Amicus Curiae Brief — Harris v. Ariz. Indenpendent Redistricting Comm'n, 135 S. Ct. 2926 (2015) (No. 14-232)

No. 14-232

IN THE

Supreme Court of the United States

WESLEY W. HARRIS, et al.,

Appellants,

v.

ARIZONA INDEPENDENT REDISTRICTING

COMMISSION, et al.,

Appellees.

ON APPEAL FROM THE UNITED STATES DISTRICT

COURT FOR THE DISTRICT OF ARIZONA

BRIEF FOR AMICI CURIAE PROFESSORS

NICHOLAS STEPHANOPOULOS AND

SIMON JACKMAN, ELECTION LAW SCHOLARS,

IN SUPPORT OF APPELLEES

JASON D. HIRSCH THOMAS G. SAUNDERS

WILMER CUTLER PICKERING Counsel of Record

HALE AND DORR LLP KEVIN LAMB*

7 World Trade Center DEREK WOODMAN**

250 Greenwich Street WILMER CUTLER PICKERING

New York, NY 10007 HALE AND DORR LLP

1875 Pennsylvania Ave., NW

Washington, DC 20006

(202) 663-6000

thomas.saunders@wilmerhale.com

TABLE OF CONTENTS

Page

TABLE OF AUTHORITIES .........cccsccsssescscsseessesssesseeeees iii

INTEREST OF AMICI CURIAE..n..ccccccccsccsscsessseesseeseeee 1

SUMMARY OF ARGUMENT ..0........cccccccescsessssssseecseeoveens 2

(ESET See a are 3

I. ARIZONA’S REDISTRICTING PLAN DOES

NoT EXHIBIT SYMMETRY SCORES INDICA-

TIVE OF PRO-DEMOCRATIC PARTISAN IN-

A. The Efficiency Gap Provides A Relia-

ble Metric Of Symmetry That Ad-

dresses Critics’ Concerns About Parti-

1. By focusing on wasted votes, the

efficiency gap goes to the essence

of partisan gerrymandering ....................-.++ 6

2. The efficiency gap addresses the

major criticisms of partisan bias.................. 7

3. Historical data on state plans’ effi-

ciency gaps provides a way to as-

sess an existing plan’s partisan im-

ee 10

B. Neither Metric Supports The Inference

That Arizona’s Redistricting Plan Was

Designed To Benefit Democratic Can-

II. POPULATION DEVIATION DOES NOT RELI-

ABLY INDICATE A PARTISAN GERRYMAN-

ii

TABLE OF CONTENTS—Continued

Page

A. There Is No Correlation Between Pop-

ulation Deviation And Partisan Sym-

B. Gerrymandering Typically Involves

Little Or No Deviation In Population

FU cstnctissnennenasntnenesnnnenesinnnesseossenanionannideninememmeccsnnents 19

ED crzcsnsereszessenesnessensnsnnncersennenteonemnnneetsnnanents 21

APPENDIX: Symmetry Scores of Current

EIT esieccserensniinecinenestininsinmninnennnsnedtininpenmatpeden la

TABLE OF AUTHORITIES

CASES

Page(s)

Alabama Legislative Black Caucus v.

Alabama, 135 S. Ct. 1257 (2015)............cccccceeeeeeseeees 19

Arizona State Legislature v. Arizona

Independent Redistricting Commission,

REE ao 4

Brown v. Thomson, 462 U.S. 835 (1983)........ccccceseesere 18

Davis v. Bandemer, 478 U.S. 109 (1986) ..........ccccccccessees 16

Karcher v. Daggett, 462 U.S. T25 (1983)............c.ccecceeseeee 15

Larios v. Cox, 300 F. Supp. 2d 1320 (N.D. Ga.

SETI issincaiciianincaiecehineeibenieasiiaaiaineinivehiisidatanaieaminteenenentaane 14

League of United Latin American Citizens v.

Perry, 548 U.S. 399 (2006) .......cce.cccccsessseneeee 4, 7,8, 19

Reynolds v. Sims, 377 U.S. 533 (1964) 20.0.0... cccecceseeees 16

Vieth v. Jubelirer, 541 U.S. 267 (2004)................. 4,10, 11

OTHER AUTHORITIES

Brief of Amici Curiae Professor King et al. in

Support of Neither Party, League of

United Latin American Citizens v. Perry,

548 U.S. 399 (2006) (No. 05-204) .00......eeececeseseeceneeees 4

Grofman, Bernard & Gary King, The Future of

Partisan Symmetry as a Judicial Test for

Partisan Gerrymandering After LULAC

v. Perry, 6 Election LJ. 1 (2007) ............ccceccceseeceee 8,9

Issacharoff, Samuel, Supreme Court Destabi-

lization of Single-Member Districts, 1995 U.

ee 20

iv

TABLE OF AUTHORITIES—Continued

Page(s)

Jackman Report, Plaintiffs’ Ex. 3, ECF No. 1-3,

Whitford v. Nichol, 3:15-cv-421-bbe (W.D.

as Gi SE ceesecensccesnsnesenscctsensmeeneennnemnnans passim

Karlan, Pamela S., The Fire Next Time: Reap-

portionment After the 2000 Census, 50 Stan.

ie, Ce nretdnttceaenonenennsieinincanemntenneeieneses 20

McGhee, Eric, Measuring Partisan Bias in

Single-Member District Electoral Systems,

39 Legis. Stud. Q. 55 (2014) ............ecseesseeeeesesssenesees i)

Pildes, Richard H., Principled Limitations on

Racial and Partisan Redistricting, 106

a iid icaianish ce cccaintiaiiaicitinienbipeenans 20

Stephanopoulos, Nicholas O. & Eric M.

