Full opinion text
FINDINGS OF' FACT AND CONCLUSIONS OF LAW; JUDGMENT BREWSTER, District Judge. This ease involves a challenge by Hispanic plaintiffs to the at-large election system of the El Centro Elementary School District Board of Trustees (“El Centro”) under Section 2 of the federal Voting Rights Act, 42 U.S.C. Section 1973. Hispanic plaintiffs allege that El Centro’s at-large election system violates Section 2 of the Voting Rights Act, both because it “dilutes” the ability of Hispanic voters to elect candidates of their choice and also because it impairs their ability to “influence” elections. Vote dilution occurs when a minority group is unable to elect candidates at-large whom it could elect within a single member district where the group is a majority. Thornburg v. Gingles, 478 U.S. 30, 68, 106 S.Ct. 2752, 2775, 92 L.Ed.2d 25 (1986). Thornburg, the leading case interpreting Section 2, sets out the following three preconditions which minority plaintiffs must establish to prove vote dilution: 1. The minority group is sufficiently large and geographically compact to constitute a majority in a single member district. 2. The minority group is politically cohesive. 3. The white majority votes sufficiently as a bloc to enable it, in the absence of special circumstances, to usually defeat the minority’s preferred candidate. Id. at 48-50, 106 S.Ct. at 2765-66. If all three preconditions are established, the Court then must consider, under a “totality of circumstances” analysis, whether a current condition of vote dilution exists. After considering the evidence at trial as well as the legal memoranda submitted by the parties and arguments of counsel, the Court, having filed its written Decision on November 1, 1994, now hereby makes the following findings of fact and conclusions of law. FINDINGS OF FACT 1. El Centro Elementary School District is located in Imperial County. Since approximately 1906, it has employed an at-large electoral system to select its Board of Trustees, which now has five members. Every voter has the opportunity to cast a ballot for all five Trustee seats. There are no districts. This system is authorized by state law. Cal. Educ.Code §§ 35012, 5019, 5020,5030. 2. Elections are held every two years in odd-numbered years and staggered so that either two or three seats are elected every two years. The two or three candidates receiving the two or three highest vote totals of all candidates running are elected to four year terms of office. There is no majority vote requirement and no runoff election. 3. Voters may cast two or three ballots in each election for the two or three seats up for election, but there is no “anti-single shot” requirement that a voter must cast as many ballots as seats up for election in order for his or her vote to count. They may “bullet vote,” or east just one ballot if they choose to do so. Thus, Hispanics may vote once for a sole Hispanic candidate and deny Anglo candidates the other vote, thereby improving the chances of electing an Hispanic candidate. 9/22/94 Trial Transcript (“Tr.”), Dr. Rabinovitz, p. 16; 10/12/94 Tr., Dr. Brischet-to, p. 85. 4.The Hispanic population of El Centro has increased substantially since 1970. The 1990 Census indicated that the City of El Centro had a total population of 31,384 people, 65% of whom were Hispanic. Its demographic composition was as follows: 28.3% Anglo 3.8% Black 2.0% Asian/PI 0.6% Other Race 65.3% Hispanies 100.0% Total Ex. EU. The boundaries of the El Centro Elementary School District (“ECESD”) are larger than but include the entire City. The Elementary School District has a total population of 32,811, of whom 64.9% are Hispanic, 28.7% are Anglo, 3.7% are Black and 3.5% other. Ex. EV. The boundaries of the Central Union High School District include all of the City, all of the Elementary School District and other territory as well. The City, Elementary District and High School District all employ the same type of five member, staggered term, plurality win, at-large election system that permits bullet voting. 5. The Hispanic population is disproportionately under the age of 18 and disproportionately noncitizen. Hispanies were 59% of the voting age population in El Centro School District in 1990 and 47% of its voting age citizens. By 1994, Hispanies had become a majority of the voting age citizens in El Centro Elementary School District. 6. In 1975 Raul Aragon was the first Hispanic candidate to be elected to the El Centro School District in a contested election. He was elected to a seat common to the El Centro Elementary School District and the Central Union High School District. In 1979 the two Boards separated. In 1983 the El Centro School District Board was expanded from three seats to the current five Board seats. 7. Since 1983, there have been 17 Trustee seats elected on the Elementary Board. Seven were filled by minorities, three by Hispanies and four by Blacks. The Elementary School Board election history since 1983 is as follows: a. In 1983 two (2) Hispanies — Rae Perez and Rebecca Montiel — were elected to the ECSD’s Board. The 1983 election was basically an uncontested election. Montiel lost in 1985 and Perez lost in 1987 when they ran for re-election in contested elections. (Admitted Fact #6 — Pretrial Order of 1/11/93, p. 3). b. Since 1983 through 1991 there have been eight (8) Hispanic sumamed candidates who have run for the ÉCSD Board in contested elections in 1985, 1987, 1989, 1990, 1991. None of the Hispanic surnamed candidates won in any of those elections. (Admitted Fact # 7 — Pretrial Order of 1/11/93, p. 3). c. In 1993, three (3) Hispanic sumamed candidates ran for the ECSD Board in a contested election. One of the Hispanic sur-named candidates won one of the two seats. d. In the ESCD’s 87 year history, only 2 Mexican American candidates have ever won in a racially contested election: Mr. Raul Aragon in 1975 and Mr. Efrain Silva in 1993. Exs. EA P- I» and FB. Thirteen Hispanics have run for the El Centro Elementary School District (“ECESD”) Board since it became a separate district. Two were elected in 1983, in basically uncontested elections. Efrain Silva won in 1993. 8. Hispanics and Blacks also have been elected to the City Council and High School District Board of Trustees. Since the 1960’s, twelve Hispanics and five Blacks have been elected to the City Council. Four Hispanics but no Blacks have been elected to the High School District Board. Minorities elected to the City Council and High School District Board are as follows: Exs. ER; FB. 9. Plaintiffs have created a single member district in which Hispanics were 58% of all registrants there as of 1990. Ex. EZ. Although Hispanics could not have formed the required single member district for the 1985 and 1987 elections, there is no dispute that Hispanics satisfy the first Thornburg precondition currently or that Hispanics are a majority of the population, voting age population, and voting age citizen population in plaintiffs’ proposed single member district, both currently and at least since 1985. 10. Hispanics in El Centro generally, though not always, are politically cohesive in support of Hispanic candidates and there is no dispute that plaintiffs satisfy the second Thornburg precondition. Hispanics in El Centro are predominantly of Mexican descent. 