Plaintiffs brought this action alleging that the electoral structure for selecting the members of the governing commission of the City of Miami Beach, Florida, violated § 2 of the Voting Rights Act by diluting Hispanic voting power. After a bench trial, the district court concluded that there was no § 2 violation and entered judgment for the defendants. We affirm.

The City of Miami Beach is located on a seven mile stretch of beach between the Atlantic Ocean and Biscayne Bay. According to the 1990 census, 92,639 people live within its 87 blocks. The racial characteristics of the population are as follows:

              White             44,721     48.27%
              Black              3,358            3.62%
              Hispanic          43,342           46.79%
              Other              1,218            1.31%
              -----------------------------------------
              Total             92,639

The voting age population ("VAP"), defined as those aged 18 and over, is as follows:

              White VAP         40,106           50.41%
              Black VAP          2,548            3.20%
              Hispanic VAP      35,947           45.18%
              Other VAP            957            1.20%
              -----------------------------------------
              Total VAP         79,558

However, according to sample data released by the Census Bureau, only 50.16% of the Hispanic residents of Miami Beach are citizens, while 88.18% of the non-Hispanic residents are citizens.

Miami Beach is governed by the City Commission, consisting of a mayor and six commissioners. All seven members (the mayor and the six commissioners) are elected in at-large elections. The six commissioner positions are numbered so that candidates must run for a particular seat. The mayor presides over City Commission meetings but otherwise has no greater authority than the other six commissioners. The City Commission hires a city manager to run the city on a day-to-day basis.

A number of Hispanic citizens who reside in Miami Beach brought this action against the City, its mayor, the city commissioners, and the city clerk.2 The complaint alleged that the defendants violated § 2 of the Voting Rights Act of 1965, 42 U.S.C. § 1973, and further violated 42 U.S.C. §§ 1981 and 1983. Plaintiffs contended that the at-large method of electing the mayor and city commissioners impermissibly diluted Hispanic voting strength in Miami Beach.

The district court held a five-day bench trial beginning May 1, 1995. The court took judicial notice of a number of census tables, which included information from the 1990 census of the entire population and information from questions posed to a smaller sample population. The plaintiffs presented several witnesses, including four experts. The defendants called no expert witnesses. On the last day of trial, plaintiff Victor Diaz, proceeding pro se, moved in open court that the district court dismiss him as a party. The district court denied that motion.

The district court concluded that the plaintiffs had failed to carry their burden of proof on the § 2 vote dilution claim because they failed to establish any of the three preconditions and to satisfy the totality of the circumstances prong, as required by Thornburg v. Gingles, 478 U.S. 30, 106 S. Ct. 2752, 92 L. Ed. 2d 25 (1986). The court further concluded that the plaintiffs had failed to show intentional discrimination, which was fatal to their § 1981 and § 1983 claims. Accordingly, the court entered judgment in favor of the defendants.

On appeal, the plaintiffs have abandoned their § 1981 and § 1983 claims, raising only the question of whether the district court correctly held that the plaintiffs failed to prove a § 2 violation. Additionally, Diaz appeals from the district court's refusal to dismiss him from the action.

We review a district court's findings of § 2 vote dilution for clear error. See Gingles, 478 U.S. at 79, 106 S.Ct. at 2781. Deference is afforded the district court's findings "due to its 'special vantage point' and ability to conduct an 'intensely local appraisal of the design and impact of' a voting system." Lucas v. Townsend, 967 F.2d 549, 551 (11th Cir.1992) (quoting White v. Regester, 412 U.S. 755, 769, 93 S. Ct. 2332, 2341, 37 L. Ed. 2d 314 (1973)). However, we will correct a district court's errors of law and its findings of fact based upon misconceptions of law. See United States v. Jones, 57 F.3d 1020, 1022 (11th Cir.1995) (citing Meek v. Metropolitan Dade County, Fla., 985 F.2d 1471, 1481 (11th Cir.1993)).

We review for abuse of discretion a district court's decision whether to grant a voluntary dismissal. See Fisher v. Puerto Rico Marine Management, Inc., 940 F.2d 1502, 1502-03 (11th Cir.1991) (citation omitted).

Section 2(a) of the Voting Rights Act states:

No voting qualification or prerequisite to voting or standard, practice, or procedure shall be imposed or applied by any State or political subdivision in a manner which results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color....

42 U.S.C. § 1973(a). Section 2(b) further explains:

A violation of [§ 2(a) ] is established if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by [§ 2(a) ] 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.

Id. at § 1973(b). As Justice O'Connor explained in Thornburg v. Gingles, "the essence of a vote dilution claim is that the State has created single-member or multimember districts that unacceptably impair the minority group's ability to elect the candidates its members prefer." 478 U.S. at 88, 106 S.Ct. at 2786 (O'Connor, J., concurring in the judgment).

