Clerveaux v. E. Ramapo Cent. Sch. Dist.

20-1668
Clerveaux v. E. Ramapo Cent. Sch. Dist.
1
UNITED STATES COURT OF APPEALS 2
FOR THE SECOND CIRCUIT 3
____________________ 4
5
August Term, 2020 6
7
(Argued: August 19, 2020 Decided: January 6, 2021) 8
9
Docket No. 20-1668 10
11
____________________ 12
13
JULIO CLERVEAUX, CHEVON DOS REIS, ERIC GOODWIN, JOSE VITELIO 14
GREGORIO, DOROTHY MILLER, HILLARY MOREAU, NATIONAL 15
ASSOCIATION FOR THE ADVANCEMENT OF COLORED PEOPLE, SPRING 16
VALLEY BRANCH, 17
18
Plaintiffs-Appellees, 19
20
v. 21
22
EAST RAMAPO CENTRAL SCHOOL DISTRICT, 23
24
Defendant-Appellant.1 25
26
____________________ 27
1 The Clerk of Court is directed to amend the caption to the above.

-- 1 of 69 --

2
Before: POOLER, HALL, and CHIN, Circuit Judges. 1
2
Defendant-Appellant East Ramapo Central School District 3
(“District”) appeals from the May 25, 2020 decision and order of the United 4
States District Court for the Southern District of New York (Seibel, J.), issued 5
after a bench trial, holding that the at-large election system used by the District to 6
elect members to its Board of Education (“Board”) resulted in dilution of black 7
and Latino residents’ votes in violation of Section 2 of the Voting Rights Act of 8
1965, 52 U.S.C. § 10301. On appeal, the District argues that: Section 2 requires a 9
finding that racial motivations caused election results; the district court abused 10
its discretion in admitting and relying on Plaintiffs’ expert’s findings, which used 11
data derived through Bayesian Improved Surname Geocoding (“BISG”) rather 12
than the more traditional Citizen Voting Age Population (“CVAP”) data; and the 13
totality of the circumstances does not support a finding of impermissible vote 14
dilution. 15
We reject these arguments. We hold that Section 2 does not require racial 16
causation, though the existence or absence of such causation is a relevant factor 17
for consideration. We further hold that the district court did not err in 18
concluding that the analysis using BISG is reliable and superior to analysis using 19

-- 2 of 69 --

3
CVAP. Lastly, we hold that the totality of the circumstances supports the finding 1
of a Section 2 violation given the near-perfect correlation between race and 2
school-type; the scant evidence supporting the District’s claim that policy 3
preferences, not race, caused election results; the Board’s blatant neglect of 4
minority needs; the lack of minority-preferred success in elections; the exclusive, 5
white-dominated slating organization; and evidence suggesting the District acted 6
in bad faith throughout the litigation. 7
AFFIRMED. 8
____________________ 9
RANDALL M. LEVINE, Morgan, Lewis & Bockius LLP 10
(David J. Butler, William S.D. Cravens, Clara Kollm, 11
David B. Salmons, Bryan Killian, Stephanie Schuster, on 12
the brief), Washington, D.C., for Defendant-Appellant East 13
Ramapo Central School District. 14
15
CHARLES S. DAMERON, Latham & Watkins LLP 16
(Andrew Clubok, Claudia T. Salomon, Corey A. 17
Calabrese, Marc N. Zubick, Russell D. Mangas, on the 18
brief), Washington, D.C., for Plaintiffs-Appellees. 19
20
Arthur N. Eisenberg, Perry M. Grossman, New York 21
Civil Liberties Union Foundation (on the brief), New 22
York, N.Y., for Plaintiffs-Appellees. 23
24
Nathan Lewin, Lewin & Lewin, LLP, Washington, D.C., 25
for Agudath Israel of America, amicus curiae. 26

-- 3 of 69 --

4
1
POOLER, Circuit Judge: 2
Few rights are more sacred than the right to vote. Indeed, the right to vote 3
is preservative of all other rights, see Yick Wo v. Hopkins, 118 U.S. 356, 370 (1986), 4
but historically it has not been granted equally in this country. To rectify this 5
deprivation, Congress passed the Voting Rights Act. Section 2 of that statute 6
prohibits states or political subdivisions from structuring elections “in a manner 7
which results in a denial or abridgement of the right of any citizen of the United 8
States to vote on account of race or color,” such that minority citizens “have less 9
opportunity than other members of the electorate to participate in the political 10
process and to elect representatives of their choice.” 52 U.S.C. § 10301. In doing 11
so, the Voting Rights Act fulfills the promise of the Fifteenth Amendment—that 12
no citizen shall be denied the right to vote based on “race, color, or previous 13
condition of servitude.” U.S. Const. amend. XV. 14
East Ramapo Central School District (“District”) appeals from the May 25, 15
2020 decision and order of the United States District Court for the Southern 16
District of New York (Seibel, J.), issued following a bench trial, holding that the 17
at-large election system used by the District to elect members to its Board of 18

-- 4 of 69 --

5
Education (“Board”) resulted in dilution of black and Latino residents’ votes in 1
violation of Section 2 of the Voting Rights Act, 52 U.S.C. § 10301. On appeal, the 2
District argues that: Section 2 requires a finding that racial motivations caused 3
the election results; the district court abused its discretion in admitting and 4
relying on Plaintiffs’ expert’s findings, which used data derived through 5
Bayesian Improved Surname Geocoding (“BISG”) rather than the more 6
traditional Citizen Voting Age Population (“CVAP”) data; and the totality of the 7
circumstances does not support a finding of impermissible vote dilution. 8
We reject these arguments. We hold that Section 2 does not require racial 9
causation, though the existence or absence of such causation is a relevant factor 10
for consideration. We further hold that the district court did not err in 11
concluding that the analysis using BISG is reliable and superior to analysis using 12
CVAP. Lastly, we hold that the totality of the circumstances supports the finding 13
of a Section 2 violation given the near-perfect correlation between race and 14
school-type; the scant evidence supporting the District’s claim that policy 15
preferences, not race, caused election results; the Board’s blatant neglect of 16
minority needs; the lack of minority-preferred success in elections; the exclusive, 17

-- 5 of 69 --

6
white-dominated slating organization2; and evidence suggesting the District 1
acted in bad faith throughout the litigation. 2
The order of the district court enjoining Board elections until the District 3
proposes and executes a remedial plan, before us through an interlocutory 4
appeal, is accordingly affirmed. On December 23, 2020, the District moved to 5
stay the district court’s injunction pending the resolution of this appeal. As this 6
opinion resolves the appeal and affirms the district court’s order, the District’s 7
motion is denied as moot. 8
BACKGROUND 9
I. Factual Background3 10
A. The District and Board Elections 11
Plaintiffs-Appellees are the Spring Valley Branch of the National 12
Association for the Advancement of Colored People and Julio Clerveaux, 13
Chevon Dos Reis, Eric Goodwin, and Dorothy Miller, who are minorities and 14
registered voters in the District. Since 2008, every candidate these individuals 15
2 A slate is a list of candidates for nomination or election. We use the term
“slating organization” to refer to a group that designates candidates for a slate.
3 These facts are drawn from the district court’s decision and order,
supplemented as necessary by the record.

-- 6 of 69 --

7
have voted for has lost. In addition, Dos Reis and Goodwin unsuccessfully ran 1
for the Board in 2017 with the perceived support of the public-school 2
community, a group of residents interested in improving the conditions of public 3
schools in the District. 4
The District is a highly segregated political subdivision of New York State 5
located in Rockland County. The population in the District is approximately 6
65.7% white, 19.1% black, 10.7% Latino, and 3.3% Asian. During the 2017-2018 7
school year, approximately 8,843 students attended public schools, while 29,279 8
students attended private schools. The private-school community consists 9
primarily of white Orthodox and Hasidic Jewish residents who educate their 10
children in yeshivas, while the public-school community consists of primarily 11
black and Latino residents whose children attend public schools. The correlation 12
between race and school attended in the District is near perfect: 92% of public- 13
school students are black or Latino, while 98% of private-school students are 14
white. 15
The District is governed by a Board, which consists of nine members 16
whose responsibilities include selecting the Superintendent of Schools and 17
approving other personnel, setting the budget and levying taxes, establishing 18

-- 7 of 69 --

8
policies, and evaluating and communicating the progress and needs of the 1
District to the public and others. As of early 2020, the Board members were: 2
Harry Grossman, president; Sabrina Charles-Pierre, vice president; Mark 3
Berkowitz; Carole Anderson, appointed on an interim basis due to the 4
resignation of Bernard L. Charles, Jr.; Joel Freilich; Ashley Leveille; Yoel T. 5
Trieger; Ephraim Weissmandl; and Yehuda Weissmandl. 6
Board elections are staggered so that three seats, each carrying a three-year 7
term, are open every year (absent special circumstances, such as death or 8
resignation of a member, in which case an extra seat may be available that year). 9
Candidates run for a specific, individually numbered seat. The elections are at- 10
large, meaning that all eligible voters in the District vote in each race. The 11
following table summarizes the results of Board elections from 2008-2018 (with 12
“W” designating white candidates, “B” designating black candidates, and L” 13
designating Latino candidates).4 14
4 The summary table was referenced in the district court’s decision and order.