11. Anglos in El Centro generally, though not always, are politically cohesive in support of Anglo candidates. Blacks too are politically cohesive in support of Black candidates. Thus, voting in El Centro elections is racially polarized in the sense that racial and ethnic groups generally vote differently from each other, each typically preferring candidates from their own group to candidates from the other two groups. 12. White bloc voting, however, did not defeat the preferred candidate of Hispanic voters in the elections of 1989, 1990, 1991 and 1993. Only in 1985 and 1987 did Anglo bloc voting defeat the preferred candidate of Hispanic voters. In those two elections, however, there was no vote dilution because Hispanics were not able to create a single member district in which they had the ability to elect. Hispanic population, registration and turnout has been increasing rapidly. Indeed, Hispanics are now a majority of the voting age citizens in El Centro, and currently have the abilify to elect at-large. 9/21/94 Tr., Dr. Morrison, p. Ill; 9/22/94 Tr., Dr. Rabinovitz, pp. 49, 59; 9/20/94 Tr., Dr. Klein, pp. 102-103. Plaintiffs therefore did not satisfy the third Thornburg precondition nor demonstrate a current condition of vote dilution or inability to elect at-large. 13. Group voting preferences were determined primarily through bivariate ecological regression analyses, multivariate ecological regression analyses, and an exit poll for the November, 1993 elections. All experts for both sides agreed that these statistical methodologies were not precise and provided only “estimates” of each group’s voting behavior. 14. Plaintiffs initially relied on a bivariate ecological regression analysis to estimate Hispanic and nonHispanic voting behavior, but this method of analysis sometimes resulted in Hispanics preferring Black candidates who were elected over Hispanic candidates who were not elected. See, e.g., Ex. 53, p. 66 (Hispanics prefer Newton to Hernandez in 1990), p. 74 (Newton first preference in 1991); Ex. EA, pp. 4-5 (Hispanic second preference in 1993 was Rutledge). Plaintiffs’ experts then concluded that the bivariate analyses produced biased results. 1/22/93 Tr., Burton, p. 22. Ecological regression makes a key assumption that a group’s voting preferences are unrelated to where they live, that Hispanic voters will support Hispanic candidates at the same level, whether Hispanics are 10% of the voters in a precinct or 60%. Plaintiffs’ experts labeled this assumption “minimal correlated error” (any variation is not systematically related to race); defense experts labeled it the “constancy assumption.” 1/22/93 Tr., Dr. Burton, pp. 29, 85, 95; 9/16/94 Tr., Dr. Klein, pp. 86-88. The bivariate analyses which combine Blacks and Anglos into a single “nonHispanic” category assumes that Blacks and Anglos vote alike. Blacks, however, often voted differently than Anglos, violating the principal assumption of ecological regression and producing biased estimates. 9/16/94 Tr., Dr. Klein, pp. 176-177; 9/20/94 Tr., Dr. Klein, pp. 21-29; 2/2/93 Tr., Dr. Loewen, pp. 183-189; 2/3/93 Tr., Dr. Loewen, pp. 81-83. 15. Plaintiffs’ experts then developed a multivariate analysis that divided the electorate into three groups: (1) Hispanics, (2) Blacks and (3) Anglos and all others (Asians, Native Americans — everyone not Hispanic or Black). Plaintiffs’ experts regarded multivariate analysis as more accurate than bivariate analysis for El Centro elections. Defense expert Dr. Klein also agreed that, if one relies on ecological regression, multivariate is better than bivariate. In both the bivariate and multivariate analyses, Hispanic voters were identified by comparing registration and voting lists to the U.S. Census Bureau’s Index of Spanish surnames. In the multivariate analyses, nonHispanics were divided into Blacks and Anglos based on then’ respective share of the voting age population in each precinct based on the 1990 Census. 9/16/94 Tr., Dr. Klein, pp. 135-136. 16. Plaintiffs’ expert Dr. Burton produced two tables of primary importance for each election studied from the ecological regression analyses. Ex. 53, pp. 38-79. One table (the “f ’ series) estimates the percentage of all Hispanics (or Anglos or Blacks) who signed in at the polls who voted for a particular candidate. This table includes those who signed in at the polls and may have voted in other elections on the ballot but did not vote (“rolled off’) in the Elementary School Board election. The other table (the “g” series) estimates the percentage of all votes actually cast by a group that were voted for a particular candidate, thus eliminating rolloff. Plaintiffs’ multivariate analy-ses did not consider absentee ballots and for the elections in 1985 through 1991 used only the 19 precincts common to both the City and the Elementary School District (thus excluding three precincts outside the City boundaries in which only some of the voters were eligible to vote in the ECESD elections). Ecological regression analyses were done for City Council and High School District elections as well as for Elementary District elections, using only the 19 common precincts. Ex. El. Dr. Brischetto’s exit poll for the November, 1993 election also used these 19 precincts in analyzing School Board and City Council elections for that year. 17.Plaintiffs’ experts also presented some secondary statistical evidence that they did not emphasize, namely homogenous precincts, overlapping percentages, complementary percentages and correlation analysis. This evidence does not lead to different conclusions about group voting preferences. The precinct and percentages analysis are mainly useful for undisputed Anglo voting patterns as there are no homogenous Hispanic precincts. The correlation statistics were produced only for the Hispanic-nonHis-panic bivariate relationship which plaintiffs’ experts testified was skewed. 2/2/93 Tr., Dr. Loewen, p. 183; 2/3/93 Tr. Dr. Loewen, pp. 81, 83. Plaintiffs’ expert Dr. Loewen also admitted that correlations are insufficient to establish racially polarized voting or voting preferences, and not the best way to prove racial bloc voting. 2/2/93 Tr., Dr. Loewen, pp. 20, 25, 175-178, 183; 2/3/93 Tr., Dr. Loewen, pp. 81-83. The 1985 and 1987 Elections 18.In 1985, two Anglos, John Anderson and Jim Hamilton, and one Hispanic, Rebecca Montiel, ran for the two seats up for election on the Elementary School Board. The two Anglos won: ■ 19.In the multivariate analysis, Hispanic voters preferred Montiel and Hamilton. 58.8% of all Hispanics who went to the polls cast a ballot for Montiel, 33.4% for Hamilton. Ex. 53, Table If, p. 45. The bivariate analysis generated the same preferences. Black voters preferred Hamilton (52%) and Montiel (40.7%). Anglo voters preferred Anderson (70.9%) and Hamilton (62.2%). Thus, Anglos elected both their preferences who were Anglo, but Hispanics did not elect their first preference Montiel who was Hispanic. White bloc voting defeated the preferred candidate of Hispanic voters. 20.In the 1987 election, three seats were up for election. Six candidates ran, four Anglos (Arnold, Gotti, Le Pere and Lester), one Black (Berryman) and one Hispanic (Perez). Arnold, Le Pere and Berryman won. Perez finished fourth, 190 votes behind Berryman: 21. Hispanics preferred Rae Perez, Charles Le Pere and Margaret Gotti. Blacks preferred Orchard Berryman, Rae Perez and Kathy Arnold. Anglos preferred Arnold, Le Pere and Berryman, all of whom won. Ex. 53, Table 2f, p. 53. Thus, white bloc voting defeated the first preference of Hispanic voters. 