in many states, a party can place its candidate on the ballot only if the party has a certain quantum of support, 353 one can have a rule that an individual or group can participate in redistricting only if that entity can demonstrate a certain quantum of support in the state for that entity's participation. A simple standard would require more than a not-too-burdensome number of petition signatures by state residents or registered voters, 354 a somewhat smaller number of registered group members, 355 or sufficient past participation as a participant in redistricting. 356 Admittedly, these participation access requirements are somewhat imprecise proxies for gauging the support garnered by a 352. Cf. Paul Brest, The Substance of Process, 42 OHIO ST. L.J. 131, 131, 142 (1981) (emphasizing the unavoidable connection between a legal process and substantive values underlying that process).

*p. 69*
353. See, e.g., Munro v. Socialist Workers Party, 479 U.S. 189, 190 (1986). "The state interest in access restrictions is primarily framed in terms of preventing voter confusion, deterring frivolous candidacies, and promoting political stability." The Appleseed Center for Electoral Reform & The Harvard Legislative Research Bureau, A Model Act for the Democratization of Ballot Access, 36 HARV. J. ON LEGIS. 451, 453 (1999) [hereinafter Model Act].

*p. 69*
354. Cf. Model Act, supra note 353, at 458, 460 (proposing a requirement of signatures of a mere 0.1% of a state's registered voters to gain a space on the state's ballot). It arguably makes sense for the number of signatures to be lower for individuals than for groups. It may also make sense for some groups or individuals, such as major political parties, minority racial groups, and incumbents, to participate automatically without collecting signatures to avoid wasting time and money collecting petition signatures when these groups or individuals could easily meet the requirement.

*p. 69*
355. It is suggested that the number of required petition signatures be greater than the number of registered members because there is more difficulty in getting individuals to commit to group membership than to affix a signature for a petition. Specifically, some details are in order with regard to the implementation of qualification by petition. 358 It is preferable to allow an individual to sign more than one group's or individual's petition, in light of the signee's presumed multitude of interests. 359 To campaign for signatures, groups and individuals should have an ample but not unlimited period of time, such as up to one year, directly preceding redistricting. 360 To level the playing field between moneyed and less moneyed interests, the state might make limited funds available to aspiring participants for advertising themselves and securing the requisite number of signatures. 361 There are well-documented concerns that the number of petition signatures gathered typically reflects how much money a petitioner has to spend on collecting signatures rather than the breadth or depth of support for the petitioner. 362 As one stab at a solution to this problem, the state itself might take a ability to pay, which is often not an accurate indicator of public support." Id. It would be inadvisable to require more than a minimal filing fee to partake in redistricting for analogous reasons.

*p. 70*
357. Satisfaction of these requirements (or failure to do so) might instead be an indicator of individuals' general indifference, the amount of money possessed by the aspiring participant, interest group dynamics, and the like. See, e.g., Richard J. Ellis, Signature Gathering in the Initiative Process: How Democratic Is It?, 64 MONT. L. REV. 35, 37 (2003) ("Gathering signatures has increasingly become a business, and like any other business it is run for profit.").

*p. 70*
358. It should be considered fraudulent to mislead or misrepresent the nature or interests of an aspiring participant. Prohibition of such fraudulent petitions would not run afoul of the First Amendment. Cf. McIntyre v. Ohio Elections Comm'n, 514 U.S. 334, 343-44 (1995) (holding that a state's prohibition on distribution of anonymous campaign literature could not be justified under the First Amendment as a means of preventing fraud or lies, because the statute was not limited to those situations).

*p. 70*
359. See infra text accompanying note 367. 360. Cf. Ellis, supra note 357, at 45 (listing extended time periods in some states for filing ballot initiatives).

*p. 70*
361. Securing ballot access is not typically cheap for a third-party candidate. Eugene McCarthy spent eighty percent of his campaign funds in 1976 to obtain ballot access. See Issacharoff & Pildes, supra note 96, at 687. More recently, Ross Perot expended eighteen million dollars on securing ballot access. See id.

