The events relevant to this appeal began in 1987 when the Bibb County Board of Education (the Board) began planning to call for a school bond referendum to provide funds to build a new high school and middle school and to air condition existing school buildings. The facts and procedural history of this case are set forth more fully in our vacated opinion at 908 F.2d 851, 852-54 (11th Cir.1990).
The Board considered submitting the issues as separate questions to be voted on individually. It also considered holding the referendum on Super Tuesday, March 8, 1988, the date of the presidential primary in which Jesse Jackson was a candidate. Ultimately, however, the Board decided to submit the bond issue to the electorate as a single question during the November 8, 1988, general election.
Plaintiffs, representing black citizens, have challenged the Board’s actions as a violation of the First, Thirteenth, Fourteenth, and Fifteenth Amendments of the Constitution and Section 2 of the Voting Rights Act. Plaintiffs’ Section 2 claim has previously been before this Court when we *551held that the Board’s actions did constitute a standard, practice, or procedure under Section 2 of the Voting Rights Act. Lucas v. Townsend, 908 F.2d 851 (11th Cir.1990). The United States Supreme Court vacated our opinion, however, and remanded for a determination of whether appellate jurisdiction was proper in light of possible outstanding constitutional claims that had not been decided by the district court. Board of Public Education and Orphanage for Bibb County v. Lucas, — U.S.-, 111 S.Ct. 2845, 115 L.Ed.2d 1013 (1991).
Upon being advised by the district court that plaintiffs’ constitutional claims remained to be decided, we remanded to the district court for resolution of those claims. Lucas v. Townsend, 943 F.2d 38 (11th Cir.1991). The district court held an evidentia-ry hearing and entered written findings of fact and conclusions of law on January 23, 1992. The court rested on its prior rejection of plaintiffs’ Section 2 claims, and further determined that plaintiffs failed to prove their claims of constitutional violations. Lucas v. Townsend, 783 F.Supp. 605, 619 (M.D.Ga.1992).
Based upon the conclusion that the district court’s findings of fact control the outcome of this case and are not clearly erroneous, we affirm the judgment for the defendants.
To prevail on their claims of violations of the Fifteenth Amendment and the Equal Protection Clause of the Fourteenth Amendment, plaintiffs had to prove first that vote dilution, as a special form of discriminatory effect, exists and second, that it results from a racially discriminatory purpose chargeable to the state. See Washington v. Finlay, 664 F.2d 913, 919 (4th Cir.1981), cert, denied 457 U.S. 1120, 102 S.Ct. 2933, 73 L.Ed.2d 1333 (1982).
After examining the evidence, briefs, and arguments of the parties, the district court found that plaintiffs
failed to show any such adverse or- discriminatory effects with respect to their claim that racial polarization infects voting on referenda in Bibb County. They have not pointed to any other evidence which they contend shows any adverse impact against blacks with respect to the timing or content of the November 8 bond election. Neither have they offered evidence that any person who wanted to vote for or against the referendum was denied or inhibited in any way in casting his ballot.
Lucas v. Townsend, 783 F.Supp. 605, 619 (M.D.Ga.1992). We review the findings of the district court only for clear error, Thornburg v. Gingles, 478 U.S. 30, 78-79, 106 S.Ct. 2752, 2781, 92 L.Ed.2d 25 (1986), affording special deference to the district court due to its “special vantage point” and ability to conduct an “intensely local appraisal of the design and impact of” a voting system. White v. Regester, 412 U.S. 755, 769, 93 S.Ct. 2332, 2341, 37 L.Ed.2d 314 (1973).
The district court carefully analyzed the evidence concerning polarized voting and concluded that plaintiffs failed to meet the standard enunciated by their expert. This failure was fatal to their claim of discriminatory effect. Absent a showing of racially polarized voting, the challenged practice does not affect minority voting rights and cannot cause a discriminatory result. See Gingles, 478 U.S. at 48-49 n. 15, 50-51, 106 S.Ct. at 2765 n. 15, 2766 (racially polarized voting is one of the most important factors a Section 2 plaintiff must establish).
The court considered the testimony of Dr. Michael Binford, who testified for plaintiffs in an effort to show that voting in Bibb County is racially polarized. Dr. Binford set forth three criteria necessary to show racially polarized voting: (1) a strong correlation coefficient between voting behavior and the racial composition of the-electorate, (2) a majority of white voters must vote in one direction or for one candidate while a majority of black voters vote in the other direction, and (3) it must be difficult or impossible for the minority population to win an election in an at-large contest.