-- 8 of 69 --

9
1

-- 9 of 69 --

10
B. The Slating Organization 1
Influential members of the white, private-school community have an 2
informal slating process by which preferred Board candidates are selected, 3
endorsed, promoted, and elected. Rabbi Yehuda Oshry, an influential Orthodox 4
community leader, selects and approves candidates, controls access to the slating 5
process, and submits petitions on behalf of candidates. Private-school advocate 6
Shaya Glick also helps select candidates and publicizes their candidacy. Yakov 7
Horowitz, a leader in the Orthodox community, connects potential candidates to 8
Rabbi Oshry and approves candidates. The slating organization (the 9
“Organization”) has secured victory for the white community’s preferred 10
candidate in each contested election. Although some minority candidates have 11
been slated by the Organization and have won seats on the Board, minority 12
voters did not prefer these candidates. 13
The Organization does not hold an open call for candidates, and only those 14
with connections to the Organization or its leaders have been introduced, vetted, 15
and selected. When vetted, candidates were not asked about their policy views. 16
Multiple successful private-school candidates did not campaign or spend money 17

-- 10 of 69 --

11
to get elected; rather, they simply were approved by Rabbi Rosenfeld, Rabbi 1
Oshry, or Glick. 2
For instance, Charles, a black man who won multiple Board elections after 3
being slated by the Organization, was connected to Rabbi Rosenfeld through a 4
mutual acquaintance. Rabbi Rosenfeld met with Charles, but he did not ask 5
Charles about his policy platform. Rabbi Rosenfeld did, however, require 6
interviews with Charles’s running mates, and Charles needed Rabbi Rosenfeld’s 7
approval to add his running mates to his slate. 8
Charles’s situation contrasts with that of public-school candidate and 9
plaintiff Goodwin. Goodwin “genuinely impressed” a former Board member 10
from the private-school community, but he was nonetheless not introduced to 11
anyone in the Orthodox community for endorsement. Supp. App’x at 94. Charles 12
believed that “when it comes to running for the school board . . . you’re either 13
working with [the] white community or you’re working with the other 14
community.” Supp. App’x at AA 146. 15
C. Minority Board Members 16
Minority candidates have won seven out of thirty-two contested Board 17
elections in the District from 2005 to 2018. However, from 2008 to 2018, no 18

-- 11 of 69 --

12
minority-preferred candidate won a contested Board election, and each minority 1
candidate who did win did so with the approval and support of the 2
Organization. 3
Most notable are Charles and Pierre Germain, both black men who won 4
four of the elections analyzed. Both were vetted and endorsed by the 5
Organization. Neither candidate campaigned in, nor sought to appeal to the 6
public-school community given that they had already secured the Organization’s 7
endorsement and with it the support of the white community. Minority voters 8
did not support Charles and Germain. Once elected, Charles and Germain allied 9
with private-school interests and against public-school interests. For example, 10
Charles did not support the appointment of Charles-Pierre, a black woman, to 11
the Board because he believed she was aligned with public-school advocates. 12
Charles testified that Charles-Pierre was “on the opposing side” and “the lamb 13
who will certainly lead to a slaughter of this board.” Trial Tr. at 1851:5-8. While 14
Germain supported Charles-Pierre, he told his fellow Board members that “we 15
can have better control of Sabrina than the Spanish girl,” a reference to another 16
public-school community candidate who Germain considered “aggressive.” 17
Supp. App’x at 96-97. In addition, Charles “went along” with the decision of 18

-- 12 of 69 --

13
other Board members to appoint a less experienced, white individual to the 1
Board over a black retired District principal who held two master’s degrees, 2
despite the pervasive spelling and grammatical errors in the white candidate’s 3
two-paragraph application. Supp. App’x at 150. 4
Two other minority candidates, Maraluz Corado and Juan Pablo Ramirez, 5
won with the support of the white community in 2013 and 2015, respectively. 6
Both resigned from the Board shortly thereafter. The Board appointed Grossman 7
to fill Corado’s seat, choosing him over a minority candidate who had also 8
applied to fill the seat. In appointing Grossman, the Board did not interview him 9
publicly, as was required by the District’s protocol. 10
The Board appointed Charles-Pierre to fill Ramirez’s seat after a state- 11
appointed monitor pressed the Board to include a public-school parent. Charles- 12
Pierre subsequently won an uncontested election in 2016. Grossman told 13
Charles-Pierre that Yehuda Weissmandl had said, “The only reason [Charles- 14
Pierre] is there and ran unopposed is because the board wants to do what [the 15
state-appointed monitor] said.” NAACP v. E. Ramapo Cent. Sch. Dist., 462 F. Supp. 16
3d 368, 393 (S.D.N.Y. 2020) (alterations in original) (internal quotation marks 17
omitted). Charles-Pierre was reminded repeatedly that her presence on the Board 18

-- 13 of 69 --

14
was up to the whims of Grossman and the Orthodox community. For instance, 1
when admonishing Charles-Pierre for supporting a candidate he did not like to 2
fill a Board vacancy, Grossman told Charles-Pierre, “If there really was any 3
desire by anybody to remove you from the [B]oard, all that would need to be 4
done was to run a candidate against you in May . . . . [The] Orthodox community 5
could just have voted you out in May. WE [sic] told them that you were good 6
and not to run a candidate.” Supp. App’x at 248. Charles-Pierre complained that 7
the Board kept her in the dark, made her look stupid, and criticized her. In a 8
message to other Board members, Grossman said that while Charles-Pierre “has 9
a voice,” “[r]ealistically, . . . she has zero control or influence on direction.” Supp. 10
App’x at 169. 11
In 2019, Ashley Leveille, a black public-school candidate, won a contested 12
race. This occurred after Grossman sent a message to Horowitz in April 2018 13
regarding this very litigation, saying, “Spoke to David Butler today. He asked me 14
to convey message that it would be good for the case to have a minority to run 15
against Sabrina that the community could support.” Supp. App’x at 128. David 16
Butler is counsel for the District in the present litigation. Originally, Leveille ran 17
unopposed. At that time, another candidate, Pastor Jose Cintron, was collecting 18

-- 14 of 69 --

15
signatures to run for Yehuda Weissmandl’s seat. Yehuda Weissmandl then 1
decided to run again, so Cintron instead ran for the same seat Leveille sought to 2
fill. After Cintron switched seats, Rabbi Oshry collected signatures for his 3
campaign, and Grossman reached out to the Organization members in support of 4
the campaign. Cintron told Leveille that if he ran against her, “they’re going to 5
give me the seat.” Trial Tr. at 1776:1. Leveille understood “they” to refer to the 6
Jewish community because Cintron told her he had been meeting with the 7
rabbis. Accordingly, Leveille and Cintron both believed Leveille would lose. But 8
on election day and to her surprise, Leveille won because voter turnout among 9
the white population in the District was unusually low. 10
D. The Board’s Favoritism Towards Private-School Interests 11
In 2014, a state-appointed monitor investigated the Board’s activities and 12
made the “[m]ost disturbing” finding that the “Board appears to favor the 13
interests of private schools over public schools.” Supp. App’x at 280. “Beginning 14
in 2009[,] [the] Board made draconian spending cuts to public school programs 15
and services in order to balance its budgets.” Supp. App’x at 280. Meanwhile, 16
“spending on programs benefitting private schools increased.” Supp. App’x at 17
284. “No meaningful effort [was] made to distribute [the] pain of deep budget 18

-- 15 of 69 --

16
cuts fairly among private and public schools.” Supp. App’x at 284. The monitor 1
found the problem of private-school bias to be “compounded by the Board’s 2
failure to conduct meetings in an open and transparent way.” Supp. App’x at 3
285. The monitor further observed that the District’s leaders responded poorly to 4
disapproval, branding critics as “anti-Semitic” and “political opponents.” Supp. 5
App’x at 287. 6
The Board’s actions support the monitor’s findings of favoritism towards 7
private schools. For example, the Board closed two public schools over minority 8
opposition and made a sweetheart deal with a yeshiva to sell it one of the closed 9
schools at a discounted price. In addition, the Board increased nonmandated 10
private-school services, such as transportation, without restoring public schools’ 11
budgets to pre-cut levels. In fact, from 2017 to 2019, the District paid yeshiva 12
contractors to bus 1,172 more private-school students than were even registered 13
to use private-school transportation services, creating $832,584 in 14
unsubstantiated costs. The Board also made accommodations at Board meetings 15
for Yiddish-speaking parents but did not do so for Spanish-speaking parents, 16
resulting in New York State issuing a corrective action plan. 17

-- 16 of 69 --

17
The Board also repeatedly failed to respond to public-school concerns. 1
Olivia Castor, a former public-school student now attending law school, attested 2
to the fact that around March 2013, she had gathered and presented responses 3
from her classmates on the quality of their education to the Board during a Board 4
meeting. The Board did not respond, and during the presentation itself, Board 5
members ignored Castor and gave her no attention. When Castor attempted to 6
discuss inadequacies in students’ schedules using her classmates’ genuine but 7
redacted schedules as examples, the Board accused Castor of falsifying the 8
schedules. 9
In a hallway during the meeting, the Board’s attorney, without 10
provocation, directed profanity and threatening language at a seventeen-year-old 11
black student on the honor roll. The student was distressed by the incident, and 12
Castor informed the Board of the occurrence and requested that the attorney be 13
removed from his position. The Board, however, did not address the incident. 14
Similarly, the Board failed to act timely after the former District 15
Superintendent, Joel Klein, made derogatory comments about immigrant 16
students. Klein stated that the influx of “illegal” immigrants from “the southern 17
border” would skew the District’s graduation rates “because we know everyone 18

-- 17 of 69 --

18
[sic] of these kids are dropping out.” Supp. App’x at 238-39. Klein proposed 1
flimsy, nonsubstantive programming for the immigrant students because 2
according to him, the students “want to learn the language, they want free lunch, 3
breakfast and whatever else they can get.” Supp. App’x at 239. The Board 4
nevertheless left Klein in place for more than a year after the comments were 5
made until state-appointed monitors worked with the Board president to find a 6
replacement. 7
II. Procedural History 8
A. Commencement of the Lawsuit 9
Plaintiffs filed suit on November 16, 2017. Plaintiffs challenged the 10
District’s use of at-large voting for Board elections on the ground that it denies 11
black and Latino citizens equal opportunity to participate in the political process 12
and elect candidates of their choice, in violation of Section 2 of the Voting Rights 13
Act, 52 U.S.C. § 10301. Plaintiffs requested the district court enter an order 14
declaring the at-large method violated Section 2; enjoining the District from 15
conducting further elections using at-large voting; ordering the implementation 16
of a single-member ward election system; and requiring compliance with Section 17
2 for all future elections. 18

-- 18 of 69 --

19
B. Expert Findings and Admissibility Challenges 1
1. Plaintiffs’ Expert: Dr. Matthew Barreto 2
Dr. Matthew A. Barreto, a professor of political science at the University of 3
California, Los Angeles, served as Plaintiffs’ expert. Dr. Barreto, working with 4
his colleague Dr. Loren Collingwood, sought “to examine whether evidence of 5
racially polarized voting exists in elections for East Ramapo and to determine if 6
black and Latino eligible voters have their electoral interest blocked by a 7
combination of institutional arrangements and white bloc-voting.” Supp. App’x 8
at 351. Dr. Barreto concluded that there was “very strong evidence of racially 9
polarized voting” in Board elections from 2013 to 2018; that minority-preferred 10
candidates had not won a single contested election during that time; and that the 11
electoral system in the District contains many features known to reduce minority 12
voter participation and the opportunities to elect minority-preferred candidates. 13
Supp. App’x at 351-52. Dr. Barreto found that blacks and Latinos voted 14
cohesively and that whites voted in a bloc in favor of the winning candidate in 15
each election. 16
To determine how groups voted, Dr. Barreto and Dr. Collingwood used 17
ecological inference (“EI”) models, which draw an inference of how groups vote 18