22. In these two elections, however, Hispanics were not an effective voting majority able to elect in plaintiffs’ single member Thornburg district. Hispanics were not a registered voter majority in 1985 (46.4%) and did not become a registration majority until 1987. Ex. EZ. Hispanics were only 36% of those turning out to vote in plaintiffs’ Thorn-burg district in 1985, 39% in 1987. Ex. EZ. Plaintiffs’ expert Dr. Loewen admitted that in 1985 the most Hispanic precinct was only 40% Hispanic in turnout in that election. 2/3/93 Tr., Dr. Loewen, pp. 78-79. Hispanics were not sufficiently large nor geographically concentrated to be able to elect their preferred candidate in these two elections. Thus, factually, there was no vote dilution in the 1985 and 1987 elections. These elections are useful only to the extent that they demonstrate Hispanic political cohesion and white bloc voting. They do not show vote dilution then, nor are they an accurate measure of the ability of Hispanics to elect at-large today, in view of the dynamic demographic change that has occurred in El Cen-tro since 1985. The 1989 Election 23.In the 1989 election, five candidates, including incumbents Anderson and Hamilton, ran for the two seats up for election. Four were Anglo, and one was Hispanic, Larry Osa. The two winning candidates were Lucille Duggan and Jim Hamilton, both Anglo: Hispanies did not prefer Osa but Duggan and Hamilton, both of whom won. Blacks preferred Hamilton and Duggan. Anglos, on the other hand, preferred Hamilton and Douthitt. Ex. 53, Table 3f, p. 61. Thus, Hispanies elected both their preferences, while Anglos elected but one of theirs. Had only Anglo votes been counted, Duggan would have lost. She won because of strong Hispanic support and mild Black support. Douthitt lost because he received no Hispanic support. Anderson lost because he received no Black or Hispanic support. As an Hispanic ran, this was a racially contested election in which the Hispanic first (and second) preference won. There was no vote dilution in this election. 24. Plaintiffs contend that special circumstances were present in this election that warrant disregarding it. They contend that Larry Osa, although Hispanic, was not Mexican-American but Spanish-Basque. Larry Osa, however, is Hispanic, has a Spanish surname, speaks fluent Spanish, held himself out at all times as Hispanic including in this election and always has self-identified himself as Hispanic on the Census and for employment and other purposes. 9/27/94 Tr., Osa, pp. 140-143. ■ When plaintiffs’ counsel asked him to sign a declaration stating that he was not Hispanic, he refused. Id. at 147-148. Social scientists agree that self identification is the accepted method for determining a person’s race or ethnicity, and is the method utilized by the Census. Other courts have not divided Hispanies into subgroups. Garza v. County of Los Angeles, 756 F.Supp. 1298, 1326-27 (C.D.Cal.1990). Plaintiffs did not subdivide Hispanies in their evidence. There is no evidentiary or legal basis to disregard this election. It represents an Hispanic success where Hispanic and Anglo preferences differed and voting was polarized, and in any event no vote dilution occurred. The 1990 Election 25. In 1990, a special election was held to fill a vacancy created by the absence of Orchard Berryman from school board meetings. Four candidates filed to run, two Anglos, one Black and one Hispanic. Both Anglos (Randall Carson and Hannah Gonzalez, an Anglo married to an Hispanic) withdrew from the race, although their names remained on the ballot. The Black candidate Dianna Newton, had lived in El Centro most of her life, been the Municipal Courts Administrator, and was politically active. Her husband was a well-known city fire battalion chief, and she ran a vigorous campaign, spending over $5,000. Refugio Hernandez, the Hispanic candidate, had moved to El Centro only two years earlier, spent no money, ran virtually no campaign, put up no signs or did any advertising, and did not appear at a candidate’s forum. 26.Newton, the Black candidate won the election handily, with 2,663 votes. Hernandez received but 753 votes. 27.Newton was the preference of Black voters (90%) and Anglo voters (47%). While Hernandez received more Hispanic votes (27.3%) than the other candidates, more His-panies “rolled off” or did not vote at all (31%) in this election than voted for Hernandez (27%). Ex. 53, Table 4f, p. 69. Thus, Hernandez cannot be regarded as a clear Hispanic preference, nor were Hispanics cohesive in their support of Hernandez. Hispanics, moreover, gave enough support to Newton (19%) that the combination of Black and Hispanic votes were sufficient to defeat Hernandez, even if no Anglo had voted. Ex. EM; 1/22/93 Tr., Dr. Burton, p. 179. Nor did Hernandez finish first in plaintiffs’ Thombwrg district. Ex. ES. With only one vote to east and no opportunity therefore to bullet vote as in two or three seat elections, this particular election within the Thornburg district was comparable to the conditions that would have occurred if the election had been held on a district basis. 28.Plaintiffs’ experts did not regard this election as an election at all, nor Hernandez as a viable or serious candidate. Plaintiffs’ expert Dr. Burton agreed that there was no viable Anglo candidate, that Hernandez was a “nonserious” candidate, and that Newton was the “only serious candidate.” He described the election as “barely an election” and agreed that “special circumstances” were present in this election. 1/21/93 Tr., Dr. Burton, pp. 134-138. Dr. Loewen agreed that it is not a fair test of racial bloc voting to use an election with a minority candidate who was not a viable candidate and then assert that whites are bloc voting against a minority candidate. He too agreed that special circumstances were present in this election. 2/2/93 Tr., Dr. Loewen, pp. 132-135. Thus, in this election, neither the second Thom- burg precondition (cohesion) was present, nor the third precondition, as there was no viable Hispanic candidate that could be regarded as a preferred candidate in any meaningful sense. Essentially, this was not a contested election, as Newton was the only serious candidate. There was no vote dilution in this election, which should be disregarded or given little weight under the totality of the circumstances. The 1991 Election 29. In the 1991 election, nine candidates ran for three Trustee seats. Three were Anglo (Kathy Arnold, Ronald Hull, Ronald Yates), one was Asian (Carl Whisler), one was Black (Dianna Newton) and four were Hispanic (Fernando Esperanza, Rae Perez, Larry Osa and Rose Mary Banks). Two Anglos and one Black won — Ron Hull, Kathy Arnold and Dianna Newton. Newton and Arnold were incumbents. Hull had been elected to the City Council twice and once had been mayor. 30. Hispanics preferred Esperanza, Perez and Osa, all of whom lost. Blacks preferred Newton, Perez and Banks. Anglos preferred Hull, Arnold and Newton, all of whom won. Ex. 58, Table 5f, p. 77. Newton defeated Perez, who finished fourth, by 160 votes. 