Spatial diversity

Harvard Law Review

Volumn 125, Issue 8, 2012, Pages 1909-2010

  Stephanopoulos, Nicholas O ^a  

^a The Earth Institute at Columbia University   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 84860620550     PISSN: 0017811X     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Article

References (502)

1. Vieth v. Jubelirer, 541 U.S. 267, 340 (2004) (Stevens, J., dissenting) (quoting Appendix to Ju-risdictional Statement at 135a, Vieth, 541 U.S. 267 (No. 02-1580)).
2. See Brief for Appellants at 13, 42-44, 47-49, Vieth, 541 U.S. 267 (No. 02-1580) [hereinafter Vieth Brief]; Pennsylvania 6th District: Chester and Montgomery Counties, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/pa/06 (last visited May 3, 2012).
3. See Vieth Brief, supra note 2, at 13, 47-49; Pennsylvania 18th District: Pittsburgh Metro Area, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/pa/18 (last visited May 3, 2012).
4. Vieth, 541 U.S. at 330-31 (Stevens, J., dissenting); see also id. at 349 (Souter, J., dissenting) (noting that Sixth District might support valid political gerrymandering claim); id. at 272-73 (plu-rality opinion) (describing Pennsylvania's 2002 redistricting).
5. See League of United Latin Am. Citizens v. Perry, 548 U.S. 399, 423-25 (2006) [hereinafter LULAC] (describing the Twenty-Third and Twenty-Fifth Districts).
6. Id. at 425-443.
7. Thornburg v. Gingles, 478 U.S. 30, 50-51 (1986).
8. More specifically, I calculate two slightly different kinds of spatial diversity: one for all of the residents in the districts at issue, and another solely for the districts' minority residents. As I explain in Part II, infra pp. 1924-35, the Court's redistricting case law implicates both variants of spatial diversity.
9. Voter roll-off measures the proportion of voters who cast a ballot for a top-ticket (e.g., pres-idential) race, but who do not cast a ballot for a lower-ticket (e.g., congressional) race.
10. This debate has raged in a series of Supreme Court cases, most notably LULAC, Vieth, and Davis v. Bandemer, 478 U.S. 109 (1986). It has also received extensive attention from legal aca-demics and political scientists. See Nicholas O. Stephanopoulos, Redistricting and the Territorial Community, 160 U. PA. L. REV. 1379, 1383-84 nn.10-16 (2012) (summarizing literature on political gerrymandering).
11. Partisan bias refers to the divergence in the share of seats that each party would win given the same share of the statewide vote. Electoral responsiveness refers to the rate at which a party gains or loses seats given changes in its statewide vote share.
12. Andrew Gelman & Gary King, Enhancing Democracy Through Legislative Redistricting, 88 AM. POL. SCI. REV. 541, 544-45 (1994).
13. Defining bias and responsiveness.
14. As expected, given the literature on the negative conse-quences of spatial diversity, the measure is positively correlated with bias (though only for higher levels of diversity) and negatively correlated with responsiveness.
15. See, e.g., Bush v. Vera, 517 U.S. 952, 976-83 (1996) (plurality opinion) (considering whether Voting Rights Act required majority-minority districts challenged under Equal Protection Clause); Shaw v. Hunt, 517 U.S. 899, 911-18 (1996) (same); see also Bush, 517 U.S. at 1037 (Stevens, J., dissenting) (discussing 'schizophrenic second-guessing' that occurs when redistricters seek to comply with both Constitution and Voting Rights Act).
16. Stephanopoulos, supra note 10.
17. Richard H. Pildes & Richard G. Niemi, Expressive Harms, "Bizarre Districts," and Vot-ing Rights: Evaluating Election-District Appearances After Shaw v. Reno, 92 MICH. L. REV. 483 (1993).
18. Grutter v. Bollinger, 539 U.S. 306, 350 (2003) (Thomas, J., concurring in part and dissenting in part).
19. Id. at 354 n.3.
20. Joseph A. Aistrup, Constituency Diversity and Party Competition: A County and State Lev-el Analysis, 57 POL. RES. Q. 267, 268 (2004).
21. Thomas L. Brunell & Bernard Grofman, Evaluating the Impact of Redistricting on District Homogeneity, Political Competition, and Political Extremism in the U.S. House of Representatives, 1962 to 2006, in DESIGNING DEMOCRAT-IC GOVERNMENT 117, 119 (Margaret Levi et al. eds., 2008).
22. "District-level homogeneity is a concept that needs precise explication....".
23. Heather K. Gerken, Second-Order Diversity, 118 HARV. L. REV. 1099, 1102 (2005).
24. Gerken refers to this sort of diversity as "first-order diversity."
25. Id. at 1104.
26. Heather K. Gerken, Keynote Address: What Election Law Has to Say to Constitutional Law, 44 IND. L. REV. 7, 15 (2010).
27. Further discussing this notion of diversity.
28. SANFORD LEVINSON, WRESTLING WITH DIVERSITY 24 (2003).
29. PETER H. SCHUCK, DIVERSITY IN AMERICA 22-23 (2003).
30. Noting that there is 'a strong tendency to look to proportionality [between the entity and the broader population] as the measure of [diversity]'.
31. LEVINSON, supra note 20, at 24.
32. Jeffrey F. Milem, The Educational Benefits of Diversity: Evidence from Multiple Sectors, in COMPELLING INTEREST 126, 132 (Mitchell J. Chang et al. eds., 2003).
33. Defining structural diversity as 'the numerical and proportional represen-tation of students from different racial/ethnic groups in the student body'.
34. Patrick S. Shin & Mitu Gulati, Showcasing Diversity, 89 N.C. L. REV. 1017, 1028 (2011).
35. Assessing workforce diversity by examining whether 'the proportion of whites to racial minority groups... reflect[s] the proportion observed in the broader community'.
36. Parents Involved in Cmty. Sch. v. Seattle Sch. Dist. No. 1, 551 U.S. 701, 727 (2007) (plurali-ty opinion) (quoting Joint Appendix at 42a, Parents Involved, 551 U.S. 701 (No. 05-908)) (internal quotation mark omitted); see also id. at 732 (equating 'racial diversity' and 'racial balance').
37. Metro Broad., Inc. v. FCC, 497 U.S. 547, 614 (1990) (O'Connor, J., dissenting), overruled by Adarand Constructors, Inc. v. Pena, 515 U.S. 200, 227 (1995); see also, e.g., Holland v. Illinois, 493 U.S. 474, 513 n.10 (1990) (Stevens, J., dissenting) (noting that 'the diversity within our society is reflected on our juries' when 'each population group is represented... in proportion to its strength in the population' (quoting JON M. VAN DYKE, JURY SELECTION PROCEDURES 18 (1977))).
38. Doe v. Kamehameha Sch./Bernice Pauahi Bishop Estate, 470 F.3d 827, 842 (9th Cir. 2006) (en banc) (praising the 'laudable goal of achieving diversity and proportional representation in the workplace'); Lutheran Church-Mo. Synod v. FCC, 141 F.3d 344, 356 (D.C. Cir. 1998) (de-scribing diversity as a 'justification for policies seeking racial proportionality' and a 'synonym for proportional representation itself').
39. Aistrup, supra note 17, at 268.
40. Michael Bailey & David W. Brady, Heterogeneity and Representation: The Senate and Free Trade, 42 AM. J. POL. SCI. 524, 536 (1998).
41. Jon R. Bond, The Influence of Constituency Diversity on Electoral Competition in Voting for Congress, 1974-1978, 8 LEGIS. STUD. Q. 201, 202 (1983).
42. Paul S. Herrnson & James G. Gimpel, District Conditions and Primary Divisiveness in Congressional Elections, 48 POL. RES. Q. 117, 121 (1995).
43. DAVID E. CAMPBELL, WHY WE VOTE 224 n.6 (2006).
44. W. Mark Crain, Districts, Diversity, and Fiscal Biases: Evidence from the American States, 42 J.L. & ECON. 675, 684 (1999).
45. PAUL GRONKE, THE ELECTORATE, THE CAMPAIGN, AND THE OFFICE 158 (2000).
46. While this Article focuses on the spatial diversity of districts, not states, the methods it employs can be applied easily to the latter. For example, as Professor Gronke expected, Illinois is indeed about fifty percent more spatially heterogeneous than Virginia with respect to the African Ameri-can composite factor, despite the states' similar shares of African American residents. See infra section III.A, pp. 1936-41 (discussing factors that emerged from nationwide statistical analysis).
47. Infra Part III, pp. 1935-1980.
48. Spatial diversity is thus spatial because it assesses the variability of the geographic subunits that comprise larger geographic entities.
49. Sean F. Reardon & Glenn Firebaugh, Response: Seg-regation and Social Distance - A Generalized Approach to Segregation Measurement, 32 SOC. METHODOLOGY 85, 86 (2002)
50. Arguing that analogous measures of residential segregation 'in-corporate implicit notions of social distance'). There do exist indices that incorporate actual geo-graphic data (e.g., latitude and longitude, distance between points, etc.), and I make some use of them below. See infra notes 186-88 and accompanying text (discussing my use of Global Moran's I). For the most part, though, those indices capture concepts such as centralization and concen-tration that are not particularly relevant to this Article's project.
51. The color of the small constituent blocks in the diagrams corresponds to their proportion (of people, say) that is black. An entirely white block is zero percent black, a medium gray block is fifty percent black, an entirely black block is one hundred percent black, etc.
52. Lawrence A. Brown & Su-Yeul Chung, Spatial Segregation, Segregation Indices and the Geographical Perspective, 12 POPULATION, SPACE & PLACE 125, 127 (2006).
53. Displaying set of diagrams with identical segregation scores but different spatial patterns.
54. Sean F. Reardon & David O'Sullivan, Measures of Spatial Segregation, 34 SOC. METH-ODOLOGY 121, 124 & n.1 (2004).
55. J.A. Wiens, Spatial Scaling in Ecology, 3 FUNCTIONAL ECOLOGY 385, 386 (1989).
56. Noting that 'different patterns emerge at different scales of investiga-tion of virtually any... system'). As a general matter, spatial heterogeneity increases as the scale of the subunits under consideration decreases.