*p. 70*
362. See, e.g., Ellis, supra note 357, at 53 ("Among the groups [in Oregon] for whom money presented no obstacle, the success rate in qualifying for the ballot approached one hundred percent."). That is not to say that opening up the petition process to moneyed interests does not serve a democratic function. As Richard Ellis observes, in the context of petitioning to put an initiative on the ballot:

*p. 70*
In opening up the initiative process to issues and interests that could not otherwise play . . . paid petitioners arguably democratize the initiative process, making it more inclusive. Banning paid signature gatherers penalizes those individuals or groups who possess money but lack spare time, while advantaging those short on money but long on time. But why should individuals with lots of spare time be privileged over people with lots of money? Are those groups who have difficulty raising money but no trouble harnessing a cadre of fanatics any more deserving of a place on the ballot than those who have an abundance of cash but lack fervent enthusiasm? Are a small group of denture makers less virtuous than a large band of religious zealots? From this vantage point, the rise of paid petitioners and professional signature gathering firms promotes democracy by increasing the involvement of a wider diversity of groups.

*p. 70*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL role in canvassing individuals to measure the support for an aspiring participant. States might allow an aspiring participant to advertise freely but require the petition to be left with state officials, to whom individuals would come to sign it. 363 In this scenario, individuals would have to take affirmative steps to sign the petition, which is arguably more indicative of support for the aspiring participant than is the signing of a petition conveniently placed before an individual, possibly by a paid signature collector. Another proposal along similar lines, although more streamlined, is to require all aspiring participants to register with the state, after which individuals would either come to a polling station to choose which subset of these aspirers they support or individuals would be sent a list of these aspirers from which they would choose the subset they support. Under either proposal, the requisite number of signatures or votes would obviously need to be lowered dramatically, because fewer individuals would take the applicable affirmative step. 364 Even though "ballot access" and like standards are objective, they are still fraught with political choices. However, the new problem of political choices impacting who gets to participate is more manageable than the old problem of the political choices present in the redistricting criteria that it would replace. That is, the procedural model "solves" an extremely difficult, if not impossible, problem by replacing it with a somewhat easier one. It is more manageable because it is more susceptible to judicial monitoring without wading too far into the political thicket, 365 at least in the same sense as with ballot access Id. at 58. Nonetheless, it remains true that aspiring participants who are moneyed interests can buy access through petitions even when their support is neither broad nor deep, while aspiring participants who are supported in full by volunteers and garner the requisite number of signatures demonstrate that their support is broad and perhaps deep. See id. at 58-59.

*p. 71*
A related concern is that, as in the context of ballot access, diffuse groups are less likely to participate (or attempt to participate) in redistricting because of the greater difficulty in organizing as compared with other groups. See, e.g., Ethan Klingsberg, Contextualizing the Calculus of Consent: Judicial Review of Legislative Wealth Transfers in a Transition to Democracy and Beyond, 27 CORNELL INT'L L.J. 303, 312 (1994).

*p. 71*
363. Cf. Ellis, supra note 357, at 94-95 (proposing something similar in the context of the initiative process, although leaving the petition with county officials).

*p. 71*
364. Cf. id. at 95 (making a similar point in the context of the initiative process). One other potential problem worth noting is that it might be more difficult to collect signatures some years, whether, for example, due to inclement weather or terrorist threats that keep people home. A solution to this problem would be to replace a requirement of a determinate number of petition signatures with a relative one. That is, to qualify as a participant, an aspirer must derive some set fraction of the signatures collected by some reference point-possibly, the aspirer with the highest number of petition signatures or the aspirer with the average number of petition signatures.

*p. 71*
365. One possible problem could be a majority party in power "locking up" access to participation by setting ballot access requirements so that only it qualifies to participate. For instance, if the access requirement is signatures of 10% of the state's voters, only well-moneyed interests can garner signatures of 10% of the voters, and if the party in power is the only well-moneyed interest in the state, then only the majority party would get to participate. This could presumably happen with regard to ballot access, and the courts have developed standards for addressing it. See, e.g., Munro v. Socialist Workers Party, 479 U.S. 189, 193 (1986) 366 In addition, because of the transparency and simplicity of ballot-accesslike rules of allowing access to participation, choices of who gets to participate are easier for the public at large to judge as fair or unfair than are the political choices inherent in the opaque and complex context of redistricting criteria.