The Gingles Court set forth three prerequisites for establishing a § 2 vote dilution claim. Plaintiffs must establish that:

(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"; and

(3) "the white majority group votes sufficiently as a bloc to enable it ... to defeat the minority's preferred candidate."

Id. at 50-51, 106 S.Ct. at 2766-67 (citations omitted). These three Gingles factors are "necessary preconditions" to a § 2 vote dilution claim. Id. at 50, 106 S.Ct. at 2766.

Proving the three preconditions is not the end of the story, however. As the Supreme Court explained in a later case:

if Gingles so clearly identified the three [preconditions] as generally necessary to prove a § 2 claim, it just as clearly declined to hold them sufficient in combination, either in the sense that a court's examination of relevant circumstances was complete once the three factors were found to exist, or in the sense that the three in combination necessarily and in all circumstances demonstrated dilution.

Johnson v. De Grandy, 512 U.S. 997, 1011, 114 S. Ct. 2647, 2657, 129 L. Ed. 2d 775 (1994). As § 2 mandates, a court must look to the totality of the circumstances to determine whether there is impermissible vote dilution. Gingles, 478 U.S. at 79, 106 S.Ct. at 2781; De Grandy, 512 U.S. at 1011, 114 S.Ct. at 2657. Borrowing from the Senate report accompanying the 1982 Amendments to the Voting Rights Act, Gingles identified a list of factors that may under the totality of the circumstances support a claim of vote dilution. See Gingles, 478 U.S. at 37, 106 S.Ct. at 2759 (quoting S.Rep. No. 97-417, at 28-29 (1982), reprinted in 1982 U.S.C.C.A.N. 177, 206-07).

The district court correctly utilized the Gingles framework in analyzing plaintiffs' § 2 claim. After examining the evidence, the court concluded that the plaintiffs had failed to establish any of the Gingles preconditions: numerosity and compactness, minority cohesion, or bloc voting. The court further concluded that, under the totality of the circumstances, plaintiffs had failed to prove their § 2 vote dilution claim. Because we hold that the district court correctly determined that plaintiffs had failed to establish the first Gingles precondition, numerosity and compactness, it is unnecessary for us to review its other determinations.

In their attempt to establish the first Gingles precondition, the plaintiffs offered Jerry Wilson, who testified as to the demographics of Miami Beach. Based on 1990 census data, Wilson explained that the Hispanic population of Miami Beach was concentrated in the southern and northern ends of the city. He also produced a plan, called Plan 7-C, that divides Miami Beach into seven districts of approximately equal population:

NOTE--Some parts of this form are wider than one screen. To view

material that exceeds the width of this screen, use the right arrow

key. To return to the original screen, use the left arrow key.

                                          Plan 7"C
          Total   White   Black   Hisp.   Hisp.         Total   White   Black  Hisp.   Hisp.
District   Pop     Pop     Pop     Pop      %            VAP     VAP     VAP    VAP      %
---------------------------------------------------------------------------------------------
   1      13,594   4,125    705    8,669  63.77%        11,817   3,850    527   7,362  62.30%
   2      13,509   4,340    600    8,408  62.24%        11,420   3,820    452   7,021  61.48%
   3      13,213   7,366    351    5,374  40.67%        12,195   7,043    307   4,747  38.93%
   4      13,224   8,149    169    4,723  35.72%        11,547   7,184    153   4,060  35.16%
   5      13,027  10,010    139    2,700  20.73%        11,288   8,770    123   2,263  20.05%
   6      13,432   6,455    464    6,289  46.82%        11,665   5,931    375   5,184  44.44%
   7      12,640   4,276    930    7,179  56.80%         9,626   3,508    611   5,310  55.16%
---------------------------------------------------------------------------------------------
 Total    92,639  44,721   3,358  43,342                79,558  40,106  2,548  35,947
----------

As the table indicates, in three of the districts, Districts 1, 2, and 7, Hispanics constitute a majority of the population and of the voting age population. Based on this proposed plan, Wilson concluded that the Hispanic population of Miami Beach is large enough and compact enough to constitute a majority in a single-member district.

As the district court found and Wilson admitted, however, he did not take into account the significant disparity between Hispanic and non-Hispanic citizenship rates. Citizenship data is available for the City of Miami Beach from the Census Bureau through a separate special tabulation. Synthesizing data from Table 182 and Table 167, contained in the 1990 Census of Population, Social and Economic Characteristics, Florida, produces the following citizenship rates, which demonstrate a significant disparity between Hispanic and non-Hispanic citizenship rates:

      Hispanic citizenship rate           50.16%
      Non"Hispanic citizenship rate       88.18%

Applying these rates to Districts 1, 2, and 7 of Plan 7-C yields the following results:

NOTE--Some parts of this form are wider than one screen. To view

material that exceeds the width of this screen, use the right arrow

key. To return to the original screen, use the left arrow key.