57. David W.S. Wong, Spatial Dependency of Seg-regation Indices, 41 CANADIAN GEOGRAPHER 128, 130-31 (1997).
58. In this Article, the only sub-units that I examine are Census tracts. See infra notes 168-72 and accompanying text (explaining reasons for focus on tracts).
59. Spatial heterogeneity is also a function of the scale of the entity that is considered. In gen-eral, the larger the entity that is taken into account, the higher the level of spatial heterogeneity. See Wiens, supra, at 388 (noting this effect in ecological context). In this Article, I hold the scale of the entity constant by analyzing only congressional districts.
60. Harvey J. Miller, Tobler's First Law and Spatial Analysis, 94 ANNALS ASS'N AM. GEOG-RAPHERS 284, 284 (2004).
61. W.R. Tobler, A Computer Movie Simulating Urban Growth in the Detroit Region, 46 ECON. GEOGRAPHY 234, 236 (1970).
62. '[E]verything is related to everything else, but near things are more related than distant things.'); see also infra p. 1940 (reporting empirical findings that all factors examined in this Article exhibit extremely high levels of spatial autocorrelation).
63. To be more precise, clusters are common and uniform distributions are unusual when the area under examination is reasonably large (the size of a state, say). If we 'zoom in' to a smaller area that happens to contain a cluster, then we will, of course, find a zone with a relatively uniform spatial composition.
64. Wiens, supra note 32, at 388.
65. Wiens, supra note 32, at 130-131.
66. However, random geographic distributions are rare no matter what the scope of the inquiry.
67. My calculations indicate that about sixty percent of the variance in congressional districts' spatial diversity scores is explained by their scores on the Sullivan Index (a common measure of top-line diversity). See supra note 25 and accompanying text (discussing Sullivan Index); infra section III.A, pp. 1936-41 (discussing methodology for calculating spatial diversity scores).
68. Robert R. Brischetto, Latino Voters and Redistricting in the New Millennium, in REDISTRICTING AND MINORITY REPRESENTATION 43, 56 (David A. Bositis ed., 1998).
69. Edward L. Glaeser & Jacob L. Vigdor, Racial Segregation: Promising News, in 1 REDEFINING URBAN AND SUBURBAN AMERICA 211, 215 (Bruce Katz & Robert E. Lang eds., 2003).
70. Philip A. Klinkner, Red and Blue Scare: The Continuing Diversity of the American Electoral Landscape, FORUM, June 2004, art. 2, at 6.
71. Brischetto, supra note 36, at 58.
72. Klinkner, supra note 36, at 7.
73. Discussing closely related Index of Exposure). Segregation scholars have also proposed other indices (less widely used at present) that measure the clustering, concentration, and centrality of different groups.
74. Brown & Chung, supra note 31, at 126.
75. Douglas S. Massey & Nancy A. Denton, Hypersegregation in U.S. Metropolitan Areas: Black and Hispanic Segregation Along Five Dimen-sions, 26 DEMOGRAPHY 373, 373 (1989).
76. Paul A. Jargowsky, Take the Money and Run: Economic Segregation in U.S. Metropolitan Areas, 61 AM. SOC. REV. 984, 987 (1996).
77. Emphasis added.
78. Reardon & Firebaugh, supra note 29, at 99.
79. Noting that "all existing segregation measures require that the groups of interest be mutually exclusive and unordered".
80. Karl E. Taeuber & Alma F. Taeuber, A Practitioner's Perspective on the Index of Dissimilarity, 41 AM. SOC. REV. 884, 886 (1976).
81. "In order to calculate any of the segregation indexes that have been proposed, it is necessary to know the minority proportion.".
82. Jargowsky, supra note 38, at 987.
83. Scott D. Cooper et al., Quantifying Spatial Heterogeneity in Streams, 16 J. N. AM. BENTHOLOGICAL SOC'Y 174, 174 (1997).
84. Id. at 175-177.
85. P.B. Adler et al., The Effect of Grazing on the Spatial Het-erogeneity of Vegetation, 128 OECOLOGIA 465, 466 (2001).
86. Habin Li & James F. Reynolds, A Simulation Experiment to Quantify Spatial Heterogeneity in Categorical Maps, 75 ECOLOGY 2446, 2446 (1994).
87. Joan L. Riera et al., Analysis of Large-Scale Spatial Heterogeneity in Vegetation Indices Among North American Landscapes, 1 ECOSYSTEMS 268, 268-269 (1998).
88. Infra section III.A, pp. 1936-1941.
89. Gerken, supra note 18, at 1104.
90. Using "diversity" and "heterogeneity" interchange-ably.
91. MERRIAM-WEBSTER'S COLLEGIATE DICTIONARY 584 (11th ed. 2006).
92. Defining "heterogeneous" as "consisting of dissimilar or diverse ingredients or constituents" (emphasis added).
93. And perhaps an even weightier reason is that the Supreme Court seems to agree that spatial diversity is an undesirable district attribute.
94. Infra Part II, pp. 1924-1935.
95. GREGORY R. WEIHER, THE FRACTURED METROPOLIS 190 (1991).
96. Explaining that political boundaries cause the 'sorting of the population... by salient characteristics such as race and socioeconomic status'); id. at 100-08 (reporting empirical findings that municipal boundaries explain high proportion of variance with respect to age, education, and profession in several met-ropolitan areas).
97. Richard Briffault, Our Localism: Part II - Localism and Legal Theory, 90 COLUM. L. REV. 346, 353-54 (1990).
98. Infra Part III, pp. 1935-1980.
99. Stephanopoulos, supra note 10, at 1389-1404.
100. Dean Alfange, Jr., Gerrymandering and the Constitution: Into the Thorns of the Thicket at Last, 1986 SUP. CT. REV. 175, 216.
101. Richard Briffault, Lani Guinier and the Dilemmas of American Democracy, 95 COLUM. L. REV. 418, 431-432+443 (1995) (book review).
102. Bernard Grofman, Would Vince Lombardi Have Been Right if He Had Said: "When It Comes to Redistricting, Race Isn't Everything, It's the Only Thing"?, 14 CARDOZO L. REV. 1237, 1262 (1993).
103. NANCY MAVEETY, REPRESENTATION RIGHTS AND THE BURGER YEARS 39 (1991)
104. '[V]oters are less disoriented and more capable of collective political organiza-tion when they are not fractured among a variety of districts.'.
105. Richard Morrill, A Geographer's Perspective, in POLITICAL GERRYMANDERING AND THE COURTS 212, 217 (Bernard Grofman ed., 1990).
106. "[C]itizens in... isolated parts of a district may come to feel that their community is unrepresented....".
107. Richard G. Niemi et al., The Effects of Congruity Between Community and District on Salience of U.S. House Candidates, 11 LEGIS. STUD. Q. 187, 193 (1986).
108. Jonathan Winburn & Michael W. Wagner, Carving Voters Out: Redistricting's Influence on Political Information, Turn-out, and Voting Behavior, 63 POL. RES. Q. 373, 379 (2010).
109. Voters are also less informed about candidates in districts that do not correspond closely to media markets (and thus are likely as well to be spatially diverse).
110. James E. Campbell et al., Television Markets and Congressional Elections, 9 LEGIS. STUD. Q. 665, 671 (1984).
111. Dena Levy & Peverill Squire, Television Markets and the Competitiveness of U.S. House Elections, 25 LEGIS. STUD. Q. 313, 319 (2000).
112. James M. Snyder Jr. & David Strömberg, Press Coverage and Political Accountability, 118 J. POL. ECON. 355, 379-80 (2010).
113. CAMPBELL, supra note 26, at 23-24.
114. Dora L. Costa & Matthew E. Kahn, Civic Engagement and Community Heterogeneity: An Economist's Perspective, 1 PERSP. ON POL. 103, 107 (2003).
115. Alberto Alesina & Eliana La Ferrara, Participation in Heterogeneous Communities, 115 Q.J. ECON. 847, 850 (2000).
116. Costa & Kahn, supra note 52, at 105.
117. J. ERIC OLIVER, DEMOCRACY IN SUBURBIA 87-88, 131 (2001).
118. Finding that economic homogeneity of cities is negatively correlated with civic participation, but that racial homogeneity is positively correlated.
119. Alberto Alesina & Eliana La Ferrara, Who Trusts Others?, 85 J. PUB. ECON. 207, 222 (2002).
120. Costa & Kahn, supra note 52, at 105-106.
121. Robert D. Putnam, E Pluribus Unum: Diversity and Community in the Twenty-First Century, 30 SCANDINAVIAN POL. STUD. 137, 147-149 (2007).
122. Rene R. Rocha & Rodolfo Espino, Racial Threat, Residential Segregation, and the Policy Attitudes of Anglos, 62 POL. RES. Q. 415, 422-23 (2009).
123. Finding that, for a given Hispanic share of the population in a metropolitan area, Anglos had more hostile views toward Hispanics if the area was ethnically segregated (i.e., spatially diverse) than if it was ethnically integrated); Note.
124. Donald R. Kinder & Tali Mendelberg, Cracks in American Apartheid: The Political Impact of Prejudice Among Desegregated Whites, 57 J. POL. 402, 419 (1995).
125. Finding that whites are less prejudiced toward blacks in racially integrated areas.
126. Jacob L. Vigdor, Community Composition and Collective Action: Analyzing Initial Mail Response to the 2000 Census, 86 REV. ECON. & STAT. 303, 307 (2004).
127. Stephanopoulos, supra note 10, at 1464-1467.
128. BRUCE E. CAIN, THE REAPPORTIONMENT PUZZLE 63 (1984).
129. Id. at 39.
130. Morrill, supra note 50, at 216-217.
131. THOMAS L. BRUNELL, REDISTRICTING AND REPRESENTATION 26 (2008).
132. Referring primarily to political homogeneity.
133. JONATHAN S. KRASNO, CHALLENGERS, COMPE-TITION, AND REELECTION 38, 59 (1994).
134. Charles Backstrom et al., Establishing a Statewide Electoral Effects Baseline, in POLITICAL GERRYMANDERING AND THE COURTS, supra note 50, at 145+153.
135. However, while it may be difficult for representatives to serve as delegates for spatially diverse districts, it may be easier for them to serve as trustees. According to the tradi-tional delegate/trustee dichotomy, representatives who are delegates abide by the expressed preferences of their constituents, while representatives who are trustees make their own autonomous policy decisions.