*p. 72*
There are a number of material ways in which the principles behind ballot access differ from those underlying participant selection. For one thing, almost no individual thinks of himself as a member of just a single group. Knowing that an individual is, say, a stamp-collecting African-American female banker with three children indicates that that person has at least five groups with which she might want to identify. In the case of ballot access, that is immaterial. Once candidates qualify for the ballot-whether as the representative candidate of a group or as an independent candidate-individuals can reasonably select among them. By contrast, once groups or individuals qualify as participants in redistricting, there is no selection among them. Each participant gets a unique say in the redistricting process, even if multiple participants are representing some aspect of the same individuals. 367 However, that should typically be seen as a benefit, in that numerous aspects of an individual's interests are thus captured in the redistricting process by the various participants that represent him or her.

*p. 72*
The flipside of this point is that a concocted group that is really another existing group in disguise could qualify as a participant alongside the existing group. This would effectively double the group's input into the redistricting results as compared with any other participant. Similarly, an individual member of a group could attempt to make it onto the participant roll for the sole purpose of weighting the redistricting process in the group's favor. At the same time, who is to say that the Judean People's Front is the same group as the People's Front of Judea? 368 It is obviously very difficult to judge when two groups are really the same, given that groups are self-defining and do not exist fully formed in nature. 369 Relatedly and far more worrisome is that the preferred rule of equal input regardless of group size 370 taken together with allowing split groups can lead to a skewed process. Large groups can split into numerous smaller groups to dominate the redistricting process with more "equal input" by collusively speaking in unison. Indeed, it might be foolish for a large group not to split if minor-party or independent candidate upon a showing of a modicum of support among the potential voters for the office.").

*p. 72*
366. At the very least, we know that the criteria by which redistricting is done is hard to police satisfactorily. Who can participate is at least a fresh problem, on which there could conceivably be greater agreement on solutions and standards.

*p. 72*
367. Not only is there no one-to-one mapping from individuals to participants, but there may also be significant overlap between the memberships of groups participating in the redistricting process.

*p. 72*
368. See MONTY PYTHON'S LIFE OF BRIAN (Anchor Bay Entertainment 1979). 369. One heuristic for unmasking impostor group participants is to see if their group membership or petition signatures overlap with the true group too significantly. This heuristic would fail whenever groups would split rather than replicate. 370. See supra Part IV.B.1.c.ii.

*p. 72*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL allowed. Small groups, by virtue of their size, do not have that same power to split. Fracturing of large groups and the inability of small groups to fracture thus leads back to the problem of proportional input, creating an exponential effect on the redistricting results. To the extent that this would occur, it becomes necessary to calibrate a minimum group size so that it does not exclude too many small groups, while also not permitting large groups to split ad infinitum. 371

## B. STRATEGIC GAMING CONCERNS

*p. 73*
A serious concern with the procedural model, as designed, is that participants will attempt to game the system, and will possibly succeed. A participant can game the system if the participant has knowledge of other participants' preferences for redistricting configurations. 372 Needless to say, gaming behavior is generally minimized by the complexity of the model at work. Because there are a colossal number of principles by which participants can district and a glut of possible districting configurations to rank, most participants will recognize that they cannot know how other participants will behave and will refuse to risk gaming behavior. That said, participants can still guess some of the districting configurations that other participants will want to eliminate early in the game (for configuration elimination) or will rank highly (for meta-election). But overall, the complexity of the game will minimize gaming behavior. The complexity of the game and the risk of gaming failure together give sufficient incentive to participants to be honest in their rankings. 373 There is also a concern that otherwise competitive participants will collude to increase their chances of arriving at a desired districting configuration. 374 For example, in the context of configuration elimination, participants might collude 371. It should be nonetheless noted that under the predominant existing approach to redistricting, the political party with a majority of the legislative power is primarily the one with input into a redistricting plan and can generally have a greater impact on the plan than when there is input from a suboptimal set of participants. Cf. Samuel Issacharoff, Collateral Damage: The Endangered Center in American Politics, 46 WM. & MARY L. REV. 415, 431 (2004) (noting the disproportionate effects of having incumbents control redistricting).

*p. 73*
372. Cf. THEODORE W. KHEEL, THE KEYS TO CONFLICT RESOLUTION: PROVEN METHODS FOR RESOLVING DISPUTES VOLUNTARILY 14 (2001) ("Negotiation is predominantly a game of strategy even though it requires skill and there can be elements of chance that affect the outcome. The moves each side makes towards reaching an agreement on terms that satisfy their respective interests necessarily turn on the moves the other side makes.").