The district court accepted Dr. Binford’s expert conclusion that racially polarized voting existed in Bibb County candidate *552elections. The court ruled that evidence of racially polarized voting in candidate elections, however, does not demonstrate racially polarized voting in non-candidate elections such as the bond referendum. The district court considered Dr. Binford’s affidavit stating that while the correlation between the racial composition of a precinct and the vote for black and white candidates is very clear and strong, candidate and referenda elections are not readily interchangeable. Dr. Binford stated that “[v]oting in referenda is often a more complex process than voting for candidates in office. The issues are usually more complex and policy positions of the electorate may be influenced by a variety of factors.” Indeed, Dr. Binford conceded that in analyzing the 1988 school bond referendum, “it is impossible to ascertain the impact of a number of additional factors, such as having school age children in a voter’s family, on the voting behavior of a precinct.”
Plaintiffs assert that the district court erred in refusing to equate candidate and referenda elections for purposes of determining whether plaintiffs had made a showing of racially polarized voting. Plaintiffs contend that this requirement was satisfied through their showing of polarized candidate elections and testimony indicating that race is a factor in Bibb County elections. We reject this argument. Drawing valid inferences and conclusions about voter behavior from statistics requires the selection of relevant elections for analysis. As Dr. Binford noted, candidate and referenda elections are not readily interchangeable since factors other than race can weigh differently in the two types of elections. Thus, it was appropriate for the district court to analyze only referenda elections in determining whether plaintiffs had proved racially polarized voting.
The district court applied Dr. Binsford’s three criteria and concluded that plaintiffs failed to prove racially polarized voting in Bibb County referenda generally or in the November 1988 referendum.
Dr. Binford analyzed a total of 30 candidate races and two referenda. He calculated a correlation coefficient, represented as “R,” to determine the relationship between the result for a particular candidate or proposition and the racial composition of the voters who turned out for the election. R is then squared to yield “R2,” which is a measure of how much of the variation in the vote can be statistically accounted for by the racial composition of the electorate. Dr. Binford testified that a correlation coefficient of .7 or greater would be considered “strong.” Any correlation coefficient of less than .7 would not be strong because over 51% of the variance between white and black voting would not be attributable to the race of the voters.
In his initial affidavit, Dr. Binford analyzed two referenda (the 1988 school bond issue and the 1988 sales tax proposal) and 13 candidate elections. The racial composition of the precincts explained only 44% of the variance in voting on the school bond referendum and 46% of the variance on the sales tax referendum. In contrast, in every candidate election analyzed, at least 84% of the variance was explained by race. In 10 of those candidate elections more than 95% of the variance in precinct voting was explained by the racial makeup of the precinct.
In his second affidavit, Dr. Binford examined an additional 17 candidate elections. The R2 values for each of the 15 candidate elections for which he obtained statistically significant results was higher than the R2 for the 1988 school bond referendum, indicating that the race of the voters accounted for more of the variance in those elections than in the 1988 referendum. In addition, although the variance explained by the race of the voters exceeded 90% for 20 of the 30 candidate elections Dr. Binford studied, it never exceeded 90% for any of the 31 refer-enda held in Bibb County between 1970 and 1989. Indeed, a strong correlation coefficient was present in only 5 of 31 Bibb County referenda. The parties have not indicated whether there was a racial context to any of these referenda. In addition, the correlation coefficient did not reach .7 in the school bond referendum held on November 8.
*553Evidence also was presented to show that some white citizens were in favor of the 1988 school bond issue and some were opposed, and that some black citizens were in favor of the bond issue and some were opposed. This evidence included the testimony of defendants' witnesse~ and other evidence in the record indicating that there was no cohesive racial position on the referendum. For example, the recommendation to the Support Services Committee of the Board, from Dr. Thomas Haggler, the Bibb County Superintendent of Schools, stated:
Black leaders in the community have been active and outspoken in their support of the combined bond issue. Mr. William P. Randall serves as a co-chairman of the steering committee. Mr. Harris Walker has published an editorial in the Macon Courier supporting the bond issue. Many of our black principals, teachers and staff have worked long and hard for the bond issue. On the other hand, there are other persons who happen to be black, such as Mr. Brown, who opposed the fifth high school and have worked against the bond issue.
The district court further concluded that the second and third criteria set forth by Dr. Binford were not satisfied. The second criteria requires a showing that a majority of white voters vote in one direction while a majority of black voters vote in the other direction. The court found that plaintiffs had not made this showing because evidence introduced by the defendants demonstrated that in nearly all Bibb County re-ferenda a plurality of black and white voters voted the same way. The third criteria requires a showing that it is difficult or impossible for the minority population to win an election in an at-large contest. The court found that plaintiffs failed to make this showing because evidence tended to show that it would not have been impossible for black voters to change the outcome of the bond referendum:
Six hundred nineteen additional voters would have defeated the subject bond issue. Ten times that number of voters turned out but did not vote on the school bonds. 70.9% of registered white voters turned out and 60.5% of registered black voters turned out. A greater number of black voters could have turned out and/or voted in opposition to the bond issue. It was not difficult and certainly not impossible for the minority-assuming the minority was opposed-to have prevailed in the subject bond election.