-- 19 of 69 --

20
based on an analysis of aggregate ecological data, such as precinct vote totals. Dr. 1
Barreto used two EI models. The first is King’s Ecological Inference (“King’s EI”), 2
and the second is row by column Ecological Inference (“EI:RxC”). Both work by 3
“regressing candidate choice against racial demographics within the aggregate 4
precinct” to find voting patterns by race. Supp. App’x at 409. King’s EI does this 5
by running “a 2-by-2 analysis of each candidate and each racial group, in 6
iterations, whereas [EI:RxC] allows multiple rows and multiple columns to be 7
estimated simultaneously.” Supp. App’x at 409. Both methods were used “in 8
tandem to provide greater confidence in results.” Supp. App’x at 409. 9
The aggregate ecological data input into King’s EI or EI:RxC can come 10
from a few source methodologies. Dr. Barreto used BISG. In broad strokes, BISG 11
can provide a probability assessment of an individual’s race based on the 12
individual’s surname and location. BISG does this by using Census Bureau data 13
to determine what percentage of the national population with the individual’s 14
surname is black, white, Latino, Asian, or other. That national data is then 15
combined with Census Bureau data pertaining to the individual’s geographic 16
“block” (which covers the geographic distance of roughly one city block) to see 17
what percentage of the residents in that block area is black, white, Latino, Asian, 18

-- 20 of 69 --

21
or other. Combining these datapoints “provides a probabilistic prediction of 1
individual ethnicity.” App’x at 1368. “Concordance between self-reported 2
race/ethnicity and BISG estimates is typically 90 to 96 percent for the four largest 3
racial/ethnic groups.” App’x at 1377. 4
Take the following illustrative example. If we sought to determine the 5
probability that an individual with the surname Smith living in New York City 6
was black, white, Latino, Asian, or other, we would first look at the Census 7
Bureau data to see what percentage of individuals in the United States with the 8
surname Smith fall into one of these groups. We then look at our specific Smith’s 9
residential block in New York City to determine what percentage of residents on 10
this block are black, white, Latino, Asian, or other. Cross-referencing these 11
percentages can provide a probability estimate as to whether Smith is black, 12
white, Latino, Asian, or other. 13
BISG allowed Dr. Barreto to compile and input actual voter data, as 14
opposed to using CVAP data, which is data of all eligible voters.5 Dr. Barreto 15
started with each voter’s “file,” which contains the voter’s name and address. 16
5 Barreto did, however, use CVAP data to cross-check his BISG analysis.

-- 21 of 69 --

22
Supp. App’x at 44-45. He then used the “Who Are You” or “WRU” software, 1
created by scholars Kosuke Imai and Kabir Khanna, “to estimate the probability 2
that a voter is white, black, Latino, or other, using a combination of surname and 3
geolocation.” Supp. App’x at 352. This yielded a reliable estimate of each actual 4
voter’s race or ethnicity, which was aggregated at the precinct level to assess the 5
racial composition of each precinct. Once Dr. Barreto estimated the racial 6
composition at the precinct level, Dr. Barreto used both King’s EI and EI:RxC to 7
estimate what percentage of the white vote, the black vote, and the Latino vote 8
each candidate in Board elections received. 9
2. The District’s Expert: Dr. John Alford 10
Dr. John Alford, a professor of political science at Rice University, served 11
as the District’s primary expert. Contrary to Dr. Barreto, Dr. Alford concluded 12
that the evidence does not support a finding of minority voter cohesion or legally 13
significant, racially polarized voting in the District. While Dr. Alford also used 14
EI, he used CVAP for the source data. As mentioned above, CVAP data consists 15
of “a precinct-level summary of the racial breakdown of the eligible voter 16
population.” App’x at 1073. CVAP data comes from the American Community 17
Survey, which is completed by two percent of the population, for each of five 18

-- 22 of 69 --

23
years, creating “what’s called a 10 percent sample.” Trial Tr. at 256:14-257:2 1
(internal quotation marks omitted). CVAP data provides racial composition 2
within Census Bureau geographic blocks, but these geographic blocks do not 3
align with precinct boundaries. 4
3. The District Court’s Admissibility Determination 5
Before trial, the District moved to exclude Dr. Barreto’s expert testimony. 6
The district court issued an oral decision finding the testimony admissible and 7
denying the motion. While the District made numerous arguments as to the 8
reliability of BISG data, the district court repeatedly noted that these arguments 9
went to the weight of the evidence as opposed to its admissibility. The district 10
court further informed the parties that it would make its decision as to reliability 11
following trial. 12
As to admissibility, the district court addressed each of the four factors set 13
forth in Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993). With 14
respect to the first factor—whether the theory or technique could be tested—the 15
district court concluded that it could. The District’s primary argument was that 16
Dr. Barreto had not turned over all the materials necessary to replicate his 17
analysis. But the district court dismissed this argument, pointing out that “one of 18

-- 23 of 69 --

24
the District’s experts was able to get the script to run on the 2015 voter data but 1
was never asked to complete the EI analysis.” App’x at 896-97; see also Supp. 2
App’x at 15-16 (one of the District’s experts testifying that he “ran the script” and 3
“got it to run”). The district court also found that Dr. Alford admitted to being 4
able to independently run the analysis. 5
With respect to the second factor—whether the theory or technique has 6
been subject to peer review—the district court found that the use of BISG on 7
precinct-level voter data had been peer reviewed. It cited to an article authored 8
by scholars Imai and Khanna, which proposed the use of BISG for voting rights 9
litigation. With respect to the third factor—the error rate of the methodology— 10
the district court found the methodology to be admissible given the strong 11
concordance, from ninety to ninety-six percent, between self-reported race and 12
BISG estimates. 13
With respect to the fourth factor—whether the methodology has been 14
generally accepted by the academic or scientific community—the district court 15
did not explicitly address the factor. The district court discussed academic 16
articles involving the use of BISG; the District argued these articles did not 17

-- 24 of 69 --

25
support Plaintiffs’ use of BISG, but the district court found that the helpfulness of 1
the articles goes to their weight of persuasiveness, not their admissibility. 2
Ultimately, the district court concluded that there “are indications of 3
scientific reliability supporting the opinions’ admissibility,” and “Plaintiffs have 4
made a sufficient showing that I should hear the testimony and give it whatever 5
weight I find it deserves.” App’x at 893. Accordingly, the district court admitted 6
Dr. Barreto’s testimony and determined that it would revisit how much weight 7
to give the testimony following the trial. 8
C. The Trial 9
A bench trial was held across the span of seventeen days. The witnesses at 10
trial included the parties’ experts, Dr. Barreto and Dr. Alford; members of the 11
public-school community; Plaintiffs; former and current Board members; and 12
influential private-school community leaders involved in slating, such as Hersh 13
Horowitz and Rabbi Oshry.6 14
6 According to the district court, “leading up to and during trial, Rabbi Oshry
went to great lengths to avoid testifying.” NAACP, 462 F. Supp. 3d at 404 n.49. In
fact, the district court initially found Oshry in contempt for his failure to appear
and testify at trial as instructed. Id. The district court purged the contempt order
and vacated the warrant when Oshry finally appeared. Id.

-- 25 of 69 --

26
D. The District Court’s Decision 1
Following the bench trial, the district court ruled in favor of Plaintiffs, 2
holding that the at-large Board elections violated Section 2 and resulted in 3
impermissible vote dilution for black and Latino residents. 4
In reaching this conclusion, the district court relied on the testimony of Dr. 5
Barreto and discounted the testimony of Dr. Alford, primarily because it 6
considered the analysis using BISG to be superior to that using CVAP. 7
Accordingly, the district court concluded that, as Dr. Barreto found, the District’s 8
black and Latino communities were politically cohesive and that the white 9
majority votes as a bloc in Board elections such that no minority-preferred 10
candidate won a contested election since 2008. 11
The district court also found several Board members and witnesses 12
associated with the private school community not credible. For instance, the 13
district court concluded that Grossman “seems to have no compunction about 14
compromising his legal obligations when it suits his purposes.” NAACP, 462 F. 15
Supp. 3d at 396. Grossman was impeached at least three times. He also testified 16
that he was not aware of any slating organization in the District, but numerous 17
pieces of testimony indicated that he clearly participated in slating with 18

-- 26 of 69 --

27
Horowitz, Oshry, and Glick. In addition, the district court found Yehuda 1
Weissmandl not credible. He denied the existence of a slating organization and 2
testified unconvincingly that a text message he sent saying that he “personally 3
got the blessing for [their] slate” through an influential Rabbi’s son was not about 4
the Board slate. Supp. App’x at 85. Weissmandl testified that he “[didn’t] know 5
what [he] was referring to.” Supp. App’x at 86. Weissmandl similarly testified 6
that, despite having sent an email with the note, “[P]lease respond ASAP as we 7
discussed” and “one choice” in connection with filling a Board vacancy, he did 8
not know what he meant by “one choice.” Trial Tr. at 1125:1-22. Weissmandl was 9
impeached twice. 10
The district court also found the totality of the circumstances weighed 11
firmly in Plaintiffs’ favor. In relevant part, while the district court found no 12
evidence of official discrimination in the District and no overt or subtle racial 13
appeals in campaigning, it concluded that the racially polarized voting in the 14
District was not explained by policy preferences. The district court also found 15
that an exclusive slating process tightly controlled by a few white individuals 16
existed in the District. The district court further concluded that while there had 17
been some minority success in elections, the only minority candidates who had 18

-- 27 of 69 --

28
won were those perceived as “safe” by the Organization, while other minority- 1
preferred candidates had not succeeded. NAACP, 462 F. Supp. 3d at 409-10. The 2
district court also found that white Board members had lied to minority Board 3
members about settlement negotiations regarding this lawsuit and considered 4
this evidence of some Board members’ bad faith in wanting to maintain the at- 5
large voting system. 6
As a result of Plaintiffs’ success in proving a Section 2 violation, the district 7
court enjoined the District from holding any further elections under its at-large 8
system, including the election that was scheduled to take place in June 2020. The 9
District timely appealed. 10
DISCUSSION 11
“[The] resolution of the question of vote dilution is a fact intensive 12
enterprise to be undertaken by the district court. And while we are required to 13
see to the proper application of governing legal principles under a de novo 14
standard of review, we are constrained to apply a clearly erroneous standard of 15
review to the district court’s ultimate findings of vote dilution, thereby 16
preserving the benefit of the trial court’s particular familiarity with the 17
indigenous political reality without endangering the rule of law.” Goosby v. Town 18