31. Plaintiffs’ multivariate analysis revealed that the Anglo bloc vote did not defeat Perez. Anglos preferred Newton third (26.2%), giving her 200 votes more than they gave to Perez. Hispanics gave 239 more votes to Perez (47%) than they gave to Newton (23.7%). Thus, if only Hispanic and Anglo votes were counted, Perez would have won by 39 votes and Newton would have lost: Anglo Hispanic Black Newton 681 243 275 Perez 481 482 76 200 239 199 Ex. EM; 1/22/93 Tr. Dr. Burton, pp. 180-lSl. Thus, Anglo bloc voting did not defeat Perez. She was defeated by strong Black bullet voting for Newton coupled with a lack of cohesion by Hispanic voters who split their votes among more Hispanic candidates (four) than there were seats up for election (three), and for Newton. This can be seen by how close the Hispanic votes were for these candidates. Estimated Hispanic Candidate Votes Esperanza 586 Perez 482 Osa 319 Banks 250 Ex. EM. Dr. Burton conceded that Hispanics split their vote in 1991, and if they had not done so, effectively could have elected Esperanza or Perez, or possibly both. 1/22/93 Tr., Dr. Burton, pp. 147-153. This election is not an example of the third precondition or inability of Hispanics to elect at-large or vote dilution. It is an example of a lack of cohesion and special or totality of circumstances. The 1998 Election 32.In the November, 1993 election, there were six candidates for three seats — two Anglos (Cheryl Von Flue and Alan Luke), one Black (Sherman Rutledge) and three Hispanics (Efrain Silva, Frank Rodriguez and Jose Bustamante). Von Flue and Silva won: Von Flue (A) 1,948 Silva (H) 1,536 Luke (A) 1,346 Rodriguez (H) 1,168 Rutledge (B) 738 Bustamante (H) 665 Silva defeated Anglo third place finisher Luke by 190 votes. Ex. EA, p. 1. 33. Group voting estimates for the 1993 election are problematic and illustrative of the range of error inherent in the estimates generated by the statistical techniques. Dr. Burton’s multivariate ecological regression estimates for 1993 are as follow: PERCENT OF VOTERS VOTING FOR CANDIDATE Candidate Among Among Among (Elected = *) Whites Blacks Hispanics Luke (W) 49.7 21.2 -13.7 Von Flue* (W) 48.0 0.4 39.6 Rutledge (B) 10.7 79.8 14.1 Bustamante (H) 9.1 13.9 26.9 Rodriguez (H) 25.1 14.2 32.6 Silva* (H) 22.9 5.7 63.3 Totals 165.5 135.2 162.8 Estimated ballots cast (votes divided by voters [*100]) Total Possible 200 200 200 Turnout = 2446 236 1219 (total = 3,901) 34. Hispanics favored Silva and Von Flue, and Anglos preferred Luke and Von Flue, while Blacks preferred Rutledge. Ex. EA, p. 2. One of the estimates, however, is physically impossible, i.e., the -13% estimate of the percentage of Hispanics who cast a ballot for Anglo candidate Alan Luke. This physically impossible estimate means that the key assumption of ecological regression was violated. 9/16/94 Tr., Dr. Klein, p. 145. Because the total number of actual votes cast is fixed, the negative estimate for Luke necessarily means that ecological regression estimated too few votes for Luke and the estimates for other candidates are also wrong, i.e., too high. Id. at 153. Dr. Burton attempted to correct for this problem by setting the Hispanic estimate for Luke at 0 and reducing the estimates pro rata of all other candidates, a procedure he described as “norming.” Below are his revised estimates of the percent of each group s voters easting ballots for each candidate: Ex. EA, p. 3. However, defense expert Dr. Klein, testified that the norming procedure that Dr. Burton utilized was inappropriate, 9/16/94 Tr., Dr. Klein, pp. 145-155, 163-166; 9/20/94 Tr., Dr. Klein, pp. 37-38. 35. Plaintiffs also did an exit poll for this election, which produced the following estimates: Asian/ Anglo Black Other 11.9 16.0 22.8 53.7 27.2 40.0 23.0 15.2 19.4 19.9 79.4 13.3 25.5 24.8 39.9 39.4 4.2 26.0 173.4 166.8 161.4 682 61 65 Ex. EG, p. 1. 36. Dr. Brischetto’s exit poll generated estimates at odds with Dr. Burton’s multivariate analyses in several significant respects. Of greatest significance, the exit poll indicates that the Hispanic second preference is Rodriguez, not Von Flue: Exit Poll Ecological Regression Rodriguez 37 © CO Von Flue 25 CO Other important differences are Hispanic support for Luke at 8% in the exit poll and - 13% (or 0) in ecological regression, Black support for Von Flue at 27% in the exit poll and 0 in ecological regression, and Black support for Silva at 25% in the exit poll and 7% in ecological regression. The exit poll, however, predicted that Black candidate Sherman Rutledge received 112 votes more than he actually did, which is obviously wrong. The exit poll also predicted 49 fewer votes for Luke than he received, 43 more votes for Bustamante than he received and 90 total votes more than actually cast overall. Ex. 143. 37. Dr. Brischetto testified in his deposition several months before trial that, where the exit poll and ecological regression differ, he did not know which was right. Yet at trial, after the defense presented its opposition case, Dr. Brischetto changed his opinion and testified that he believed that the multivariate estimates for Rutledge were more rehable than his exit poll estimate, but his exit poll estimates were more rehable than Dr. Burton’s multivariate analyses for the Hispanic estimates for Von Flue, Rodriguez and Luke, and the Black estimate for Von Flue. Yet the actual votes cast for many candidates and overall is known and if the exit poll is wrong for Rutledge, other exit poh estimates for other candidates also have to be wrong and the same is true if the multivariate estimates are wrong. 9/29/94 Tr., Dr. Brischetto, pp. 38-62. On the other hand, Dr. Brischetto testified that, using his exit poll to test whether the assumption of ecological regression was violated in Dr. Burton’s 1993 ecological regression, he found little or no correlated error. 9/29/94 Tr., Dr. Brischetto, pp. 67-68; Ex. 138. Dr. Bris-chetto, in other words, both validated Dr. Burton’s ecological regression and yet claimed many of its estimates were wrong. Dr. Brischetto conceded that either there was correlated error in Dr. Burton’s 1993 multivariate regression analyses, or a -13% estimate can be created with little correlated error or the error was created by some other factor not known to him. 9/29/94 Tr., Dr. Brischetto, p. 88. 38. Neither the ecological regression nor the exit poll is necessarily more reliable. Both have advantages and disadvantages relative to the other. Exit polls are based on individual level data, not inferences about individuals from group precinct data as ecological regression is. On the other hand, exit polls are based on a sample and ecological regression is based on all voters except absentees. 9/24/94 Tr., Dr. Klein, pp. 48-51; 9/29/94 Tr., Dr. Brischetto, pp. 88-89. 39. Exit polls are particularly prone to high nonresponse rates which can seriously bias estimates and distort inferences, because people who do not respond may vote differently than those who do. 9/29/94 Tr., Dr. Brischetto, pp. 89-94, 107; 9/24/94 Tr., Dr. Klein, pp. 51. There were high nonres-ponse rates for Anglos (49%), Blacks (43%) and Asians (46%) in Dr. Brischetto’s exit poll, much higher than Hispanics (27%). Another problem with exit polls is that people may lie. 9/29/94 Tr., Dr. Brischetto, p. 103; 9/20/94 Tr., Dr. Klein, pp. 49. Response rates may be affected by race or ethnicity of interviewers. In Dr. Brischetto’s exit poll, 26 of 31 exit pollsters were Hispanic. Ex. DY. 40. Summarizing the results of these techniques, Silva was the Hispanic first preference in all analyses. Under bivariate analysis, the two Hispanic preferences were Silva and Rutledge. In the multivariate analysis, the two Hispanic preferences were Silva and Von Flue. In the exit poll, the two Hispanic preferences were Silva and Rodriguez. 