McGhee, Partisan Gerrymandering and

the Efficiency Gap, 82 U. Chi. L. Rev. 831

et ercnsecicnesessnentncenennesersseummianssnnanenscenmeneen passim

Stephanopoulos, Nicholas O., Spatial Diversity,

125 Harv. L. Rev. 1903 (2012)..............ccccccerscescsenceeess 8

Toobin, Jeffrey, The Great Election Grab, The

eB ee 20

INTEREST OF AMICI CURIAE'

Professor Nicholas Stephanopoulos and Professor

Simon Jackman are election scholars who study parti-

san gerrymandering. They have a professional interest

in the proper disposition of challenges to redistricting

plans and believe that such challenges should be decid-

ed with the benefit of a reliable measure of a given

plan’s partisan symmetry.

Professor Stephanopoulos is Assistant Professor of

Law at the University of Chicago Law School. His re-

search and teaching include the fields of election law,

constitutional law, legislation, administrative law, com-

parative law, and local government law. Together with

Dr. Eric McGhee, he has developed a quantitative

measure of partisan symmetry called the “efficiency

gap” and has computed this gap for both congressional

and state house plans over the entire modern redis-

tricting era, from 1972 to 2012. See Stephanopoulos &

McGhee, Partisan Gerrymandering and the Efficiency

Gap, 82 U. Chi. L. Rev. 831 (2015).

Professor Jackman is Professor of Political Science

and, by courtesy, Professor of Statistics at Stanford

University. He is also a fellow of the Society for Politi-

cal Methodology and a member of the American Acad-

emy of Arts and Sciences. He teaches courses on

' By letters on file with the Clerk, all parties have consented

to the filing of this brief. Pursuant to Supreme Court Rule 37.6,

amici state that no counsel for a party authored this brief in whole

or in part; no counsel or party made a monetary contribution in-

tended to fund the preparation or submission of this brief; and no

person other than amici, their members, or their counsel made

such a monetary contribution.

2

American politics and statistical methods in the social

sciences. His expert report on Wisconsin’s redistrict-

ing plan, which includes efficiency gap calculations from

1972 to 2014, was recently filed in pending litigation.

See Jackman Report, Pls.’ Ex. 3, ECF No. 1-3, Whitford

v. Nichol, 3:15-ev-421-bbe (W.D. Wis. July 8, 2015).

Amici write to show that: (1) Arizona’s plan does

not exhibit symmetry scores indicative of pro-

Democratic partisan intent; and (2) small deviations in

district population size do not reliably indicate a parti-

san gerrymander.

SUMMARY OF ARGUMENT

A partisan gerrymander is a redistricting scheme

that helps one party translate its popular support into

legislative power more efficiently than its rival. This

greater efficiency in converting votes into seats is ac-

complished by creating districts in which that party

does not “waste” as many votes—either by losing seats

or by winning seats by unnecessarily high margins—as

its adversary. By denying the other side a share of

seats that fairly reflects its level of statewide backing, a

gerrymander undermines the election’s democratic le-

gitimacy and distorts the policies eventually enacted by

the legislature.

While many scholars support the use of partisan

symmetry to assess district plans, the question has

been how to measure symmetry empirically. The “effi-

ciency gap,” advanced by amici here, measures sym-

metry by comparing the parties’ respective shares of

wasted votes in actual elections. This newer metric

overcomes the major criticisms of the primary alterna-

tive metric proposed in the past: “partisan bias.”

3

Under either metric, the Arizona plan before the

Court somewhat favors Republicans, exhibiting a pro-

Republican efficiency gap of 5% and a pro-Republican

partisan bias of 8.5% over the 2012 and 2014 elections.

These partisan symmetry scores do not support the in-

ference appellants would have the Court draw that the

plan was designed with the aim of benefiting Democrat-

ic candidates.

Finally, although the Arizona plan features minor

variations in district population size, a broad statistical

comparison of state house plans reveals no meaningful

correlation between population deviation and either the

efficiency gap or partisan bias. Comprehensive elec-

toral data and sophisticated computer modeling are the

standard tools of most gerrymanders, and these tools

readily permit states to minimize deviation while ad-

vancing partisan ends.