-- 28 of 69 --

29
Bd., 180 F.3d 476, 492 (2d Cir. 1999) (internal quotation marks, brackets, and 1
citation omitted). A finding is clearly erroneous only if the reviewing court “is 2
left with the definite and firm conviction that a mistake has been committed.” 3
Atlantic Specialty Ins. Co. v. Coastal Envt’l Grp., Inc., 945 F.3d 53, 63 (2d Cir. 2019) 4
(internal quotation marks omitted). As Chief Justice Roberts recently 5
acknowledged in a concurrence: 6
The question is not whether we would reach the same findings from 7
the same record. These District Court findings entailed primarily 8
. . . factual work and therefore are reviewed only for clear error. Clear 9
error review follows from a candid appraisal of the comparative 10
advantages of trial courts and appellate courts. While we review 11
transcripts for a living, they listen to witnesses for a living. While we 12
largely read briefs for a living, they largely assess the credibility of 13
parties and witnesses for a living. 14
June Med. Servs. L.L.C. v. Russo, 140 S. Ct. 2103, 2141 (2020) (Roberts, C.J., 15
concurring) (internal alterations, citations, and quotation marks omitted). 16
Section 2 of the Voting Rights Act provides in its entirety that: 17
(a) No voting qualification or prerequisite to voting or 18
standard, practice, or procedure shall be imposed or applied by any 19
State or political subdivision in a manner which results in a denial or 20
abridgement of the right of any citizen of the United States to vote on 21
account of race or color, or in contravention of the guarantees set forth 22
in section 10303(f)(2) of this title, as provided in subsection (b). 23
(b) A violation of subsection (a) is established if, based on the 24
totality of circumstances, it is shown that the political processes 25

-- 29 of 69 --

30
leading to nomination or election in the State or political subdivision 1
are not equally open to participation by members of a class of citizens 2
protected by subsection (a) in that its members have less opportunity 3
than other members of the electorate to participate in the political 4
process and to elect representatives of their choice. The extent to 5
which members of a protected class have been elected to office in the 6
State or political subdivision is one circumstance which may be 7
considered: Provided, That nothing in this section establishes a right to 8
have members of a protected class elected in numbers equal to their 9
proportion in the population. 10
11
52 U.S.C. § 10301. 12
Congress amended the Voting Rights Act in 1982 to clarify that plaintiffs 13
need not prove intent to discriminate. Id.; see also Thornburg v. Gingles, 478 U.S. 14
30, 35 (1986). The Supreme Court, in a decision two years before the amendment, 15
held that minority voters must prove that state actors had adopted or maintained 16
the challenged electoral mechanism with discriminatory intent. See Mobile v. 17
Bolden, 446 U.S. 55 (1980). Congress acted in 1982 to “make clear that a violation 18
could be proved by showing discriminatory effect alone” and adopting a “results 19
test” as opposed to the intent test used in Bolden. Gingles, 478 U.S. at 35; see also S. 20
Rep. No. 97-417, at 28 (1982), as reprinted in 1982 U.S.C.C.A.N. 177, 179, 205 21
(hereinafter “1982 Senate Report”) (“[T]he specific intent of this amendment is that 22
the plaintiffs may choose to establish discriminatory results without proving any 23

-- 30 of 69 --

31
kind of discriminatory purpose. . . . If as a result of the challenged practice or 1
structure plaintiffs do not have an equal opportunity to participate in the 2
political processes and to elect candidates of their choice, there is a violation of 3
this section.” (emphasis added) (footnotes omitted)). 4
The Supreme Court set forth the framework for analyzing claims of 5
unlawful vote dilution in Thornburg v. Gingles, 478 U.S. 30. There, a majority of 6
the Court held that a plaintiff must establish three “necessary preconditions” 7
when bringing a Section 2 vote dilution claim: (1) “that [the minority group] is 8
sufficiently large and geographically compact to constitute a majority in a single 9
member district”; (2) “that it is politically cohesive”; and (3) “that the white 10
majority votes sufficiently as a bloc to enable it—in the absence of special 11
circumstances, such as the minority candidate running unopposed—usually to 12
defeat the minority’s preferred candidate.” Id. at 50-51. 13
If a plaintiff successfully shows the Gingles preconditions, the court must 14
next assess whether the totality of the circumstances, based on the following 15
factors, supports the plaintiff’s claim: 16
the history of voting-related discrimination in the State or political 17
subdivision; the extent to which voting in the elections of the State or 18
political subdivision is racially polarized; the extent to which the State 19

-- 31 of 69 --

32
or political subdivision has used voting practices or procedures that 1
tend to enhance the opportunity for discrimination against the 2
minority group, such as unusually large election districts, majority 3
vote requirements, and prohibitions against bullet voting; the 4
exclusion of members of the minority group from candidate slating 5
processes; the extent to which minority group members bear the 6
effects of past discrimination in areas such as education, employment, 7
and health, which hinder their ability to participate effectively in the 8
political process; the use of overt or subtle racial appeals in political 9
campaigns; and the extent to which members of the minority group 10
have been elected to public office in the jurisdiction. 11
12
Id. at 44-45 (citing 1982 Senate Report at 28-29). These factors (the “Senate 13
Factors”) come from the Senate Judiciary Committee Report accompanying the 14
passage of the 1982 amendment. Two other factors in the Senate Judiciary 15
Committee Report (the “Additional Factors”) are also probative in some cases: 16
(1) “evidence demonstrating that elected officials are unresponsive to the 17
particularized needs of the members of the minority group”; and (2) evidence 18
“that the policy underlying the State’s or the political subdivision’s use of the 19
contested practice or structure is tenuous.” Id. at 45. 20
This list of nine factors is “neither exclusive nor comprehensive.” Goosby, 21
180 F.3d at 492. “[N]o specified number of factors need be proved, and [] it is not 22
necessary for a majority of the factors to favor one position or another.” Id. “[T]he 23
ultimate conclusions about equality or inequality of opportunity were intended 24

-- 32 of 69 --

33
by Congress to be judgments resting on comprehensive, not limited, canvassing 1
of relevant facts.” Id. (internal quotation marks and citation omitted). 2
In the present case, the District first argues that Section 2 requires a 3
showing that racial motivations caused the election results at issue. Second, the 4
District contends that Plaintiffs have failed to show the second and third Gingles 5
preconditions, which includes a challenge to the district court’s decision to admit 6
and find reliable Dr. Barreto’s expert analysis and testimony. Third, the District 7
argues that Plaintiffs failed to demonstrate that a totality of the circumstances 8
supports their claim because Senate Factors 2, 4, 7, and Additional Factor 9 weigh 9
in favor of the District, not Plaintiffs. We address each argument in turn. 10
I. There is no requirement that a plaintiff must prove racial animus to 11
establish a Section 2 vote-dilution claim. 12
Our Circuit considered an argument similar to that the District makes here 13
in Goosby v. Town Board, 180 F.3d at 493. There, the defendant Town Board of the 14
Town of Hempstead, New York (“the Town Board”) argued that political 15
partisanship, and not race, explained the adverse election results. Id. Therefore, 16
the Town Board claimed that the white bloc voting patterns in the elections could 17
not be legally significant under the third Gingles precondition. Id. Our Circuit 18

-- 33 of 69 --

34
acknowledged that the splintered Gingles Court left open the issue of what role 1
causation plays in Section 2 analyses following Congress’s 1982 amendment. We 2
held that the best reading of the Court’s various opinions in Gingles counsels for 3
consideration of alternative causal explanations like the Town Board’s political- 4
partisanship claim “under the ‘totality of the circumstances’ analysis rather than 5
as part of the third Gingles precondition.” Id. 6
The District, however, is not concerned with where in the analysis 7
causation is analyzed. Instead, the District urges us to hold that a plaintiff 8
bringing a Section 2 Voting Rights Act claim must prove, at some point, that 9
racial animus caused the challenged election result. Appellants’ Br. at 37 (“There 10
must be evidence that racial animus is a but-for cause of election results . . . .” 11
(emphasis in original)). 12
That is not so. The District’s argument rests on a fundamental 13
misunderstanding of our precedent, Supreme Court precedent, and the 14
framework for Section 2 claims. See Gingles, 478 U.S. at 471; id. at 100-01 15
(O’Connor, J., concurring); Goosby, 180 F.3d at 491-92; 1982 Senate Report at 15- 16
16, 27-28, 36-37. The only facts that must be proven without exception for a 17
Section 2 claim are the Gingles preconditions; this is why they are termed 18

-- 34 of 69 --

35
“preconditions.” All three preconditions are necessary, but are not sufficient, for 1
a Section 2 violation. Those factors considered under the totality-of-the- 2
circumstances stage of the analysis, on the other hand, are not strict 3
requirements. The factors on the whole must support a vote-dilution finding. But 4
no single specific factor or definite number of factors must be proven. See Goosby, 5
180 F.3d at 492. Nor must a majority of the factors favor the plaintiff for a Section 6
2 claim to succeed. Id. When combined with the Gingles preconditions, any one 7
factor or combination of factors may be sufficient, but are not necessary in whole 8
or part, for a Section 2 violation. 9
Goosby holds that the absence or existence of racial causation is a factor 10
properly considered at the totality-of-the-circumstances step. Id. at 493 (“We 11
think the best reading of the several opinions in Gingles, however, is one that 12
treats causation as irrelevant in the inquiry into the three Gingles preconditions 13
but relevant in the totality of circumstances inquiry.” (citation omitted)); see also 14
Lewis v. Alamance County, N.C., 99 F.3d 600, 615 n.12 (4th Cir. 1996) (causation is 15
relevant to the totality of the circumstances inquiry and irrelevant when 16
considering the Gingles preconditions); Uno v. City of Holyoke, 72 F.3d 973, 983 (1st 17
Cir. 1995) (non-racial reasons for divergent voting patterns to be considered 18

-- 35 of 69 --

36
under totality of circumstances test); Nipper v. Smith, 39 F.3d 1494, 1524 (11th Cir. 1
1994) (courts consider racial and non-racial explanations for community voting 2
patterns under the totality of the circumstances). Causation is just one of many 3
factors courts consider in determining whether “the political processes leading to 4
nomination or election in the State or political subdivision are not equally open 5
to participation by members of a class of citizens protected by [Section 2] in that 6
its members have less opportunity than other members of the electorate to 7
participate in the political process and to elect representatives of their choice.” 52 8
U.S.C. § 10301. Section 2 claims do not rise or fall on racial causation.7 Still, racial 9
causation may be sufficient—though not necessary—to find a Section 2 violation. 10
7 Some of our sister circuits have similarly expressed that courts may consider
evidence of racial causation in resolving Section 2 claims but that such causation
is not required to succeed on Section 2 claims. See Lewis, 99 F.3d at 615 n.12
(explaining that causation is “relevant” to the totality of the circumstances
inquiry); Nipper, 39 F.3d at 1524 (explaining that “a violation of Section 2 may be
established . . . without proof of discriminatory intent” but that an inquiry into
causation may be relevant when the record suggests that “disparate electoral
results [are] principally caused by a factor other than race” (footnote omitted)).
Legal scholars have categorized our Circuit as one of the many that permit, but
do not require, consideration of racial causation. See Christopher S. Elmendorf,
Kevin M. Quinn & Marisa A. Abrajano, Racially Polarized Voting, 83 U. Chi. L.
Rev. 587, 614-15 (2016) (explaining that, in many circuits, racial causation “is only
one consideration among many to be weighed, either as part of the Gingles
analysis or at the totality-of-the-circumstances stage.” (footnotes omitted)).