*p. 73*
373. The risk of ending up with a disfavored districting plan may induce participants to reveal their true preferences. See YOUNG, supra note 271, at 130 ("Preferences are usually private information, and we cannot expect people to honestly reveal them unless it is in their interest to do so. The challenge, therefore, is to design procedures that induce the claimants to reveal enough information about their preferences so that an equitable and efficient solution can be implemented."). Even if a participant chooses to game the system, one might view that goal as legitimately as ranking by, say, protection of incumbents or symmetry. See supra note 284 and accompanying text.

*p. 73*
374. For instance, it is hypothesized that Republicans and African-American Democrats colluded in the southern states during the 1980s and 1990s to effectuate districting schemes that pack African-American voters into a small number of districts, while increasing the number of Republican-controlled 2005] 1619 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING by comparing rankings in advance of the redistricting process and then eliminating any overlap in rankings. It is unlikely, however, that otherwise competitive participants will collude. Collusion in the context of the procedural model requires a strong degree of trust among participants that probably does not exist in the political world. Participants could simply share their rankings and then one or both could readjust their rankings to trick those participants that shared information, not to help them. This is particularly true when all of the participants' rankings are acted on simultaneously to derive a redistricting scheme, as in the meta-election model, because there is no possibility for a participant to realize that another participant who promised to collude opted out of the collusion until after the ultimate districting scheme has been determined, as the participant would realize if turn-taking were involved.

*p. 74*
C. OPPORTUNITY COSTS Among other things, the procedural model requires the selection of participants, their ranking of districting configurations, and the mapping of these rankings onto a unique districting configuration. This redistricting process will therefore absorb a lot of time, political energy, and money. The financial and other expenditures necessitated by the procedural model will, by definition, take money, time, attention, and so forth away from other causes and interests of a state's populace. Even if persuaded by the worth of the procedural model in the abstract, the question lingers: is it worth the cost? Despite all of the negative comments that can be made about the current methods of redistricting, 375 they are generally free of delay and expense, 376 at least when compared with the procedural model. Nonetheless, there is almost nothing more important than setting the district boundaries appropriately to protect the right to vote-the right preservative of all other rights. In fact, a concerted choice to funnel increased resources to the redistricting process and to require public involvement-whether in choosing participants or in partaking as participants-should itself enhance awareness of the political importance of redistricting.

## D. DOES PROCEDURAL FAIRNESS GUARANTEE SUBSTANTIVE FAIRNESS?

*p. 74*
One may acknowledge the procedural fairness of the model, but object to the particular redistricting scheme that results from applying the model. Arguably, the redistricting scheme may have an unfair effect, depending on one's definition of substantive fairness. In fact, it is even possible that the application of the districts, with white Democrats emerging as the losers of this collusive redistricting. See Toobin, supra note 2, at 74-75.

*p. 74*
375. See supra Parts II and III. 376. But see supra Part I (describing the grandstanding that occurred in the most recent rounds of redistricting in Texas when, for instance, a number of Democrats fled the state to deprive the state legislature of a quorum).

*p. 74*
[Vol. 93:1547 THE GEORGETOWN LAW JOURNAL procedural model over time might lead to identical redistricting schemes as under the current models. Nonetheless, it would still be preferable to employ the procedural model because it would be perceived by the public as fair, would incorporate a variety of viewpoints, and would be policeable by courts, even if the application of the model would render the redistricting process more complex and cumbersome. Moreover, as discussed above, by plucking a redistricting scheme from the overlap of participants' positions, a degree of substantive fairness is achieved that is lacking from the existing models. 377 In any case, this Article argues that most redistricting criteria, when implemented to derive a districting plan, have some political effect that results in unfairness for some group or individual. In a sense, any districting scheme is unfair to someone. Therefore, this model rejects the possibility of elaborating a redistricting method or standard that guarantees substantive fairness to everyone. The procedural model is thus preferable to the redistricting criteria described above because it guarantees at least procedural fairness. 378

## CONCLUSION

*p. 75*
In light of this Article, the Vieth plurality's conclusion that "'[f]airness' does not seem to [be] a judicially manageable standard," 379 and that "[s]ome criterion more solid and more demonstrably met than [fairness] seems to us necessary to enable the state legislatures to discern the limits of their districting discretion, to meaningfully constrain the discretion of the courts, and to win public acceptance for the courts' intrusion into a process that is the very foundation of democratic decisionmaking," 380 is sensible only when viewing "fairness" as a substantive measure. When examining fairness through the lens of procedure, however, there are ready methods for constructing fair districting schemes and standards for adjudging the fairness of redistricting plans.