ARGUMENT

I ARIZONA'S REDISTRICTING PLAN DOES NOT EXHIBIT

SYMMETRY SCORES INDICATIVE OF PRO-DEMOCRATIC

PARTISAN INTENT

In any election, there are two kinds of votes that do

not help a party win additional seats: votes that are

cast for a losing candidate and votes that are cast for a

winning candidate beyond the threshold required for

victory. In that sense, these are “wasted” votes. A

partisan gerrymander favors a party in statewide elec-

tions by reducing the number of votes the party wastes

compared to its adversary. Under a neutral plan, nei-

ther side wastes significantly more votes than the oth-

er, and consequently a given percentage of the

statewide vote yields about the same share of seats, re-

gardless of the party for whom the votes are cast. As

4

this Court has acknowledged, severe partisan gerry-

manders entrench a party in power and thereby violate

“democratic principles.” Arizona State Legislature v.

Arizona Indep. Redistricting Comm'n, 135 S. Ct. 2652,

2658 (2015) (quoting Vieth v. Jwbelirer, 541 U.S. 267,

292 (2004) (plurality opinion); id. at 316 (Kennedy, J.,

concurring in the judgment)).

Although the Court has held gerrymandering

claims to be justiciable, it has yet to adopt a substantive

standard to apply in these cases. See League of United

Latin Am. Citizens v. Perry, 548 U.S. 399, 413-414

(2006) (“LULAC”). The concept of symmetry has gar-

nered a strong consensus among scholars and is sup-

ported by decades of research. See Brief of Amici Curi-

ae Professor King et al. in Support of Neither Party 6-

7, LULAC, 548 U.S. 399 (No. 05-204) (citing sources).

Members of the LULAC Court expressed an interest in

symmetry as a promising theory. 548 U.S. at 466-468

(Stevens, J., concurring in part and dissenting in part),

483-484 (Souter, J., joined by Ginsburg, J., concurring

in part and dissenting in part), 491-492 (Breyer, J., con-

curring in part and dissenting in part). The question

remains how to measure symmetry.

Justice Kennedy, while not “altogether discounting

[symmetry’s] utility in redistricting planning and litiga-

tion,” has identified a number of concerns about using

the “partisan bias” metric proposed by amici in LU-

LAC. 548 U.S. at 419-420 (plurality opinion). Partisan

bias measures symmetry by reference to the results of

a hypothetical election in which the parties evenly split

the statewide vote. The major criticisms of this metric

concern assumptions about shifting voter preferences

that are made in order to generate this imaginary tie.

5

The “efficiency gap,” a newer metric advanced by

amici here, answers these criticisms by comparing the

parties’ respective shares of wasted votes in actual

elections. By either metric, there is no indication that

the Arizona plan before the Court favors Democratic

candidates—and, hence, no support for the inference

that the plan’s motivation was Democratic partisan ad-

vantage.

A. The Efficiency Gap Provides A Reliable Met-

ric Of Symmetry That Addresses Critics’

Concerns About Partisan Bias

As noted, a partisan gerrymander unfairly enhanc-

es one party’s ability to translate its votes into electoral

victories. The typical gerrymander works by a combi-

nation of “packing,” i.e., clustering voters for the disfa-

vored party into a few safe districts for that party, and

“cracking,” i.e., dispersing the remainder of those vot-

ers throughout districts won by the favored party. Ina

two-candidate race, the losing candidate’s votes—as

well as the winning candidate’s votes in excess of the

50% (plus one) threshold required to carry the seat—

could be put to better use securing victories in other

districts. For that reason, if a party’s goal is to maxim-

ize the seats it wins in any given election, both lost and

surplus votes are “wasted.” At bottom, a partisan ger-

rymander is a redistricting plan that ensures that one

party will exploit its popular support more efficiently

than its rival, wasting fewer votes and thus securing

more seats with the votes it receives.

6

1. By focusing on wasted votes, the effi-

ciency gap goes to the essence of parti-

san gerrymandering

The value of analyzing partisan gerrymandering in

terms of the “efficiency gap” lies in combining into a

single figure a comparison of the parties’ respective

shares of wasted votes. This comparison reflects the

series of packing and cracking decisions built into a

state’s redistricting plan. Because the core issue in

gerrymandering cases is the difference between the

two parties’ abilities to convert their votes into seats,

the efficiency gap provides a metric that reflects the

actual mechanism that produces asymmetry.

The efficiency gap is computed in two steps—by

first summing the votes each party wastes in an elec-

tion and then dividing the difference between the two

sums by the total number of votes cast. See Jackman

Report 15-19. To take a simple example, suppose that

in a state with 10 districts of 100 voters each, Party A

receives 70 votes in districts 1-3, 54 votes in districts 4-

8, and 35 votes in districts 9-10, with the remaining

votes being won by Party B. Party A wins the election

but, as in any single-member district system, still ends

up wasting a substantial number of votes—150 votes in

all (20 in districts 1-3, 4 in districts 4-8, and 35 in dis-

tricts 9-10). Party B, however, wastes many more

votes—350 votes in al! (30 in districts 1-3, 46 in districts

4-8, and 15 in districts 9-10). The resulting efficiency

gap, i.e., the difference between the two parties’ sums

of wasted votes divided by the total number of votes

cast, is 0.2, or 20% ((350-150)/1000). For a detailed

chart, see Stephanopoulos & McGhee, 82 U. Chi. L.