41. Bivariate, multivariate and exit poll analyses all agree that Anglo and nonHispanic preferences were Von Flue and Luke, both Anglos. Silva was the fourth preference of Anglo voters among the six candidates. The Black preference was Rutledge. 42. Thus, Hispanics, according to multivariate analysis, elected both their preferences while Anglos elected but one of theirs. At worst, as reflected in the other two methods, Hispanics elected one of their preferences, as did Anglos. There was no vote dilution in this election, as Anglo bloc voting did not defeat the Hispanic first or perhaps even second preference. 43. Silva won despite Anglo bloc voting for Anglo candidates. As the fourth preference of Anglos and the fifth preference of Blacks, he would not have been elected if only Anglo and/or Black votes had been counted. Silva won even though he was the President of the local chapter of the Mexican-American Political Association (MAPA), and openly supported bilingual education, district elections and this lawsuit, which no doubt cost him some Anglo and/or Black support. Anglo bloc voting obviously was not sufficient to defeat Silva, the first preference of Hispanic voters. 44. Silva won despite three Hispanics vying for two seats and splitting the Hispanic vote among themselves. 9/20/94 Tr., Dr. Klein, pp. 63, 68, 70, 71; 10/12/94 Tr., Dr. Brischetto, pp. 88-90. Also, an unusual number of Hispanics (39%) bullet voted for Bus-tamante, Rodriguez or other candidates, even though there was no need for Hispanics to bullet vote, as there were as many Hispanics running as seats open. Silva’s margin of victory was depressed due to these factors. Klein, id. 45. What changed in 1993 was Hispanic voting behavior. Hispanics did not spread their votes as much as they did in 1991. 36% of all Hispanic votes were cast for Silva, compared to 18% for Rodriguez and 15% for Bustamante. Ex. EA, p. 5. In 1991, 25% of all Hispanic votes were cast for Esperanza, 20% for Perez, 14% for Osa, and 11% for Banks. Ex. 53, p. 78. The Hispanic vote was therefore much more concentrated behind Silva in 1993 than for Esperanza in 1991. There were more Hispanies voting who were more cohesive in favor of Silva and supported him by a wider margin over then-next preference than was true in 1991: 1991 1993 Esperanza 586 Silva 707 Perez 482 Von Flue 449 Osa 319 Rodriguez 336 46. Silva also was helped by an increase in the number and percentage of Hispanies among the total population, voting age citizens, registrants and voters in El Centro. By 1990, Hispanies had grown to 65% of the total population, and 59% of the voting age population (Exs. EU, EW). Also, 4,811 of the 12,451 Hispanies over the age of 18 were noncitizens. Ex. EX. Thus, in 1990 Hispanics and Anglos both were 46% of all voting age citizens in the City of El Centro. Ex. EU. For the elementary district, Hispanies were 47% of all voting age citizens, Anglos 44%. 9/21/94 Tr., Dr. Morrison, p. 99; Exs. EV, EW, EX. Since 1990, the Hispanic proportion of El Centro’s voting age citizens has increased. 9/21/94 Tr., Dr. Morrison, p. 107; Dr. Estrada, Ex. EY, p. 8. Hispanies under the age of 18 are mostly citizens. Thus, while Hispanies are 65% of the total population of El Centro, they are 76% of all people in El Centro under the age of 18. 9/21/94 Tr., Morrison, pp. 106-107. Each year since 1990, as Hispanies aged 14 to 17 aged into maturity, the Hispanic percentage of all voting age citizens rose and will continue to rise. Also, 4,000 Hispanic noncitizens applied to legalize their status under the Immigration Reform and Control Act (“IRCA”), some of whom have become naturalized citizens and many others soon will. 9/21/94 Tr., Morrison, pp. 107-108. This increase in Hispanic voting age citizens in El Centro also is reflected in registration increases. 9/21/94 Tr., Dr. Morrison, p. 109; Dr. Estrada, Ex. EY, p. 8. Plaintiffs’ demographer Dr. Estrada concurred with everything Dr. Morrison said regarding Hispanic age structure, naturalization, registration and the fact that the Hispanic share of voting age citizens is higher in 1994 than in 1990. 9/21/94 Tr., Dr. Morrison, pp. 109-111, 123-124; Dr. Estrada, Ex. EY. 47. In 1985, there were 2,776 Hispanic registrants out of 10,312, or 27% of the total. Ex. ET. By 1993, there were 4,928 Hispanic registrants out of 12,018, or 41% of the total. NonHispanic registrants declined in absolute numbers from 7,536 in 1985 to 7,090 in 1993. These increases in Hispanic registration continued in 1994, as Hispanies were 42.6% of all registrants by the June primary: NOTES: Implied Hispanic ethnicity derived as 1.059 x no. of Spanish-surnamed registrants. Plaintiffs show identical numbers of registrants in 1989 and 1990. SOURCES: 1985 through 1991: Plaintiffs’ “ELCT85.PRN,” and subsequent files, Bonilla 4/28/92 Declaration, Ex. C; 1993 and 1994: Imperial County Elections Dept, official voter files. Dr. Rabinovitz expects Hispanic registration to continue to “rise and rise rapidly.” 9/22/94 Tr., Dr. Rabinovitz, p. 39. 48. Turnout also has increased. There were only 571 Spanish surnamed voters in 1985 out of 3,118 total voters, or 18.3% of the total. By 1993, there were 1,226 Spanish surnamed voters out of 3,941, or 31.1% of all voters. That percentage rose to 31.7% by the June, 1994 primary. Ex. ET. 49. Defense demographic expert Dr. Morrison testified that it is his opinion that Hispanies are a majority of the voting age citizens in the El Centro Elementary School District now and that the Hispanic proportion of all voting age citizens will continue to increase, as young Hispanies continue to age into maturity and IRCA applicants naturalize soon. 9/21/94 Tr., Morrison, p. 111. The Court accepts this unrebutted testimony as a valid, legitimate, reasonable inference from the demographic evidence. Additionally, this dynamic demographic change that has occurred in El Centro since 1985 means that elections ten years ago in 1985 are not an accurate measure of the ability of Hispanies to elect at-large today or the ability of white bloc voting usually to defeat Hispanic preferred candidates today. 50. Exhibit 144 estimates what would happen if the 1993 candidates received the same level of support as they did in 1993 but applying those support rates to the number of Hispanic and non Hispanic voters in the voter base for each election from 1985 to 1991. Silva would lose in 1985 by 91 votes and in 1987 by 24 votes, but wins by 1 vote in 1989 and in each election thereafter by wider and wider margins, paralleling the growth of the Hispanic electorate. 51. Exhibits 145 and FT estimate who would have won in 1985-91 elections if the 1993 voter turnout base was used. The 190 vote difference between Perez and Berryman in 1987 disappears, and Anglo bloc voting is no longer sufficient to defeat Perez in that election. 10/12/94 Tr., Dr. Brischetto, pp. 64r-69, 74. 52. There is no longer any advantage to district elections for Hispanies, as they have as much if not greater likelihood of electing at-large as by district. 9/22/94 Tr., Dr. Rabi-novitz, p. 25. Dr. Rabinovitz, one of defendant’s experts, testified that El Centro employs the type of at-large election system most advantageous to minorities among the three most commonly used at-large election systems — pure at-large (El Centro), at-large with a residency subdistrict, and at-large with posts or places. Ex. FK. Pure at-large systems with staggered term, plurality win features that permit bullet voting, which El Centro uses, have proven in research to elect minorities significantly more frequently than other at-large systems. 