-- 36 of 69 --

37
This understanding is in accordance with the legislative intent as well. A 1
causation requirement would entail demonstrating that racial animus caused the 2
election results at issue. Indeed, the District’s primary argument on appeal is that 3
at-large voting for Board elections does not violate Section 2 because of the 4
absence of “‘racial animus’ on the part of Orthodox Jewish voters.” Appellant’s 5
Br. at 26. But Congress rejected precisely such a showing when it drafted the 1982 6
amendments. Congress considered “the intent test [to be] unnecessarily divisive 7
because it involves charges of racism on the part of individual officials or entire 8
communities.” 1982 Senate Report at 36 (emphasis added); see also Christopher S. 9
Elmendorf, Making Sense of Section 2, 160 U. Pa. L. Rev. 377, 424 (2012) 10
(explaining that the 1982 amendments addressed what Congress “found 11
objectionable in the Bolden plurality’s intent requirement” and that “[r]elief 12
under Section 2 ought not to require ‘brand[ing] individuals as racist’” (quoting 13
1982 Senate Report at 36)). The legislative record flatly contradicts the District’s 14
assertion that Congress only rejected the intent test for officials and still sought to 15
require proof of “‘race-based motivation’ on the part of the electorate who 16
exploit the challenged practice.” Appellant’s Br. at 31 (emphasis omitted). 17

-- 37 of 69 --

38
Contrary to the District’s claims, requiring a showing of racial causation would 1
contravene Congress’s unmistakable intent. 2
For this reason, the District’s reliance on Bostock v. Clayton County, 140 S. 3
Ct. 1731 (2020), is not persuasive. The District focuses on Section 2’s language 4
prohibiting the denial or abridgement of the right to vote “on account of race or 5
color” and argues that similar language in Title VII (which prohibits 6
discrimination “because of” sex) was interpreted in Bostock to require but-for 7
causation, see Bostock, 140 S. Ct. at 1739. As an initial point, the District fails to 8
acknowledge that the Supreme Court has elsewhere interpreted “because of” 9
language to not require proof of race-based intent in the Title VII context. See 10
Griggs v. Duke Power Co., 401 U.S. 424, 432 (1971) (recognizing that Title VII 11
claims may be proved based on a disparate impact). Regardless, however, the 12
Supreme Court has “not hesitated to give a different reading to the same 13
language—whether appearing in separate statutes or in separate provisions of 14
the same statute—if there is strong evidence that Congress did not intend the 15
language to be used uniformly.” Smith v. City of Jackson, 544 U.S. 228, 260-61 16
(2005) (O’Connor, J., concurring in the judgment). Here, the unique context of the 17
Voting Rights Act and Congress’s clear desire not to require a showing of racial 18

-- 38 of 69 --

39
animus indicate that “on account of race or color” should not be interpreted to 1
require but-for causation. 2
In sum, our precedent and the legislative history make manifest that 3
Section 2 claims do not require a showing of racial causation. Racial causation is 4
one factor, of many, to be considered when assessing the totality of the 5
circumstances. Goosby, 180 F.3d at 493. The existence of such causation may be 6
sufficient for a Section 2 violation, but it is not necessary. See id. at 492, 493. 7
II. The second and third Gingles preconditions are met. 8
The District next contends that Plaintiffs failed to adduce sufficient 9
evidence to satisfy the second and third Gingles preconditions of whether black 10
and Latino residents voted cohesively and whether the white majority voted as a 11
bloc to defeat minority-preferred candidates. Specifically, the District argues that 12
the district court improperly admitted and relied on the testimony and findings 13
of Dr. Barreto, Plaintiffs’ expert, establishing that black and Latino residents were 14
politically cohesive and that white residents voted as a bloc. 15
We review the district court’s decision to admit Dr. Barreto’s expert 16
testimony for abuse of discretion. See Kumho Tire Co. v. Carmichael, 526 U.S. 137, 17

-- 39 of 69 --

40
152 (1999); Restivo v. Hessemann, 846 F.3d 547, 575 (2d Cir. 2017).8 Similarly, “[t]he 1
question of what weight to accord expert opinion is a matter committed to the 2
sound discretion of the factfinder, and we will not second guess that decision on 3
appeal absent a basis in the record to think that discretion has been abused.” Pope 4
v. County of Albany, 687 F.3d 565, 581 (2d Cir. 2012). 5
A. The district court did not abuse its discretion in admitting Dr. 6
Barreto’s testimony. 7
“In assessing reliability, the district court should consider the indicia of 8
reliability identified in [Federal Rule of Evidence] 702, namely, (1) that the 9
testimony is grounded on sufficient facts or data; (2) that the testimony is the 10
product of reliable principles and methods; and (3) that the witness has applied 11
the principles and methods reliably to the facts of the case.” United States v. 12
8 The District argues that we should apply de novo review because the district
court did not analyze the Daubert factors. The District mischaracterizes the
record. Although it did not explicitly cite Daubert by name, the district court did
in fact address the Daubert factors when it rendered an oral decision on the
District’s pretrial motion to exclude Dr. Barreto’s expert testimony. To the extent
the District considers this oral decision insufficient to satisfy the district court’s
gatekeeping role, we disagree. There is no requirement that an inquiry into
reliability take any specific form. See Restivo, 846 F.3d at 576 (explaining the
Daubert inquiry is a “flexible one” (internal quotation marks omitted)).

-- 40 of 69 --

41
Williams, 506 F.3d 151, 160 (2d Cir. 2007) (internal quotation marks and citation 1
omitted). In addition to these factors, the district court may consider those 2
enumerated in Daubert, “some or all of which might prove helpful in 3
determining the reliability of a particular scientific theory or technique.” Kumho, 4
526 U.S. at 141 (internal quotation marks omitted) (citing Daubert, 509 U.S. at 593- 5
94). These factors are: (1) whether the methodology or theory has been or can be 6
tested; (2) whether the methodology or theory has been subjected to peer review 7
and publication; (3) the methodology’s error rate; and (4) whether the 8
methodology or technique has gained general acceptance in the relevant 9
scientific community. Daubert, 509 U.S. at 593-94. 10
Although the District argues that the BISG methodology fails to satisfy the 11
Daubert test, the district court properly considered the Daubert factors and did 12
not exceed its discretion by admitting Dr. Barreto’s expert testimony. 13
1. Whether the methodology could be tested 14
Turning to the first Daubert factor, the District argues that Dr. Barreto’s 15
analysis “cannot be tested,” is “unprecedented,” and claims that he “destroyed 16
his bespoke data sets before trial.” Appellant’s Br. at 56-57. The district court 17
rejected these arguments and correctly concluded that the BISG methodology Dr. 18

-- 41 of 69 --

42
Barreto used could be tested. This is borne out by the fact that one of the 1
District’s experts, Dr. Stevenson, began replicating the analysis as to the 2015 2
voter data and was able to get the scripts to run and reproduce the results, 3
NAACP, 462 F. Supp. 3d at 391, though he was never asked to complete the full 4
EI analysis. Dr. Alford also indicated that he could have independently run the 5
BISG model. Additionally, the district court described how Dr. Barreto 6
“validated his analysis using other methodologies,” and that all five of these 7
validation methods “supported his conclusions.” Id. at 385. 8
Although the District on appeal claims that Dr. Barreto failed to preserve 9
“a spreadsheet whose rows identified voters by surname, address, and race 10
probabilities” needed to replicate his analysis, Appellant’s Br. at 55, the district 11
court found that Dr. Barreto credibly testified that no such spreadsheet exists. Dr. 12
Barreto explained that no “interim printout of BISG race estimates” existed 13
because “[t]hose are just generated in the background of the [WRU] program, 14
and as those BISG estimates get generated, they then just get plugged into the 15
precincts and then the precinct analysis is done.” Supp. App’x at 54. Dr. Barreto 16
testified that he had “turned over everything that [he] ran and detailed how the 17
script could be used to generate those race estimates.” Supp. App’x at 54. Given 18

-- 42 of 69 --

43
that the District’s experts admitted to being able to replicate the analysis and the 1
absence of evidence as to any requisite interim spreadsheet, the district court 2
properly concluded that the BISG methodology could be tested. 3
2. Whether the methodology had been peer-reviewed 4
Turning to the second Daubert factor, the district court properly concluded 5
that the use of BISG to estimate voter race for precinct-level populations has been 6
subjected to peer review. The district court supported this finding by referencing 7
several peer-reviewed articles, see NAACP, 462 F. Supp. 3d at 383, including one 8
written by Imai and Khanna that proposed the use of BISG for voting-rights 9
litigation. As Dr. Barreto explained, the purpose of this article was “to see if [the 10
BISG methodology] could improve our estimates of race and ethnicity at the 11
precinct level.” Supp. App’x at 49 (emphasis added); see also App’x at 1372 (“We 12
now estimate voter turnout by racial category and validate our estimates against 13
actual turnout by race at the precinct and congressional district levels in Florida.” 14
(emphasis added)). Imai and Khanna were successful and concluded that BISG 15
“enables academic researchers and litigators to conduct more reliable ecological 16
inference in states where registered voters are not asked to report their race.” 17

-- 43 of 69 --

44
App’x at 1374. The district court found that Dr. Barreto “applied BISG in the 1
manner proposed in the academic literature.” NAACP, 462 F. Supp. 3d at 384. 2
Other evidence, such as a footnote in an article co-authored by one of the 3
District’s experts, also supports the conclusion that BISG could “assign a race to 4
registrants in a voter file where this quantity is not present and then aggregate 5
these individuals by geographic unit such as a voting precinct.” App’x at 1276 n.21 6
(emphasis added).9 Thus, the district court did not err in finding that the BISG 7
methodology as used by Dr. Barreto had been subjected to peer review. 8
3. The error rate 9
Turning to the third Daubert factor, the District argues that the BISG 10
methodology has a high potential rate of error. This is untrue. The district court 11
discussed two studies finding that “self-reported race matched with . . . BISG 12
race estimate[s]” over 90% of the time: one study found that the probability was 13
95% for Hispanics and 93% for blacks and whites and the other found a range of 14
90-96%. NAACP, 462 F. Supp. 3d at 383. The District does not challenge this 15
9 Although the author of this footnote testified at trial that it was aspirational, the
district court did not credit this testimony because it contradicted his published
work. This determination was not an abuse of discretion. See Pope, 687 F.3d at
581.