Rev. at 852.

7

A key property of the efficiency gap is that it re-

veals a party’s excess share of seats relative to a neu-

tral plan. See Jackman Report 19. In the above exam-

ple, Party A receives 550 votes and wins 8 seats, cap-

turing 80% of the districts with 55% of the statewide

vote. An efficiency gap of 20% indicates that, out of the

10 races, Party A won 2 (0.2 x 10) it would otherwise

have lost had there been an efficiency gap of zero. A

neutral district plan would award 6 seats to Party A.

In comparison, the state’s plan confers a competitive

advantage sufficient to account for 2 extra seats. Giv-

ing one party the discrepant power to amass additional

seats relative to its share of votes is the crux of a parti-

san gerrymander.

2. The efficiency gap addresses the major

criticisms of partisan bias

The longstanding precursor to the efficiency gap is

the metric proposed by other amici in LULAC: parti-

san bias. In contrast to the efficiency gap, which focus-

es on the parties’ wasted votes in actual elections, par-

tisan bias measures asymmetry by looking at the di-

vergence in the share of seats each party would win in a

hypothetical election in which the two parties received

the same share—normally half—of the statewide vote.

See Jackman Report 11-13. For example, if Democrats

would win 55% of the seats with 50% of the vote, then

the district plan has a 5% pro-Democratic bias.

While partisan bias is unquestionably useful in

identifying asymmetry, scholars and this Court have

noted some of the metric’s potential pitfalls. LULAC,

548 U.S. at 420 (plurality opinion). The efficiency gap

avoids these pitfalls by comparing the parties’ abilities

to convert votes into seats in actual elections.

8

In the extraordinary circumstance where the two

parties split the statewide vote down the middle, the

efficiency gap and partisan bias are identical. Even in

that special case, however, the two metrics measure

different things. Partisan bias measures the skew in

outcome by reference to the results of a hypothetical

tie. It so happens that because the parties did tie, the

election results make conjecture unnecessary. The effi-

ciency gap, on the other hand, measures the manipula-

tion of wasted votes that leads to the skewed outcome

in the first place. In any election that is not tied, the

two metrics diverge, and the further an election is from

being tied, the more they diverge. See Stephanopoulos

& McGhee, 82 U. Chi. L. Rev. at 856-857.

The shortcomings of partisan bias stem from the

need to compute the results of a tied election, which in

all but the rarest cases is counterfactual. See Jackman

Report 14-15. To compute a plan’s bias, the parties’

vote shares must be shifted in some way to generate an

imaginary 50-50 split. The classic way to accomplish

this is to assume a uniform swing in votes to the losing

party. See id. at 13-14. A major criticism of partisan

bias is that this uniform swing assumption is unrealistic

and depends on “conjecture about where possible vote-

switchers will reside.” LULAC, 548 U.S. at 420 (plural-

ity opinion). Because the makeup of districts is not ho-

mogeneous, vote-shifters are unlikely to be spread

evenly throughout the state; hence, any shift in the

statewide vote is likely to be the product of uneven

shifts among districts. See Stephanopoulos, Spatial Di-

versity, 125 Harv. L. Rev. 1903, 1940-1941 (2012). Alt-

hough the assumption of uniformity can be relaxed and

made more sophisticated, see Grofman & King, The Fu-

ture of Partisan Symmetry as a Judicial Test for Par-

tisan Gerrymandering After LULAC v. Perry, 6 Elec-

9

tion LJ. 1, 4 (2007), the problem to some extent per-

sists, see Stephanopoulos & McGhee, 82 U. Chi. L. Rev.

at 859-860.

In competitive elections, in which the parties’

statewide vote shares are within 10 points of each oth-

er, partisan bias works well. McGhee, Measuring Par-

tisan Bias in Single-Member District Electoral Sys-

tems, 39 Legis. Stud. Q. 55, 67 (2014). In uncompetitive

elections, however, the uniform swing assumption is

sufficiently implausible to make the metric unreliable.

Even proponents of partisan bias admit that the meth-

od is not useful below a certain threshold of competi-

tiveness. See Grofman & King, 6 Election LJ. at 19.

But gerrymandering remains an issue in states that

strongly favor one party. Indeed, a gerrymander may

permit the already dominant party to achieve the su-

permajority needed for extraordinary legislative

measures.

The efficiency gap does not have these issues be-

cause it does not require any counterfactual analysis.

The crucial input is the difference between the parties’

wasted votes in the actual election. So long as election

results are available at the district level, calculating the

efficiency gap is straightforward. The process does not

entail unrealistic assumptions about how vote-shifters

are distributed throughout the state. And it can be

meaningfully calculated even for uncompetitive elec-

tions in states dominated by a single party.