9/22/94 Tr., Dr. Rabinovitz, pp. 12-24, 27; 107-109. The other types of at-large election systems create head to head contests where only one candidate can win and where bullet voting is not possible. Id. The same is true of district election systems where voters get but one vote in each district, there is no bullet voting and the most votes wins. Id., esp. 107-109. Indeed, under a district election system, if one Anglo and one Hispanic ran for one seat, and nonHispanics preferred the Anglo at the same level as Hispanics, supported the Hispanic candidate, a majority would be needed to prevail. In El Centro’s type of at-large election system, a minority can win by being second or third in a two or three seat election without being a majority of the voters or getting the most votes. Id. at 15, 27, 48. Staggering is advantageous because it enables a minority group to win one seat every two years, which the steady increase in Hispanic citizenship and registration makes “very likely.” Id. at 19, 42. 53. Plaintiffs’ expert Dr. Brischetto has done research indicating that, when the minority population exceeds 50% of the total population, at-large systems elect minorities more frequently than district election systems. In his research, at-large cities over 50% minority achieved near proportionality in minority representation on governing bodies, while cities that changed to district or mixed district and at-large election systems did not. The same was true if Hispanics were considered separately from other minorities. Dr. Brischetto testified that at-large systems are more dilutive in Anglo majority cities. 10/12/94 Tr., Dr. Brischetto, pp. 93-108; Exs. FL, FM, FN. El Centro was only 28% Anglo in 1990, a figure which has declined since then. 54. Plaintiffs have the ability to elect at-large, with as much or more ability than would exist with districts, as demonstrated by Silva’s victory, rising voting age citizen and registration levels, and by the fact that Hispanics have been elected frequently to the City Council and the High School District Board. 9/20/94 Tr., Dr. Klein, pp. 102-103; 9/22/94 Tr., Dr. Rabinovitz, p. 25, 49, 59. Hispanic successes in City Council and High School District elections mean that the more mixed record in Elementary District elections is not a structural issue but has to do with differences among the candidates and the quality of their campaigns. 9/22/94 Tr., Dr. Rabinovitz, pp. 31-32; 9/21/94 Tr., Dr. Klein, pp. 56-57. Silva, for example, “campaigned hardest” in 1993, was the “best qualified,” ran a “very thorough campaign” and “got out there and did the work.” Fernando Esperanza was not surprised Silva won. 9/21/94 Tr., Taylor, pp. 140-141; 9/27/94 Tr., Von Flue, pp. 158-159; 10/13/94 Tr., Esperanza, pp. 28, 57. Also, there are racial issues affecting the City as well as the school district. 2/4/93 Tr., McFadden, pp. 21-23. Thus, Dr. Rabinovitz does not believe that it is appropriate to ignore “exogenous” elections. 9/22/94 Tr., Dr. Rabinovitz, pp. 30-31. Dr. Burton and Dr. Loewen have utilized exogenous elections in other cases in which they have testified, Dr. Burton examined other elections in this case (Exs. 60; ED) and Dr. Loewen testified that he regards exogenous elections as relevant evidence. 1/22/93 Tr., Dr. Burton, pp. 109-110; 2/3/93 Tr., Dr. Loewen, pp. 104-112. Significantly, Dr. Loewen testified that this was only the second case of many he has examined where exogenous election results differed from the results in the jurisdiction under challenge, that Hispanic success in exogenous election decreases the assertion that Anglos are unwilling to vote for Hispanics, and is evidence against any thesis that the at-large system has a “chilling effect” on Hispanic efforts to elect. 2/3/93 Tr., Dr. Loewen, pp. 108, 112, 111. Special Circumstances in 1993 55. Plaintiffs contend that the November, 1993 election should be disregarded because of the special circumstances of this pending litigation. They claim that Anglo voters knew that, if they voted for and elected an Hispanic, this lawsuit might be dismissed. They claim that Anglo voters in fact did so, and that Silva’s victory was a fluke or an aberration. The evidence does not support this contention. 56. Silva was not an Anglo preference. He finished fourth out of six candidates among Anglo voters. Anglos preferred two Anglos, Von Flue and Luke. Silva would not have been elected if only Anglo votes were counted. He won despite Anglo bloc voting for Anglo candidates in an election Dr. Bris-chetto characterized as “clearly polarized.” 9/29/94 Tr., Brischetto, p. 13. Thus, there is no obvious indication of changed Anglo voting behavior such as Anglos making Silva their preference or even unusual Anglo support for Silva. 57. Only 22-25% of Anglo voters cast a ballot for Silva, depending on methodology. This Anglo support level for an Hispanic candidate is not unusual in El Centro elections for Elementary School District, City Council, or High School District elections, and less than the Hispanic support level for Yon Flue in this election (37%). 9/20/94. Tr., Dr. Klein, pp. 73, 75-77, 96; Ex. EJ. Indeed, these same Anglo voters gave greater support to the two Hispanic candidates in the City Council race the same day, 35% for Perez and 34% for Dhillon, even in the absence of a lawsuit. Id. at 73; 9/22/94 Tr. Dr. Rabinovitz, p. 57; Ex. ED. Anglo support rates for Hispanic candidates in past elections are as follows: ANGLO SUPPORT RATES FOR HISPANIC CANDIDATES Hispanic Candidate Name Date of Election Jurisdiction Anglo Support Rate Gonzalez 1987 City Council 61.8 Beltran 1984 City Council 58.9 Dhillon 1985 City Council 53.7 Terrazas 1991 City Council 50.5 Villanueva 1992 High School 43.0 Perez, G. 1993 City Council 34.8 Dhillon 1993 City Council 34.2 Aragon 1988 High School 30.9 Perez, A 1987 Elem. School 28.6 Rodriguez, F. 1993 Elem. School 26.5 Silva 1993 Elem. School 25.0 Montiel 1985 Elem. School 23.4 Osa 1991 Elem. School 21.5 Perez, A. 1991 Elem. School 18.5 Banks 1991 Elem. School 16.8 Osa 1989 Elem. School 13.4 Gonzalez 1985 City Council 13.1 Esperanza 1991 Elem. School 11.0 Bustamante 1993 Elem. School 10.0 Sesma 1993 City Council 9.7 Garcia 1987 City Council 6.2 Hernandez 1990 Elem. School 5.5 Rodriguez 1985 City Council 4.9 Alva 1990 High School 3.4 Marquez 1991 City Council 1.9 Beltran Ex. EJ. 1985 City Council -0.2 58. Plaintiffs, however, contend that Ang-los gave less support to Anglo candidates making it possible for an Hispanic to win. Yet in 1991, Anglos cast 52% of all their votes for Anglo candidates, and 48% for minority candidates. Ex. 53, p. 78. In 1993, Anglos cast 56% of their votes for Anglo candidates and 44% for minority candidates. In other words, Anglo support for Anglo candidates increased in 1993 over 1991. 9/20/94 Tr., Dr. Klein, pp. 79-80. Anglo support for Hispanic candidates in 1993 also increased: The net increase in Anglo support is 1.7%, which all experts agreed was statistically insignificant, given the margin of error in the estimates, and which Dr. Klein indicated would disappear altogether if proper norming procedures are utilized. 9/20/94 Tr., Dr. Klein, pp. 87-89; Ex. EK. Additionally, Hispanic support for Anglo candidates increased from 18% in 1991 to 22.7% in 1993. This 4.7% increase nearly offsets the 5.8% increase in Anglo support for Hispanic candidates, leaving a statistically insignificant 1.1% difference that also disappears with proper norming. 