I. BACKGROUND

A. Overview of Legal Framework on Racially Polarized Voting This case concerns enforcement of the California Voting Rights Act of 2001 (Elec. Code, §§ 14025-14032, hereafter “the Act”)1 and the interpretation of federal voting rights law upon which the Act was in part modeled. The Act provides a private right of action for members of a protected class to challenge at-large election methods in their political subdivision. Section 14027 states that “[a]n at-large method of election may not be imposed or applied in a manner that impairs the ability of a protected class to elect candidates of its choice or its ability to influence the outcome of an election, as a result of the dilution or the abridgment of the rights of voters who are members of a protected class, as defined pursuant to Section 14026.” To prove a section 14027 violation, the protected class must prove that the challenged voting method impairs its ability to elect preferred candidates or influence election outcomes because of the dilution or abridgment of its voting rights. (§ 14027.) A plaintiff must show racially polarized voting to prove a violation of section 14027. (§ 14028, subd. (a); Sanchez v. City of Modesto (2006) 145 Cal.App.4th 660, 667 (Sanchez).) Section 14028, subdivision (a) provides that a violation is established “if it is shown that racially polarized voting occurs in elections for members of the governing body of the political subdivision or in elections incorporating other electoral choices by the voters of the political subdivision.” The City in our case

1 Unspecified references are to the Elections Code.

2 challenges only the trial court’s finding of racially polarized voting under section 14028. We turn our attention accordingly to the factors and preconditions relevant to proving racially polarized voting. We begin with California’s statutory definition of racially polarized voting. “ ‘Racially polarized voting’ means voting in which there is a difference, as defined in case law regarding enforcement of the federal Voting Rights Act of 1965 (52 U.S.C. Sec. 10301 et seq.), in the choice of candidates or other electoral choices that are preferred by voters in a protected class, and in the choice of candidates and electoral choices that are preferred by voters in the rest of the electorate. The methodologies for estimating group voting behavior as approved in applicable federal cases to enforce the federal Voting Rights Act of 1965 (52 U.S.C. Sec. 10301 et seq.) to establish racially polarized voting may be used for purposes of this section to prove that elections are characterized by racially polarized voting.” (§ 14026, subd. (e).) We take note of two directions in the statutory definition. First, racially polarized voting draws a comparison between the voting preferences of the “voters in [the] protected class” and those of “voters in the rest of the electorate.” (§ 14026, subd. (e).) Second, it is the difference between the voting pattern of the two groups that determines racially polarized voting, as that difference is defined in case law based on the federal Voting Rights Act. (Ibid.) Federal cases also demonstrate the methodologies that may be used to prove that elections are characterized by racially polarized voting. (Ibid.) The landmark voting rights decision in Thornburg v. Gingles (1986) 478 U.S. 30 (Gingles) serves as our principal guide. The United States Supreme Court in Gingles construed section 2 of the federal Voting Rights Act, which like California’s voting rights law creates liability for vote dilution. (Gingles, supra, 478 U.S. at p. 34.) A section 2 violation is established by a showing, “based on the totality of the circumstances, . . . that the political processes leading to nomination or election in the State or political subdivision are not equally open 3 to participation by members of a class of citizens . . . in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” (52 U.S.C.A. § 10301, subd. (b).) The Gingles Court delineated three threshold requirements to proving a section 2 violation, sometimes referred to as the “Gingles factors” or “Gingles preconditions.” These are: “(1) The racial group is sufficiently large and geographically compact to constitute a majority in a single-member district; (2) the racial group is politically cohesive; and (3) the majority votes sufficiently as a bloc to enable it to usually defeat the minority’s preferred candidate.” (Mo. State Conf. of the NAACP v. Ferguson-Florissant Sch. Dist. (8th Cir. 2018) 894 F.3d 924, 930 (Missouri State) see Gingles, supra, 478 U.S. at pp. 50-51.) As we explain below, only the second and third Gingles factors are required to prove a violation of California’s Act. Under the federal Voting Rights Act, a plaintiff who satisfies the three Gingles preconditions must then prove the ultimate issue of vote dilution based on the totality of circumstances. (Old Person v. Cooney (9th Cir. 2000) 230 F.3d 1113, 1120 (Old Person); accord Missouri State, supra, 894 F.3d at p. 930.) The focus at this stage is on “the impact of the contested structure or practice on minority electoral opportunities . . . .” (Gingles, supra, 478 U.S. at p. 44.) In its analysis, the court considers a list of factors based on the 1982 Senate Judiciary Committee majority Report that accompanied the bill amending section 2, sometimes referred to as “the Senate Factors.” (Gingles, supra, at p. 36.) The “list is exemplary, and not exhaustive, and ‘ “there is no requirement that any particular number of factors be proved, or that a majority of them point one way or the other.” ’ [(Gingles, at p. 45.)]” (Old Person, supra, 230 F.3d at p. 1128.) The Senate Factors include “the history of voting-related discrimination in the State or political subdivision; the extent to which voting in the elections of the State or political subdivision is racially polarized; the extent to which the State or political 4 subdivision has used voting practices or procedures that tend to enhance the opportunity for discrimination against the minority group, . . . ; the exclusion of members of the minority group from candidate slating processes; the extent to which minority group members bear the effects of past discrimination in areas such as education, employment, and health, which hinder their ability to participate effectively in the political process; the use of overt or subtle racial appeals in political campaigns; and the extent to which members of the minority group have been elected to public office in the jurisdiction.” (Gingles, supra, 478 U.S. at pp. 44-45.) The most important factors in assessing a vote dilution claim “are the ‘extent to which minority group members have been elected to public office in the jurisdiction’ and the ‘extent to which voting in the elections of the state or political subdivision is racially polarized.’ ” (Id. at p. 48, fn. 15.) Use of the factors is intended to support a “flexible, fact-intensive” approach to answering the question of whether vote dilution is actionable. (Id. at p. 46.) The district court’s determination of “ ‘whether the political processes are “equally open” depends upon a searching practical evaluation of the “past and present reality” ’ . . . and on a ‘functional’ view of the political process.” (Gingles, supra, at p. 45.) The liability determination for a voting rights violation under California law in many respects mirrors the process articulated in leading federal cases. As stated above, the plaintiff must establish that racially polarized voting is present and that the method of voting impairs the protected class members’ ability to elect their choice of candidates or influence the election outcome. (§§ 14028, 14027.) And as mentioned, the Act directs California courts assessing racially polarized voting to federal case law on section 2 of the federal Voting Rights Act. (§ 14026, subd. (e).) The Act also diverges from the federal Voting Rights Act in ways consistent with the Legislature’s intent to provide a broader cause of action for vote dilution than the federal law provides. (Sanchez, supra, 145 Cal.App.4th at p. 667.) The Legislature eliminated the first Gingles precondition requiring plaintiffs to show they are sufficiently 5 large and geographically compact to enable a majority-minority district but retained geographical compactness as a consideration at the remedy stage. (Sanchez, supra, at p. 669.) Section 14028, subdivision (c) states that “[t]he fact that members of a protected class are not geographically compact or concentrated may not preclude a finding of racially polarized voting, or a violation of Section 14027 and this section . . . .” The intent behind this change, expressed in a bill analysis prepared by staff for the Assembly Judiciary Committee, was to allow a showing of dilution or abridgement of minority voting rights without proving geographical compactness, which the authors deemed less relevant to assessing dilution and more relevant to determining an appropriate remedy if racially polarized voting is shown. (Sanchez, supra, at p. 669, citing Assem. Com. on Judiciary, Analysis of Sen. Bill No. 976 (2001–2002 Reg. Sess.) as amended Apr. 9, 2002.) The Act’s legislative history further acknowledged the need to address racially polarized voting in the context of California’s unique diversity and “lack of a racial majority group.” (Sanchez, supra, 145 Cal.App.4th at p. 669, citing Assem. Com. on Judiciary, Analysis of Sen. Bill No. 976 (2001–2002 Reg. Sess.) as amended Apr. 9, 2002; see also Bill Analysis, Senate Third Reading of Sen. Bill No. 976 (2001–2002 Reg. Sess.) as amended Apr. 9, 2002, p. 3 [stating author’s intent for the legislation to “address[] the problem of racial block voting, which is particularly harmful to a state like California due to its diversity”].)2 Section 14028, subdivision (b) provides that “[t]he occurrence of racially polarized voting shall be determined from examining results of elections in which at least one candidate is a member of a protected class or elections involving ballot measures, or other electoral choices that affect the rights and privileges

We grant the City’s unopposed request for judicial notice of the Senate Floor 2

Analysis of Senate Bill No. 976, which supplements portions of the legislative history that are already in the record. Judicial notice is appropriate pursuant to Evidence Code sections 452 and 459.

6 of members of a protected class.” It further provides that in determining a violation of section 14027, one circumstance the court may consider “is the extent to which candidates who are members of a protected class and who are preferred by voters of the protected class, as determined by an analysis of voting behavior, have been elected to the governing body of a political subdivision that is the subject of” the action. (§ 14028, subd. (b).) Section 14028 also specifies that proof “of intent on the part of the voters or elected officials to discriminate against a protected class is not required” (id., subd. (d)) and that other factors (reflective of several of the Senate Factors) “are probative, but not necessary” to establish a violation (id., subd. (e)). The probative factors listed in section 14028, subdivision (e) include “the history of discrimination, the use of electoral devices or other voting practices or procedures that may enhance the dilutive effects of at-large elections, denial of access to those processes determining which groups of candidates will receive financial or other support in a given election, the extent to which members of a protected class bear the effects of past discrimination in areas such as education, employment, and health, which hinder their ability to participate effectively in the political process, and the use of overt or subtle racial appeals in political campaigns . . . .” The case law construing the Act, which took effect in 2003, is limited to a few published Court of Appeal decisions, none which address the issues in this appeal. B. City of Santa Clara The City of Santa Clara adopted its charter in 1951. Santa Clara’s charter establishes a council-manager form of government. The city council is composed of seven-members, including the mayor. Each city council office other than mayor is designated by a seat number, sometimes referred to as a “numbered post” (e.g., Council Member Seat No. 1, Council Member Seat No. 2). The charter establishes an at-large system of elections for city council, whose members hold four-year terms. The trial

7 court’s ruling in this case stated that the Act supersedes the charter provision, as discussed further post (part II.E). At the time of trial in 2018, Santa Clara had a population of approximately 125,000. Asian Americans comprised approximately 39.5 percent of the total population and 30.5 percent of its citizen-voting-age population, which we refer to as eligible voters. Latinos comprised approximately 16.9 percent of the total population and 15.0 percent of eligible voters. The remaining population, for purposes of this case, was grouped together and classified as “non-Hispanic whites and blacks.” Non-Hispanic whites and blacks comprised 46.3 percent of Santa Clara residents and 51 percent of eligible voters. The race and ethnicity classifications used at trial reflect those used by the United States Census and adopted by the federal Voting Rights Act. The City began to consider changes to its electoral system in 2011, prompted by a letter from plaintiffs’ attorney which stated that the at-large election system appeared to violate the Act. The city council convened a charter review committee and solicited proposals to analyze the election system. A demographic research firm provided several reports to the City on election demographics and voting patterns, resulting in a recommendation by the charter committee that the City abandon its numbered post system and move to a pure at-large system. The city council did not adopt the recommendation but later convened a new charter review committee. In July 2017, the city council adopted recommendations of the 2017 charter review committee to amend the city charter by splitting the City into two voting districts and allowing voters to rank their preferences. The city council approved resolutions in December 2017 and March 2018 to submit the proposed changes to the electorate on June 5, 2018. This action followed. C. Complaint for California Voting Rights Act Violation Plaintiffs and respondents LaDonna Yumori-Kaku, Wesley Kazuo Mukoyama, Umar Kamal, Michael Kaku, and Herminio Hernando (together, plaintiffs) filed the 8 operative complaint (complaint) on December 27, 2017, alleging that the at-large election system established by the City’s charter, and the proposed changes approved by the city council, violated the Act. Plaintiffs are Asian American residents of Santa Clara, registered voters, and members of a protected class under section 14026, subdivision (d).3 Plaintiffs alleged that racially polarized voting between the electoral choices of Asian American voters and those of non-Asian American voters prevented Asian American voters from electing candidates of their choice to the city council. They alleged that the City had taken no action to change its unlawful election system despite being on notice for years of potential federal and state law violations. Plaintiffs complained that the changes proposed in 2017 did nothing to remedy the voting rights violations, thus “ensuring that the City’s elections w[ould] continue to disenfranchise Asian-American voters in Santa Clara.” The complaint sought declaratory and injunctive relief to prevent the City from retaining its at-large method of electing city council members and to implement district-based elections, or other alternative relief to remedy the violation of the Act. Plaintiffs also sought attorney fees and costs. Santa Clara denied the material allegations of the complaint. D. Trial Phase I: Liability The case proceeded to a bifurcated bench trial. The trial court conducted the liability phase of the trial over several days in April 2018. It issued a proposed statement of decision on May 15, 2018. Plaintiffs and the City filed their respective objections and responses to the proposed statement of decision. On June 6, 2018, the trial court issued its statement of decision finding “the City liable for violating the [Act].”