Lucas v. Townsend, 783 F.Supp. 605, 618 (M.D.Ga.1992).
Thus, the trial court found that the evidence did not demonstrate racially polarized voting behavior in Bibb County in re-ferenda in general or in the referendum at issue. In light of the evidence supporting the district court's finding, we cannot conclude that the finding was clearly erroneous.
Although polarized voting is an essential element of a showing of discriminatory effect, plaintiffs must further prove that the totality of the circumstances indicate a discriminatory effect. See Gingles, 478 U.S. at 46, 106 S.Ct. at 2764; Zimmer v. McKeithen, 485 F.2d 1297, 1305-07 (5th Cir.1973) (en bane) aff'd sub nom. East Carroll Parish School Board v. Marshall, 424 U.S. 636, 96 S.Ct. 1083, 47 L.Ed.2d 296 (1976). The following evidence supports the district court's finding that the totality of the circumstances did not demonstrate actual discriminatory effect on minority voters in Bibb County:
1. In setting the date for the referendum, there was no evidence that the Board tried to manipulate the white or black vote.
2. During its consideration of the school bond referendum, the Board received evidence that the proposed new school was projected to have a 53% black student body and that certain black leaders in the community supported the combined bond issue.
3. During its consideration of the referendum, the Board received evidence that the purpose of the combined bond referendum was to garner enough votes to pass a package which would benefit the entire school system.
4. Four of the ten-member Board were black.
*5545. Defendants have operated the schools of Bibb County under the supervision of the court since 1964. The consent decree currently in effect was entered into in 1978. With the exception of the question of whether to rebuild an elementary school, there has never been any allegation or contention by the parties to the earlier desegregation case or this case that defendants have operated the Bibb County school system since 1978 in a discriminatory manner. See Adams v. Board of Public Educ., 585 F.Supp. 215 (M.D.Ga.1984), aff'd 770 F.2d 1562 (11th Cir.1985).
The Supreme Court has stated that “the clearly erroneous standard of review is a deferential one, [which requires] that ‘[i]f the district court’s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently.’ ” Amadeo v. Zant, 486 U.S. 214, 223, 108 S.Ct. 1771, 1777, 100 L.Ed.2d 249 (1988) (quoting Anderson v. Bessemer City, 470 U.S. 564, 573-74, 105 S.Ct. 1504, 1511, 84 L.Ed.2d 518 (1985)). The trial court’s finding with respect to plaintiffs’ failure to show discriminatory effect is affirmed.
To establish a constitutional violation, plaintiffs were also required to prove that the purpose of the challenged activity was discriminatory. See e.g., Washington v. Davis, 426 U.S. 229, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976); City of Mobile v. Bolden, 446 U.S. 55, 100 S.Ct. 1490, 64 L.Ed.2d 47 (1980). Discriminatory purpose implies that the decisionmaker chose a particular course of action because of its adverse effects upon a minority group, not merely in spite of its effects upon the minority. See Personnel Administrator of Massachusetts v. Feeney, 442 U.S. 256, 279, 99 S.Ct. 2282, 2296, 60 L.Ed.2d 870 (1979). The district court found that plaintiffs failed to prove that the combining of the issues and the timing of the referendum were motivated by discriminatory intent. Like the court’s finding of discriminatory effect, this finding is reviewable only for clear error, Gingles, 478 U.S. at 78-79, 106 S.Ct. at 2781, and is entitled to special deference, White v. Regester, 412 U.S. 755, 769, 93 S.Ct. 2332, 2341, 37 L.Ed.2d 314 (1973).
Plaintiffs assert that the district court’s finding regarding the timing of the referendum was erroneous because defendants attempted to dilute the black vote by manipulating the referendum date so that it would not be held on Super Tuesday when Jesse Jackson would be on the ballot. In light of the overall evidence that was before the district court, however, we cannot say that the court’s finding was clearly erroneous.
The sequence of events leading up to the challenged decision is relevant to an inquiry into the decisionmaker’s purposes. Village of Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 267, 97 S.Ct. 555, 564, 50 L.Ed.2d 450 (1977). The district court carefully examined the sequence of events leading up to the decision to propose the bond referendum on the general election day rather than on Super Tuesday, and found no evidence of any departure from the procedural sequence normally followed in calling for a bond election, and no substantive departure in terms of the factors usually considered important by the decisionmakers strongly favoring a decision contrary to the one reached.