136. HANNA FENICHEL PITKIN, THE CONCEPT OF REPRESENTATION (1967).
137. RICHARD F. FENNO, Jr., HOME STYLE: HOUSE MEMBERS IN THEIR DISTRICTS (1978).
138. Richard F. Fenno, Jr., U.S. House Members in Their Constituencies: An Exploration, 71 AM. POL. SCI. REV. 883 (1977).
139. MALCOLM E. JEWELL, REPRESENTATION IN STATE LEGISLATURES (1982).
140. Fenno, supra note 61, at 885.
141. FENNO, supra note 61, at 1-8.
142. JEWELL, supra note 62, at 117.
143. Id. at 55-60+115-120.
144. J. Vincent Buck & Bruce E. Cain, British MPs in Their Constituencies, 15 LEGIS. STUD. Q. 127, 138-39 (1990).
145. Describing analogous views of British Parliament members from spatially diverse districts.
146. Bailey & Brady, supra note 25, at 537.
147. Studying senators' votes on free trade issues.
148. Christopher Dennis et al., Constituency Diversity and Congress: The Case of NAFTA, 29 J. SOCIO-ECON. 349, 355 (2000) (same).
149. Elisabeth R. Gerber & Jeffrey B. Lewis, Beyond the Median: Voter Preferences, District Heterogeneity, and Political Representation, 112 J. POL. ECON. 1364, 1376-78 (2004).
150. Atudying legislators' votes in Los Angeles County.
151. Bailey & Brady, supra note 25, at 526.
152. Gerber & Lewis, supra note 66, at 1376.
153. Crain, supra note 26, at 689.
154. Finding that representatives from heterogeneous districts do not make as great an effort to win district-specific projects.
155. Snyder & Strömberg, supra note 51, at 395-399
156. Finding that representatives from districts that are more congruent with media markets are less loyal to their parties and hence less polarized).
157. Gerken, supra note 18, at 1102-1103.
158. 'Second-order diversity involves variation among decisionmaking bodies, not within them. It favors interorganizational heterogeneity, not intra-organizational heterogeneity.'.
159. Id. at 1124-1132.
160. Richard Thompson Ford, Geography and Sovereignty: Jurisdic-tional Formation and Racial Segregation, 49 STAN. L. REV. 1365, 1369 (1997).
161. Arguing for "the inversion of James Madison's extended sphere... to allow one group some degree of autonomy".
162. Gerken, supra note 18, at 1104+1161.
163. Steven Callander, Electoral Competition in Heterogeneous Districts, 113 J. POL. ECON. 1116, 1118-19 (2005).
164. Showing that parties grow more divergent as inter-district heterogeneity increases.
165. Reardon & O'Sullivan, supra note 32, at 124 n.1.
166. Noting that subunits that are 'relative-ly homogeneous internally' tend to 'exaggerate [measures of] segregation' among subunits.
167. Wiens, supra note 32, at 388.
168. Observing that as subunits become more internally diverse, 'a great-er proportion of the spatial heterogeneity of the system is contained within [the subunits] and is lost to our resolution'.
169. PITKIN, supra note 60, at 60.
170. Quoting Letter from John Adams to John Penn (Jan. 1776), reprinted in 4 JOHN ADAMS, THE WORKS OF JOHN ADAMS 205 (1851)); Note.
171. James A. Gardner, Foreword: Representation Without Party: Lessons from State Constitutional Attempts to Control Gerrymandering, 37 RUTGERS L.J. 881, 957 (2006).
172. Noting that 'where district vot- ers are [relatively] homogeneous, all conflict resolution occurs in the legislature and none in the electorate'.
173. Supra note 11 (defining partisan bias).
174. For example, adherence to both political subdivisions and geographic communities of inter-est tends to produce districts that are relatively spatially homogeneous.
175. Supra note 45 and accompanying text.
176. Infra notes 122, 217.
177. JONATHAN WINBURN, THE REALITIES OF REDISTRICTING 200-201 (2008).
178. Concluding that bars on splitting political subdivisions constrain political gerrymandering.
179. MICHAEL P. MCDONALD, MIDWEST MAPPING PROJECT 22-186 (2009), available at http://elections.gmu.edu/Midwest_Mapping_Project.pdf.
180. Providing data showing that average proportion of Democratic-leaning districts would increase, according to my calculations, from for-ty percent to forty-nine percent.
181. Todd Makse, Defining "Communities of Interest" in Redistricting Through Initiative Voting 11 tbl.4 (Dec. 2011) (unpublished manuscript), available at http://users.dickinson.edu/~makset/coi.pdf.
182. Showing that proportion of Democratic-leaning congressional districts in Ohio would increase from thirty-three percent to forty-four percent.
183. Stephanopoulos, supra note 10, at 1457-1462.
184. In the legal academy, Professors Samuel Issacharoff and Richard Pildes have sung the praises of electoral competition most loudly.
185. Samuel Issacharoff, Gerrymandering and Political Cartels, 116 HARV. L. REV. 593 (2002).
186. Samuel Issacharoff & Richard H. Pildes, Politics as Markets: Partisan Lockups of the Democratic Process, 50 STAN. L. REV. 643 (1998).
187. Richard H. Pildes, The Constitution and Political Competition, 30 NOVA L. REV. 253 (2006).
188. Aistrup, supra note 17, at 267.
189. Claiming basis in Madisonian political theory for view that "party competition develops in response to constituency diversity".
190. Herrnson & Gimpel, supra note 25, at 120 (same).
191. KRASNO, supra note 60, at 62+69.
192. Bond, supra note 25, at 206.
193. Brunell & Grofman, supra note 17, at 122-126.
194. William Koetzle, The Impact of Constituency Diversity upon the Com-petitiveness of U.S. House Elections, 1962-96, 23 LEGIS. STUD. Q. 561, 564 (1998).
195. Michael P. McDonald, Redistricting Institutions and Competition in U.S. House Districts, in DESIGNING DEMOCRATIC GOVERNMENT, supra note 17, at 141, 149-152.
196. Robert E. Hogan, Sources of Competition in State Legislative Primary Elections, 28 LEGIS. STUD. Q. 103, 115 (2003).
197. Tom W. Rice, Gubernatorial and Senatorial Primary Elections: Determinants of Competition, 13 AM. POL. Q. 427, 438 (1985).
198. GRONKE, supra note 27, at 97.
199. Jon R. Bond et al., Explaining Challenger Quality in Congressional Elections, 47 J. POL. 510, 525 (1985).
200. Michael J. Ensley et al., District Complexity as an Advantage in Congressional Elections, 53 AM. J. POL. SCI. 990, 998 (2009).
201. Aistrup, supra note 17, at 273.
202. Finding that socioeconomic diversity increases competi-tiveness but that racial diversity decreases it.
203. Herrnson & Gimpel, supra note 25, at 128.
204. Finding that heterogeneity is associated with competitiveness in Democratic but not in Republican prima-ry elections.
205. Campbell et al., supra note 51, at 672-673.
206. Levy & Squire, supra note 51, at 317.
207. Niemi et al., supra note 51, at 193.
208. Campbell et al., supra note 51, at 673.
209. Levy & Squire, supra note 51, at 321.
210. Campbell et al., supra note 51, at 667.
211. Levy & Squire, supra note 51, at 315.
212. Niemi et al., supra note 51, at 193.
213. Stephanopoulos, supra note 10, at 1457-1462.
214. DOUGLAS JOHNSON, COMPETI-TIVE DISTRICTS IN CALIFORNIA: A CASE STUDY OF CALIFORNIA'S REDISTRICTING IN THE 1990S 4, 7-9 (2005), available at http://www.claremontmckenna.edu/rose/publications/pdf/rose_ca_case_study.pdf
215. Finding high level of competitiveness when California's districts were drawn in the 1990s to correspond to geographic communities).
216. MCDONALD, supra note 76, at 22-21.
217. Finding that number of competitive districts generally increases when five Midwestern states' districts are redrawn so as to minimize splits of counties and census places.
218. Vladimir Kogan & Eric McGhee, Redistricting California: An Evaluation of the Citizens Commission Final Plans 5-6, 22-24 (Sept. 13, 2011).
219. Unpublished manuscript), available at http://igs.berkeley.edu/politics /redistricting_california.pdf (finding greater competitiveness for new California districts drawn by independent commission with preservation of geographic communities as key criterion).
220. Infra Part III, pp. 1935-1980.
221. For instance, the Court has never cited any of the empirical studies discussed above in sec-tion I.C, pp. 1917-1924.
222. In a previous work, I have made the normative argument that the Court should combat political gerrymandering by requiring districts to correspond, where possible, to organic geograph-ic communities.
223. Stephanopoulos, supra note 10.
224. Davis v. Bandemer, 478 U.S. 109, 132 (1986) (plurality opinion).
225. Vieth v. Jubelirer, 541 U.S. 267, 281-84 (2004) (plurality opinion) (criticizing the ap-proach of the Bandemer plurality).
226. Id. at 281-301.
227. LULAC, 548 U.S. 399, 414-23 (2006) (opinion of Kennedy, J.).
228. Vieth, 541 U.S. at 292-305 (plurality opinion).
229. Id. at 341 (Stevens, J., dissenting) (decrying the Court's "failure of judicial will").
230. 462 U.S. 725 (1983).
231. Id. at 744.
232. Id. at 762
233. (Stevens, J., concurring) (quoting Barry Light, New Jersey Map Imaginative Gerrymander, 40 CONG. Q. WKLY. REP. 1190, 1193 (1982)) (internal quotation mark omitted). The district was thus spatially diverse with respect to residential type (suburban versus rural).
234. Id. at 762-763.
235. Quoting Light, supra note 99, at 1193-1195.
236. Internal quotation mark omitted). The district was thus spatially diverse with respect to industrial employment, academic orienta-tion, and religion.
237. Id. at 763 n.31.
238. Quoting Light, supra note 99, at 1198.
239. Internal quotation mark omitted). As long as the two metropolitan areas' suburbs differed from each other in important respects, then this district too was spatially diverse.
240. Id. at 764 n.33.
241. Noting that in several districts 'different television and radio stations, different newspapers, and different transportation systems serve the northern and southern localities' (quoting Daggett v. Kimmelman, 535 F. Supp. 978, 984 (D.N.J. 1982) (three-judge court) (Gibbons, J., dissenting))).