10

3. Historical data on state plans’ efficiency

gaps provides a way to assess an existing

plan's partisan implications

The efficiency gap has been computed for congres-

sional and state house results over the entire modern

redistricting era, from 1972 to 2014.” See Stephanopou-

los & McGhee, 82 U. Chi. L. Rev. at 867-873; see also

Jackman Report 19-21. As a result, there is a large da-

taset from which to draw empirically based conclusions

about both the history of gerrymandering and the par-

tisanship of district plans currently in use.

By measuring the real source of competitive ad-

vantage—t.e., disparities in the parties’ respective

shares of wasted votes—the efficiency gap can provide

“clear, manageable, and politically neutral standards

for measuring the particular burden a given partisan

classification imposes on representational rights.” Vi-

eth, 541 U.S. at 307-308 (Kennedy, J., concurring in the

judgment). Although devised only in the past few

years, this metric is already being employed in gerry-

mandering litigation. See Jackman Report.

There are two related challenges in gerrymander-

ing cases that this dataset allows election experts, liti-

gants, and courts to overcome. First, since virtually

any plan will create some imbalance in an election,

there needs to be a way to set thresholds above which

an efficiency gap is truly large and problematic. Sec-

ond, and relatedly, voters’ preferences change from one

? ‘The primary dataset developed by Stephanopoulos and

McGhee uses congressional plans with at least 8 seats and single-

member state house races because the efficiency gap is more diffi-

cult to reliably calculate for smaller plans and multimember dis-

tricts. See Stephanopoulos & McGhee, 82 U. Chi. L. Rev. at 868.

11

election to the next; hence, one has to take into account

these swings to distinguish plans whose efficiency gap

indicates a durable asymmetry from plans whose gap is

likely to unravel over their own lifetimes. From the

standpoint of partisan symmetry, only those plans

whose efficiency gap is large and likely to persist over

multiple elections merit judicial intervention.

The first challenge (1.e., “determining when politi-

cal gerrymandering has gone too far”) has long been

recognized as “[tJhe central problem” in gerrymander-

ing cases. Vieth, 541 U.S. at 296-297 (plurality opinion).

For both congressional and state house races, however,

the dataset includes the historical as well as current

distribution of state plans’ efficiency gaps and thus can

greatly inform the choice of reasonable thresholds of

presumptive unconstitutionality. See Stephanopoulos

& McGhee, 82 U. Chi. L. Rev. at 885-891; Jackman Re-

port 56-59, 63-69.

As to the second challenge, a plan’s efficiency gap

in one election is not necessarily a strong predictor of

its gap in the next. Some plans nevertheless feature

large and durable gaps over several elections. Stepha-

nopoulos & McGhee, 82 U. Chi. L. Rev. at 864-865.

Sensitivity testing can be used to ascertain the likely

persistence of a large asymmetry. That testing uses

the magnitude of historical swings in voter preferences

to estimate the likelihood that a plan will remain

skewed over its lifetime. Jd. at 889-890. In addition,

the durability of a plan’s efficiency gap is linked to the

gap’s initial size. Given this initial size, the probability

that a plan will feature a consistent tilt can be accurate-

ly estimated. See Jackman Report 56-69.

In sum, the efficiency gap offers a reliable metric of

partisan symmetry that is empirical, easily calculable,

12

and impervious to the major criticisms of partisan bias.

The data already compiled on plans’ gaps over the last

four decades also helps to render tractable the chal-

lenges in operationalizing this metric in litigation.

B. Neither Metric Supports The Inference That

Arizona's Redistricting Plan Was Designed To

Benefit Democratic Candidates

The data on efficiency gaps in congressional and

state house races from 1972 to 2014 supports three

background conclusions pertinent to this case. First, as

a historical matter, most district plans are reasonably

balanced and reasonably likely to favor different par-

ties at different points during their ten-year lifespans.

Second, only in recent years has the average efficiency

gap increased noticeably. And, finally, most of the liti-

gation that has so far been brought has not targeted

state plans with truly large and durable efficiency gaps.

See Stephanopoulos & McGhee, 82 U. Chi. L. Rev. at

875-877; Jackman Report 44-47.

Where does the Arizona plan before the Court fit

into this overall picture? Figures 1 and 2 show the dis-

tributions of state house plans’ efficiency gaps and par-

tisan biases over the entire 1972-2014 period as well as

in the 2012 and 2014 elections. Despite the uptick in

the level of partisan gerrymandering nationwide, Ari-

zona belongs to a large group of 22 states whose cur-

rent plans are only modestly asymmetrical, as meas-

ured by their average absolute efficiency gap. See

App.’ Specifically, as shown in Figure 1, the Arizona

> These 22 states have an absolute efficiency gap less than or

equal to Arizona’s. See App.

13

plan has a gap of -5%. Id.‘ Republicans waste 5% few-

er votes than Democrats and, thus, win 3 seats (0.05 x

60 seats) in excess of their share under a neutral plan

with an efficiency gap of zero.

As shown in Figure 2, the Arizona plan also tilts in

Republicans’ favor as measured by partisan bias. The

plan’s bias, when averaged over 2012 and 2014, is -8.5%.

App. Thus, in a hypothetically tied election, where

each side received exactly half of the statewide vote,

Republicans would be expected to pick up 58.5% or 35

(0.585 x 60) of the state house seats.