3The Elections Code defines a “protected class” as “a class of voters who are members of a race, color, or language minority group, as this class is referenced and defined in the federal Voting Rights Act of 1965 (52 U.S.C. Sec. 10301 et seq.).” (§ 14026, subd. (d).)

9 To evaluate the trial court’s reasoning and conclusions, we believe it necessary to summarize in some detail the evidence adduced during the liability trial. This consisted principally of competing expert testimony on the analytical methodologies and validity of the statistical outcomes. Testimony pertaining to other statutory factors like historical patterns of discrimination and political exclusion supplemented the statistical evidence. Plaintiffs also introduced percipient witness testimony by the demographer who authored the reports on Santa Clara’s election demographics and voting patterns, which informed the charter review committee’s consideration of changes to the at-large election system.

1. Statistical Analyses of Election Results The trial court considered the opinions of “[t]wo prominent statistics experts.”4 Dr. Morgan Kousser, a professor of history and social science at the California Institute of Technology, testified for plaintiffs. Dr. Jeffrey B. Lewis, a professor of political science at the University of California Los Angeles, testified for the City. The trial court also relied on the testimony of Dr. S. Karthick Ramakrishnan, plaintiffs’ expert on Asian American political and civic participation. Dr. Kousser analyzed data from 10 city council elections held between 2002 and 2016 involving one or more Asian American candidates, as well as Santa Clara precinct data for nine county board of education and school district elections between 2000 and 2016.5 Dr. Kousser explained that alongside the “endogenous” city council elections

4 The trial court, in discussion with the parties, decided to treat the expert reports submitted by both sides as direct testimony. The parties supplemented the expert reports with additional direct testimony and cross-examination at the bench trial. 5 We grant the City’s unopposed request for judicial notice of the official vote

counts of the 19 elections analyzed in the trial court’s ruling. The vote counts are produced by the Santa Clara County Registrar. They are not reasonably subject to dispute and are capable of immediate and accurate determination by resort to sources of reasonably indisputable accuracy. (Evid. Code, §§ 452, 459.) Although the official vote counts were not produced as such at trial, they present in a familiar format the information already contained in plaintiffs’ expert’s report.

10 involving only the jurisdiction at issue, “exogenous” Santa Clara Unified School District and Santa Clara County Board of Education elections were “the most important . . . to analyze” because they were local, nonpartisan, and did not involve statewide issues or partisan allegiances that could complicate their relationship with nonpartisan city council elections. (See Luna v. County of Kern (E.D. Cal. 2018) 291 F.Supp.3d 1088, 1120 [defining exogenous elections as “contests for any other office aside from the” office in dispute].) We refer to the school district and county board of education elections together as “school elections.” Plaintiffs presented three statistical methods at trial: ecological regression, weighted ecological regression, and ecological inference. Dr. Kousser testified that experts in voting rights cases necessarily draw inferences from statistical analyses of aggregate data at the precinct level because they lack direct evidence about individual voting choices. He outlined the history, limitations, methodology, and results for each method he applied. He explained that ecological regression was the dominant method for nearly 40 years and was used to estimate racial polarization in the leading Supreme Court case, Gingles, supra, 478 U.S. at pages 52 through 54. Here, he used ecological regression to estimate Asian American voting behavior by correlating the precinct-level results for each candidate in the city council contest for that year with the percentage of Asian American voters. His weighted ecological regression method produced similar but more nuanced results than standard regression by accounting for the variations in precinct sizes. Lastly, he testified that ecological inference was developed by a Harvard political scientist to overcome certain limitations in the regression models. Political scientists have come to view ecological inference as the most advantageous of the three models. Dr. Kousser ran the data according to each method and tabulated the results. He “relied on surnames as a proxy for race/ethnicity classifications,” separating the Santa Clara population into three groups: non-Hispanic whites and blacks, Latinos, and Asian Americans. Dr. Kousser explained that because these “three significant ethnic groups” 11 systematically voted differently and could be distinguished by surname, it would be inappropriate to employ two-variable (bivariate) analyses (e.g., Asian American and non-Hispanic whites and blacks). Instead, he used multi-variable (multivariate) models to estimate voting patterns between the three groups. Dr. Kousser did not separate the Asian American group by country of origin because each subgroup would be too small to analyze. He opined, however, that there was “substantial cohesion” across the Asian American group. African American and non-Hispanic white voters were grouped together because their surnames are indistinguishable. Dr. Kousser noted that even though their voting patterns could not be analyzed separately, the effect was minimal since the “overwhelming proportion” of the group is non-Hispanic white. Using the regression and inference methods, he estimated rates of support among the three ethnic groups for the different candidates. Dr. Kousser ultimately concluded that “in half of the contests, the voting was racially polarized between” Asian American voters and non-Hispanic white and black voters. He opined that of the 10 city council races, only those involving a candidate named Gap Kim in 2004 and a candidate named Mohammed Nadeem, who ran unsuccessfully in 2010, 2012, 2014, and 2016, were not racially polarized or were polarized in the “ ‘wrong’ ” direction as when Nadeem received more votes from non-Hispanic white and black voters. Dr. Kousser believed that the lack of Asian American support for Nadeem could be attributed to what the Supreme Court in Gingles called “special circumstances.” (Gingles, supra, 478 U.S. at p. 57.) We discuss the special circumstances issue in detail post (part II.C.1). Dr. Kousser opined that “the most important fact, ascertained without any statistical analysis whatsoever, is that all of the Asian-American candidates lost.” Dr. Kousser suggested that the Santa Clara Unified School District was most relevant to the school elections analysis, because it handled matters significant for local schools and because its races involved several Asian American candidates who 12 consistently lost. Dr. Kousser concluded that in all but one of the school district races, the vote for the Asian American candidate was racially polarized by all three statistical techniques between Asian American voters and non-Hispanic white and black voters. Dr. Kousser opined that the statistical results and race outcomes for the 10 city council races, considered together with special circumstances like those concerning candidate Nadeem, supported a finding of what the Supreme Court in Gingles called “ ‘legally significant racial polarization.’ ” He added that while the school elections were “less important in evaluating racial polarization for the City of Santa Clara,” they nonetheless reinforced the general conclusion that Asian American candidates usually lost and that Santa Clara elections were racially polarized in a legal significant way. Dr. Lewis, whose field of political methodology focuses on quantitative analysis and political science, challenged Dr. Kousser’s methodology and conclusions though he agreed with certain premises. He explained that surname matching using the statewide database on voter registration and turnout “is a widely-used and accepted way of measuring the ethnic composition of precincts in instances in which surnames are strongly indicative of ethnicity.” And he recognized the fundamental purpose of the statistical methods identified by Dr. Kousser as a way to infer “the behavior of individuals (here voters of different ethnicities) from observations on aggregate units (here the jurisdiction and its precincts) . . . .” He also agreed that between ecological inference and ecological regression, the inference method “more fully exploits the information that’s available in the data, and . . . at least in theory should provide better estimates.” But Dr. Lewis disagreed that those methods could provide reliable estimates of voting behavior. He testified that the reliability of the regression and ecological inference models depended on the degree of racial or ethnic homogeneity of precincts. Dr. Lewis explained that despite the fact that approximately 22 percent of registered Santa Clara voters are Asian American (spread across six national origin groups), the most 13 heavily-concentrated precinct is only 42 percent Asian American. He opined that such disbursement across precincts “provides little or no direct deterministic information about the presence of racially polarized voting,” leaving the statistical models to rely on “untestable assumptions” about support for candidates among members of each racial or ethnic group. Dr. Lewis explained that under these circumstances, substantial uncertainty accompanies the use of ecological regression and ecological inference. Only by making “strong, and largely untestable, assumptions, these models narrow the range of logically possible rates of support for each candidate among the voter of each ethnic group described above down to a single ‘best guess’ (point estimate) and associate degree of uncertainty (confidence . . . interval).” Dr. Lewis illustrated how the confidence interval produced by the ecological inference and regression models may “substantially understate the possible departures of the model’s estimates from the true values of the quantities of interest” by applying them to the 2016 election for city council seat 7 and using democratic party registration data to test the validity of the underlying assumptions. Dr. Lewis explained that party registration is a “vote-like variable that can be tabulated at the individual level for the voters in the jurisdiction under study” and compared to the estimate of those values produced by applying regression or ecological inference techniques. It is “the one sort of political decision that one can directly observe at the individual-level within the jurisdiction . . . .” Dr. Lewis found that his calculation of democratic party registration using ecological regression for Asian American and other voters in Santa Clara differed from the true rate by 30 percentage points. He noted that “even the large confidence interval [did] not come close to capturing the true rate of Democratic registration among Asians.” For example, the statistical model showed “95 percent of the time that you do this you capture the truth,” while the party registration data he looked at showed it was true “less than half the time.”

14 Dr. Lewis concluded that the “massive bias” seen in the party registration exercise casts “serious doubt” on the reliability of the regression results to measure voting patterns. He acknowledged on cross-examination that it was “possible” that voting in local nonpartisan elections might not closely correlate with political party registration. Dr. Lewis further opined that despite improved estimation techniques, ecological inference could not solve the inference problem illustrated by the democratic party registration example because it relies on the same, underlying independence assumption as the regression method. He explained that ecological inference is “more robust to violations of this assumption” only “when the logical bounds are informative . . . .” Logical bounds are a factor of relative group size, becoming “increasingly tight for the larger group as the population becomes increasingly homogenous, until in the limiting case of complete homogeneity, the problem becomes trivial and the upper and lower logical bounds converge on a single value.” But he testified that this case presented “two forms of uncertainty. One is the qualifiable uncertainty, which is the uncertainty that is provided by the confidence levels that are generated by the model under the condition that the maintained assumptions hold. [¶] In addition to that qualifiable uncertainty . . . , there’s additional uncertainty that arises from the fact that it’s quite likely that those assumptions do not hold.” Dr. Lewis concluded that the ecological inference method provided “little evidence of cohesive voting by either Asians or non-Asians in the City of Santa Clara” and no evidence of polarization. Dr. Lewis acknowledged on cross-examination that while he was not able to arrive at a sufficiently reliable result using the data made available to him, there was “more information that could have been gleaned” from analyzing other contests in addition to the 2016 city council races. He could not offer a direct opinion based on his own analysis as to whether racially polarized voting occurred in any earlier city council elections or other local nonpartisan elections. He also could not reliably conclude for the 2016 elections that racially polarized voting occurred or did not occur. He confirmed that in 15 three of the four 2016 elections that he analyzed which involved an Asian American candidate, there was at least an 80 percent probability that Asian American voters supported the Asian American candidate. He also confirmed that he performed a bivariate analysis, comparing Asian American to non-Asian American voting patterns, in contrast with Dr. Kousser’s trivariate analysis. In rebuttal, Dr. Kousser disagreed with Dr. Lewis’s interpretation of party registration data. He pointed out that Asian American voters in Santa Clara County “tend more than other groups to have no party preference, and where there are particularly high proportions of Asians, they are particularly likely to have no party preference.” Dr. Kousser believed that the nonpartisan characteristic of Asian American voters in Santa Clara made any analogy between partisanship and candidate preferences in local, nonpartisan elections less informative. Dr. Kousser also generally disagreed with Dr. Lewis’s opinions on methodology. He opined that Dr. Lewis’s methodological choices tended to “cause the finding of least polarization and least cohesion.”