242. Id. at 764.
243. Supra section I.C, pp. 1917-1924.
244. Karcher, 462 U.S. at 787 (Powell, J., dissenting).
245. Id. at 789.
246. Id. at 787 n.3.
247. Arguing that when a district corresponds to a community, the 'Representative... knows the needs of his district and is more responsive to them'); id. at 789 (criticizing 'contorted Districts' that do not 'reflect any attempt to follow natural, historical, or local political boundaries'.
248. 478 U.S. 109 (1986).
249. Id. at 177.
250. Powell, J., concurring in part and dissenting in part) (quoting Bandemer v. Davis, 603 F. Supp. 1479, 1494 (S.D. Ind. 1984) (three-judge court)) (internal quotation marks omitted).
251. Id. at 173 n.13.
252. Id. at 174 n.13.
253. "[I]rrational lines themselves affect the ability of all voters to exercise their political influence....".
254. Id. at 167.
255. Referring to districts that adhere to community boundaries.
256. Id. at 177.
257. Quoting Bandemer, 603 F. Supp. at 1487) (internal quotation mark omitted). The district was thus spatially diverse with respect to residential type (urban versus rural).
258. Id.
259. Quoting Bandemer, 603 F. Supp. at 1487) (internal quotation mark omitted). The district was thus spatially diverse with respect to race.
260. Id. at 180 n.21.
261. Quoting Bandemer, 603 F. Supp. at 1487) (internal quotation mark omitted). The district was thus spatially diverse with respect to residential type (urban versus suburban). See also id. at 176-77 (criticizing district that 'plac[ed] the seat of one county in a voting district composed of townships from other counties'); id. at 180 (noting that 'the mapmakers split Fort Wayne... and associated each of the halves with areas from outlying counties').
262. Id. at 169, 173 n.13.
263. The appellants in Vieth focused on statewide rather than district-specific claims. See Vieth v. Jubelirer, 541 U.S. 267, 355 (2004) (Souter, J., dissenting).
264. 548 U.S. 399 (2006). The appellants in LULAC emphasized the mid-decade timing of Texas's redistricting. See id. at 413-14; id. at 414-23 (opinion of Kennedy, J.). As discussed be-low, spatial diversity did play a crucial role in the vote dilution portion of the Court's decision. See infra pp. 1931-32.
265. Vieth, 541 U.S. at 340 (Stevens, J., dissenting) (quoting Appendix to Jurisdictional State-ment, supra note 1, at 136a) (internal quotation mark omitted); id. at 349 (Souter, J., dissenting) (same); see also id. (noting that Montgomery County was divided into six different districts); Erfer v. Commonwealth, 794 A.2d 325, 342 (Pa. 2002) (Zappala, C.J., dissenting) (noting that Sixth Dis-trict 'combines the relatively unrelated rural parts of Chester and Berks County with the 'densely settled suburban' communities of... Montgomery County' (quoting trial transcript)).
266. Vieth, 541 U.S. at 331 (Stevens, J., dissenting); see also id. at 328 (noting the 'more individ-ualized representation injury to [the plaintiff] as a resident of District 6').
267. LULAC, 548 U.S. at 454 (Stevens, J., concurring in part and dissenting in part).
268. Id. at 479.
269. Quoting Dep't of Justice, Section 5 Recommendation Memorandum 67 (Dec. 12, 2003), available at http://www.washingtonpost.com/wp-srv/nation/documents/texasDOJmemo.pdf) (internal quotation mark omitted). Justice Stevens would have invalidated these districts on polit-ical gerrymandering grounds, not because of racial vote dilution. See id. at 481.
270. Id. at 470.
271. See Stephanopoulos, supra note 10, at 1421-24 (discussing political gerrymandering cases through prism of community disruption rather than spatial diversity).
272. The concept of spatial diversity is even more salient in the case law on whether state com-munity-of-interest provisions have been followed. Five states include such provisions in their constitutions, seven more have analogous statutory requirements, and a further twelve adopted non-binding guidelines along similar lines during the last redistricting cycle. The provisions are aimed at preventing gerrymandering, and they typically require districts to correspond, where possible, to geographic communities of interest, i.e., 'the shared social, cultural, and economic in-terests of people living in a particular area.' Id. at 1425; see also id. at 1424-1428.
273. State courts enforcing community-of-interest requirements have repeatedly invalidated dis-tricts that were highly spatially diverse along key dimensions. The spatial diversity of these dis-tricts was a strong indication that they did not in fact correspond to geographic communities of interest. See, e.g., Hickel v. Se. Conference, 846 P.2d 38, 51 (Alaska 1992) (striking down district that 'mixes the small, rural, Native communities with the urban areas of Ketchikan and Sitka').
274. In re Reapportionment of Colo. Gen. Assembly, 647 P.2d 209, 212 (Colo. 1982) (striking down dis-trict due to 'cultural differences of religion, ethnicity, age and life-style between those parts of northwest Denver and east-central Denver encompassed within district'); In re Apportionment of Towns of Hartland, Windsor & W. Windsor, 624 A.2d 323, 331 (Vt. 1993) (striking down district that merged socially and economically distinct towns on opposite sides of mountain range).
275. Con-versely, state courts have frequently upheld districts that were more spatially homogeneous in terms of important variables. These districts' greater geographic uniformity suggested that they were indeed congruent with underlying territorial communities. See, e.g., In re 2001 Redistricting Cases, 44 P.3d 141, 145 (Alaska 2002) (upholding district whose various regions shared 'their in-volvement in the commercial fishing industry'); Beauprez v. Avalos, 42 P.3d 642, 652 (Colo. 2002) Note.
276. Upholding district composed of 'western slope counties' containing a 'historical Hispanic com-munity of interest'); Hartung v. Bradbury, 33 P.3d 972, 981 (Or. 2001) (upholding district made up of 'small communities that have a rural resource economy and share agricultural and timber in-terests' (quoting Oregon Secretary of State Bill Bradbury)).
277. 42 U.S.C. § 1973 (2006). Under the Court's canonical construction of section 2, minority groups are entitled to districts in which they can elect the representatives of their choice if (1) the groups are sufficiently large and geographically compact to constitute majorities in single-member districts; (2) they are politically cohesive; and (3) significant racial polarization in voting exists - and if a totality-of-the-circumstances inquiry also supports the groups' claims. See Thornburg v. Gingles, 478 U.S. 30, 49-51, 79-80 (1986).
278. T. Alexander Aleinikoff & Samuel Issacharoff, Race and Redistricting: Drawing Constitutional Lines After Shaw v. Reno, 92 MICH. L. REV. 588, 601 (1993).
279. 412 U.S. 755 (1973).
280. Id. at 768; see also id. at 767-68 (observing that the 'Mexican-American community... is concentrated for the most part on the west side of the city of San Antonio').
281. Id. at 768.
282. See id. at 767-70 (discussing history of discrimination against Hispanics in San Antonio ar-ea); see also Rogers v. Lodge, 458 U.S. 613, 626-28 (1982) (ruling in favor of constitutional vote dilution claim brought by spatially homogeneous African American community in rural Georgia).
283. 478 U.S. 30, 50-51 (1986); see also id. at 49 (requiring a 'politically cohesive, geographically insular minority group'); id. at 50 (stating that risk of vote dilution is higher when 'electoral mi-nority is homogeneous and insular' (quoting City of Mobile v. Bolden, 446 U.S. 55, 105 n.3 (1980) (Marshall, J., dissenting))); id. at 51 (noting that 'distinctive minority group interests' must exist before a group can prevail).
284. I should note that the Gingles prongs themselves do not amount to a spatial homogeneity requirement; rather, it is the other key language in Gingles, as well as the Court's gloss on the prongs in subsequent decisions (particularly LULAC), that have effectively created such a re-quirement. I should also note that spatial homogeneity is not the only element that plaintiffs must establish in order to prevail on a vote dilution claim. The third prong of Gingles (i.e., racial polar-ization) and its totality-of-the-circumstances inquiry are both largely unrelated to the geographic variation of the minority population.
285. Id. at 64 (plurality opinion); see also id. ('Where such characteristics are shared, race or ethnic group... functions as a shorthand notation for common social and economic characteristics.').
286. See id. at 50 n.17 (majority opinion) (noting that 'if the minority group is spread evenly' throughout an area, then it cannot prevail on a vote dilution claim).
287. Growe v. Emison, 507 U.S. 25, 41 (1993).
288. Id.
289. See Shaw v. Hunt (Shaw II), 517 U.S. 899, 916 (1996).
290. Bush v. Vera, 517 U.S. 952, 979 (1996) (plurality opinion).
291. See id. ('[District 30] could not possibly form part of a [remedial] district.'); Shaw II, 517 U.S. at 916 ('District 12 could not remedy any potential § 2 violation.').
292. LULAC, 548 U.S. 399, 435 (2006).
293. See id. at 427-29 (analyzing Gingles factors and concluding that they were satisfied); see also id. at 435 (arguing that Hispanics along Mexican border were 'cohesive' and 'had found an efficacious political identity').
294. Id. at 424.
295. 140 Id. (citations omitted) (internal quotation marks omitted); see also id. at 432 ('[T]here was a 300-mile gap between the Latino communities in District 25, and a similarly large gap between the needs and interests of the two groups.'); id. at 434 (noting the 'different characteristics, needs, and interests of the Latino community near the Mexican border and the one in and around Austin').
296. Id. at 433.
297. See id. at 434 (arguing that district's configuration 'could make it more difficult for the con-stituents in the Rio Grande Valley to control election outcomes' (quoting Session v. Perry, 298 F. Supp. 2d 451, 503 (E.D. Tex. 2004) (per curiam)) (internal quotation mark omitted)).
298. Id. (quoting Session, 298 F. Supp. 2d at 502).
299. Daniel R. Ortiz, Cultural Compactness, 105 MICH. L. REV. FIRST IMPRESSIONS 48, 50 (2006).
300. Richard H. Pildes, The Decline of Legally Mandated Minority Representation, 68 OHIO ST. L.J. 1139, 1144-1146+1159 (2007).