The court found that questions of voter turnout were irrelevant to the Board’s decision not to hold the referendum on Super Tuesday based on the following evidence:
(1) During the December 17, 1987, meeting at which the Board first considered holding the referendum on Super Tuesday, no discussion was held concerning the Super Tuesday candidates or the racial makeup of expected voter turnout on Super Tuesday.
(2) There was no discussion at the January 1988 Board meeting concerning voter turnout or the race of the candidates on the Super Tuesday ballot.
(3) There was testimony and other evidence which indicated that the decision not to hold an election on Super Tuesday *555was motivated solely by the desire to avoid a conflict with another referendum which had been previously scheduled for that day. A date other than Super Tuesday was recommended because members of the Board knew that the county commissioners, who have sole authority to call the bond election and select the election date, were opposed to the addition of the school bond referendum to the Super Tuesday ballot. The opposition of the commissioners stemmed from the fact that the Super Tuesday ballot already included a referendum on a 1% increase in Bibb County sales tax to pay for county construction projects.
The district court’s finding that the Board’s decision to combine the issues into a single referendum was not motivated by a discriminatory purpose is similarly supported by the evidence. The record indicates that a majority of the Board favored and voted for a package bond issue because in their opinion it would benefit the entire school system and not just particular portions thereof. For example, the district court quoted a letter to the Support Services Commission of the Board which reads:
It is the strong recommendation of the Administration that only one question be presented to the voters on the entire package. The whole point of this entire program has been to do something for the entire system and not just one part of the county or one group which happens to be for or against parts of the package. Balkanizing the issue and presenting multiple questions to the voters for each project would encourage a “ward healer” type of mentality and lead to the sure defeat of each question.
There are some voters who are going to be against any bond issue regardless of its purpose because they simply are not going to vote for a tax increase. There are those in the community who would be directly affected by the air conditioning of a particular building but who might not anticipate immediate benefit from a new high school since they live on the other side of town. Likewise, there are those who live in the western part of the county and would attend the new school who could care less about air conditioning at Central or Southwest.
In order to overcome all of this and to structure a program which has the best chance of passing and for doing something for all of Bibb County, the Administration believes that we should keep the package together and try to get as many persons as possible who have an interest in any part of it to support it.
Lucas v. Townsend, 783 F.Supp. 605, 614 (M.D.Ga.1992).
The court also had before it the testimony of then President of the Board, Judy Townsend, who stated that she supported the combination of the issues because:
[s]o many times the Board of Education has been criticized for doing things for some schools or some parts of the community and not for others and when we were talking about all the issues here, as a package, this would practically affect every community in Bibb County and this was the opportunity for us to be able to reach and do something for everyone across the county.
[Everything we do, whether we serve in a specific post or in my case of countywide, everything we do affects every teacher and every student in this county and we have to act accordingly.
Based on this and other similar evidence, the district court concluded that the issue was submitted to the voters as a package with the hope that the package would attract enough favorable white and black votes to cause the entire twenty-nine million dollar issue to pass for the benefit of the entire school system and all of its students.
Thus, the district court concluded that plaintiffs had failed to prove their constitutional claims, stating: “In requesting the Bibb County Commissioners to call a bond referendum for November 8,1988, and submit a package bond issue to the voters of Bibb County, defendants acted without any discriminatory intent or purpose.” Although plaintiffs have pointed out evidence that may suggest a different result, the *556district court's finding, in light of the record viewed in its entirety, is not clearly erroneous. See Amadeo, 486 U.S. at 223, 108 S.Ct. at 1777; Fowler v. Blue Bell, Inc., 737 F.2d 1007, 1013 (11th Cir.1984).
The district court had previously rejected plaintiffs' Section 2 claim in its order of June 7, 1989. Lucas v. Townsend, 714 F.Supp. 525 (M.D.Ga.1989). That judgment was later reversed by a divided panel of the Eleventh Circuit, 908 F.2d 851 (11th Cir.1990), which was subsequently vacated by the United States Supreme Court. - U.S. -, 111 S.Ct. 2845, 115 L.Ed.2d 1013 (1991). In its order of January 23, 1992, the district court stated that it was "unwilling to reconsider" plaintiffs' Section 2 claim. Under Section 2 of the Voting Rights Act, plaintiffs may prevail by proving either discriminatory effect or intent. Gingles, 478 U.S. at 35, 106 S.Ct. at 2758. The district court's factual findings regarding plaintiffs' constitutional claims, however, are applicable to the Section 2 claim. Assuming without deciding that the form and timing of the bond referendum can constitute a standard, practice, or procedure as contemplated by Section 2 of the Voting Rights Act, the district court's factual findings regarding plaintiffs' failure to show either discriminatory effect or discriminatory intent are equally fatal to plaintiffs' Section 2 claim.
The other arguments of plaintiffs in the court below and on appeal do not merit reversal of the judgment of the district court.
AFFIRMED.