2. Historical Discrimination and Political Participation The trial court heard additional evidence relevant to “other factors” deemed probative for determining a voting rights violation, as set forth in section 14028, subdivision (e). Dr. Ramakrishnan, a professor of public policy and political science and associate dean at the University of California, Riverside, testified as an expert on immigrant political and civic participation and cohesiveness. He identified three areas relevant to Asian American political participation and power. Dr. Ramakrishnan first described historical patterns of discrimination and political exclusion of Asian Americans in California that inform present-day disparities in political outreach and participation. He highlighted the anti-Asian policies that existed in California since the 1850s, embodied in iterations of legislative and constitutional enactments and reinforced by federal anti-immigration policies and naturalization bans

16 that persisted until the 1950s. He described additional restrictions imposed in Santa Clara and other cities through the 1960s which limited meaningful civic engagement. And he stated that undercurrents of political and social exclusion persist today, citing broad and vocal opposition to efforts to recognize symbols of racial diversity in the community, popular perceptions of the Asian as the “perpetual foreigner,” and recent upticks in hate crimes. Dr. Ramakrishnan opined that these mechanisms of political exclusion leave legacies. He explained that “even if Asians are gaining in terms of economic mobility, they are not seen as fully part of the civic fabric of the United States.” Dr. Ramakrishnan next addressed political cohesion. He testified that the Asian American community is demographically diverse but “remarkably consistent across national origin groups in their public opinion” regarding policy and political preferences. Dr. Ramakrishnan testified that several national surveys have shown cohesion across Asian American national origin groups on issues like environmental protection, school funding, taxation, and government provision of social services. He opined that there was more political cohesion among Asian Americans than might be expected given the diversity of national origin, language, and socioeconomic status. Dr. Ramakrishnan explained that despite sharing common policy preferences, the political power of Asian Americans is limited. Factors that reduce Asian American political participation include the legacy of exclusion, the high percentage of eligible Asian American voters who are naturalized citizens and face language barriers, and the lack of outreach and political campaigning directed at Asian Americans. He reported that out of 10 city council election cycles from 2004 to 2016 in Santa Clara, there were only two instances in which an Asian American candidate received a public endorsement from a current mayor or city councilmember. Dr. Ramakrishnan also noted that the distribution of Asian American voters across the jurisdiction makes it challenging for an Asian American to succeed to citywide elected office.

17 Dr. Jeanne Gobalet testified as a percipient witness. The City retained her firm in 2011 and again in 2016 to provide demographic services to the charter review committee. Dr. Gobalet testified that on September 12, 2011, her firm sent a report titled “Report on Demographic Characteristics and Voting Patterns of Residents of the City of Santa Clara” to the acting city attorney. The report was not marked as a draft. It expressed reservations about the statistical methods used to ascertain racially polarized voting but nevertheless concluded that the City was vulnerable to a lawsuit because of its election history and because courts have accepted ecological regression as a method to determine racially polarized voting. Dr. Gobalet sent a slightly revised version of the report the next day as well as several days later, making additional revisions after an e-mail from the acting city attorney commented on the second version of the report. Dr. Gobalet testified that she had heard of the ecological inference method but was not familiar with it. She believed the use of any ecological regression or inference method was “deeply flawed” because “race ethnicity is only one of many factors that affect voting.”

3. Statement of Decision (Liability) After closing statements and the filing of posttrial briefs, the trial court issued a proposed statement of decision finding the City liable for violations of the Act. On June 6, 2018, after considering the parties’ written objections and comments, the trial court issued its final statement of decision. We highlight those aspects of the trial court’s decision that are pertinent to the issues on appeal. The trial court observed that while the percentage of Santa Clara residents who are non-Hispanic white and black (46.3 percent) is not so different from the percentage who are Asian American (39.5 percent), the percentage of actual voters is different (64.1 percent non-Hispanic white and black versus 21.2 percent Asian American). The court noted that this disparity “raises the possibility that [non-Hispanic white and black] bloc voting could impair the ability of Asians to elect preferred candidates.”

18 The trial court then considered the evidence presented by the experts at trial and the election outcomes for those races analyzed between 2002 and 2016. The court explained that the output from the ecological inference models includes the “most likely ‘point estimate’ along with a ‘standard error’ associated with the point estimate,” which “in turn, can be converted into ‘confidence intervals’ that represent a range within which there is a certain degree of confidence.” The court then made findings on the methodological disputes, noting that the few published court cases interpreting the California Voting Rights Act did not address the disputed issues. First, the trial court found that Dr. Kousser’s trivariate analysis—dividing the electorate into three groups (non-Hispanic whites and blacks, Latinos, and Asian Americans)—was consistent with the statutory language defining racially polarized voting in terms of the voting choices of the voters in the protected class and those of the “voters in the rest of the electorate.” (§ 14026, subd. (e).) The court rejected the City’s interpretation of the statutory language as requiring a bivariate analysis comparing Asian American voting preferences with those of all the other voters in the electorate, meaning in this case non-Hispanic black and white voters and Latino voters. Next, the trial court found that ecological inference provided a valid tool to assess political cohesion among Asian American voters in Santa Clara despite the uncertainty and aggregation bias highlighted by the City. In pertinent part, the court rejected the City’s claim that potential correlation errors undermined the reliability of the ecological inference results. It recognized that the distribution of Asian American voters across precincts would impact the statistical results but noted that plaintiffs had conceded this as the reason the confidence intervals were often quite large. The court reasoned that if the ecological inference results presented by Dr. Kousser were less reliable than those generated in more segregated communities, they were “nonetheless probative.” It cited federal cases applying the federal Voting Rights Act where the inference data was inexact and held that “confidence intervals less than 95 percent may be sufficient.” 19 The trial court also noted the experts’ consensus at trial in this case that there is no bright line minimum at which ecological inference results are no longer valid and no better statistical method for determining racial or ethnic group voting behavior. The experts also agreed that some information gained from statistical models is better than none. The trial court rejected the City’s argument that aggregation bias, as illustrated by Dr. Lewis’s democratic party registration analysis, rendered the ecological inference results too unreliable. It found that the party registration analysis was itself “fraught with uncertainties” and did not predict voting preferences in nonpartisan candidate races. The trial court lastly evaluated the statistical evidence and “other probative factors set forth” in the Act. It found that the parties agreed that of the 10 city council elections at issue, three were racially polarized (seat 2 in 2002; seat 3 in 2004; and seat 5 in 2014) and five were not racially polarized (seat 4 in 2004; seat 2 in 2010; seat 3 in 2012; seat 2 in 2014; and seat 6 in 2016). Of the two city council elections in dispute (seats 4 and 7 in 2016), the trial court ultimately rejected the City’s argument that plaintiffs could not show an Asian American preferred candidate due to overlapping confidence intervals. The court stated that although the 95 percent confidence intervals overlapped among the Asian American-supported candidates, “at the 80 percent confidence interval urged by [p]laintiffs in their post-trial brief, there is an Asian preferred candidate in both contests . . . .” The court reasoned that use of the 80 percent confidence interval provided “sufficiently reliable results” to support a finding of racially polarized voting in the two disputed elections. In a footnote, the trial court provided the calculations it used to determine the 80 percent confidence intervals and rejected the City’s argument that in doing so it has “assumed it was ‘acceptable’ for [the court] to ‘create its own evidence.’ ” It noted that it also could look to point estimates to assess cohesion around a preferred candidate, citing an unpublished district court case as an example. The trial court also considered how “special circumstances” as described in Gingles, supra, 478 U.S. at page 51, might affect the weight given to any of the election 20 results. The trial court did not adopt Dr. Kousser’s interpretation of Asian American candidate Nadeem’s four election losses (2010, 2012, 2014, 2016) as the likely result of special circumstances. It found the evidence too speculative to warrant disregarding those elections, which comprised four of the five city council elections in which the parties agreed that racially polarized voting was not present. The court nonetheless decided to assign less weight to those four elections because Nadeem’s poor track record as a candidate—evidenced by his diminishing returns in each subsequent contest—was “a reasonable explanation for the lack of Asian support.” The trial court similarly evaluated the school election evidence. The parties agreed that two of the nine school elections at issue were racially polarized (2004 district and 2016 county) and three were not (county in 2000, 2008, and 2012). Of the four school elections in dispute (district elections in 2008, 2010, 2012, and 2014), the trial court acknowledged that plaintiffs could not show a preferred candidate among Asian American voters at 95 percent confidence intervals. It found, however, that there was an Asian American-preferred candidate in two school elections (2008 and 2012) at the 80 percent confidence intervals. Based on these calculations, the trial court found racially polarized voting in four of nine school elections between 2000 and 2016. The trial court additionally considered other statutory factors relevant to ascertaining vote dilution. It observed, under section 14028, subdivision (b), that no Asian American had “ever won” a city council election even though Asian American candidates ran 10 times in elections from 2002 to 2016. It reviewed evidence of past discrimination and of practices that enhance vote dilution under section 14028, subdivision (e). It found that the “numbered posts” system of at-large elections was an electoral device with known disadvantages for minority voters. And it noted that the City failed to change the system despite the recommendations of the charter review committee in 2011. The court found that “[i]nstead of candidly addressing the issue, the City’s interim general counsel asked that [Dr. Gobalet’s] report be ‘stripped’ of ‘the information 21 about the council election history and the charts . . . showing racial polarization’ before it was distributed” to the city council and the charter review committee. As to California’s history of discrimination against Asian Americans, the trial court found it difficult to measure the present-day effects of the historical discrimination shown by Dr. Ramakrishnan. It concluded that the evidence at trial did not suggest any unique circumstances affecting Asian American voting patterns in Santa Clara elections. The trial court summarized the totality of its findings to conclude that plaintiffs had proven their case by a preponderance of the evidence. The trial court found that the statistical analyses showed (1) racially polarized voting in five of 10 city council elections between 2002 and 2016, and cohesive voting among Asian Americans in six of those races, and (2) racially polarized voting in four of nine school elections, though those had lower probative value than the city elections. The court also found that no members of the protected class had been elected to the governing body of the relevant political subdivision, and that the City’s response was inaction despite an “overwhelming majority” of charter committee members voting in 2011 to modify the numbered posts, at-large election system. Finally, the court found the evidence of historical discrimination had few measurable effects but still supported a finding that the City’s system of city council elections violated the Act. E. Trial Phase II: Remedy The trial court conducted the remedies phase of the trial in July 2018. The parties stipulated to join the Santa Clara County Registrar of Voters as a necessary party. After a bench trial of several days, which included expert testimony on both sides, the trial court issued a statement of decision, followed shortly after by an amended statement of

22 decision and judgment. It found, pursuant to section 14029,6 that the adoption of district-based elections based on the districting plan proposed the City would “adequately remediate the City’s violations of the [Act] and best serve its residents.” The details of the remedial plan are not at issue here. We note only the trial court’s findings that the statistics generated for the chosen district plan showed it would “remedy the dilution and abridgement of voting rights of Asians who reside in the City” including by creating one district of 51 percent eligible Asian American voters. The court ordered the City to adopt the district-based elections for six city council seats and to retain the at-large system of election for the mayor’s seat. It noted the two provisions of the city charter affected by the ruling and ordered the registrar of voters to immediately begin implementing the plan to meet its timetable for the upcoming November 2018 elections. The November 2018 elections implemented the court-ordered remedy.7 F. Judgment and Appeals On July 24, 2018, the court entered judgment in favor of plaintiffs in accordance with the statement of decision after the liability and remedies phases of the trial.