301. See Ortiz, supra note 144, at 48 ('In an unexpected turn, the Court adopted a new require-ment - cultural compactness - under Section 2 of the Voting Rights Act.'); Pildes, supra note 145, at 1143 (discussing the 'dramatic new principle' that emerged in LULAC).
302. In particular, the Court's governing standard in this area asks whether 'race was the pre-dominant factor motivating the legislature's decision to place a significant number of voters with-in or without a particular district.' Miller v. Johnson, 515 U.S. 900, 916 (1995).
303. 509 U.S. 630, 635-36 (1993) (quoting Shaw v. Barr, 808 F. Supp. 461, 476 (E.D.N.C. 1992) (Voorhees, C.J., concurring in part and dissenting in part)) (internal quotation mark omitted).
304. Id. at 647. On the other hand, the Court commented that districts that contain spatially homogeneous minority populations are much less objectionable. '[W]hen members of a racial group live together in one community, a reapportionment plan that concentrates members of the group in one district and excludes them from others may reflect wholly legitimate purposes.' Id. at 646.
305. See Shaw II, 517 U.S. 899, 901-02 (1996).
306. 515 U.S. 900 (1995).
307. Id. at 908.
308. Id. (quoting Johnson v. Miller, 864 F. Supp. 1354, 1389 (S.D. Ga. 1994)).
309. Id. at 919; see also id. at 908 ('[T]he social, political, and economic makeup of the Eleventh District tells a tale of disparity, not community.'); id. at 909 ('Geographically, [the district] is a monstrosity, stretching from Atlanta to Savannah. Its core is the plantation country in the center of the state, lightly populated, but heavily black. It links by narrow corridors the black neighbor-hoods in Augusta, Savannah and southern DeKalb County.' (quoting MICHAEL BARONE & GRANT UJIFUSA, THE ALMANAC OF AMERICAN POLITICS 1994, at 356 (1993)) (internal quo-tation marks omitted)).
310. Id. at 920.
311. Id. at 919.
312. Id. (quoting Johnson, 864 F. Supp. at 1389-90) (internal quotation marks omitted); see also id. at 919-20 ('It is apparent that it was not alleged shared interests but rather the object of max-imizing the district's black population... that in fact explained the General Assembly's actions.').
313. See also Bush v. Vera, 517 U.S. 952, 965, 979 (1996) (plurality opinion) (striking down Dal-las-area district that combined dissimilar African American communities).
314. 521 U.S. 567 (1997).
315. Id. at 581 (citation omitted); see also id. ('[T]he residents of District 21 regard themselves as a community.' (internal quotation mark omitted)).
316. See Easley v. Cromartie, 532 U.S. 234, 237 (2001); Hunt v. Cromartie, 526 U.S. 541, 543, 546 (1999).
317. Shaw v. Reno, 509 U.S. 630, 635 (1993).
318. Easley, 532 U.S. at 250 (quoting statement of State Senator Leslie Winner); see also Hunt, 526 U.S. at 544 (describing the updated district).
319. To be sure, the Court's focus in Easley was on the partisan motivations that explained the district's boundaries, not on the higher spatial uniformity of the district's minority population. See Easley, 532 U.S. at 243-44.
320. Bush v. Vera, 517 U.S. 952, 1037 (1996) (Stevens, J., dissenting).
321. The only similar effort of which I am aware is a 2001 white paper that used principal com-ponents analysis and clustering analysis to identify communities of interest in New York, on the basis of which state senate districts could be drawn. This paper did not consider any other states, nor did it attempt to assess districts' actual heterogeneity. See Todd Breitbart et al., Mapping Communities of Interest: The Revised Plaintiffs' Senate Plan (2002) (on file with the Harvard Law School Library).
322. See supra pp. 1912-1917.
323. See About the American Community Survey, U.S. CENSUS BUREAU, http://www.census.gov/acs/www/about_the_survey/american_community_survey/ (last visited May 3, 2012); see also Michael P. McDonald, Redistricting Developments of the Last Decade - And What's on the Table in This One, 10 ELECTION L.J. 313, 316 (2011) (noting that ACS data 'may be useful to establish communities of interest where this is a state requirement').
324. See 2009 Data Release, U.S. CENSUS BUREAU, http://www.census.gov/acs/www/data _documentation/2009_release/ (last visited May 3, 2012). The data covers the five-year period from 2005 through 2009. Id.
325. U.S. CENSUS BUREAU, GEOGRAPHIC AREAS REFERENCE MANUAL 10-1 (1994), availa-ble at http://www.census.gov/geo/www/garm.html.
326. Alan Berube & Benjamin Forman, Patchwork Cities: Patterns of Urban Popula-tion Growth in the 1990s, in 1 REDEFINING URBAN AND SUBURBAN AMERICA, supra note 36, at 75+77.
327. Barrett A. Lee et al., Beyond the Census Tract: Patterns and Determinants of Racial Segregation at Multiple Geographic Scales, 73 AM. SOC. REV. 766, 767 (2008).
328. Ming Wen et al., Ethnic Neighborhoods in Multi-Ethnic America, 1990-2000: Resurgent Ethnicity in the Ethnoburbs?, 88 SOC. FORCES 425, 426-427 (2009).
329. All of these studies relied on Census tract da-ta drawn from the (more limited) Census long form, not from the ACS survey.
330. The Census tract is also an attractive unit of analysis because it is neither so small that most variation is expressed among tracts, nor so large that most variation is expressed within tracts. See Wong, supra note 32, at 131 (finding that segregation at tract level in Connecticut is higher than at township level but lower than at block group level).
331. See, e.g., DANTE CHINNI & JAMES GIMPEL, OUR PATCHWORK NATION 222-23 (2010) (using similar variables in study of all U.S. counties); Bernadette Hanlon, A Typology of Inner-Ring Suburbs: Class, Race, and Ethnicity in U.S. Suburbia, 8 CITY & COMMUNITY 221, 225-26 (2009) (same in study of inner-ring suburbs); Thomas J. Vicino et al., Megalopolis 50 Years On: The Transformation of a City Region, 31 INT'L J. URB. & REGIONAL RES. 344, 352-53 (2007) (same in study of Northeast urban corridor); Breitbart et al., supra note 166, at 1-3 (same in study of New York geographic communities).
332. It would be ideal to include attitudinal data in the analysis as well. Unfortunately, there are no national opinion surveys large enough to generate reliable figures all the way down to the Census tract level. In a sequel to this Article, however, I incorpo-rate voting results from California's popular initiatives - which are available for tracts - into my spatial diversity calculations for the state.
333. Nicholas O. Stephanopoulos, Com-munities and the Commission, 23 STAN. L. & POL'Y REV. (forthcoming 2012); see also Makse, supra note 77, at 1-2 (using Ohio initiative data to draw districts).
334. LULAC, 548 U.S. 399, 424 (2006) (noting as relevant, in vote dilution context, 'differ-ences in socio-economic status, education, employment, health, and other characteristics' (quoting Session v. Perry, 298 F. Supp. 2d 451, 512 (E.D. Tex. 2004)) (internal quotation marks omitted)); Thornburgh v. Gingles, 478 U.S. 30, 64 (1986) (plurality opinion) (discussing importance, also in vote dilution context, of 'share[d] socioeconomic characteristics, such as income level, employment status, amount of education, housing and other living conditions, religion, language, and so forth').
335. See infra app. tbl.1.
336. Breitbart et al., supra note 166, app. (finding that results of analysis barely changed when twelve new variables were added or when eighteen were dropped). My own experiments with adding and dropping variables also indicate that the precise mix of variables is mostly im-material.
337. See, e.g., CHINNI & GIMPEL, supra note 173, at 223-25 (carrying out factor analysis in or-der to identify patterns in U.S. counties); Hanlon, supra note 173, at 227 (same to categorize inner-ring suburbs); Joel Lieske, Regional Subcultures of the United States, 55 J. POL. 888, 895 (1993) (same to identify U.S. regional subcultures); Thomas J. Vicino, The Spatial Transformation of First-Tier Suburbs, 1970 to 2000: The Case of Metropolitan Baltimore, 19 HOUSING POL'Y DE-BATE 479, 493-95 (2008) (same to analyze evolution of Baltimore suburbs); Breitbart et al., supra note 166, at 4 n.2 (same to map New York communities of interest); Makse, supra note 77 (same to draw Ohio districts).
338. Hanlon, supra note 173, at 227; see also Vicino, supra note 177, at 493 (describing factor analysis as an 'important tool in deciphering the spatial organization of urban places').
339. To be more specific, I employed principal factors analysis with varimax rotation but with-out Kaiser normalization, and I retained all factors with an eigenvalue greater than two (i.e., all factors that explained more than twice as much of the variance as the original set of variables). I settled on this approach because it yielded the most intelligible composite factors and captured the largest proportion of the data's original variance.
340. However, the results were nearly identical when I experimented with other common approaches (e.g., principal components analysis, princi-pal factors analysis with Kaiser normalization, and principal factors analysis with oblique rota-tion).
341. Douglas S. Massey et al., The Dimensions of Segregation Revisited, 25 SOC. METHODS & RES. 172, 178 (1996).
342. Finding 'repeated confirmation of basic findings' after conducting differ-ent kinds of factor analysis); Breitbart et al., supra note 166, at 4 n.2 (also finding little difference between different factor analytical techniques); see also Hanlon, supra note 173, at 227 (also using varimax rotation and retaining factors with eigenvalues near to or greater than two); Vicino, supra note 177, at 495 (retaining eigenvalues over two).
343. See, e.g., Hanlon, supra note 173, at 228-31 (identifying similar composite factors in study of inner-ring suburbs).
344. Brian A. Mikelbank, A Typology of U.S. Suburban Places, 15 HOUSING POL'Y DEBATE 935, 949-957 (2004) (same in study of suburban places).
345. F Vicino et al., supra note 173, at 354+356 (same in study of Census places in Northeast corridor).
346. Elvin K. Wyly, Continuity and Change in the Restless Urban Landscape, 75 ECON. GEOGRAPHY 309, 326 (1999) (same in study of Minneapolis metropolitan area).
347. See infra app. tbl.1. All raw variables are displayed in Table 1, while subsequent tables that show the results of factor analyses display only the raw variables with significant loadings (i.e., greater than 0.4 or less than -0.4).