59. The reason why there were offsetting increases in Anglo support for both Anglo and Hispanic candidates in 1993 is a change in the composition of the candidates from 1991. There was an Asian (Whisler) and a strong Black candidate in 1991 (Newton). There was no Asian and a weak Black candidate (Rutledge) in 1993. As a result, Anglo votes for Black and Asian candidates were redistributed to both Anglo and Hispanic candidates: 60. Perhaps the best evidence that Ang-los did not change their voting behavior to win the lawsuit in the 1993 ECESD election is the outcome of the 1993 City Council election that took place on the same day. 9/20/94 Tr., Dr. Klein, pp. 73, 92-94. Dr. Brischet-to’s exit poll of the same voters indicates that Anglos gave more support to Anglo candidates in the School District and less support to Hispanic candidates than in the City Council race. 72% of Anglos cast one or more ballots for Anglo candidates in the City Council race, while 78% east one or more ballots for Anglo candidates in the School Board race. Ex. FI. Also, Anglos gave less support to their second preference in the City Council race (MaMn) than they gave to Luke. 9/20/94 Tr., Dr. Klein, pp. 93-95; Ex. ED. Thus, these same Anglo voters were more united in their support of Anglo candidates in the School Board race than in the City Council race, even though no lawsuit was pending against the City. 9/24/94 Tr., Dr. Brisehetto, pp. 16-22. Luke, moreover, did not lose because he lacked Anglo support. He received nearly as much support (43%) as Yon Flue (49%), who won by a considerable margin. According to Dr. Burton’s multivariate analysis, Von Flue received 1,211 Anglo votes, Luke 1,052. Luke lost because he received virtually no minority support. Ex. FJ. Substantial Anglo support alone was not sufficient to prevail. 61.Plaintiffs argue that Anglo support for nonHispanic candidates declined 5.6%, while Anglo support for Hispanic candidates increased by 5.8%, for a net difference of 11.4%: Ex. 128. As noted above, however, lower Anglo support for all nonHispanic candidates is due to a change in the composition of the candidates and actually masks increased support for Anglo candidates. Less Anglo support for Black and Asian candidates in 1993 helped Anglo candidates nearly as much as it helped Hispanic candidates (a net 1.7% difference, which is statistically insignificant), and the increased Anglo support for Hispanic candidates did not necessarily go to Silva, as Anglos preferred Rodriguez to Silva. Also, plaintiffs’ Exhibit 128 includes Anglo non-preferences, Black and Asian candidates as well as Anglo candidates and ignores offsetting increases in Hispanic support for Anglo candidates and the fact that Hispanics preferred Von Flue to two other Hispanic candidates and at a level (37%) exceeding Anglo support of Silva. A better comparison is Ex. EN which reflects the percentage of all votes cast by Anglo and Hispanic voters for their preferred candidates in 1991 and 1993: PERCENTAGE OF TOTAL VOTES CAST BY A GROUP FOR ITS PREFERRED CANDIDATES FOR THE SEATS AT STAKE Election White Voters Hispanic Voters 1991 60 59 1993 56 58 Difference 4 1 Dr. Klein testified that this 3% difference is statistically insignificant and would disappear entirely if his alternate norming procedure were utilized. 9/20/94 Tr., Dr. Klein, p. 92. 62.Another problem with Ex. 128 is that it is based on Dr. Burton’s normed multivariate regression analysis which Dr. Brisehetto states contains five erroneous estimates, and is derived from an analysis that contained a physically impossible — 13% estimate of Hispanic support for Luke that Dr. Klein testified was normed improperly. The exit poll data is no more reliable, as Dr. Brischetto’s exit poll predicted 112 more votes for Rutledge than he actually received, 90 more than actually cast and contained high nonresponse rates. In the City Council election Dr. Bris-chetto’s exit poll predicted Perez finished second and won, even though he actually lost. Ex. FU. The statistical evidence is not precise enough to prove any change in Anglo voting behavior, particularly given the deficiencies, mistakes and differences among the techniques. The Court is also skeptical of comparing election years with different candidates. Any difference, further, would not necessarily be due to Anglos voting for Silva to end the lawsuit. 63.Plaintiffs presented no evidence of unusual, organized Anglo financial support or endorsement of Silva or other Hispanic candidates, nor did Anglos campaign to elect an Hispanic nor was Silva the handpicked candidate of the Anglo power structure. Neither incumbent Trustees nor District officials, employees or agents endorsed, gave any money or assisted in the campaigns of any of the Hispanic candidates. 9/21/94 Tr., Taylor, pp. 141, 145-147; 9/27/94 Tr., Newton, pp. 12-14. There was no testimony or evidence that any Anglo voter voted for Silva to win the lawsuit. Fernando Esperanza, who testified for plaintiffs, testified that he never even considered voting for an Anglo in order to help plaintiffs win the lawsuit. 10/13/94 Tr., Esperanza, pp. 56-57. 64. Dr. Francine Rabinovitz, a political scientist at the University of Southern California, testified that voting for an Hispanic that Anglos otherwise would not support, just to win the lawsuit, involves a more complicated chain of reasoning than political scientists know to be true of most voters when they go to the polls — Anglos, many of whom, have no children in the elementary schools, would have to be aware of the lawsuit, understand it, and care about the outcome, be aware that an Hispanic success might enable the District to win the lawsuit, understand that the result if plaintiffs won the lawsuit would be district elections and believe that they somehow would be disadvantaged by district elections. Dr. Rabinovitz believes that the likelihood this occurred is very small. 9/22/94 Tr., Dr. Rabinovitz, pp. 51-56. Plaintiffs did not controvert this testimony. Indeed, Dr. Rabinovitz believed that Anglos would vote their normal self-interest even more strongly. Id. at 54-55. Dr. Klein, moreover, believed that Hispanics had the same incentive to “throw” the election as Anglos. Indeed, consistent with that hypothesis, Hispanics preferred Von Flue over Rodriguez and Bustamante, and gave her more support (37%) than Anglos gave Silva (25%). Dr. Klein, however, did not believe any evidence exists that either Anglos or Hispanics changed the way they voted because of the lawsuit. 9/20/94 Tr., Dr. Klein, p. 96. 65. This trial initially began in January 1993 but on February 9, 1993 was continued following El Centro’s Rule 52(c) motion for judgment when plaintiffs rested their case, eventually resuming September 14, 1994, because the Court wanted the benefit of another election before ruling on the case. Rising Hispanic population, registration and turnout, coupled with recent successes in City Council and High School District elections caused the Court to suspect that Hispanics already had the ability to elect at-large. On February 10, the Court’s ruling was reported in the local newspaper, the Imperial Valley Press, along with a statement by plaintiffs’ witness Francis Beope that the Court had denied ECESD’s motion, which ECESD believed was untrue. Ex. FC; 9/21/94 Tr., Taylor, pp. 142-143. 