-- 44 of 69 --

45
concordance finding but instead argues that Dr. Barreto did not calculate error 1
ranges when using BISG to estimate racial probability of District voters. 2
The District fails to understand how the BISG methodology works. As the 3
district court recognized, error rates are not necessary to calculate separately in 4
the BISG analysis because BISG provides racial probabilities—that is, the 5
likelihood that an individual is black, white, Latino, or other. As such, error rates 6
are “built into the model.” App’x at 969.10 While miscoding (i.e., improperly 7
coding a white person as having a higher probability of being black, or vice 8
versa) can occur, Dr. Barreto said that this had “no impact on the conclusions [he 9
drew] at all” because he followed “the prescribed methodology of aggregating 10
those probabilities down to a precinct.” Supp. App’x at 56. In other words, Dr. 11
Barreto was “not attempting to look at one individual on the file and say this 12
person is black, white, or Hispanic. . . . So where there might be an error, the 13
literature suggests that those things often cancel out and that’s why you 14
aggregate the probabilities and then your estimates are extremely accurate.” 15
10 The District points to the Imai and Khanna article as evidence that BISG studies
in the literature include error rates. But the Imai and Khanna study was meant to
validate whether BISG gave accurate predictions, and it was for this end that
error rates were generated in that study.

-- 45 of 69 --

46
Supp. App’x at 56. Accordingly, the District’s assertion that Dr. Barreto was 1
required to generate error rates for his analysis is unfounded. The district court 2
properly found that the strong concordance between BISG results and self- 3
reported race or ethnicity supports the reliability of Dr. Barreto’s methodology. 4
4. Whether the methodology has been generally accepted by the 5
scientific or academic community 6
Finally, while the use of BISG may be novel in voting-rights litigation, it 7
certainly is not otherwise novel. The record is replete with studies validating the 8
use of BISG, which the district court cited in finding the methodology reliable. 9
The district court found that BISG “has been extensively validated by experts,” 10
and that “[m]any respected scholars have used and validated BISG in the 11
political science context and across a variety of disciplines.” NAACP, 462 F. Supp. 12
3d at 383; see also id. at 392 (“The method has been endorsed by respected social 13
scientists in leading publications.”). Although the district court acknowledged 14
“[t]his may be the first time that voter-preference estimates based on BISG have 15
been admitted into evidence at a VRA trial,” id. at 392, the court thoughtfully 16
considered how this case was uniquely suited for the use of BISG data based on 17
the fact that the District was very diverse and highly segregated: “BISG is 18

-- 46 of 69 --

47
particularly reliable for use in the District because of its unique characteristics,” 1
id. at 384. While this may indeed be the first time that the BISG methodology was 2
admitted at a VRA trial,11 as the district court aptly explained, “[t]here must 3
always be a first time.” Id. at 392. This is especially true when the new method is 4
superior to the old method with respect to the case at hand. See id. at 387 5
(“[G]iven the unique characteristics of the District, BISG is a better data set than 6
CVAP for use as an input for ecological inference, and Dr. Barreto therefore used 7
the superior methodology.”). There is more than enough evidence indicating the 8
acceptance of the BISG methodology in the scientific or academic community. 9
In conclusion, because the district court has significant latitude in deciding 10
how to determine reliability, see Restivo, 846 F.3d at 575-56, and because the 11
district court conducted a thorough review of the Daubert factors, see NAACP, 462 12
F. Supp. 3d at 382-92; see also App’x at 885-903 (district court’s oral ruling on 13
11 At least one other court has found such evidence reliable enough to be
admitted in a case involving a Section 2 challenge to an at-large voting system.
See United States v. City of Eastpointe, 378 F. Supp. 3d 589, 612-13 (E.D. Mich.
2019). This case, however, was resolved by consent decree before trial, see No.
17-CV-10079, 2019 WL 2647355 (E.D. Mich. June 6, 2019), motion for relief from
judgment denied, 2020 WL 127953, at *1 (E.D. Mich. Jan. 10, 2020).

-- 47 of 69 --

48
Daubert motions), the district court did not abuse its discretion in admitting Dr. 1
Barreto’s expert testimony. 2
B. The district court did not abuse its discretion by according more 3
weight to Dr. Barreto’s expert testimony than it did to Dr. Alford’s 4
expert testimony. 5
The district court ultimately relied on Dr. Barreto’s testimony and 6
discounted Dr. Alford’s, finding that “given the unique characteristics of the 7
District, BISG is a better data set than CVAP for use as an input for ecological 8
inference, and Dr. Barreto therefore used the superior methodology.” NAACP, 9
462 F. Supp. 3d at 387. As discussed above, we review for abuse of discretion the 10
weight a district court assigns to expert testimony. Pope, 687 F.3d at 581. 11
Here, the district court’s ultimate decision to give greater weight to Dr. 12
Barreto’s testimony was not an abuse of discretion, and the factual findings 13
underpinning its determination were not clearly erroneous. As the district court 14
explained, it found the BISG data set more reliable than CVAP for three main 15
reasons. NAACP, 462 F. Supp. 3d at 387-88. 16
First, CVAP data is less precise than BISG data because CVAP data comes 17
from the American Community Survey which contains all eligible voters in a 18

-- 48 of 69 --

49
district, whereas BISG data is pulled from the actual voter file. CVAP data is 1
derived from a sample of only approximately ten percent of the population, see 2
Trial Tr. 256:14-257:2, thus, applying that data to the entire district population 3
requires extrapolation. That may yield inaccurate results. Dr. Barreto’s BISG 4
methodology uses data as to the actual voters in the District, so no such 5
extrapolation would be required. This makes BISG a more reliable and precise 6
data set. 7
Second, CVAP data is less precise than BISG data because of geographic 8
misalignment between CVAP data and the data needed for analyzing voting 9
patterns in a precinct. CVAP data provides racial proportions within census 10
block groups, but census blocks are smaller geographically than precincts, see 11
Brian Amos, Michael P. McDonald, Russell Watkins, When Boundaries Collide, 81 12
Pub. Opinion Q. 385, 387 (2017), thereby causing a misalignment. Because BISG 13
data, however, uses information on actual voters in the precinct, the BISG 14
methodology generates racial probabilities at the precinct level, making it 15
superior for analyzing voting patterns in a precinct. 16
Finally, CVAP data is overinclusive. Because CVAP data reflects 17
information on all eligible voters, rather than actual voters like BISG does, CVAP 18

-- 49 of 69 --

50
data overestimates voter turnout. Even Dr. Alford acknowledged flaws with 1
using CVAP data as a proxy for voter turnout. Dr. Alford cited Dr. Barreto’s 2
research on the turnout-estimation issues with CVAP data and explained that 3
using some form of double equation regression or double equation EI could have 4
addressed those issues. However, there is no evidence that Dr. Alford applied 5
any such double equation regression or EI. Even if he had, Dr. Barreto testified 6
that CVAP “still start[s] with an incorrect input variable” of “all citizen adults,” 7
so that issue would not have been cured. Supp. App’x at 175. 8
Any one of these three reasons, which are all supported by sound factual 9
findings drawn from sufficient record evidence, suffices to support the district 10
court’s conclusion that BISG is the superior data set, at least in this case. 11
Therefore, the district court did not abuse its discretion by concluding Dr. 12
Barreto’s expert testimony and analysis was more reliable than that of Dr. Alford. 13
In sum, because Dr. Barreto’s expert findings of political cohesion amongst 14
black and Latino voters and of a white voting bloc are admissible and reliable 15
and were afforded greater weight than those of Dr. Alford, the district court did 16
not clearly err in concluding that Plaintiffs have adduced sufficient evidence to 17
establish the second and third Gingles preconditions. 18

-- 50 of 69 --

51
III. Senate Factors 2, 4, 7, and Additional Factor 9 weigh in favor of 1
Plaintiffs. 2
The District’s final contention is that Plaintiffs failed to demonstrate that a 3
totality of the circumstances supports their vote-dilution claim. On appeal, the 4
District does not challenge the district court’s findings as to Senate Factors 1, 3, 5, 5
6, and Additional Factor 8. Those findings, in any event, are fully supported in 6
the record. The District challenges only the district court’s findings that Senate 7
Factors 2, 4, 7, and Additional Factor 9 weigh in Plaintiffs’ favor. 8
A. Senate Factor 2 9
Senate Factor 2 requires courts to consider “the extent to which voting in 10
the elections of the State or political subdivisions is racially polarized.” Goosby, 11
180 F.3d at 491 (quoting Gingles, 478 U.S. at 44-45). It is under this factor that 12
racial causation or alternative explanations, such as partisanship, are properly 13
considered. Id. at 493. 14
The District argues, as it did before the district court, that the primary 15
driver of election results is not race, but rather, the public-private school divide. 16
The District claims that a majority of voters prefer policies, such as lower 17