348. See, e.g., CHINNI & GIMPEL, supra note 173, at 223-25 (calculating factor scores for U.S. counties); Vicino, supra note 177, at 498 (same for Baltimore suburbs); Makse, supra note 77, at 7 (same for Ohio voting precincts).
349. See 110th Congressional Districts Geographic Relationship Tables, U.S. CENSUS BUREAU (Mar. 14, 2007), http://www.census.gov/geo/www/cd110th/tables110.html. While the vast majority of tracts are located entirely within a single district, approximately eight percent of the country's population lives in tracts that are divided among two or more districts. Because tracts are de-signed to be internally homogeneous, I simply included split tracts in my calculations for all of the districts that contain them. But my results were essentially identical when I excluded split tracts altogether from my analysis.
350. See, e.g., CAMPBELL, supra note 26, at 67 (analyzing distribution of ideological preferences using standard deviation); GRONKE, supra note 27, at 40 ('Variance is an appropriate measure of dispersion for continuous variables....').
351. P.J. TAYLOR & R.J. JOHNSTON, GEOGRAPHY OF ELECTIONS 150 (1979).
352. Paul DiMaggio et al., Have Americans' Social Attitudes Become More Polarized?, 102 AM. J. SOC. 690, 696 (1996).
353. I calculated weighted standard deviations in order to take into account tracts' differing populations. See Breitbart et al., supra note 166, at 4 (also weighing variables by population). I could not compute meaningful coefficients of variance (i.e., the standard deviation divided by the mean) for my data because the factor scores produced by the analysis all have means of zero.
354. Not surprisingly, my metric of spatial diversity has only a moderate relationship to measures of districts' top-line diversity. See, e.g., Bond, supra note 25, at 202 (presenting one such meas-ure); Herrnson & Gimpel, supra note 25, at 120-21 (presenting another); see also supra note 35 (noting that districts' scores on Sullivan Index explain about sixty percent of variance in their spa-tial diversity scores).
355. See Lee et al., supra note 171, at 770 (noting that most measures of residential segregation cannot distinguish between checkerboard patterns and clustered arrangements); Reardon & O'Sullivan, supra note 32, at 123 (same).
356. See supra note 33 and accompanying text (discussing Tobler's First Law).
357. I measured spatial autocorrelation by calculating the Global Moran's I for a wide range of districts and with respect to all of the composite factors. Global Moran's I is 'the most commonly employed method of assessing the significance and/or degree of spatial autocorrelation in the data.'
358. Wendy K. Tam Cho, Contagion Effects and Ethnic Contribution Networks, 47 AM. J. POL. SCI. 368, 372 (2003).
359. Su-Yeul Chung & Lawrence A. Brown, Racial/Ethnic Residential Sorting in Spatial Context: Testing the Explanatory Frameworks, 28 URB. GEO. 312, 322 (2007)
360. Using Global Moran's I to evaluate spatial clustering of racial groups in the Columbus, Ohio, metropolitan area.
361. Jowei Chen & Jonathan Rodden, Using Legislative Districting Simulations to Measure Electoral Bias in Legislatures 8-11 (July 15, 2010) (unpublished manuscript), available at http://www-personal.umich.edu/~jowei/florida.pdf (same for spatial clustering of Democratic and Republican voters in Florida). The Global Moran's I scores that I calculated were almost always positive (indicating spatial autocorrelation) and highly statistically significant.
362. See Breitbart et al., supra note 166, at 2 (carrying out separate factor analyses for different New York regions).
363. See supra Part II, pp. 1924-35. These points were made most forcefully by Justice Powell in Karcher and Bandemer, by Justice Stevens in Vieth and LULAC, and by Justice Kennedy in Miller and LULAC. See id. Of course, Justices Powell and Stevens are no longer on the Court, meaning that the current Justices may be less receptive to these arguments than were some of their predecessors.
364. See supra section I.C, pp. 1917-1924.
365. Adam B. Cox & Richard T. Holden, Reconsidering Racial and Partisan Gerrymander-ing, 78 U. CHI. L. REV. 553, 584-585 (2011).
366. Observing that scholars have not yet determined whether 'the composition of an electoral district influence[s] voters' decisions... about whether to go to the polls').
367. I obtained congressional election results from the Clerk of the House of Representatives, Statistics of the Presidential and Congressional Elections of November 4, 2008, OFFICE OF THE CLERK, U.S. HOUSE OF REPRESENTATIVES (July 10, 2009), http://clerk.house.gov/member _info/electioninfo/2008/2008Stat.htm, and presidential election results from Presidential Results by Congressional District, 2000-2008, SWING STATE PROJECT (Dec. 12, 2008, 4:10 PM), http://www.swingstateproject.com/diary/4161/.
368. I used two-party congressional turnout figures because they are both more accurate and more relevant to the claim that spatially diverse districts cause voters to become disillusioned with the major parties. I omitted districts in which the 2008 congressional elections were either uncontested or extremely uncompetitive (i.e., one party's can-didate received more than ninety percent of the vote) because the voter roll-off in such districts cannot be attributed to spatial diversity.
369. I also omitted Alabama, Illinois, and Massachusetts dis-tricts because the presidential election results by district were not reliable for those states. I fo-cused on the 2008 election because it enables an easy comparison between top-ticket (i.e., presi-dential) and lower-ticket (i.e., congressional) voting patterns.
370. Danny Hayes & Seth C. McKee, The Participatory Effects of Redistricting, 53 AM. J. POL. SCI. 1006, 1009 (2009).
371. See id. (noting that 'a wide variety of studies have documented a strong relationship be-tween voter information and roll-off'); see also Danny Hayes & Seth C. McKee, The Intersection of Redistricting, Race, and Participation, 56 AM. J. POL. SCI. 115, 120 (2012) ('We focus on roll-off, rather than turnout, because it allows us to more closely tie redistricting to House election participation.').
372. A simple bivariate regression (t = 7.99, P = 0.000) confirms that spatial diversity is a signifi-cant predictor of voter roll-off. All of the regressions that I ran for this Article used ordinary least squares.
373. See Juliana Menasce Horowitz, Divided Government and Voter Turnout in Gubernatorial Elections 8 (unpublished manuscript) (on file with the Department of Government and Politics, University of Maryland), available at http://www.bsos.umd.edu/gvpt/apworkshop/menasce.pdf (listing variables linked to turnout).
374. See infra app. tbl.2 (showing multivariate regression results; t = 3.14 and P = 0.002 for spa-tial diversity). Of the individual factors that are included in my aggregate measure of spatial di-versity, heterogeneity in terms of African American population, Asian American population, age, and agricultural employment had the greatest impact on voter roll-off.
375. See supra note 25 and accompanying text (discussing Sullivan Index). I included the follow-ing eight categories in my computation of districts' Sullivan Index scores: race (white, black, Asian, other); ethnicity (Hispanic, non-Hispanic); age (under 19, 20 to 44, 45 to 64, over 65); household income (under Ani15,000, Ani15,000 to Ani50,000, Ani50,000 to Ani150,000, over Ani150,000).
376. Educa-tion (less than high school, high school but less than college, college but less than graduate school, graduate school); profession (management, service, sales, agriculture, construction, production); household type (married family, nonfamily, other); and housing status (owner, renter).
377. See infra app. tbl.2 (showing multivariate regression results; t = 5.59 and P = 0.000 for spa-tial diversity). As noted earlier, political scientists have obtained similar results when they studied the consequences of top-line diversity for voter turnout. See supra note 52 and accompanying text.
378. See Royce Carroll et al., DW-NOMINATE Scores with Bootstrapped Standard Errors, VOTEVIEW.COM (Feb. 3, 2011), http://www.voteview.org/dwnominate.asp. I used an average of each district's scores over the last three Congresses (i.e., 2005-2010). For the handful of districts that had more than one representative in a given Congress, I averaged the members' scores. The results for each individual Congress are very similar to the results for the entire six-year period.
379. See id. (noting that in post-1980 period only one dimension is required to capture most of variance in House members' voting patterns).
380. See supra note 182 and accompanying text (discussing calculation of factor scores).
381. As with the standard deviations, I calculated weighted averages in order to take into ac-count tracts' differing populations. See supra note 184. The results of my analysis are essentially unchanged if I use districts' median (rather than average) factor scores.
382. See infra app. tbl.3 (showing multivariate regression results); see also Bailey & Brady, supra note 25, at 537 (using analogous research design). Like political scientists who have conducted similar studies, I included interaction terms in my regressions between the variables of interest and corresponding measures of spatial diversity. See id. at 540 n.11; Gerber & Lewis, supra note 66, at 1376. However, the results of my analysis are very similar if I omit the interaction terms.
383. However, this is not a perfect control since the level of top-line diversity still varies between (as opposed to within) the two samples.
384. These results are consistent with the findings of political scientists who have investigated the voting records of representatives from top-line diverse and non-diverse constituencies. See supra notes 65-66 and accompanying text. Of course, it is only troubling for elected officials to vote in ways not predicted by their constituents' attributes if one subscribes, at least in part, to a delegate model of representation.
385. Under a pure trustee model, it is largely irrelevant how well officials' voting records are tied to their constituents' characteristics. See generally PITKIN, supra note 60 (describing delegate and trustee models of representation).
386. See infra app. tbl.3 (showing multivariate regression results). I again included an interac-tion term between Cook PVI and spatial diversity, though again, my results are almost identical if I omit the interaction term. In a separate set of regressions, I used legislators' partisan affiliation as a measure of partisanship.
387. The legislators' party membership explained ninety-one percent of the voting record variance in the most spatially diverse districts, and eighty-eight percent of the variance in the least spatially diverse districts. These results are consistent with (though much less dramatic than) those produced by the Cook PVI analysis.
388. See Introducing the Cook Political Report Partisan Voting Index (PVI) for the 111th Con-gress, COOK POL. REP. (Apr. 9, 2009), http://cookpolitical.com/node/4201.
389. Again, these results are consistent with the findings of political scientists who have studied elected officials' responsiveness to partisan pressures in top-line diverse and non-diverse constit-uencies. See supra note 67 and accompanying text.
390. Richard H. Pildes, Why the Center Does Not Hold: The Causes of Hyperpolarized Democra-cy in America, 99 CALIF. L. REV. 273, 276 (2011).