6 Section 14029 states, “Upon a finding of a violation of Section 14027 and Section 14028, the court shall implement appropriate remedies, including the imposition of district-based elections, that are tailored to remedy the violation.” 7 We grant plaintiffs’ unopposed request for judicial notice of the election results

of the November 2018 City of Santa Clara district 2 election. The election took place after the liability and remedies phases of trial in this case and after implementation of the judgment ordering the adoption of district-based elections based on the City’s districting plan. The result of a public election is a fact not reasonably subject to dispute and capable of immediate and accurate determination by resort to sources of reasonably indisputable accuracy. (Evid. Code, §§ 452, subd. (h), 459.) On the same basis, we grant plaintiffs’ recently filed, unopposed request for judicial notice of the election results of the November 2020 City of Santa Clara district 4 and district 5 elections. The November 2018 election resulted in the election of Raj Chahal as council member for district 2, making him the first Asian American candidate elected to the city council in Santa Clara’s history. The November 2020 district 4 and district 5 elections resulted in the election of two more Asian American candidates to Santa Clara’s city council since implementation of the trial court’s judgment in this case.

23 Plaintiffs moved for an award of reasonable attorney fees, expenses, and costs as the prevailing party pursuant to section 14030. The City opposed the motion, and the matter was litigated extensively. After an initial hearing and supplemental briefing, the trial court issued an order on January 22, 2019, granting the motion for attorney fees. The court struck time listed for certain activities deemed not recoverable, reduced in part the lodestar and lodestar multiplier, and ordered the City to pay attorney fees in the amount of $3,164,955.61. On January 22, 2019, the trial court amended the judgment to include the attorney fees award and costs order. The City timely appealed from both the July 24, 2018 judgment after trial and from the January 22, 2019 amended judgment stating the attorney fees and costs award.

II. DISCUSSION

The City claims that plaintiffs failed to make the necessary showing to establish legally cognizable racially polarized voting under the Act. At issue is whether plaintiffs proved that Santa Clara’s non-Hispanic white and black majority “votes sufficiently as a bloc to enable it . . . usually to defeat” (Gingles, supra, 478 U.S. at p. 51, italics added) the preferred candidate of Asian American voters in Santa Clara city council elections. The City contends that the trial court stretched the meaning of the word “usually” beyond even its most generous judicial interpretation to make a liability finding based on racially polarized voting in only five of 10 city council elections. The City also contends that the trial court improperly conducted its own statistical analysis to find liability, rather than rely on the expert testimony presented through the adversarial process at trial. A. Standard of Review The parties dispute the standard of review that applies to our analysis. We must decide whether the trial court erred as a matter of law in finding racially polarized voting in five of 10 city elections sufficient to satisfy the “usually” standard of the third Gingles precondition—a question that the City asserts calls for de novo review, or whether the

24 determination of racially polarized voting is a question of fact subject to deferential review, as plaintiffs have asserted. Our Supreme Court has explained the general principles governing the standards of appellate review as follows. “ ‘Questions of fact concern the establishment of historical or physical facts; their resolution is reviewed under the substantial-evidence test. Questions of law relate to the selection of a rule; their resolution is reviewed independently. Mixed questions of law and fact concern the application of the rule to the facts and the consequent determination whether the rule is satisfied. If the pertinent inquiry requires application of experience with human affairs, the question is predominantly factual and its determination is reviewed under the substantial-evidence test. If, by contrast, the inquiry requires a critical consideration, in a factual context, of legal principles and their underlying values, the question is predominantly legal and its determination is reviewed independently.’ ” (Haworth v. Superior Court (2010) 50 Cal.4th 372, 384 (Haworth).) One reason that mixed questions of law and fact are reviewed de novo in most cases is “ ‘ “because usually the application of law to fact will require the consideration of legal concepts and involve the exercise of judgment about the values underlying legal principles.” ’ ” (Id. at p. 385.) We find the main issue on appeal in this case—whether an equal ratio of polarized to non-polarized elections precludes liability for racially polarized voting and vote dilution—involves a mixed question of law and fact that is best addressed by novo review. This is not to ignore the intensely factual nature of the inquiry. Plaintiffs are correct that federal voting rights law treats the ultimate finding of racially polarized voting and vote dilution as a question of fact subject to the clearly-erroneous standard of appellate review under rule 52(a) of the Federal Rules of Civil Procedure. (Gingles, supra, 478 U.S. at pp. 78-79.) The district court’s findings of fact are reviewed for clear error because the determination of vote dilution “ ‘ “is peculiarly dependent upon the facts of each case,” ’ . . . and requires ‘an intensely local appraisal of the design and 25 impact’ of the contested electoral mechanisms.” (Id. at p. 79.) Under the federal standard, reviewing courts defer to the district court’s “superior fact-finding capabilities” (Smith v. Salt River Project Agr. Imp. & Power Dist. (9th Cir. 1997) 109 F.3d 586, 591) and review the ultimate finding of vote dilution “only for clear error . . . .” (Ibid.) Of course, deferential review also applies to factual findings under California law. (Ghirardo v. Antonioli (1994) 8 Cal.4th 791, 800 [“ ‘Questions of fact are reviewed by giving deference to the trial court’s decision.’ ”].) But as the Supreme Court recognized in Gingles, deferential review of ultimately factual findings “ ‘does not inhibit an appellate court’s power to correct errors of law, including those that may infect a so-called mixed finding of law and fact, or a finding of fact that is predicated on a misunderstanding of the governing rule of law.’ ” (Gingles, supra, 478 U.S. at p. 79.) Here, the central issue on appeal is whether findings of racially polarized voting are legally cognizable under California voting rights law if they do not appear to meet the “usually” standard for the third Gingles factor. This requires the application of law to facts. (Haworth, supra, 50 Cal.4th at p. 384.) What is more, in the context of state voting rights, it requires careful consideration of legal concepts developed under the rubric of the federal law and inevitably involves the exercise of judgment about the values expressed therein. (See ibid.) We apply our independent judgment in accordance with these principles.

B. The Third Gingles Factor (Requiring a Showing that Majority Block Voting “Usually” Enables Defeat of the Minority Preferred Candidate) Did Not Preclude the Trial Court’s Finding of Racially Polarized Voting The City submits that despite articulating the correct standards of proof at the liability phase of trial, the trial court never applied the “usually” requirement of the third Gingles factor to its findings of fact on the number of racially polarized elections. The City argues that having failed to prove “that the white majority votes sufficiently as a bloc to enable it—in the absence of special circumstances, . . . usually to defeat the

26 minority’s preferred candidate” (Gingles, supra, 478 U.S. at p. 51, italics added), plaintiffs could not establish legally cognizable racially polarized voting. Plaintiffs do not dispute that their burden at trial required them to satisfy both the second and third Gingles factors. But they argue that the third Gingles factor does not impose a strict mathematical formula, and that in any event they proved occurrences of racially polarized voting sufficient to support the trial court’s judgment. In an amicus brief filed by Asian Americans Advancing Justice—Asian Law Caucus (Advancing Justice—ALC) and joined by several legal services and community-based organizations, amici Advancing Justice—ALC, et al., emphasize the historic foundations for the Act and related reasons that the City’s attempt to establish a bright-line threshold for legally cognizable racially polarized voting contravenes the Act’s purpose. We stated in our introductory discussion of the overarching legal framework (see ante, part I.A.) that section 14026 defines racially polarized voting by reference to federal case law. (§ 14026, subd. (e).)8 It is the difference between the voting preferences of the “voters in [the] protected class” and those of “voters in the rest of the electorate” (ibid.) that determines racially polarized voting, as that difference is defined in case law based on the federal Voting Rights Act. (Ibid.) Following the United States Supreme Court’s decision in Gingles, three factors are prerequisite to establishing liability under the federal law for a claim of vote dilution. (Gingles, supra, 478 U.S. at pp. 50-51.) Only

8 We repeat the definition here. “ ‘Racially polarized voting’ means voting in which there is a difference, as defined in case law regarding enforcement of the federal Voting Rights Act of 1965 (52 U.S.C. Sec. 10301 et seq.), in the choice of candidates or other electoral choices that are preferred by voters in a protected class, and in the choice of candidates and electoral choices that are preferred by voters in the rest of the electorate. The methodologies for estimating group voting behavior as approved in applicable federal cases to enforce the federal Voting Rights Act of 1965 (52 U.S.C. Sec. 10301 et seq.) to establish racially polarized voting may be used for purposes of this section to prove that elections are characterized by racially polarized voting.” (§ 14026, subd. (e).)