Regardless of the metric used, there is no indica-

tion that the Arizona plan favors Democratic candi-

dates, either taken on its own or in comparison with

most other state house plans. Again, the plan’s average

efficiency gap is modestly pro-Republican, and its aver-

age partisan bias is noticeably (if not egregiously) so.

Similarly, of the 40 state plans surveyed, at least 25 are

more pro-Democratic by each metric. Republicans con-

* In many states, including Arizona, there are a number of un-

contested elections. These pose a problem for any measure of par-

tisan symmetry because the parties’ vote shares are unknown and

must be estimated. Vote shares are typically imputed to districts

with uncontested elections by using presidential races, the results

of previous years when the seats were contested, or other varia-

bles shown to be good predictors. See Stephanopoulos & McGhee,

82 U. Chi. L. Rev. at 865-867; Jackman Report 24-32. Elections in

multimember districts pose a similar problem when a party does

not field enough candidates to fill the available seats.

Arizona, which has 30 two-member state house districts, had

several such uncontested or “under-contested” races in 2012 and

2014. In these cases, vote shares have been estimated using a

model based on the presidential vote in each state house district.

See Jackman Report 26-28.

14

sequently fare better under Arizona’s plan than under

most states’ plans. See App.

These partisan symmetry scores do not “speak for

themselves” as signs of pro-Democratic intent. Larios

v. Cox, 300 F. Supp. 2d 1320, 1330 (N.D. Ga.), aff'd, 542

U.S. 947 (2004). On the contrary, insofar as partisan

effect is probative of partisan intent, the scores under-

mine appellants’ claim that the Arizona plan was de-

signed with the motive of assisting Democrats.

Fig. 1. State House Plans’ Efficiency Gap Scores

he

15

Fig. 2. State House Plans’ Partisan Bias Scores

tw 4

a

@ «

—s

Il. POPULATION DEVIATION DOES NOT RELIABLY INDI-

CATE A PARTISAN GERRYMANDER

Members of this Court have long recognized that

redistricting plans can harm political opponents even

when those plans scrupulously adhere to the rule of

“one person, one vote.” See Davis v. Bandemer, 478

U.S. 109, 174 (1986) (Powell, J., concurring in part and

dissenting in part); Karcher v. Daggett, 462 U.S. 725,

752 n.10 (1983) (Stevens, J., concurring); id. at 776

(White, J., dissenting); id. at 785 (Powell, J., dissenting).

Analysis of state plans’ efficiency gap and partisan bias

scores supports a further conclusion: deviation in dis-

trict population size is not at all statistically correlated

with any increase in partisan asymmetry. Rather, in

recent decades, block-by-block census data and more

advanced computer software have provided the sophis-

ticated, readily available tools by which modern district

16

plans are able to accomplish partisan ends without sub-

stantially varying district population.

A. There Is No Correlation Between Population

Deviation And Partisan Asymmetry

Departures from perfect equality among voters

may be constitutionally suspect independent of a plan’s

partisan effects, at least if those departures are suffi-

ciently large to implicate the requirement of “one per-

son, one vote.” Reynolds v. Sims, 377 U.S. 533, 579

(1964) (“(Tjhe overriding objective must be substantial

equality of population among the various districts[.!”).

But deviations from numerical equality in no way relate

to partisan asymmetry under either metric. Figures 3

and 4 illustrate the point.

Fig. 3. Population Deviation vs. Efficiency Gap

_

orm se

evan onc

om

= ne = our

3 : ’ en @wv eon

eno or

i ——_

co , eur 0 gees

= eo ar

oo ox ae -

ewv

ec Cal

eh cor -

o 4

0 05 1 15 2

Population Deviation

17

Fig. 4. Population Deviation vs. Partisan Bias

"

me

2)

can = eno em oon

} on

te

«4 our om

Bn

ae on

a em

— Ppa

8 ) eon ow

a i a

ie i

oJ ii 4

0 06 4 16 2

Population Deviation

In each figure, population inequality is plotted hori-

zontally, while partisan asymmetry is plotted vertical-

ly, with both variables being measured in absolute val-

ues.” The best fit line drawn through the middle of the

respective scatterplots expresses the relationship be-

tween the two variables. How well the line fits the da-

ta indicates the strength of the correlation between

population inequality and partisan asymmetry—from

not correlated at all (0) to perfectly correlated (1 or -1).

The more tightly the points are clustered along the line,

the stronger the correlation; conversely, the more ran-

dom the distribution, the weaker the correlation. In

addition, the line’s slope reveals how change in popula-

> For purposes of identifying a correlation, the sign (-/+) of

the deviation from the median population (under/over) and of the

partisanship (Republican/Democrat) is immaterial.

18

tion inequality is linked to change in partisan asym-

metry in general.