391. ALAN I. ABRAMOWITZ, THE DISAPPEARING CENTER 139-157 (2010).
392. NOLAN MCCARTY et al., POLARIZED AMERICA 15-70 (2006)
393. I considered the hundred most and least spatially diverse districts in the country, rather than the fifty, simply in order to obtain more data for the histograms. The results are almost identical if smaller numbers of districts are taken into account.
394. Arithmetically, the gap between the respective DW-Nominate averages for the two parties is 1.05 for the hundred most spatially diverse districts and 0.77 for the hundred least spatially di-verse districts. See infra figs.5 & 6.
395. Some of the voting records in the center of the distribution are attributable to my averaging DW-Nominate scores over three Congresses. See supra note 201. The histograms for very spa-tially homogeneous districts for each individual Congress reveal somewhat greater polarization - though still significantly less than do the histograms for very spatially diverse districts. Once again, these results are consistent with those of political scientists. See sources cited supra note 67.
396. Pildes, supra note 211, at 277+281.
397. See supra section II.A, pp. 1925-1928.
398. See supra note 122; see also Stephanopoulos, supra note 173 (using spatial diversity scores to analyze California districts' congruence with geographic communities). Of course, spatial diversi-ty is only a proxy for the failure to respect communities of interest. It is possible (though not very likely) that objectively dissimilar groups of people nevertheless think of themselves subjectively as a unified community.
399. Analogously, it is possible (though again unlikely) that objectively similar groups of people feel subjectively that they belong to different communities. However, even if spatial diversity is an imperfect proxy for community disruption, it is clearly superior to the usual approach: simply counting the numbers of towns and counties that a given district plan splits. See BRUNELL, supra note 60, at 66-67 (summarizing the standard approach); Note.
400. John C. Courtney, From Gerrymanders to Independence: District Boundary Readjustments in Canada, in REDISTRICTING IN COMPARATIVE PERSPECTIVE 11, 17-18.
401. Lisa Handley & Bernard Grofman eds., 2008) (explaining that in Quebec the socioeconomic homogeneity of electoral pre-cincts is used to assess how well district plans respect communities of interest).
402. See LULAC, 548 U.S. 399, 416-23 (2006) (opinion of Kennedy, J.); Vieth v. Jubelirer, 541 U.S. 267, 281-301 (2004) (plurality opinion).
403. Similarly, the Court's vote-dilution case law requires racial polarization in voting to be cal-culated, see Thornburg v. Gingles, 478 U.S. 30, 55-58 (1986), and the VRA's preclearance provi-sion focuses on the number of districts in covered states in which minority groups can successfully elect the representatives of their choice, see 42 U.S.C. § 1973c (2006). Complex quantitative anal-ysis cannot be avoided in the election law domain.
404. See infra app. tbl.4. I did not include all 435 districts simply because the full list (which is on file with the Harvard Law School Library) is somewhat unwieldy.
405. See Illinois 7th District: Chicago; Downtown, West Side, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/il/07 (last visited May 3, 2012).
406. See Pennsylvania 9th District: South Central Pennsylvania; Altoona, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/pa/09 (last visited May 3, 2012).
407. All district maps were created using Caliper Corporation's Maptitude for Redistricting software. Deeper shades of grey generally indicate higher tract scores, though I occasionally switch the usual color scheme in order to improve intelligibility. Tracts for which data is missing are blank. For information on which raw variables are included in each composite factor, see in-fra app. tbl.1. For other examples of geographic units' factor scores being mapped, see Vicino, supra note 177, at 500-01; Wyly, supra note 180, at 330-31; Breitbart et al., supra note 166; Makse, supra note 77, at 20-23.
408. Factor score maps may also be useful to those who are entrusted with drawing district lines in the first instance. Even if they are under no obligation to do so, line-drawers may choose to design districts that are spatially homogeneous with regard to some or all of the factors. See, e.g., Breitbart et al., supra note 166, at 24 ('The maps derived from... the principal components anal-ysis... show coherent geographic patterns that can be used as the basis for drawing Senate districts.').
409. See infra app. tbl.4.
410. See infra app. tbl.4; New York 8th District: West Manhattan, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/ny/08 (last visited May 3, 2012).
411. See infra app. tbl.4; New York 16th District: South Bronx, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/ny/16 (last visited May 3, 2012).
412. See infra app. tbl.4; California 30th District: Part Los Angeles, Santa Monica, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/ca/30 (last visited May 3, 2012).
413. See infra app. tbl.4; New York 15th District: North Manhattan; Harlem, NAT'L J. ALMA-NAC, http://www.nationaljournal.com/almanac/area/ny/15 (last visited May 3, 2012).
414. See infra app. tbl.4; Illinois 1st District: Chicago; South Side, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/il/01 (last visited May 3, 2012).
415. See infra app. tbl.4; Ohio 5th District: Northwest Ohio; Bowling Green, NAT'L J. ALMA-NAC, http://www.nationaljournal.com/almanac/area/oh/05 (last visited May 3, 2012).
416. See infra app. tbl.4; Florida 19th District: South Florida; Part Boca Raton, NAT'L J. AL-MANAC, http://www.nationaljournal.com/almanac/area/fl/19 (last visited May 3, 2012).
417. See infra app. tbl.4; California 43rd District: Inland Empire; Ontario, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/ca/43 (last visited May 3, 2012).
418. Cf. Pildes & Niemi, supra note 14, at 571-73 tbl.6 (calculating statewide averages of dis-tricts' compactness scores), cited in Bush v. Vera, 517 U.S. 952, 960 (1996) (plurality opinion).
419. In a sequel to this Article, I introduce a more complex - but also potentially more accu- rate - alternative to using states' raw spatial diversity averages: the calculation of states' residu-als based on a regression of their raw averages against their intrinsic levels of heterogeneity and their numbers of districts. The higher the residual, the more problematic the district plan, and vice versa. See Stephanopoulos, supra note 173 (manuscript at 19-20) (detailing this methodology).
420. For instance, the average spatial diversity of the congressional districts that California's Cit-izens Redistricting Commission recently drew is somewhat lower than the average spatial diversi-ty of California's existing districts. California's high current average thus is not inevitable. See id. (manuscript at 12-14); see also infra pp. 1963-65 (finding relationships between statewide spa-tial diversity averages and measures of gerrymandering such as partisan bias and electoral responsiveness).
421. See infra app. tbl.5. For states with just one congressional district, of course, redistrict- ing can have no impact on the statewide average (which is the same as the average for the one district).
422. See infra app. tbl.5.
423. See infra app. tbls.4 & 5.
424. See infra app. tbl.5.
425. See supra pp. 1925, 1928.
426. See 541 U.S. 267, 281-301 (2004) (plurality opinion).
427. See id. at 328, 331, 339 (Stevens, J., dissenting); id. at 349 (Souter, J., dissenting) (noting that Sixth District might support valid political gerrymandering claim).
428. See infra app. tbl.5.
429. Id.
430. See supra section III.A, pp. 1936-41 (explaining methodology for nationwide analysis).
431. Though the state-specific procedure is more accurate, its results do not differ greatly from the national analysis. In fact, the ordinal ranking of Pennsylvania's districts by their spatial di-versity is almost identical under both approaches.
432. See infra app. tbl.6.
433. Compare infra app. tbl.6, with infra app. tbl.5.
434. See infra app. tbl.7.
435. The nationwide scores computed earlier confirm this ranking. The Sixth District is the 136th most spatially diverse district in the country, while the First and Second Districts, respec-tively, are 41st and 37th. See infra app. tbl.4. This is further evidence that the Sixth District is not an outlier, by national standards, in terms of spatial diversity.
436. See Vieth Brief, supra note 2, at 13, 47-49; Pennsylvania 18th District: Pittsburgh Metro Area, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/pa/18 (last visited May 3, 2012).
437. See infra app. tbl.7.
438. See Pennsylvania 6th District: Chester and Montgomery Counties, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/pa/06 (last visited May 3, 2012). While interesting, the Sixth District's heterogeneity in terms of socioeconomic status is not as significant, in my view, as its aggregate spatial diversity incorporating all of the composite factors.
439. See supra note 11 and accompanying text.
440. See Gelman & King, supra note 11, at 544-45 (defining bias and responsiveness).
441. Reducing bias all the way to zero is unproblematic. However, very high rates of respon-siveness are undesirable because they result in large changes in seat shares despite only small shifts in vote shares. Fortunately, the responsiveness scores reported here are not high enough to raise such concerns.
442. See Election Statistics, OFF. OF THE CLERK OF THE U.S. HOUSE OF REPRESENTA-TIVES, http://clerk.house.gov/member_info/electionInfo/index.aspx (last visited May 3, 2012). I used congressional rather than presidential election data because, unlike some researchers, see, e.g., Chen & Rodden, supra note 188, I am interested in how the incumbency advantage may af-fect district plans' levels of bias and responsiveness. Line-drawers, of course, take incumbency heavily into account when they devise new district boundaries.
443. Janet Campagna & Bernard Grofman, Party Control and Bias in 1980s Congres-sional Redistricting, 52 J. POL. 1242, 1247-1248 (1990).
444. Simon Jackman, Measuring Electoral Bias: Australia, 1949-93, 24 BRIT. J. POL. SCI. 319, 331-337 (1994).
445. Chen & Rodden, supra note 188, at 14-15.
446. See, e.g., Campagna & Grofman, supra note 258, at 1247 ('[T]he assumption of uniform swing... appropriately models the types of change, i.e. partisan shifts, that we are interested in....').
447. Bernard Grofman & Gary King, The Future of Partisan Symmetry as a Judicial Test for Partisan Gerrymandering After LULAC v. Perry, 6 ELECTION L.J. 2, 11 (2007)
448. '[I]t is remark-able that the uniform partisan swing assumption does hold approximately in a vast array of dem-ocratic elections in the U.S.....'). There is also software for calculating bias and responsiveness that loosens the uniform swing assumption. See Andrew Gelman et al., JudgeIt II: A Program for Evaluating Electoral Systems and Redistricting Plans, available at http://gking.harvard.edu/judgeit. The results generated by the software typically do not differ greatly from those produced through the conventional method.