27 the second and third Gingles factors—showing the minority group is politically cohesive and that majority bloc voting enables it to usually defeat the minority’s preferred candidate (Missouri State, supra, 894 F.3d at p. 930)—are required to prove a violation at the liability stage under California’s Act. (Sanchez, supra, 145 Cal.App.4th at p. 667; see § 14028, subd. (c).) The City, despite criticizing the statistical evidence behind the trial court’s finding of Asian American political cohesion in the disputed elections, does not challenge the application of the second Gingles factor. We focus our attention on the third Gingles factor and whether the trial court misapplied it in this case. The third factor states that the minority group “must be able to demonstrate that the white majority votes sufficiently as a bloc to enable it—in the absence of special circumstances, . . . usually to defeat the minority’s preferred candidate.” (Gingles, supra, 478 U.S. at p. 51.) The Supreme Court expanded on this principle by explaining that whether a political subdivision “experiences legally significant racially polarized voting requires discrete inquiries into minority and white voting practices.” (Id. at p. 56.) Where the minority group also shows its political cohesiveness (the second factor), “a white bloc vote that normally will defeat the combined strength of minority support plus white ‘crossover’ votes rises to the level of legally significant white bloc voting.” (Ibid.) The Supreme Court emphasized the significance of “a pattern of racial bloc voting that extends over a period of time” (id. at p. 57) to distinguish between the “loss of political power through vote dilution” (ibid.) and “the mere inability to win a particular election . . . .” (Ibid.) It is “the usual predictability of the majority’s success [that] distinguishes structural dilution from the mere loss of an occasional election.” (Id. at p. 51.) The City contends that federal cases have uniformly applied the third Gingles factor to require a plaintiff to prove that a majority voting bloc defeats the minority’s preferred candidates in more than 50 percent of the relevant elections. The City points to several federal appellate decisions in support of its more-than-50-percent proposition. It 28 avers that the Ninth Circuit in Old Person, supra, 230 F.3d at page 1122 endorsed the definition of “usually” as “more than half the time.” And it argues that other federal circuits have required an even greater showing to satisfy the “usually” requirement. For example, in Lewis v. Alamance County (4th Cir. 1996) 99 F.3d 600 (Lewis), the court expounded in a footnote its understanding that the third Gingles element is not satisfied “if plaintiffs merely show that white bloc voting defeats the minority-preferred candidate more often than not.” (Id. at p. 606, fn. 4.) The court reasoned that the Supreme Court’s use of the terms “ ‘usually,’ ” “ ‘normally,’ ” and “ ‘generally’ ” to describe the third Gingles factor “mean something more than just 51%.” (Ibid., citing Gingles, supra, 478 U.S. at pp. 49, 51, 56.) The City argues that apart from case law, simple logic contravenes the trial court’s application of the third Gingles factor. The City asserts that just as “[n]o one would say that a flipped coin ‘usually’ lands on heads, because it is equally likely to land on tails,” it cannot be said that Santa Clara’s city council elections are “usually” characterized by racially polarized voting after the trial court found that to be true in only five of 10 elections. We find the City’s reasoning is sound in theory but flawed in practice. It ignores that whether a majority voting block is “usually” able to defeat a cohesive minority group’s preferred candidate per Gingles third factor is not measured by mathematical formula but by the trial court’s searching assessment of statistical and other evidence presented. It indeed would be strange to say that a flipped coin “usually” lands on heads when odds are 50-50 that it lands on tails. But the City’s analogy to a coin toss here fails because it ignores that the legal standard requires a consideration of local circumstances and weighing of factors, not just a simplistic arithmetic exercise. Whether repeated occurrences of racially polarized voting cross the “usually” threshold articulated in Gingles depends on context. It is “the ‘extent to which voting in the elections of the state or political subdivision is racially polarized,’ . . . [that] is relevant to a vote dilution 29 claim.” (Gingles, supra, 478 U.S. at p. 55, italics added, quoting the 1982 Senate Judiciary Committee majority Report, at p. 206.) As the Supreme Court explained, “[b]ecause . . . the extent of bloc voting necessary to demonstrate that a minority’s ability to elect its preferred representatives is impaired varies according to several factual circumstances, the degree of bloc voting which constitutes the threshold of legal significance will vary from district to district.” (Gingles, supra, at pp. 55-56.) A closer look at the cases referenced by the City reveals that they are not inconsistent with this approach. In Old Person, the Ninth Circuit found that the district court erred in applying the third Gingles factor to conclude that white bloc voting “did not usually (i.e., more than half of the time) vote to defeat the preferred candidate of Indian voters.” (Old Person, supra, 230 F.3d at p. 1122.) The decision, however, was not based on the number of occurrences of white block voting but on the district court’s erroneous failure to distinguish electoral success in majority-Indian jurisdictions from the results in majority-white jurisdictions. (Id. at pp. 1122, 1127.) The court notably rejected the suggestion of a “bright-line test” (id. at p. 1127) regarding the standard for determining legally significant white bloc voting in instances when “white voters ‘cross over’ and vote for the minority-preferred candidate.” (Ibid.) Instead, the court reiterated the Gingles court’s observation that “ ‘there is no simple doctrinal test for the existence of legally significant racial bloc voting.’ ” (Ibid., quoting Gingles, supra, 478 U.S. at p. 58.) Similarly in Lewis, though the Fourth Circuit’s decision addressed the third Gingles factor, the subject of the court’s inquiry did not depend on the legal standard for racially polarized elections sufficient to satisfy the “usually” requirement, as that concept is challenged in this case. (See Lewis, supra, 99 F.3d at pp. 605-606.) In fact, the “overarching error” identified in Lewis was that the plaintiffs only introduced evidence of elections in which a black candidate was on the ballot (id. at p. 605), a consideration that would be treated differently under California’s Act, which expressly directs the court to 30 ascertain racially polarized voting by “examining results of elections in which at least one candidate is a member of a protected class . . . .” (§ 14028, subd. (b).) We believe it would be inappropriate to derive a bright-line rule for the minimum frequency of legally significant racially polarized voting based on the “usually” standards recited in Old Person and Lewis, which in each case are not necessary to the decision rendered and may be considered dictum. (City of San Diego v. Board of Trustees of California State University (2015) 61 Cal.4th 945, 958 [“ ‘ “Dictum is the ‘statement of a principle not necessary to the decision.’ ” ’ ”].) What is more, as emphasized in the analysis in Old Person, the analysis of the third Gingles factor does not lend itself to a “ ‘simple doctrinal test’ ” (Old Person, supra, 230 F.3d at p. 1127) because “the degree of racial bloc voting that is cognizable as an element of a . . . vote dilution claim will vary according to a variety of factual circumstances.” (Gingles, supra, 478 U.S. at pp. 57-58.) Indeed, we agree with plaintiffs and their amici that federal decisions consistently espouse a flexible approach to the third Gingles factor. For example, in Vecinos De Barrio Uno v. City of Holyoke (1st Cir. 1995) 72 F.3d 973, the First Circuit highlighted the rigorous, yet flexible approach, required to assess a vote dilution claim. The court compared racially polarized voting to “a silent, shadowy thief” of minority voting rights (id. at p. 984) whose “process of detection typically involves resort to a multifaceted array of evidence including demographics, election results, voting patterns, campaign conduct, and the like” (ibid.). Uno explained that “the question whether a given electoral district experiences racially polarized voting to a legally significant extent demands a series of discrete inquiries not only into election results but also into minority and white voting practices over time.” (Ibid.) The court used familiar language to describe the third Gingles factor, which it said “embodies a showing that the majority votes sufficiently as a bloc to enable it, in the ordinary course, to trounce minority-preferred candidates most of the time . . . .” (Id. at p. 980.) Yet in reviewing the district court’s findings, the court emphasized that “determining whether 31 racial bloc voting exists is not merely an arithmetic exercise that consists of toting up columns of numbers, and nothing more. To the contrary, the district court should not confine itself to raw numbers, but must make a practical, commonsense assay of all the evidence.” (Id. at p. 989.) The court concluded, however, that the district court’s factual findings “reflecting racially polarized voting in at most three or four elections (out of eleven)” (ibid.) did not justify a finding of vote dilution, particularly where the district court “offered no explanation of this seeming contradiction” (ibid.). In Gomez v. Watsonville (9th Cir. 1988) 863 F.2d 1407, Hispanic residents challenged the city of Watsonville’s at-large system of mayoral and city council elections under the federal Voting Rights Act. The Ninth Circuit viewed Watsonville’s at-large election scheme as “the functional equivalent of the electoral scheme at issue in Gingles” (id. at p. 1413) and applied the Gingles factors. Among the facts cited by the reviewing court, voting-age Hispanics comprised about 40 percent of Watsonville’s residents and 37.0 percent of citizens, and no Hispanic had been elected as mayor or city council member prior to trial, though eight Hispanic candidates had run for city council positions and one Hispanic had run for mayor. (Id. at pp. 1409-1410.) The court did not take a formulaic approach to assessing the third Gingles factor. Rather, it accepted the trial court’s factual finding “that Hispanics and Anglos supported different candidates” (id. at p. 1417) based on average support for Hispanic candidates by voters in predominantly white precincts as compared to in heavily Hispanic precincts and noted that a “pattern over time of minority electoral failure” was also probative under Gingles. These combined facts supported the district court’s determination that “the non-Hispanic majority in Watsonville usually vote[d] sufficiently as a bloc to defeat the minority votes plus any crossover votes” (ibid.). In Pope v. County of Albany (2d Cir. 2012) 687 F.3d 565, 578, the Second Circuit explained that the law guiding application of the third Gingles factor “recognizes the need for some flexibility.” Though Pope upheld the district court’s finding that black and 32 Hispanic voters failed to demonstrate a likelihood of success at the preliminary injunction stage on the third Gingles factor, because data omitted from the plaintiffs’ expert’s bloc voting analysis raised questions about the patterns presented (id. at pp. 578-582), the court reiterated that flexibility is warranted depending on the data available (id. at p. 581). Similarly in Flores v. Town of Islip (E.D.N.Y. 2019) 382 F.Supp.3d 197, the district court relied on Pope and other Second Circuit cases for its pronouncement that determining whether the evidence of white bloc voting satisfies the third Gingles factor “is largely a fact-driven inquiry” (Flores, supra, at p. 231) that requires flexibility and for which reason “courts have deviated from a bright-line rule” (ibid.). In Flores, the district court found that because the Hispanic-preferred candidates lost in 13 of 14 town elections, the evidence showed “sufficient white bloc voting to usually defeat the minority-preferred candidate for Town Board.” (Id. at p. 232.) The court explained that while the showing of white voter cohesion may have been weaker than in other cases, “the particular percentage of bloc voting is significantly less important than whether the white bloc regularly defeats the minority-preferred candidate.” (Ibid.) These cases evince a flexible approach to ascertaining the third Gingles factor, grounded in the recognition that what constitutes legally significant racial bloc voting will vary depending on a range of factual circumstances. As such, whether majority bloc voting usually enables defeat of the minority preferred candidate cannot be reduced to a simple mathematical or doctrinal test. (Gingles, supra, 478 U.S. at pp. 56-58.) It follows that the “usually” threshold stated in the third Gingles factor does not as a matter of law preclude a determination of racially polarized voting when the factual findings point to an equal number of polarized and non-polarized elections over time.

C. The Evidence of Racially Polarized Voting Was Adequate to Support the Trial Court’s Determination Under the Third Gingles Factor We next consider whether the trial court erred in finding that plaintiffs’ showing of racially polarized voting in five of 10 city council elections satisfied the third Gingles

33 factor. Plaintiffs argue that even applying a strict interpretation of the “usually” standard, the record supports the trial court’s judgment. They assert that after factoring in the reduced weight attributed to several of the non-racially polarized elections, the trial court’s findings showed racially polarized voting in a majority of the more heavily weighted elections. We agree that the trial court’s determination based on the extent of racially polarized voting in this case was consistent with a section 14028 violation because the threshold of legal significance varies according to factual circumstances. (Gingles, supra, 478 U.S. at p. 56.) The City’s arguments to the contrary pertain to (1) the trial court’s application of the “special circumstance” doctrine articulated in federal case law and (2) the validity of its statistical analysis. We reject both contentions for reasons we explain herein.

1. The Trial Court Did Not Err in Assigning Less Weight to Certain Elections As described above, plaintiffs’ expert, Dr. Kousser, performed a series of trivariate statistical analyses using the three statistical methods to estimate voting patterns among Asian American, Latino, and non-Hispanic white and black voters in Santa Clara. Dr. Kousser noted that for three of the five city council elections in which voting was not clearly polarized, Asian American support for Asian American candidate Mohammed Nadeem dwindled each time. He estimated based on the ecological inference method that Nadeem’s percentage of the Asian American vote was 63.2 percent in 2010; 47.2 percent in 2012; 19.8 percent in 2014; and 16.7 percent in 2016. He observed that Nadeem’s declining popularity could be attributed to his shifting allegiances on controversial issues that dominated Santa Clara politics at the time, particularly support for development of the football stadium, or to the fact that he was one of only four members of the charter review committee who voted against recommending changes to the at-large electoral system. Dr. Kousser posited that Nadeem’s inconsistent stance on the dominant issue in the City’s politics seemed to be a