Here, the variables are, for all practical purposes,

unrelated. In both cases, the correlations are virtually

indistinguishable from zero: 0.07 for the efficiency gap

and 0.02 for partisan bias. The slope of each best fit line

is also essentially flat. In other words, statistically

speaking, population deviation has no bearing on a

plan’s efficiency gap or partisan bias. With respect to

the efficiency gap, some states with little variation in

population have large efficiency gaps (e.g., Florida, In-

diana, Kansas, Virginia, and Wisconsin), while others

with wide variance have small gaps (e.g., Delaware,

Kentucky, Maine, and West Virginia). Fig. 3. The

same non-relation holds true for partisan bias. There

are states with high bias scores that preserve almost

uniform population (e.g., Georgia, Indiana, Virginia, and

Wisconsin) as well as states with low to moderate bias

but large population deviation (e.g., Kentucky, Maine,

Massachusetts, Vermont, and Wyoming). Fig. 4.

Many state house plans have populations that devi-

ate by about 10%—the threshold below which such

plans are presumptively constitutional. See Brown v.

Thomson, 462 U.S. 835, 842 (1983). States in that situa-

tion, however, run the entire gamut of symmetry

scores. Figs. 3 & 4. For example, Delaware and North

Carolina have the same population deviation (9.9%), but

Delaware has a modest efficiency gap and partisan bias,

while North Carolina scores very poorly on the effi-

ciency gap and partisan bias alike. App. Virginia and

Wisconsin are similar to North Carolina in having sig-

nificant partisan asymmetry under both metrics, but,

unlike North Carolina, Virginia has relatively little

population deviation, and Wisconsin almost none. Id.

19

In sum, equality among district voters—however

desirable in its own right—does not track partisan ger-

rymandering in any meaningful way.

B. Gerrymandering Typically Involves Little Or

No Deviation In Population Size

The lack of correlation between population devia-

tion and either the efficiency gap or partisan bias comes

as no surprise. State plans designed to achieve parti-

san ends can do so without substantially varying dis-

trict population. Instead, they shift district boundaries,

bringing in and excluding selected voters in equal

measure with a high degree of precision. As the Court

explained just this past Term, equal population is not

itself a conventional line-drawing criterion but, rather,

“part of the redistricting background, taken as a given,

when determining whether race, or other factors, pre-

dominate in a legislator’s determination as to how equal

population objectives will be met.” Alabama Legisla-

tive Black Caucus v. Alabama, 135 S. Ct. 1257, 1270

(2015).

The Texas plan struck down in LULAC under Sec-

tion 2 of the Voting Rights Act demonstrates the basic

process by which voters may be packed and cracked to

dilute a group’s influence in a district without altering

population size. 548 U.S. 399. In District 23, whose

electorate comprised 57.5% Latinos, the incumbent was

at risk of losing his state seat because he received only

8% of the Latino vote. To protect the seat, the legisla-

ture redrew the district—by first shifting 100,000 Lati-

nos to nearby District 28, itself heavily Latino, and then

offsetting the lost population in District 23 with addi-

tional voters from largely white Republican counties in

central Texas. Jd. at 423-425, 438-439. District 23 re-

20

tained the same population, but had an entirely differ-

ent racial and partisan composition.

Precise demographic shifts of this sort are possible

because of the immense data and computing power that

are now readily available. The political parties and

other organizations keep detailed information on vot-

ers—including party registration and voting history—

down to the census block.® See Toobin, The Great Elec-

tion Grab, The New Yorker, Dec. 8, 2003, at 63, 75; see

also Issacharoff, Supreme Court Destabilization of

Single-Member Districts, 1995 U. Chi. Legal F’. 205, 232

& n.132 (1995). With this information, the right soft-

ware enables one to view block-by-block breakdowns

and to redraw district boundaries while tracking

changes in population and likely electoral outcome.

Toobin, The Great Election Grab, at 75-76. Of course,

these tools can also be used to further the goal of neu-

trality. The fact, however, remains that such tools—

not population deviation—are the staple of modern ger-

rymandering and, indeed, make deviation much easier

for the intentional gerrymanderer to avoid. See Issa-

charoff, 1995 U. Chi. Legal F. at 233; see also Karlan,

The Fire Next Time: Reapportionment After the 2000

Census, 50 Stan. L. Rev. 731, 736 (1998); Pildes, Prin-

cipled Limitations on Racial and Partisan Redistrict-

ing, 106 Yale L.J. 2505, 2553 (1997).

Population deviation, then, is no substitute for ac-

curate measures of a plan’s partisan asymmetry. Popu-

lation disparities do not indicate partisan favoritism.

° Census blocks, which are the smallest level of geography for

which the Census Bureau provides demographic data, often consist

of a single city block, with voter populations from less than a dozen

to around a thousand. Toobin, The Great Election Grab, at 75.

21

Moreover, because states can carefully select voters

with tools that monitor for population change, gerry-

mandering does not depend on such disparities in prac-

tice.

CONCLUSION

The judgment of the district court should be af-

firmed.

Respectfully submitted.