-- 51 of 69 --

52
property taxes and increased benefits to private-school students, that private- 1
school candidates support. 2
The district court did not clearly err when it found that the Board election 3
results were caused by race and not policy preferences. There is a near-perfect 4
correlation between race and schooltype. While correlation is not necessarily 5
causation, the circumstances indicate that schooltype is a proxy for race. Those 6
policies favorable to the private-school community come at the cost of the public- 7
school community. This is apparent from facts in the record, including the 8
Board’s closure of two public schools over minority opposition; the Board’s 9
subsequent attempt to sell one of the school’s buildings to a yeshiva at a deep 10
discount; and the increase in nonmandated private-school transportation while 11
public school cuts were left unrestored. It defies reality to say that those who 12
vote for private-school-friendly policies would be ignorant that the brunt of these 13
policies is borne by minority children. And a finding of vote dilution “depends 14
upon a searching practical evaluation of the past and present reality.” Gingles, 478 15
U.S. at 45 (emphasis added) (internal quotation marks and citation omitted). 16
Given this evidence, the public-school community “can be viewed as a vehicle 17

-- 52 of 69 --

53
for advancing distinctively minority interests.” League of United Latin Am. Citizens 1
v. Clements, 999 F.2d 831, 861 (5th Cir. 1993). 2
In addition, as the district court found, there is scant—if any—evidence 3
that Board candidates campaigned on specific policies, a fact that seriously 4
undermines the District’s policy-preferences argument. There is no evidence in 5
the record to suggest that private-school candidates campaigned on specific pro- 6
private-school policies, to the extent that they campaigned at all, and there is no 7
evidence that public-school candidates advocated for raising taxes or cutting 8
private-school services. The District does not dispute the absence of this evidence 9
and instead argues that campaigning on particular policies was unnecessary 10
because the voters knew what policies the private-school candidates supported. 11
Even if the voters assumed what the private-school candidates stood for, the 12
leaders who slated these candidates did not ensure that candidates aligned with 13
their policy views. Multiple candidates slated by the Organization testified that 14
they were not asked about their policy positions; the Organization essentially 15
selects election winners by virtue of adding them to the private-school slate. The 16
fact that candidates were not asked about policy positions casts serious doubt on 17
the argument that policy preferences drove election results. 18

-- 53 of 69 --

54
More broadly, the existence of an organization that has precluded 1
minority-preferred candidates itself supports the finding that Factor 2 weighs in 2
favor of Plaintiffs.12 In Goosby, we found it probative that “blacks simply are 3
unable to have any preferred candidate elected to the Town Board, given the 4
historical success of the Republican party in all Town Board elections” because 5
no black candidate, besides a “black crony” of the County Chairman, had been 6
slated by the Republican party. 180 F.3d at 496. Here, too, the only minority 7
candidates put forward by the Organization are those the private-school 8
community believed would be easy to control. See, e.g., Supp. App’x at 96 9
(Germain reading an email he wrote in which he indicated he would support 10
Charles-Pierre because “we can have better control of Sabrina than the Spanish 11
girl”); Supp. App’x at 248 (Grossman telling Charles-Pierre that “[i]f there really 12
was any desire by anybody to remove you from the [B]oard, all that would need 13
to be done was to run a candidate against you in May”). 14
12 Although consideration of slating organizations is the focus of Senate Factor 4,
discussed below, the presence of a well-established slatting organization can be
probative of racial polarization.

-- 54 of 69 --

55
Furthermore, the record is replete with evidence that the private-school- 1
run Board was chronically unresponsive to public-school concerns. When a 2
public-school student approached the Board about deficiencies in public-school 3
students’ schedules, she was accused of lying and ignored. Supp. App’x at 474- 4
78. The Board also neglected to act after its attorney threatened an innocent 5
student. It likewise dilly-dallied when the former District Superintendent made 6
seemingly derisive comments about immigrant students, leaving the 7
Superintendent in place for over a year despite condemnation from the public- 8
school community and taking action only when the monitor intervened. 9
Moreover, the Board routinely favored private-school students over public- 10
school students. And though the Board cut public-school services in the name of 11
budget balancing, it paid yeshiva contractors to bus 1,172 more students than 12
registered, yielding unsubstantiated expenses to the tune of $832,584. We must 13
agree with the state-imposed monitor that such blatant favoritism on the part of 14
the Board is “[m]ost disturbing.” Supp. App’x at 280. The Board’s lack of 15
responsiveness further supports rejecting the policy-preferences argument. See 16
Goosby, 180 F.3d at 497. 17

-- 55 of 69 --

56
When taken together, the evidence supports the district court’s finding in 1
favor of Plaintiffs on Senate Factor 2. The near-perfect correlation between race 2
and school-type, along with the political realities in the District, suggest that “the 3
majority is voting against candidates for reasons of race.” See NAACP v. City of 4
Niagara Falls, 65 F.3d 1002, 1015 (2d Cir. 1995).13 Additionally, the dearth of 5
evidence that candidates campaigned on specific policies, the Organization’s 6
slating process that excludes minority-preferred candidates, and the Board’s 7
unresponsiveness to public school concerns all support the conclusion of the 8
district court that the election results are “not best explained by” policy 9
preferences. Goosby, 180 F.3d at 497 (internal quotation marks omitted).14 10
13 We note, as the district court explicitly did, that this finding does not suggest
“that the white bloc voters harbor conscious racial animus.” NAACP, 462 F.
Supp. 3d at 400. As we explained previously, a finding of racial animus on the
part of individuals or communities is not necessary for a Section 2 violation.
Accordingly, to the extent the District argues that its “Jewish residents are not
White supremacists,” Appellant’s Br. at 81, that fact has no bearing on our
analysis.
14 We also reject the District’s suggestion that purportedly anti-Semitic comments
made by Plaintiff Goodwin and public-school candidate Steven White explain
why voters rejected public-school candidates. While the state-imposed monitor
found it typical of the Board to brand critics as anti-Semitic, we pass no judgment
on whether that is what the District is doing in making this argument here. Even
assuming arguendo their comments could be interpreted as anti-Semitic,
Goodwin ran for a Board seat once and White ran twice. At most, then, these

-- 56 of 69 --

57
B. Senate Factor 4 1
The fourth Senate Factor assesses, “if there is a candidate slating process, 2
whether the members of the minority group have been denied access to that 3
process.” Gingles, 478 U.S. at 37. The district court found that leaders from the 4
private-school, Orthodox community run an exclusive slating process with no 5
input from minorities. There is no open call for candidates, and only candidates 6
with some personal connection to the Organization are introduced, vetted, and 7
supported by it. Even public-school candidates considered impressive by 8
private-school community members were denied an opportunity to be slated. 9
And to the extent any minority candidates were slated, they were either not 10
minority-preferred, perceived to be “safe,” or the result of unusual 11
circumstances, as discussed with respect to Senate Factor 7. The slated 12
candidates always prevailed in contested elections. As a result, the district court 13
concluded that “blacks and Latinos did not have the opportunity to participate in 14
the private school slating process, which was tightly controlled by a few white 15
individuals.” NAACP, 462 F. Supp. 3d at 406. 16
comments contributed to the defeat of minority-preferred candidates in three out
of thirty-two contested races analyzed.

-- 57 of 69 --

58
The District does not challenge the existence of the slating process on 1
appeal. Instead, the District argues that minority candidates have appeared on 2
the private-school slate, and that this is all that is required. According to the 3
District, the district court’s focus on the lack of input from minorities and the 4
failure to slate minority-preferred candidates was erroneous. The District points 5
to Bernard Charles as evidence that minority candidates were actively involved 6
in the slating process, and it reiterates its policy-preferences argument to justify 7
the failure to slate minority-preferred candidates. 8
We first take up the District’s argument that the lack of minority input and 9
the failure to slate minority-preferred candidates are irrelevant facts. In light of 10
the governing precedent, this argument falls flat. Our Circuit made clear in 11
Goosby that the focus is properly on whether minority-preferred, not simply 12
minority, candidates have been slated. See 180 F.3d at 496 (stressing that the 13
failure to slate black candidates meant that “blacks simply are unable to have any 14
preferred candidate elected to the Town Board” (emphasis added)). 15
Additionally, Supreme Court precedent indicates that consideration of the 16
lack of minority input is both appropriate and important. In White v. Regester, 412 17
U.S. 755 (1973), the Court focused on evidence showing that black residents were 18

-- 58 of 69 --

59
“generally not permitted to enter into the political process in a reliable and 1
meaningful manner” such that a finding of vote dilution was proper. Id. at 767. 2
And the Gingles Court emphasized that courts reviewing vote-dilution claims 3
must take a functional view of the political process. Gingles, 478 U.S. at 45. For 4
these reasons, it is not enough that minority candidates were occasionally slated. 5
Courts must instead assess, as the district court did in this case, whether 6
minorities were “permitted to enter into the political process in a reliable and 7
meaningful manner.” White, 412 U.S. at 767 (emphases added). 8
Here, the largely uncontested facts that the district court relied on, such as 9
the failure to conduct open calls for candidates, the inside connections necessary 10
to be slated, the vetting process, and the use of only “safe” or politically or 11
legally expedient minority candidates, all support the finding that minority 12
candidates and residents were denied meaningful access to the slating process. 13
These facts underscore the significant control of the white private-school leaders 14
over the slating process, which worked to exclude minority interests and 15
viewpoints from the slate and ultimately the Board. 16
Bernard Charles’s situation does not alter this conclusion. Although 17
Charles was accepted by the Organization after being vetted, Charles testified 18

-- 59 of 69 --

60
that Rabbi Rosenfeld’s approval to add running mates to the slate was required. 1
Thus, Charles’s involvement in slating was nominal. Regardless, however, any 2
say Charles had in the selection of his running mates is a slender reed upon 3
which to lean. The evidence suggests that the white, private-school community 4
leaders found Charles acceptable because he would further their interests. 5
Charles’s situation thus does not undermine the finding that minorities were 6
effectively excluded from the slating process. In Goosby as well, we discounted 7
the slating of a black candidate because he was selected over a minority- 8
preferred candidate and was a “crony” of the chairman. Goosby, 180 F.3d at 496; 9
cf. Velasquez v. City of Abilene, 725 F.2d 1017, 1022-23 (5th Cir. 1984) (considering 10
“relevant and substantial” any evidence “that the minority candidates slated by 11
[an organization] were not true representatives of the minority population in the 12
city of Abilene”). 13
As for the District’s claim that policy preferences explain why minority- 14
preferred candidates were not slated, we reject this argument for the same 15
reasons given with respect to Senate Factor 2. The candidates approved for the 16
private-school slate, such as Charles, were not asked about any specific policy 17
platforms before being slated, and there is no evidence that public-school 18