34 “special circumstance” that accounted for the lack of racial polarization in those contests. Yet Dr. Kousser pointed out that the last three city council elections he analyzed (seats 5 in 2014; seats 4 and 7 in 2016) were racially polarized regardless of the candidates’ preferences on those same, controversial issues, suggesting that the Asian American candidate losses could not be attributed to political stance. Dr. Kousser opined that “whichever factional white candidate they opposed, the Asian candidates always lost.” In its statement of decision, the trial court discounted Dr. Kousser’s proffered explanation for Nadeem’s poor performance with Asian American voters as “speculative” but found that Nadeem’s track record for losing Asian American support still justified giving less weight at least to the 2012, 2014, and 2016 election data. The City contends that in doing so the trial court misconstrued the “special circumstances” doctrine—which under Gingles may be applied to explain instances of minority electoral success in a polarized environment, not electoral failure. (Gingles, supra, 478 U.S. at p. 57, fn. 26.) The City argues that in the absence of a “special circumstances” finding, the trial court’s decision to give “less weight” to those election results is not supported by case law governing what constitutes racially polarized voting under section 14026, subdivision (e). Plaintiffs respond that courts applying the fact-intensive and flexible standards under federal case law and section 14028 may properly increase or decrease the weight given to individual elections for reasons not limited to the “special circumstances” discussed in Gingles. The City is correct that the Supreme Court in Gingles framed “special circumstances, such as the absence of an opponent, incumbency, or the utilization of bullet voting” as a way to explain instances of minority candidate electoral success in a polarized contest. (Gingles, supra, 478 U.S. at p. 57.) The City points out that neither plaintiffs nor the trial court in this case cited a single case in which “special circumstances” applied to a non-polarized election involving an unsuccessful minority candidate. We find the point moot, however, because the trial court in this case rejected 35 Dr. Kousser’s suggestion to treat the Nadeem races as a special circumstance and instead applied its broader factfinding discretion to give less credit to those elections. In other words, the trial court’s finding that Dr. Kousser’s “speculation about Dr. Nadeem’s voting record” on issues salient to Santa Clara voters did not “rise[] to the level of ‘special circumstances’ that [would] warrant disregarding Dr. Nadeem’s election losses” did not preclude the trial court from deciding to give those elections less weight based on the peculiar circumstance of an Asian American candidate losing support from Asian American voters in each subsequent race. By its plain language, section 14028 suggests that courts look to “the extent to which candidates who are members of a protected class and who are preferred by voters of the protected class, as determined by an analysis of voting behavior, have been elected to the governing body of a political subdivision that is the subject of” an action for vote dilution under the Act. (§ 14028, subd. (b), italics added.) The trial court properly considered the 10 city council elections in which Asian American candidates participated. The fact that one of those Asian American candidates lost ground each time as a preferred candidate of the protected class, as demonstrated by Dr. Kousser’s analysis, is a valid circumstance for consideration. (Ibid.) As explained in Gingles, the court performs “discrete inquiries into minority and white voting practices” (Gingles, supra, 478 U.S. at p. 56) to ascertain whether the district experiences legally significant racially polarized voting. As we understand Gingles’s elucidation of legally significant racially polarized voting, the extent to which racially polarized voting impairs the minority group’s political power depends on case-specific circumstances. (Gingles, supra, 478 U.S. at pp. 56-57.) At times, the court may need to extend its inquiry to consider factors likely to have influenced the electoral outcomes, including features of the local election system affecting cohesion levels (id. at p. 56) and election results that fall outside of the dominant pattern of polarization due to “special circumstances” (id. at p. 57). For each 36 example in which the Supreme Court listed such factors, it noted they were “illustrative” and not comprehensive or exclusive. (Id. at p. 56, fn. 24 [factors affecting degree of minority cohesion]; id. at p. 57, fn. 26 [special circumstance factors].) Gingles thus expressed a broad view of the factors relevant to the assessment of racially polarized voting, even as it defined “special circumstances” in relation to atypical instances of minority electoral success. We are not convinced that the trial court erred in weighing certain elections differently. We find support for this conclusion by looking to federal cases in which courts gave more (or less) weight to certain elections based on factors other than “special circumstances.” One such case is Ruiz v. City of Santa Maria (9th Cir. 1998) 160 F.3d 543, 552, in which the Ninth Circuit considered whether the district court “should have placed less evidentiary weight on Hispanic voters’ ability to elect . . . a white candidate[] than on their inability to elect a Hispanic candidate.” After reviewing how circuit courts addressed similar arguments in other cases, the court reasoned that “the inability of Hispanic voters to elect a Hispanic candidate is more probative in a Gingles prong three analysis than the ability of Hispanic voters to elect a non-minority candidate.” (Id. at p. 553.) The court held accordingly that the minority group’s ability to elect a non-minority candidate warranted less weight in the Gingles prong three analysis. (Id. at pp. 553-554). It explained that the district court’s “mechanical approach” (id. at p. 554) to analyzing the election data failed to fulfill its duty “to make ‘a searching practical evaluation of the past and present reality’ with ‘a functional view of the political process.’ ” (Ibid., quoting Gingles, supra, 478 U.S. at p. 45.) Other cases similarly assign more probative value to elections involving a minority candidate, as opposed to elections involving only majority-representative candidates. (See, e.g., Jenkins v. Red Clay Consol. Sch. Dist. Bd. of Educ. (3d Cir. 1993) 4 F.3d 1103, 1128-1129; League of United Latin Am. Citizens, Council No. 4434 v. Clements (5th Cir. 1993) 986 F.2d 728, 748 (Clements) [“[I]n ascertaining whether a 37 community’s elections are characterized by racially polarized voting, a court may properly give more weight to elections in which the minority-preferred candidate is a member of the minority group”].) In another example, also from Clements, the Fifth Circuit accepted the district court’s weighting of certain elections as more probative than others, noting the district court must retain flexibility when faced with a sparsity of data. (Clements, supra, at p. 792.) The City disputes the relevance of Ruiz and the other cases cited by plaintiffs, none of which address the probative value of non-polarized elections involving a minority candidate. We recognize that the cited cases have limited applicability to the factual circumstance presented here. What they do provide is direction as far as the legal principles established by case law that guide the trial court’s analysis. (See § 14026, subd. (e).) These cases are “driven by the facts.” (Clements, supra, 986 F.2d at p. 736.) It is appropriate under such circumstances for the existence of baseline prerequisites, like the three Gingles factors, to embody the “flexible, fact-intensive” (Gingles, supra, 478 U.S. at p. 46) nature of the vote dilution inquiry. We conclude that a court’s analysis of racially polarized voting, in accordance with Gingles’ third factor and consistent with section 14028, invariably depends on its ability to weigh the usefulness of the election evidence presented and to assign probative value where appropriate. We decline to penalize the trial court for assigning less evidentiary significance to the Nadeem races. To impose an overly restrictive interpretation on the trial court’s reasonable discretion to assign probative value would contravene the flexible, factfinding approach indicated in cases enforcing the federal Voting Rights Act (Gingles, supra, 478 U.S. at p. 62) and suggested by the language of section 14028.

38 2. The Trial Court Did Not Abuse its Discretion in Considering a Lower Confidence Interval for the Racially Polarized Voting Analysis Having found no error of law in the trial court’s application of the third Gingles factor, we turn to the City’s contention that the trial court abused its discretion by conducting a posttrial statistical analysis to find racially polarized voting in five of 10 city council elections.9 As previously summarized, the trial court credited Dr. Kousser’s analytical methodology and found the ecological inference results “probative” despite the uncertainties highlighted by the City’s expert, Dr. Lewis. It rejected the City’s contention that overlapping 95 percent confidence intervals made it impossible to identify the Asian preferred candidate in the disputed, 2016 elections city council elections for seat 4 and seat 7. The trial court found that applying an 80 percent confidence interval enabled identification of an Asian preferred candidate and provided “sufficiently reliable results.” The City attacks the trial court’s use of a lower confidence interval to ascertain which candidate was preferred by Asian American voters and its reference to point estimates to bolster its findings. It contends that the trial court stepped outside of its gatekeeping role by substituting its own methodology for the analyses offered by the expert witnesses at trial, arriving at a result that was unsupported by the evidence and inadequate to satisfy the third Gingles factor.

9 The City states in its reply brief that it “does not challenge” the trial court’s finding that five out of 10 elections exhibited racially polarized voting. We do not construe this statement as a concession to the trial court’s factual findings, which the City does challenge in its opening brief. Instead, we understand the arguments in the alternative. The City appears to challenge the trial court’s application of the third Gingles factor, arguing that even if supported by the evidence, the statistical results cannot as a matter of law satisfy plaintiffs’ burden to prove that racially polarized voting enables the white majority to “usually” defeat Asian American preferred candidates. Alternatively, the City challenges the trial court’s factual finding of racially polarized voting in five out of 10 city council elections.

39 Plaintiffs dispute any error in the trial court’s methods. They argue that the trial court soundly rejected the City’s claim that the candidates preferred by Asian American voters could not be shown with sufficient reliability in the two disputed city council elections. Plaintiffs assert that the law does not require use of a 95 percent confidence level to determine racially polarized voting, because the test for legally significant racially polarized voting is not tied to a specified level of statistical significance. We evaluate the contentions by two-part inquiry. First, did the trial court violate its gatekeeping responsibility by adopting a statistical method that was unvetted by the adversarial process or expert witness testimony? Second, did the trial court’s process undermine or invalidate its substantive findings of racially polarized voting in satisfaction of section 14028? As to the first question, the City relies on case authority that delineates the procedures for admitting and evaluating expert testimony. “Under California law, trial courts have a substantial ‘gatekeeping’ responsibility.” (Sargon Enterprises, Inc. v. University of Southern California (2012) 55 Cal.4th 747, 769.) Consistent with statutory and decisional law, the trial court determines whether the expert opinion testimony is admissible. (Id. at pp. 771-772.) The court at the admissibility stage does “not weigh an opinion’s probative value or substitute its own opinion for the expert’s opinion.” (Id. at p. 772.) Its gatekeeping responsibility “is simply to exclude ‘clearly invalid and unreliable’ expert opinion.” (Ibid.) In a bench trial, of course, the court as trier of fact also weighs the evidence. (See Code Civ. Proc., § 631.8, subd. (a).) The City contends, however, that the judge’s role as the trier of fact in a bench trial does not permit the court to conduct an independent “expert” analysis. The City argues that expert opinion must be vetted through the adversarial process, especially in complex matters dependent on statistical methods, like those to estimate group voting behavior. The City cites Duran v. U.S. Bank National Assn. (2014) 59 Cal.4th 1, 49 (Duran) for the proposition that a trial court may not substitute its own statistical methods for the 40 analyses offered by the expert witnesses at trial. The trial court in Duran devised a statistical sampling plan, without input from the parties’ experts, for use in the liability and damages phases of a wage and hour class action bench trial. (Id. at pp. 12, 38.) The court’s method extrapolated from the testimony of a small sample group, restricted to 21 of the 260 class members, to determine liability as to the entire class, and to prove damages. (Id. at p. 12.) The court “adamantly adhered” to its “invented” methodology (id. at p. 49) despite “substantial expert criticism” (ibid.). The California Supreme Court rejected the trial court’s approach and reversed the judgment, explaining that “[a] trial plan that relies on statistical sampling must be developed with expert input and must afford the defendant an opportunity to impeach the model or otherwise show its liability is reduced.” (Id. at p. 13.) The high court found the court’s sampling method was “profoundly flawed” (ibid.) for numerous reasons related to the limited sample size, selection bias, and resulting margin of error. (Id. at pp. 42-46.) The City likens the trial court’s use of a lowered confidence interval in this case to the statistical sampling plan devised by the trial court in Duran. It contends that the application of an “80 percent confidence interval” lacked any support in expert evidence or input from the adversarial process and, like the sampling plan in Duran, supra, 59 Cal.4th at page 13, deprived the City of any opportunity to impeach it. The City further contends that by selecting an untested confidence interval and applying it to the two 2016 elections at issue, the trial court (perhaps inadvertently) grossly increased the margin of error while ignoring evidence in the record that expert statisticians do not use such low confidence intervals in their ordinary work. It argues that because a 95 percent confidence interval means the true answer is not in the identified range five times out of 100, as Dr. Kousser explained, and an 80 percent confidence interval means the true answer is outside the range 20 times in 100, the difference marks a 400 percent increase in the likelihood of error. The City notes that the trial court repeated the “same unsupported statistical analysis” in its review of the school elections. 41 We find that the trial court’s decision to use 80 percent confidence intervals to ascertain Asian American cohesion behind a preferred candidate fell well within the bounds of its discretion. Duran is distinguishable because the “invented” and unvetted statistical sampling plan (Duran, supra, 59 Cal.4th at p. 49) defined all manner of proof of liability at trial (id. at pp. 39-40) and deprived the defendant of the ability to litigate certain affirmative defenses (id. at p. 35). Here, the selection of an alternative confidence interval did not dictate the methods of proving liability or restrict the City from litigating its defenses or the specific issue of how the confidence interval affected the results of the inference analyses. On the contrary, the question of confidence intervals—their application and interpretation—was thoroughly litigated. In his testimony, Dr. Kousser described the 95 percent confidence interval as “a standard convention” and “the most usual one” seen in political science literature. He testified that it “means that if the null hypothesis is that there is no difference between one point estimate and the other point estimate, that five times out of 100 we would say that there was a difference at some level. But we would be wrong.” Though he used the 95 percent confidence interval (also referred to, conversely, as .05) for his calculations, he explained that other confidence levels could be used: “If you look at the standard errors, you can use any confidence interval that you want to. . . . [¶] . . . I used .05 [95 percent confidence] because that’s a standard convention.” As plaintiffs point out, Dr. Kousser described in detail during his redirect examination how a lower confidence interval could apply. He explained that his calculations showed an Asian American preferred candidate in six of the 10 city council elections at the 95 percent confidence interval.10 But at an alternative measure of statistical significance, such as 90 percent confidence, “[y]ou could reject the null

10This finding was expressed in the relevant trial exhibit as “[s]tatistically significantly preferred over number two candidate at .05 level . . . .”