*p. 75*
The model of redistricting presented herein turns on its head the previous scholarly literature of fairness in redistricting. Instead of searching for particular substantive norms of fairness to use in constructing or measuring redistricting schemes, the procedural model offers a wide and public group of participants an opportunity to shape redistricting schemes by utilizing any, all, or none of these norms and effectively "compromising" on a fair districting scheme. Rather than unrealistically pretending that politics has no place in redistricting, participants can use any criteria-political or apolitical-within constitutional limit, to rank potential districting configurations to be mapped onto a unique districting scheme. Fairness is thus achievable procedurally via collective compromise 377. See supra text accompanying notes 324-26. Furthermore, there is another sense in which the procedural model does offer a (weak) version of substantive fairness: the agreement among the participants to invoke the procedural model with its rules and bounds is a substantive compromise. 378. Cf. McCleskey v. Kemp, 481 U.S. 279, 313 (1987) (concluding that as long as procedures are fair, a racially discriminatory impact in capital sentencing does not violate the Constitution).

*p. 75*
379. Vieth v. Jubelirer, 541 U.S. 267, 291 (2004) (plurality opinion). 380. Id.

*p. 75*
2005] 1621 DERIVING AND MEASURING FAIRNESS IN REDISTRICTING rather than through an impossible-to-agree-upon substantive standard. The model also allows a vast number of groups and individuals to participate plausibly in the process of compromise, allowing their voices to be heard on voting-the basic right preservative of all other rights-more than any existing fair redistricting model. This model can also be used by courts and scholars to measure the fairness of existing districting schemes. This Article's goal is to spark further development of procedurally fair methods for constructing districting schemes and standards for judging schemes' procedural fairness.

## Footnotes

> 16. See id. 17. Id. (quoting the Texas Republican Party leadership). 18. See id. 19. See Balderas v. State, No. 6:01CV158, 2001 WL 34104833, at *3 (E.D. Tex. Nov. 28, 2001) (three-judge court) (per curiam); see also Toobin, supra note 2, at 63.

> 26. See id. at 66 (detailing such redistricting in the hands of Democrats and Republicans alike). 27. See infra Parts II and III.

> 33. 377 U.S. 533 (1964). 34. See id. at 562-66. 35. See id. at 545. 36. See id. at 542 n.7. 37. See id. 38. As detailed by the Supreme Court: Under the existing provisions, applying 1960 census figures, only 25.1% of the State's total population resided in districts represented by a majority of the members of the Senate, and only 25.40. Samuel Issacharoff, Judging Politics: The Elusive Quest for Judicial Review of Political Fairness, 71 TEX. L.REV. 1643, 1650 (1997). "Lurking in [the early voting rights] cases was a deeper concern that the manipulation of the districting agenda could cause systemic harms to the political process." Issacharoff, supra note 2, at 596. 41. See Issacharoff, supra note 40, at 1649.

> 55. Issacharoff, supra note 40, at 1676. 56. See Reno v. Bossier Parish School Bd., 528 U.S. 320, 333-34 (1999). 57. 42 U.S.C. § § 1973 to 1973bb-1.

> (discussing, but disagreeing with, this school of thought). 147. See infra Part III.B.4. 148. See infra Part III.B.5-6. 149.

> 161. See Ansolabehere & Snyder, supra note 159, at 448. 162. See id. (suggesting that this is precisely what happened during post-Baker-reapportionment in the Northeast and North-Central United States).

> 173. See King et al., supra note 167, at 93. 174. Proportional representation is a more specific form of symmetry. See, e.g., Hirsch, supra note 47, at 190-91. 175. Lowenstein & Steinberg, supra note 152, at 55. A "balloon effect" is "[t]he . . . gap between a party's vote share and its share of legislative seats." Lewyn, supra note 170, at 444.

> 183. See Lewyn, supra note 170, at 475. 184. See, e.g., T. Alexander Aleinikoff & Samuel Issacharoff, Race and Redistricting: Drawing Constitutional Lines After Shaw v. Reno, 92 MICH. L.

> ´,HOW NATIONS NEGOTIATE 206 (photo. reprint 1985) (1967) (quoting diplomat Edward R. Stettinius, Jr.); see also SELTSER,

> 246. Kuflik, supra note 240, at 41. 247. See, e.g., id. at 39.