-- 60 of 69 --

61
candidates advocated for increasing taxes or decreasing private-school services 1
while campaigning. 2
For these reasons, the district court did not err in concluding that the 3
Organization excludes minority-preferred candidates and minority voices from 4
the slating process. Senate Factor 4 weighs in favor of Plaintiffs. 5
C. Senate Factor 7 6
Senate Factor 7 considers “the extent to which members of the minority 7
group have been elected to public office in the jurisdiction.” Gingles, 478 U.S. at 8
37. We focus primarily on the elected office at issue. See Goosby, 180 F.3d at 497. 9
We consider not only whether minority candidates have been elected, but 10
whether minority residents can “elect their preferred candidates.” Id. at 495-97 11
(emphases added) (discounting the placement of a black attorney on Town Board 12
and minority success in exogenous elections because the candidates were not 13
preferred by black residents). Similarly, “the election of a few minority 14
candidates does not necessarily foreclose the possibility of dilution of the black 15
vote, in violation of this section,” because “majority citizens might evade the 16
section . . . by manipulating the election of a ‘safe’ minority candidate.” 1982 17
Senate Report at 29 n.115 (internal quotation marks and citation omitted). The 18

-- 61 of 69 --

62
Senate Report cites favorably to Zimmer v. McKeithen, 485 F.2d 1297, 1307 (5th 1
Cir. 1973), which held that minority success may be discounted if it results from 2
“politicians, who, apprehending that the support of a black candidate would be 3
politically expedient, campaign to insure his election” or efforts to “thwart 4
successful challenges to electoral schemes on dilution grounds.” 5
The district court concluded that this factor weighs in favor of Plaintiffs 6
because, although minority candidates have won some contested races, between 7
2008 and 2018, no minority-preferred candidate won a contested election. The 8
court held that “every candidate of color who won was either perceived as ‘safe’ 9
by the white slating [O]rganization or affected by special circumstances.” 10
NAACP, 462 F. Supp. 3d at 409. 11
The District challenges this conclusion on two grounds. First, the District 12
argues that the district court erroneously focused on whether minority-preferred 13
candidates, as opposed to minority candidates, have succeeded. Second, it argues 14
that the district court misapplied the safe-candidate doctrine. We reject both 15
grounds. 16
The District’s belief that the subject of our inquiry is minority candidates, 17
not minority-preferred candidates, is wrong as a matter of law. “Were we to hold 18

-- 62 of 69 --

63
that a minority candidate’s success at the polls is conclusive proof of a minority 1
group’s access to the political process, we would merely be inviting attempts to 2
circumvent the Constitution.” Zimmer, 485 F.3d at 1307. And contrary to the 3
District’s assertion, the text of Section 2 expressly focuses on whether “members 4
have less opportunity than other members of the electorate to participate in the 5
political process and to elect representatives of their choice.” 52 U.S.C. § 10301(b) 6
(emphasis added). Consistent with this language, our Circuit in Goosby 7
discounted election results when minority, but not minority-preferred, 8
candidates prevailed. 180 F.3d at 495-97. Therefore, the district court properly 9
focused on whether minority-preferred candidates were successful. 10
The district court also did not err in finding that the successful minority 11
candidates have been perceived as “safe” or otherwise resulted from unusual 12
circumstances. The district court primarily discussed four candidates in its 13
analysis: Charles, Germain, Charles-Pierre, and Leveille. 14
Charles and Germain, two black men who won four of six contested 15
elections, were heavily vetted and slated by the Organization. They were not 16
minority-preferred candidates. Once elected, they aligned with the white 17
majority and took positions counter to minority interests. For instance, Charles 18

-- 63 of 69 --

64
did not support the appointment of Charles-Pierre because she was “on the 1
opposing side,” Trial Tr. at 1851:5-8, and Germain only supported her because he 2
thought the Board could maintain “better control” of Charles-Pierre than the 3
other public-school candidate, whom he dubbed “the Spanish girl,” Supp. App’x 4
at 96-97. In addition, Charles went along with Board members’ choice to appoint 5
a less qualified white applicant with an error-riddled application over a black 6
candidate with two master’s degrees and extensive experience. The district court 7
thus properly discounted Charles and Germain’s election, just as the Goosby 8
Court discounted the election of a slated black candidate not preferred by the 9
black community who then failed to respond to the community’s needs. 10
Charles-Pierre was selected after Board members recognized that they 11
needed to follow the state-imposed monitor’s instructions of having at least one 12
public-school parent on the Board. After her election, Grossman constantly 13
reminded Charles-Pierre that she could be removed at the Orthodox 14
community’s will, and he believed she had “zero control or influence.” Supp. 15
App’x at 169. Because this evidence indicates that Charles-Pierre was selected to 16
assuage the state monitor, it was appropriate to discount her election as well. See 17
Zimmer, 485 F.2d at 1307 (explaining that minority success attributable to 18

-- 64 of 69 --

65
politicians who support minority candidates because it is “politically expedient” 1
would not undermine a vote-dilution finding). 2
The latest successful minority candidate, Leveille, was elected in 2019. The 3
district court found that her election was “engineered,” NAACP, 462 F. Supp. 3d 4
at 406, a finding supported by the record. In 2018, counsel for the District 5
suggested that “it would be good for th[is] case to have a minority to run against 6
[Charles-Pierre] that the community could support.” Supp. App’x at 128.15 7
Originally, Leveille and another minority candidate, Pastor Cintron, were 8
running unopposed for different seats. When Yehuda Weissmandl decided to 9
run again for the seat Cintron was running for, Cintron switched to run against 10
Leveille. The Organization purported to support Cintron. Cintron told Leveille 11
that the rabbis promised him the election if he ran against Leveille, and both he 12
and Leveille believed he would win. But because voter turnout was unusually 13
15 The fact that the District’s counsel purportedly gave this “advice” to the
District is deeply troubling. Considering Section 2 case law directs courts to look
past such disingenuous ploys, it is bad legal advice. More disturbing, however,
is that the advice appears to be directed at aiding the District in flouting the well-
established and clear intent of the Voting Rights Act. Such deceptive posturing
has no place in the legal profession.

-- 65 of 69 --

66
low at polling places in white areas, Leveille won. These facts more than 1
adequately support the district court’s finding that the victory was engineered. 2
The District does not challenge these facts in arguing that the district court 3
erred. Instead, it argues that the district court misapplied the safe-candidate 4
doctrine, which it says “allows courts to discount suspicious elections of a 5
minority candidate after a Section 2 case has been filed.” Appellant’s Br. at 42. But 6
this assertion has no basis in the case law. Zimmer, the precedent upon which 7
Congress and the Supreme Court relied for the safe-candidate doctrine, contains 8
no such requirement. Zimmer states that minority success may be discounted 9
when it is meant to “thwart successful challenges to electoral schemes on 10
dilution grounds” without any requirement that such efforts postdate litigation. 11
See Zimmer, 485 F.2d at 1307. 12
For these reasons, the district court did not err in focusing on minority- 13
preferred candidates’ success and in finding that the successful minority 14
candidates were perceived as safe. Thus, Senate Factor 7 weighs in favor of 15
Plaintiffs. 16
D. Additional Factor 9 17

-- 66 of 69 --

67
The final factor at issue is Additional Factor 9, which asks “whether the 1
policy underlying the . . . political subdivision’s use of . . . [the challenged] 2
practice or procedure is tenuous.” Gingles, 478 U.S. at 37. The district court held 3
that this factor favors Plaintiffs because, although the District reasonably 4
believed that it was required by state law to use at-large voting, “there is 5
evidence that the dominant Board members and the Organization have a desire 6
to adhere to the current system despite its discriminatory effect and went to 7
extraordinary lengths to preserve that system to maintain political power.” 8
NAACP, 462 F. Supp. 3d at 416. The district court relied on facts such as the 9
Districts’ witnesses’ disingenuity, the apparent engineering of Leveille’s victory, 10
and Board members’ failure to give the Board’s public-school representatives 11
accurate settlement information. Id. at 416-17. As to the settlement discussions, 12
Grossman affirmatively misled minority Board members such as Charles-Pierre 13
and Leveille. When asked why Board members needed to go to court, Grossman 14
said it was because “Judge [Seibel] wants to talk/yell at us” and that the district 15
court “wants to force us to do what the NAACP wants,” when in fact it was to 16
attend a settlement conference. D. Ct. Dkt. No. 553-1 ¶ 9. Leveille was also left off 17

-- 67 of 69 --

68
emails and other communications pertaining to settlement proposals. D. Ct. Dkt. 1
No. 551-3 ¶¶ 6-8. 2
The District does not challenge the district court’s credibility findings or 3
reliance on Leveille’s election, facts which in any event are substantiated in the 4
record. Nor does the District challenge the factual finding that Board members 5
were affirmatively misled about settlement negotiations. The District’s only 6
argument as to Additional Factor 9 is that the district court improperly relied on 7
out-of-court settlement negotiations to reach speculative presumptions about 8
Board members’ motivations. 9
There was nothing improper about the district court’s reliance on the 10
Board members’ actions in failing to provide minority Board members with 11
accurate settlement information. Evidence pertaining to settlement negotiations 12
may be used when “offered for a purpose other than to prove the validity, 13
invalidity, or amount of a disputed claim.” Fed. R. Evid. 408 advisory 14
committee’s note to 2006 amendment. Accordingly, evidence of settlement 15
negotiations may be used to demonstrate bad faith. Id. (citing Athey v. Farmer’s 16
Ins. Exch., 234 F.3d 357, 362 (8th Cir. 2000)). And bad faith on the District’s part is 17
probative of whether the District’s reasons for maintaining at-large voting are 18

-- 68 of 69 --

69
tenuous. The district court thus appropriately relied on this evidence and 1
correctly concluded that Additional Factor 9 favors Plaintiffs. 16 2
CONCLUSION 3
For the foregoing reasons, the order of the district court is affirmed. As 4
mentioned above, the District’s motion for a stay of the district court’s injunction 5
is dismissed as moot. 6
16 To the extent the District argues that its reasons are not tenuous because it has
a legitimate basis for at-large elections, that argument is meritless. In Goosby,
there was no evidence that at-large voting was implemented for discriminatory
purposes, and some legitimate bases for maintaining that system were offered at
trial. 180 F.3d at 488, 490. But we nonetheless agreed with the district court that
the Town Board sought to “cling” to at-large voting for improper reasons. Id. at
497.

-- 69 of 69 --