42 hypothesis that there was no preferred candidate” and would be wrong “ten times out of 100 rather than five times out of 100.” Dr. Kousser stepped plaintiffs’ counsel through the process of lowering the confidence interval, explaining that “[y]ou can do the same thing for 80 percent . . . . And the way to get those figures to make that determination is to multiply [the standard errors] by 1.65 for 90 percent or 1.28 for 80 percent rather than 1.96 . . . for 95 percent.” In the re-cross examination that followed, the City’s counsel asked no questions about lowering the confidence interval. This pattern repeated the next day when plaintiffs’ counsel pressed the City’s expert, Dr. Lewis, on his opinion that “sufficient uncertainty in the estimates” meant he could not reliably conclude whether racially polarized voting had occurred. Plaintiffs’ counsel asked if Dr. Lewis had attempted to apply 80 percent or 90 percent confidence intervals to the data to see if that would allow him to draw a conclusion. Dr. Lewis explained that in the 2016 elections he analyzed, he “reported estimates of the probability” of each candidate being the preferred candidate among each group without identifying a threshold confidence interval. The trial court interjected with its own questions about Dr. Lewis’s data, including about his preferred candidate analysis for the disputed 2016 election (seat 7). Dr. Lewis estimated that there was an 80 percent probability that candidate Park was the preferred candidate of Asian Americans. The trial court inquired, “[s]o when we see a .80 here, do you believe that it is unreliable to assume that Park was not preferred candidate of Asian Americans?” Dr. Lewis answered that assuming the other model assumptions were correct, he “couldn’t say that with . . . 95 percent certainty” but “could say it with 80 percent certainty.” The City’s counsel followed in redirect without addressing confidence intervals or Asian American cohesion. Plaintiffs revisited the subject again in Dr. Kousser’s rebuttal testimony. Dr. Kousser agreed that there could be a statistical correlation at less than 95 percent confidence interval or at the .05 uncertainty level, as Dr. Lewis had himself done in his 43 tables on preferred candidates, and testified extensively about how overlapping confidence intervals may be interpreted according to different statistical theories. The City declined to examine Dr. Kousser further. Plaintiffs reiterated the point in their posttrial brief, charging that the City’s attempt to refute plaintiffs’ showing of racially polarized voting by claiming the statistical inference analyses did not show Asian American cohesion behind a preferred candidate was flawed in several respects. The thrust of plaintiffs’ argument was that federal case authority following Gingles did not require cohesion to be shown by a statistically significant preference for a single candidate, that it was error to conflate statistical significance with plaintiffs’ burden to prove politically cohesive voting at the preponderance of the evidence standard, and that in any event confidence levels of 80 percent were probative to show correlation between Asian American voters and a preferred candidate. The record thus reflects extensive expert testimony—and pointed questions by the trial court—about calculating Asian American voter cohesion behind a preferred candidate at 80 percent confidence interval as opposed to the standard convention of 95 percent, from both technical and interpretive perspectives. This record undercuts the City’s claim that the trial court usurped the role of the expert witness and applied a methodology that was not vetted by the adversarial process. Nor did the trial court create its own evidence, as the City suggests. Since Dr. Kousser reported the standard errors for each election that he analyzed, including the city council elections that the City disputes showed statistically significant cohesion behind a preferred candidate (2016 seats 4 and 7), the trial court’s decision to adopt the alternate convention at 80 percent confidence interval for those elections can hardly be compared to conjuring its own statistical model. (Cf. Duran, supra, 59 Cal.4th at p. 49.) Instead, the trial court credited Dr. Kousser’s analysis of preferred candidates but found that in the few disputed elections in which preferred status could not be confirmed at 95 percent confidence, it could be confirmed at 80 percent. We find this was a valid exercise of discretion by the trial court as the finder 44 of fact when faced with competing opinions on the need to identify the candidate preferred by Asian American voters at a stated confidence level. Turning to the second question, the City contends that because the trial court’s calculations were not supported by evidence in the record or vetted through the adversarial process, they cannot support the finding that five of 10 city council elections, or four of nine school elections, involved racially polarized voting. We find the issue largely resolved by our analysis of the record above, which contains ample evidentiary support for the trial court’s use of a lower confidence interval in assessing the disputed evidence of voter cohesion. We nevertheless briefly address the implication that calculating voter support for a preferred candidate at 80 percent confidence interval invalidated the trial court’s substantive findings of racially polarized voting. The City suggests that the trial court likely did not understand the effect of the lower confidence level on the reliability of the conclusion. We believe this supposition is contradicted by the record. The trial court’s colloquies with counsel and the experts reveal that the court was conversant with the relevant statistical principles. More importantly, the court’s determination that assessing the protected class’s preferred candidate at 80 percent confidence interval was “sufficiently reliable” under the circumstances to support a finding of racially polarized voting must be viewed in context of the operating framework of California’s Voting Rights Act. This framework, as previously discussed (see ante, parts I.A and II.C.1) was patterned after the federal Voting Rights Act but “provide[s] a broader cause of action for vote dilution.” (Sanchez, supra, 145 Cal.App.4th at p. 669.) As plaintiffs and amici Advancing Justice—ALC have variously pointed out, the Act recognizes “[t]he methodologies for estimating group voting behavior” (§ 14026, subd. (e)) approved in federal case law to establish racially polarized voting but does not mandate that plaintiffs seeking to prove a vote dilution claim under California law adopt those same

45 methodologies. (Ibid. [methodologies approved in federal case law “may” be used to prove that elections are characterized by racially polarized voting].) Even if California’s definition of racially polarized voting had limited state trial courts to the methodologies approved in federal Voting Rights Act case law to estimate group voting behavior, we are unaware of federal case authority that prescribes a bright-line rule tying legal sufficiency of cohesion to a mathematical formula or statistical method. For example, in United States v. City of Euclid (N.D. Ohio 2008) 580 F.Supp.2d 584, 596 (Euclid), the district court expressly rejected the contention that a “failure to satisfy . . . proposed statistical benchmarks” precluded the plaintiffs’ vote dilution claim. The court emphasized that when assessing statistical evidence of racially polarized voting, courts should keep in mind “the broader legal principles described in Gingles” and be “neither to be wedded to, nor hamstrung by, blind adherence to statistical outcomes. . . . [W]hile courts have found certain particular statistical or mathematical outcomes to be compelling evidence in the context of the cases before them, no decision out of either the Supreme Court or the Sixth Circuit (or any other Circuit for that matter) requires the use of a particular statistical methodology, or demands a particular statistical outcome before a court may conclude that racial bloc voting exists.” (Ibid.) It bears repeating that voting rights claims as interpreted according to federal case authority “are inherently fact-intensive” (Nipper v. Smith (11th Cir. 1994) 39 F.3d 1494, 1498), requiring “a ‘searching practical evaluation of the “past and present reality” ’ of the electoral system’s operation” (ibid., quoting Gingles, supra, 478 U.S. at p. 45). California’s statute demands an equally fact-intensive expedition through the factors for ascertaining racially polarized voting while also enabling greater flexibility around variables like geographic compactness (§ 14028, subd. (c)) and “[o]ther factors” deemed “probative” but not necessary to establish a violation (id., subd. (e)). The evidence presented “must be evaluated with a functional, rather than a formalistic, view of the political process . . . .” (Nipper, supra, at p. 1498; see Gingles, supra, at pp. 45, 79.) The 46 City’s restrictive view of the trial court’s ability to make findings at 80 instead of 95 percent confidence intervals clashes with this flexible and functional approach. We decline to edge toward a bright-line rule for substantially similar reasons as we stated in our rejection of a percentile-based “usually” requirement for the third Gingles factor. (See ante, part II.B.) In sum, where the outcome depends to some degree on evidence produced by statistical analysis, courts must be able to exercise discretion in weighing the probative value against the uncertainties and limitations inherent in statistical methods. We agree to that end with the district court’s observation in Euclid that a statistical approach “might yield an inexact result for purposes of a hypothetical mathematical challenge, but could still be correlative, probative, and sufficiently accurate to bear on the ultimate issue of racial bloc voting. The standard of proof . . . is preponderance, not mathematical certainty.” (Euclid, supra, 580 F.Supp.2d at p. 602; cf. Turpin v. Merrell Dow Pharmaceuticals, Inc. (6th Cir. 1992) 959 F.2d 1349, 1357, fn. 2 [distinguishing scientists’ use of confidence intervals “as a common-sense device to give professional weight to their results” from the preponderance of the evidence standard of proof, which “requires proving one’s case by the greater weight of the evidence”].) We conclude that statistical tools for expressing degrees of certainty should not eclipse the factfinder’s ability to weigh the evidence and decide whether it meets the legal standard of proof, as occurred here.11 D. The District-Based Remedy Did Not Violate Equal Protection The City contends that the trial court’s judgment violated the Equal Protection Clause of the Fourteenth Amendment to the United States Constitution by imposing a

11Because we find no abuse of discretion in the trial court’s identification of the candidate preferred by Asian American voters for the disputed elections at a lowered confidence interval, we need not address the use of point estimates as an alternative to confidence intervals.