449. For an example of another study that analyzes bias and responsiveness using scatter plots.
450. Thomas R. Belin et al., Using a Density-Variation/Compactness Measure to Evaluate Redis-tricting Plans for Partisan Bias and Electoral Responsiveness, 2 STAT. POL. & POL'Y 1, 10-13 (2011).
451. The results are nevertheless similar if all states with at least five congressional districts are included in the analysis. Spatial diversity continues to have a curvilinear relationship with bias and to correlate negatively with responsiveness, though the patterns are not quite as pronounced. The results are also similar if states' regression residuals, see supra note 234, are used instead of their raw spatial diversity averages.
452. I use the quadratic best fit line here because a quadratic regression model (R2 = 0.66) cap-tures much more of the data's variance than a linear regression model (R2 = 0.01). Cf. CAMPBELL, supra note 26, at 38-39 (also using quadratic model in investigation of county heterogeneity). For all of the other scatter plots in this Article, the linear and quadratic models did not differ signifi-cantly, and I therefore used the linear best fit line.
453. See supra pp. 1921-1922.
454. In a recent article, Professors Adam Cox and Richard Holden argue that 'matching slices' is the optimal gerrymandering strategy because it makes the most efficient use of a party's most committed supporters. See Cox & Holden, supra note 192, at 567. The districts that the strategy produces tend to be spatially heterogeneous since they deliberately combine dissimilar groups of voters. See id.
455. This conclusion is consistent with the positive correlation that I found between district-level spatial diversity and electoral margin of victory. Spatially diverse districts tend to have larger margins of victory than spatially homogeneous districts; that is, they are less competitive. See also supra notes 85-88 and accompanying text (discussing existing scholarship that found higher competitiveness in districts that are more congruent with political subdivisions and media markets).
456. See supra note 124 and accompanying text.
457. See supra section II.B, pp. 1929-33 (discussing Court's vote dilution case law).
458. See supra section II.C, pp. 1933-35 (discussing Court's racial gerrymandering case law).
459. Other racial minority groups are not large or concentrated enough to warrant separate ex-amination. Asian Americans, for example, constitute less than five percent of the U.S. population, USA Quick Facts, U.S. CENSUS BUREAU, http://quickfacts.census.gov/qfd/states/00000.html (last updated Jan. 17, 2012, 4:42 PM), and are numerous enough to constitute congressional district majorities in only a handful of areas nationwide.
460. Out of 65,273 total tracts, 7960 are at least forty percent African American, and 7048 are at least forty percent Hispanic. Fifty percent of all African Americans and fifty-two percent of all Hispanics, respectively, live in these heavily minority tracts. The numbers of tracts incorporated into the various factor analyses are slightly smaller because tracts with incomplete data could not be used.
461. It might be preferable to focus not on tracts that have large minority populations (but that are also home to many other residents) but rather on the minority populations themselves. Unfor-tunately, the ACS data does not allow this sort of analysis. Still, my results were not appreciably different when I examined tracts that were thirty, fifty, or sixty percent African American or His-panic.
462. John R. Logan & Wenquan Zhang, Identifying Ethnic Neighborhoods with Census Data, in SPATIALLY INTEGRATED SOCIAL SCIENCE 113, 116-117
463. Michael F. Goodchild & Donald G. Janelle eds., 2004) (also analyzing minority populations by examining tracts with large minority population shares).
464. See supra section III.C.2, pp. 1960-1963.
465. See infra app. tbl.8.
466. See infra app. tbl.9.
467. See infra app. tbls.10-11.
468. If anything, the relevant universe is smaller, because the Court has never struck down a non-majority-minority district as a racial gerrymander, and because a cause of action can only be stated under section 2 of the VRA if the relevant minority group could constitute the majority of a single-member district. See Bartlett v. Strickland, 129 S. Ct. 1231, 1246, 1249 (2009). Moreover, I consider all districts whose whole populations are at least forty percent African American or His-panic, while the more relevant figure may be the racial share of the citizen voting-age population.
469. See infra Figures 14-15.
470. 278 See supra section II.C, pp. 1933-35; see also Richard H. Pildes, The Supreme Court, 2003 Term - Foreword: The Constitutionalization of Democratic Politics, 118 HARV. L. REV. 28, 67-68 (2004) (noting the failure of almost every racial gerrymandering lawsuit in the 2000s).
471. See supra Figure 7 (spatial diversity histogram of all congressional districts). This is admit-tedly a bit of an apples-to-oranges comparison, but it still indicates that minority populations in heavily minority districts are not exceptionally spatially heterogeneous.
472. See infra app. tbls.10-11. The districts' minority populations are also the most spatially heterogeneous in the country with respect to each of the composite factors displayed in the maps.
473. See Florida 18th District: South Florida; Miami, NAT'L J. ALMANAC, http://nationaljournal.com/almanac/area/fl/18 (last visited May 3, 2012).
474. For ease of presentation, only the northern half of this district (which includes most of its population) is displayed.
475. See Florida 21st District: South Florida; Hialeah, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/fl/21 (last visited May 3, 2012).
476. See Maryland 4th District: Prince George's County, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/md/04 (last visited May 3, 2012).
477. Cf. Breitbart et al., supra note 166, at 25 (using similar factor-analytic approach to 'ensure that the black or Hispanic majorities in the resulting districts would have shared characteristics and interests extending far beyond race or Hispanic origin'); id. at 26-31.
478. See Illinois 4th District: Chicago; North and Southwest Sides, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/il/04 (last visited May 3, 2012).
479. Interestingly, the Fourth District forms a sheath around the Seventh, which is the country's most spatially diverse constituency overall. See supra section III.C.1, pp. 1949-60.
480. See infra app. tbl.11.
481. See King v. State Bd. of Elections (King II), 979 F. Supp. 619 (N.D. Ill. 1997) (three-judge court); King v. State Bd. of Elections (King I), 979 F. Supp. 582 (N.D. Ill. 1996) (three-judge court), vacated, 519 U.S. 978 (1996). These decisions involved the Fourth District as it was drawn in the 1990s, but that earlier rendition had the same basic shape as the current constituency.
482. King I, 979 F. Supp. at 613 n.59 (citing the testimony of Congressman Luis Gutierrez); see also id. (referring to the 'historical as well as present-day solidarity and cohesiveness of the Chi-cago Latino community'); Illinois 4th District: Chicago; North and Southwest Sides, supra note 285 (explaining the geographic distribution of the Fourth District's Hispanic community).
483. See North Carolina 1st District: Northern Region; Part Goldsboro, NAT'L J. ALMANAC, http://www.nationaljournal.com/almanac/area/nc/01 (last visited May 3, 2012).
484. See infra app. tbl.10.
485. Cromartie v. Hunt, 133 F. Supp. 2d 407, 432 (E.D.N.C. 2000) (Thornburg, J., concurring in part and dissenting in part), rev'd on other grounds sub nom., Easley v. Cromartie, 532 U.S. 234 (2001). This decision involved the 1990s version of the First District, but the 2000s version is not very different.
486. As these examples suggest, and as my empirical analysis confirms, districts' geographic compactness has little connection to their level of spatial diversity. A district may be shaped very strangely but be very spatially homogeneous, and vice versa.
487. See supra notes 137-46 and accompanying text.
488. See LULAC, 548 U.S. 399, 435 (2006).
489. See id. at 501 (Roberts, C.J., concurring in part, concurring in the judgment in part, and dissenting in part) ('What is blushingly ironic is that the district preferred by the majority - for-mer District 23 - suffers from the same 'flaw' the majority ascribes to District 25, except to a greater degree.').
490. See League of United Latin Am. Citizens (LULAC) v. Perry, 457 F. Supp. 2d 716, 720-21 (E.D. Tex. 2006) (three-judge court).
491. See infra app. tbl.12.
492. See infra app. tbl.13. The Twenty-Third District was not majority-Hispanic when LULAC was decided, but both its original and final versions did have Hispanic majorities. See LULAC, 548 U.S. at 427; LULAC, 457 F. Supp. 2d at 721. I consider only districts with Hispanic majori-ties (in terms of citizen voting-age population) because they are the only ones that the Court took into account in LULAC. See LULAC, 548 U.S. at 428-30.
493. Interestingly, Texan Hispanics who are married homeowners tend to be poorer than their counterparts who are unmarried apartment-dwellers. See infra app. tbl.12 (showing negative loading for median household income for urban/suburban factor). The opposite relationship holds in the general population. See infra app. tbl.1.
494. LULAC, 548 U.S. at 499 (Roberts, C.J., concurring in part, concurring in the judgment in part, and dissenting in part) (quoting Session v. Perry, 293 F. Supp. 2d 451, 503 (E.D. Tex. 2004)).
495. See infra app. tbl.13. Even the current version of the Twenty-Third has a Hispanic popula-tion that is more spatially heterogeneous than average compared to its peers around the country. See infra app. tbl.11.
496. See LULAC, 548 U.S. at 507, 511 (Roberts, C.J., concurring in part, concurring in the judg-ment in part, and dissenting in part).
497. LULAC also involved a political gerrymandering challenge to Texas's entire statewide plan. See id. at 413-14 (majority opinion); id. at 414-23 (opinion of Kennedy, J.). As measured by the average spatial diversity of its districts, the plan was quite problematic, ranking seventh in the country overall, and third among the states with ten or more districts. See infra app. tbl.5. Had it been available at the time, this information might have weighed in favor of the invalidation of the plan on constitutional grounds.
498. Specifically, the Hispanic population of the new Twenty-Fifth District had a spatial diversi-ty score of 0.78. See infra app. tbl.11 (listing spatial diversity scores of current heavily Hispanic districts).
499. See infra app. tbl.10 (listing spatial diversity scores of current heavily African American districts).
500. See Pildes, supra note 145, at 1146 (noting that LULAC might eliminate states' 'VRA obli-gations to create districts that, for example, bring together urban and rural minorities, or subur-ban and city ones').
501. See LULAC, 548 U.S. at 435 (emphasizing that it is not only the Hispanic population's spa-tial heterogeneity, but also the 'enormous geographical distance separating the Austin and Mexican-border communities,' that 'renders District 25 noncompact for § 2 purposes').
502. See generally Stephanopoulos, supra note 10; Stephanopoulos, supra note 173. In future work, I also hope to apply the concept of spatial diversity to geographically defined entities other than districts, e.g., political subdivisions such as towns and counties.