JASON D. HIRSCH THOMAS G. SAUNDERS

WILMER CUTLER PICKERING Counsel of Record

HALE AND DORR LLP KEVIN LAMB*

7 World Trade Center DEREK WOODMAN**

250 Greenwich Street WILMER CUTLER PICKERING

New York, NY 10007 HALE AND DORR LLP

1875 Pennsylvania Ave., NW

Washington, DC 20006

(202) 663-6000

thomas.saunders@wilmerhale.com

*Admitted to practice in the

State of Connecticut. Super-

vision by members of the firm

who are members of the Dis-

trict of Columbia Bar.

** Admitted to practice in the

State of Michigan. Supervi-

sion by members of the firm

who are members of the Dis-

trict of Columbia Bar.

NOVEMBER 2015

APPENDIX

4

APPENDIX

“xdse'a[qe}-uoTeLAep- 3unoLtysIpel-[sou U-O1OZ/SuTPLystpas/qoreasal/B10 Teor anna //. 4434

‘aqD] wong Sunnusipey sarynjsie] amg pun pouorssesbu0) TSIN 0108 “ISON 9S ("ISON) 8en7eIs2]

ayEIS JO soUaIEZUOD [BUOTEN sy} WO. seuI0O suOT}ELAep UOrzErNdod [e403 Sued a7878 WUaLMd JO; EYP BY], ,

“UOTPIIIP J9YzIe Ul Paziqryxe sey uBld ey} yey} drysuesnyred Jo yuNoUTE 9} samnseeUl 19}78] 9Y} 9TYA

‘uoremp refnorred & ur 1 weyd asnoy 23078 v UBETpTEd AjZuoISs MOY SPeal JULIO; GY], “G pu’ g SUUMIOO UT aneA

aqnjosgE ay} WO s1aQIP } pue Z SUUIMJOO UT aNTEA MEI BY} ‘sUOTPeIe UseMjaq paddy uss ay} a1ayA seBED Mj B U]

‘samBy PT0Z PUB ZI0Z OY} Jo aBBIOAe MBI B SE PazETNTeo WSIY aiem seg UEsTLEd pus des Aouapyye ay} Y0g |

%0'F %9'EI %9'I- %L'EI %L"ST- Td

%6'6 %9'2 %9°2 %L2 %E°2- ad

%0°9 %ET %P'T- %E% %T'T- LO

%0°9 % 9 %e'9 %2 %& b 09

%F'0 %E'9 %E'9 %0°S ET VO

DFS %61 %L'0° %VZ %& 0- uv

%8°8 %F'8 TE %0'S %0°S- ZV

%0°6 %F°0 %F°0 20° %60°F WV

MOH BAed wor | evi_ UBEAIET| «BIG dep Auepwysd CL) aqeig

‘<emndog Te0q, | eynjosqy UBS eynjosqy AUP A

gUB[Y EsNOH 07487g JUSLIND Jo seroog Arjeumusg

POT SPST %P'ST- %'6 %P'6- HO

%6'6 %0°ST %0'ST- %6'2I %6'2I- ON

%6'L %8'T %8'T- 2°01 %E'0T- AN

%L'9 %E'S %E'¢- %9'T %69°T- WN

%E'T LT %9'0- %8'Z % 0- AN

%P'S SZ %8'I- %29 %69°0 LW

%8'L %I'SI ®T'SI- %0°6 260°6- ON

%P'T %F'L %S'L- %1°9 %1°9- NW

0°01 LL MLL LSI %L'S1- IN

%1'6 %ET %E'T %1°S %9°0 VA

%6'6 %S'T %9'T- %9' %69°2 aN

60°01 %8'2 %8'Z %9'Z %9'Z x¥

%6'% %0°6 %0°6- 66°01 %6'0I- S¥

%6'T 1S %1°9- P'S %I°E- VI

%L'T Lal L2I- 69°01 %9'0I- NI

%60°0 FT PT %60°E %9°0- Tl

%9'1Z PTT PTT 8" %L'% IH

0° %e SI %% SI- %% 9 %% 9- ie)

MORBAST UOT | eI UNEG | (SEI dep Asuepwya 1985) ang

-eindog [8}0], | enjosqy UvsTIBg eqnjosqy Aue T

%8'6 %e I %9'°0- %9'6 %S'6- AM

%8'0 %T al %1al- %S TT %9'TT- IM

%0°01 %0°E %1'T %T %PT- AM

%1'°0 %0'T %0'T %I'2 %6 '0- VM

%0°% %¥ al %P'SI- TST %TSI- VA

%6 81 %8'F %0'% %0°6 %0°6 LA

%9'T %9'01 %9'01- %0% %E0 Ln

g %66 %9'01 %9'0I- %L'S LE XL

%26 %6 91 %6 9T- 2 %1'S NIL

%0°S %&'8 %E 8 bE %P'E- aS

%0°S %9'% %O2- %9'6 %9'6 ies

%6 L %8'S %8'S- %0°9 %0°9- Vd

%T'S %LP %LY SZ %S (8)

%8'T %6'L %6 L- %SF %9'F- 30

MORSTAeG uoT | slg UBSHIEY svIg der AuepysT 84) aqnig

eznjosqy UBSsT}IE eynjosqy Aue A

“seindog [#40],