47 “draconian race-conscious remedy” without an adequate showing that structural vote dilution existed in the jurisdiction, or that abolishing at-large elections for Santa Clara city council seats would remedy any such vote dilution. The City’s constitutional claim is premised on the argument that by erroneously failing to enforce the “usually” requirement of the third Gingles factor in its liability determination, the trial court imposed a race-conscious remedy without meeting strict scrutiny standards. The City’s argument may be summarized as follows. The City generally asserts that the Act uses race-based classifications both to authorize a challenge by a member of a protected class to an at-large election system (§§ 14032, 14026, subd. (d)) and to confer liability on the basis of racially polarized voting (§§ 14028, 14026, subd. (e)). It contends that racially polarized voting under the Act distinguishes between individuals on racial grounds and accordingly “falls within the core prohibition of the Equal Protection Clause.” It claims that because the “usually” test required by the third Gingles factor enables courts to “distinguish[] structural dilution from the mere loss of an occasional election” (Gingles, supra, 478 U.S. at p. 51), its enforcement by courts serves as a crucial safeguard to the constitutional application of the Act. The City avers that without the “usually” test, the trial court’s imposition of a district-based electoral remedy that takes race into account cannot survive strict scrutiny review. The City’s equal protection argument fails on all fronts. The City invokes strict scrutiny review without squarely addressing settled California authority holding that the race-conscious provisions of the Act do not trigger strict scrutiny. In Sanchez, the Fifth District Court of Appeal rejected the City of Modesto’s attempt to show the Act was facially invalid because any possible application of it would necessarily involve unconstitutional racial discrimination by using “race” to identify the polarized voting that causes vote dilution. (Sanchez, supra, 145 Cal.App.4th at pp. 665, 666.) Sanchez found that the race-related provisions of the Act do not trigger strict scrutiny because the Act does not favor any race over others or allocate benefits or impose burdens on the basis of 48 race. (Id. at pp. 680-681, 687-688.) Having rejected the argument that strict scrutiny should apply, Sanchez held that the Act “readily passes” rational basis review. (Id. at p. 680.) We reject the City’s attempt to revive arguments that were rejected over a decade ago in Sanchez, supra, 145 Cal.App.4th at pages 665, 680 through 681. The City suggests no case authority or reasoned argument that would lead us to depart from the thoroughly supported ruling in Sanchez. It does not explain how the Act “distributes burdens or benefits on the basis of individual racial classifications” such as would trigger strict scrutiny review. (Parents Involved in Cmty. Sch. v. Seattle Sch. Dist. No. 1 (2007) 551 U.S. 701, 720.) Instead, the City attempts to frame its equal protection argument as an as-applied challenge based on the purportedly race-conscious remedy ordered by the trial court without application of the “usually” test. In theory, the City has a valid basis for trying to raise an as-applied challenge. Sanchez addressed only the facial validity of the Act, leaving room for a defendant in a vote dilution case “to attempt to show as-applied invalidity . . . if liability is proven and a specific application or remedy is considered that warrants the attempt.” (Sanchez, supra, 145 Cal.App.4th at p. 665.) The court stated, for example, that a defendant faced with “a remedy that uses race, such as a district-based election system in which race is a factor in establishing district boundaries . . . may again assert the meaty constitutional issues” raised by the City of Modesto in that case. (Ibid.) The City’s as-applied attempt here, however, fails on the merits. It states, without citation to the record, that the trial court forced the City to adopt a district-based system and to choose “among proposed maps that all took race into account in drawing the proposed boundaries between districts.” Given that “race-conscious redistricting is not always unconstitutional” (Shaw v. Reno (1993) 509 U.S. 630, 642), more than a conclusory statement that the trial court took race into account is required.

49 This is true even insofar as the City’s invocation of strict scrutiny for the as-applied challenge. (See Bush v. Vera (1996) 517 U.S. 952, 958 [“Strict scrutiny does not apply merely because redistricting is performed with consciousness of race.”].) The City fails to point this court to even a single example from the record that would show the trial court’s selection of a district-based remedy made race “ ‘the predominant factor motivating the . . . [redistricting] decision.’ ” (Id. at p. 959; see also Higginson v. Becerra (S.D. Cal. 2019) 363 F.Supp.3d 1118, 1125 [holding that a plaintiff seeking to state “a racial gerrymandering claim subject to strict scrutiny under the Equal Protection Clause . . . must allege facts to support the inference that a districting decision was made ‘on the basis of race’ ”], aff’d (9th Cir. 2019) 786 Fed. Appx. 705.) In Higginson, the district court dismissed a challenge similar to that which the City outlines here, finding that the complaint’s allegations under California’s Voting Rights Act did not support the inference that passage of the Act, or its implementation through the City of Poway’s ordinance implementing district-based elections, “classified [the plaintiff] into a district because of his membership in a particular racial group.” (Higginson, supra, at p. 1127.) We find, based on these principles, that the unsupported reference to race-based considerations does not support the City’s call for strict scrutiny review. Its arguments based on the “usually” test also do not assist its case because, as explained ante in our discussion of the third Gingles factor, the trial court’s determination of racially polarized voting did not fail to apply the “usually” requirement. We accordingly reject the City’s claim to strict scrutiny review. Since the City does not attempt to assert an as-applied challenge under rational basis review, we conclude that its equal protection claim fails.

E. Application of the California Voting Rights Act to Santa Clara Did Not Violate Charter City Plenary Authority Under the California Constitution The City’s final argument is based on article XI, section 5, subdivision (b) of the California Constitution, which grants charter cities, in relevant part, “plenary authority”

50 to decide “the manner in which” their municipal officers are elected. (Cal. Const., art. XI, § 5.) In its statement of decision, the trial court summarily relied on Jauregui v. City of Palmdale (2014) 226 Cal.App.4th 781, 802 (Jauregui) for the proposition that California’s Voting Rights Act preempts city charter provisions that establish at-large election of city council members. In Jauregui, the Court of Appeal held that the driving forces behind adoption of the Act, including the implementation of equal protection and voting rights and the integrity in the conduct of local elections, constituted an issue of statewide concern, enabling the statute to override the charter city’s otherwise plenary power over its municipal elections. (Id. at pp. 799-802.) On appeal, the City claims that the Jauregui decision focused only on the question of the Act’s statewide interest in preventing race-based voter dilution as justification to supersede the charter city’s authority over its municipal affairs, based on article XI, section 5, subdivision (a) of the California Constitution, and failed to fully consider the charter city’s “plenary” authority under article XI, section 5, subdivision (b)(4) of the California Constitution. An amicus curiae brief filed by John K. Haggerty, a resident of Santa Clara, expands on the City’s “plenary authority” arguments and urges this court to part ways with Jauregui. Haggerty argues that the Court of Appeal in Jauregui did not adequately weigh the statewide interest held by California’s citizens to protect charter cities’ exclusive control over their own municipal affairs, especially their own local elections. He asserts that the Jauregui decision failed to adequately consider the “plenary authority” language used in article XI, section 5(b) of the California Constitution. Haggerty also raises the question of an as-applied equal protection violation, discussed above, arguing that the trial court’s race-conscious remedy of imposing district-based elections should be reviewed both for its potential impingement on equal protection guarantees for affected citizens and for its imposition of a state authorized electoral remedy over a charter city’s chosen manner of election. 51 Plaintiffs defend the trial court’s remedy and Jauregui’s interpretation of statewide interest. Plaintiffs assert, contrary to amicus Haggerty’s depiction of the unwanted imposition of the district-based remedy, that a wide majority of Santa Clara voters actually supported an advisory ballot measure in Santa Clara’s November 2018 election to engage citizens in a process to draft a charter amendment to elect council members, other than mayor, by district.12 The vote on the advisory ballot measure took place after the liability and remedies phases of trial in this case. We observe that Jauregui directly addressed the “plenary authority” provision. (Jauregui, supra, 226 Cal.App.4th at pp. 802-803.) The appellate court concluded that the plenary authority identified in article XI, section 5, subdivision (b) of the California Constitution and expressly applied to municipal elections “can be preempted by a statewide law after engaging in the four-step evaluation process specified by our Supreme Court” to ascertain whether the subject matter of the general law is of statewide concern. (Jauregui, supra, at p. 803.) Although Haggerty contends that Jauregui erroneously determined that the Act’s treatment of racial dilution in elections is a matter of statewide concern, the City’s appeal does not challenge this point. The Legislature moreover

12 We grant plaintiffs’ request for judicial notice of the following items: (1) City of Santa Clara Resolution No. 18-8574, which placed on the November 2018 ballot an Advisory Measure N, asking “Shall the City of Santa Clara engage the voters in a public process to draft a Charter Amendment ballot measure to elect its Council Members, other than the Mayor, by district?”; (2) November 2018 election results for Advisory Measure N as produced by the Santa Clara County Registrar showing 70.4 percent “Yes” votes; and (3) Meeting Agenda of the City of Santa Clara Charter Review Committee for September 26, 2019, which includes the results of a City-administered survey revealing that more than 60 percent of voters prefer a six-district election system for City Council. The City does not oppose the request for judicial notice. The ballot measure, the election results on the ballot measure produced by the county registrar, and the charter review committee meeting agenda and City survey results are “[f]acts and propositions that are not reasonably subject to dispute and are capable of immediate and accurate determination by resort to sources of reasonably indisputable accuracy.” (Evid. Code, § 452, subd. (h); id., § 459.)

52 declared its intent to codify the holding in Jauregui in amendments to the statute in 2015. (Stats. 2015, ch. 724 § 1, subd. (a)-(d) [declaring the dilution of votes of a protected class to be “a matter of statewide concern” for which the provisions of the Act “constitute a narrowly-drawn remedy that does not unnecessarily interfere with municipal governance” and stating the intent to apply the Act to charter cities and counties].) As amended, the Act expressly includes charter cities among the “political subdivisions” subject to the Act’s provisions. (§ 14026, subd. (c).) Such legislative declarations of statewide concern are not determinative but are relevant and entitled to “great weight” by the court in deciding whether the general law supersedes conflicting charter enactments. (Anderson v. City of San Jose (2019) 42 Cal.App.5th 683, 703, 707; see State Building & Construction Trades Council of California v. City of Vista (2012) 54 Cal.4th 547, 565.) We decline the invitation to depart from the Jauregui court’s reasoning and holding. The City acknowledges that its charter is subject to the Act’s implementation of equal protection guarantees securing the ability of members of a protected class to exercise their voting rights. (See §§ 14027, 14031.) It “agrees that its charter must yield if the City’s method of holding elections violates a protected class’s right to equal protection of the laws, as implemented” in the Act. But it asserts there “can be no such violation unless” the at-large system of elections met the “usually” standard for racially polarized voting. It asks this court to consider its plenary power argument only in that “limited and specific context.” Our conclusion that the trial court satisfied the applicable standard in determining racially polarized voting resulting in vote dilution, however, renders the argument moot. We conclude that the application of the Act to Santa Clara’s charter did not impinge unlawfully on the City’s plenary authority to control the manner and method of electing its officers. F. Appeal of the Award of Attorney Fees and Costs The City premises its appeal of the trial court’s award of attorney fees and costs solely on the anticipated reversal of the trial court’s liability judgment. Because we find 53 no error requiring reversal of the judgment of liability under the Act, we affirm the award of attorney fees and costs.13

III. DISPOSITION

The judgment of liability under the California Voting Rights Act is affirmed. The award of attorney fees and costs to plaintiffs is also affirmed. Plaintiffs are entitled to their costs on appeal.

13 Plaintiffs assert that since the entry of the award of attorney fees and costs, they have expended “substantial time” and some costs on the appeal and “other case-related work.” They ask this court to remand the matter with instructions to determine the amounts of reasonable costs and fees due for that work. Plaintiffs may pursue their costs claim in the superior court after the issuance of the remittitur in this matter, pursuant to California Rules of Court, rule 8.278(c).

54 Premo, Acting P.J.

WE CONCUR:

Elia, J.

Danner, J.

Yumori-Kaku et al. v. City of Santa Clara

H046105; H046696

Trial Court: Santa Clara County Superior Court Superior Court No. 17CV319862

Trial Judge: Hon. Thomas E. Kuhnle

Counsel for Defendant/Appellant: Churchwell White Steven G. Churchwell City of Santa Clara Karl A. Schweikert J. Scott Miller Liah Burnley

Downey Brand John C. McCarron

Boersch & Illovsky Kevin A. Calia

Santa Clara City Attorney’s Office Brian L. Doyle Sujata Reuter

Amicus Curiae in support of John K. Haggerty Defendant/Appellant In Propria Persona City of Santa Clara

Counsel for Plaintiffs/Respondents: GOLDSTEIN, BORGEN,

DARDARIAN & HO

LaDonna Yumori-Kaku Morris J. Baller, Wesley Kazuo Mukoyama Laura L. Ho Umar Kamal Anne P. Bellows Michael Kaku Ginger L. Grimes Herminio Hernando

LAW OFFICE OF ROBERT RUBIN

Robert Rubin

ASIAN LAW ALLIANCE

Richard Konda

Yumori-Kaku et al. v. City of Santa Clara

H046105; H046696