66. The School Board then issued a press release on the lawsuit, stating its interpretation of the Court’s action and disagreeing with Ms. Beope. The press release quoted School District legal counsel Frank Oswalt as follows: It was clear to me that Judge Brewster will dismiss this case if there is Híspame electoral success in next November’s election. On February 11, 1993, the Imperial Valley Press reported Mr. Oswalt’s statement. Ex. CY. Dr. Rabinovitz did not believe that voters were likely to remember Oswalt’s statement eight months later at the time of the November election. 9/22/94 Tr., Dr. Ra-binovitz, p. 50. Dr. Rabinovitz testified that: (1) fewer people read newspapers than watch TV, (2) people forget what they read, (3) politicians concentrate their advertising just before an election because people don’t focus on elections until then, and (4) the chain of reasoning to connect Oswalt’s statement in February to electoral behavior in November is too complicated to reflect what political scientists know to be typical of how voters behave on election day. 9/22/94 Tr., Dr. Rabinovitz, pp. 49-52, 54r-56. None of the incumbent Trustees nor the District’s legal counsel ever discussed the lawsuit in public after Oswalt’s statement. This was the only statement ever made by anyone in El Centro that the Court “will” dismiss the case, as opposed to “may.” Subsequent news stories all used the word “may.” 67. The candidate filing deadline for the November, 1993 election was August 6. Only two candidates filed — Von Flue and Rutledge, none of whom were Hispanic. A newspaper article reported that the absence of Hispanic candidates might “stymie” the Court’s objective to evaluate the impact of demographic change on the ability of Hispanics to elect at-large. Ex. 85-2. Thus, Hispanics arguably knew that by not running a candidate or by not voting for Hispanic candidates, their position in the lawsuit perhaps would be enhanced. Because no incumbents filed to run, the candidate filing deadline was extended to August 11. Four more candidates filed by the extended deadline, three of them Hispanic, indicating that publicity and the lawsuit had little effect on the November election. 68. Plaintiffs claim that publicity about the costs of the lawsuit and possible teacher layoffs and program cuts might have caused El Centro teachers who are predominantly Anglo to support Hispanic candidates. Yet no teachers were laid off in 1998 or 1994, and El Centro teachers knew there would be no layoffs by May of 1993, well before the election. The layoff notices given in March of 1998 were required by state law for following year layoffs, and were precautionary only because the District does not finalize its budget nor know how much state revenue it will receive until June each year. 9/21/94 Tr., Taylor, pp. 134-136. A1 Dempsey, the President of the union, the El Centro Elementary Teachers’ Association (“ECETA”), testified that the union understood that the notices were precautionary and did not take them seriously, telling members receiving the notices not to worry. 9/21/94 Tr., Dempsey, p. 189. Exhibit FD, a May 5, 1993 newspaper article, announced that the notice deadline for teacher layoffs had passed without any layoffs occurring. 69. Another special circumstance alleged by plaintiffs is that ECETA, which is predominantly Anglo, did not endorse any candidates in the 1993 election. The reason that there was no endorsement is that ECETA, like other California Teachers Association (“CTA”) affiliate unions in the state, were diverting their • energies to defeating the school voucher initiative that was on the November, 1993 ballot. 9/21/94 Tr., Dempsey, p. 182; 9/27/94 Tr., Waters, p. 128. Why the lack of an endorsement in the School Trustee election voting rights is of any voting rights consequence is unclear. Apparently, plaintiffs believe that the lack of an ECETA endorsement of Anglo candidates meant less Anglo support than in past elections because ECETA members are predominantly Anglo. Yet there is no reason to believe that ECE-TA would not have endorsed Silva — in fact, ECETA President A1 Dempsey voted for Silva, and was surprised that an Hispanic was not recommended by the ECETA Political Action Committee. 9/21/94 Tr., Dempsey, p. 199. The first candidate ECETA ever endorsed was an Hispanic, Raul Aragon in 1975. ECETA frequently supported Hispanic and Black candidates in past elections— Perez and Berryman, in 1987, Newton in 1990, and Newton and Perez in 1991. Nor do ECETA candidates necessarily win if they are endorsed. Thus, it would not have been unusual if ECETA had endorsed and supported Silva, nor would it have been unusual or disabling if ECETA had endorsed Anglo candidates. Additionally, Dempsey testified that it would be “stupid” for members to vote for someone just to end the lawsuit, as the union would have to deal with the new Trustee on many other issues for the next four years. 9/21/94 Tr., Dempsey, p. 188. He personally did not prefer Luke because Luke spoke of fiscal controls which generally means that he would be unfavorable to teacher salary demands, Von Flue because she was “management” in her current job, and Rutledge and Bustamante because they were not serious candidates. He thought Silva and Rodriguez would be more sympathetic on teacher salary issues. 70. Still another special circumstance alleged by plaintiffs is the lack of incumbents in the November, 1993 election. Anglo incumbents Lucille Duggan and Jim Hamilton did not run for reelection. Plaintiffs contend that this made it easier for an Hispanic to win. Yet other elections have included open seats — 1990 (one seat) and 1991 (one seat). The lack of an incumbent, moreover, provides a truer test of the relative voting strength of Hispanies, Anglos and Blacks, without the special advantage of incumbency. 9/22/94 Tr., Dr. Rabinovitz, pp. 56-57. Nor was there any evidence that Duggan or Hamilton were urged or pressured not to run to make it easier for an Hispanic to get elected. Hamilton had served two terms, and Duggan was 70 years old. Taylor, pp. 148-149. Both had advised people that they would not run for reelection in November, 1993 before this Court’s February 9,1993 ruling. 9/27/94 Tr., Newton, p. 15. 71. On the basis of the totality of the evidence, there are no special circumstances that warrant disregarding the Hispanic electoral success in the 1993 election. Not one witness expressed the opinion that any votes were switched because of the lawsuit. The Court does not believe that the newspaper article or the lawsuit affected the outcome of the election, nor did the objective or statistical evidence prove that. The 1993 election, together with recent Hispanic successes in City Council and High School District elections and rapidly increasing Hispanic population, registration and turnout, convince the Court that Hispanics are able to elect at-large currently and that white bloc voting does not usually defeat Hispanic preferred candidates at the present time. Summary of Elections 72. To summarize the elections above, there was no vote dilution in any of the elections analyzed: 1993 — 3d Thornburg precondition not present; Hispanic candidate wins. 1991 — 2d and 3d Thornburg preconditions not present (Anglo bloc voting not sufficient to defeat Hispanic candidate who lost due to lack of cohesion); also disregard under totality of circumstances because of Hispanic vote splitting. 1990 — 2d and 3