            Non       Non         Non                                 Hispanic     % of
          Hispanic  Hispanic   Hispanic          Hispanic  Hispanic   Citizen    Hispanic
District    VAP     Cit.Rate  Citizen VAP          VAP     Cit.Rate     VAP     Citizen VAP
-------------------------------------------------------------------------------------------
   1        4455     88.18%      3928              7362     50.16%      3692      48.45%
   2        4399     88.18%      3879              7021     50.16%      3521      47.58%
   7        4316     88.18%      3805              5310     50.16%      2663      41.17%

As the final column of this table indicates, when citizenship is taken into account, there is no Hispanic majority in any of the districts. Accordingly, the district court concluded that the plaintiffs failed to meet the first Gingles precondition because, among other reasons, they failed to consider citizenship rates in drawing Plan 7-C.

Plaintiffs contend that citizenship should not be taken into account. To support that position, plaintiffs rely on Solomon v. Liberty County, Florida, 899 F.2d 1012 (11th Cir.1990) (en banc). In Solomon, this Court unanimously held that the plaintiffs had established the three Gingles preconditions. The court split evenly, however, on the question of the legal significance of that accomplishment. Judge Kravitch wrote a special concurrence, in which four judges joined. Then-Chief Judge Tjoflat also wrote a special concurrence in which four judges joined.3 For present purposes, the differences that split the Court in Solomon are not relevant.

What is relevant is that every member of the Solomon Court agreed that the district court had erred in focusing on registered voter statistics instead of voting age population statistics. See Solomon, 899 F.2d at 1018 (Kravitch, J., specially concurring); Solomon v. Liberty County, Florida, 865 F.2d 1566, 1574 (11th Cir.1988) (Tjoflat, J.), vacated and reh'g en banc granted, 873 F.2d 248 (11th Cir.1989). Because blacks made up 51% of the voting age population of the proposed district, the first Gingles precondition was satisfied even though blacks made up only 46% of the registered voters of that district. See Solomon, 899 F.2d at 1018 (Kravitch, J., specially concurring) and Solomon, 865 F.2d at 1574 (Tjoflat, J.)

The plaintiffs in this case contend that Solomon requires that voting age population statistics be used in § 2 analysis instead of voting age citizenship population statistics. Of course, Solomon did not address that issue, because there was no indication in that case that there was any disparity between black and white citizenship rates. Nor is there likely to be any disparity in citizenship rates, except in a case, such as this one, where the minority population includes a substantial number of immigrants.

Because the question was not presented in Solomon, the holding in that case could not and does not preclude a holding in this case that refinement with citizenship data is appropriate, where that data is available and indicates a significant difference in the citizenship rates of the majority and minority populations. The reasoning in Solomon does not preclude such a holding, either. Both opinions in that case reasoned that voting age population statistics were preferable to registered voter statistics because minority voter registration might well have been suppressed by the absence of any prospect of electing representatives of the minority's choice. 899 F.2d at 1018 (Kravitch, J., specially concurring) ("An at-large election system that frustrates the ability of minorities to elect their chosen representatives will naturally reduce the incentive for blacks to register to vote."); 865 F.2d at 1574 (Tjoflat, J.) ("Minority voter registration figures are inherently unreliable measures in vote dilution cases because the very lack of minority political power responsible for bringing of the section 2 action may also act to depress voter registration."). The same cannot be said of citizenship. It is far more difficult for an immigrant to become a citizen than it is for a citizen to register to vote. Given all the other benefits citizenship entails, it is unlikely that a significant number of the non-citizen Hispanic residents of the City of Miami Beach would have become citizens if only the City had been districted in a manner making it possible for Hispanics to elect a representative of their choice to the City Commission.

We find support for consideration of citizenship data in Supreme Court opinions. The Court has thus far declined to answer directly the question of "which characteristic of minority populations (e.g., age, citizenship) ought to be the touchstone for proving a dilution claim." De Grandy, 512 U.S. at 1008, 114 S.Ct. at 2656. However, the Court has said that the foundational inquiry for the first Gingles precondition is whether "the minority has the potential to elect a representative of its own choice in some single-member district." Growe v. Emison, 507 U.S. 25, 39, 113 S. Ct. 1075, 1084, 122 L. Ed. 2d 388 (1993) (citation omitted and emphasis added). As the Gingles court explained:

The reason that a minority group making [a vote dilution] challenge must show, as a threshold matter, that it is sufficiently large and geographically compact to constitute a majority in a single-member district is this: Unless minority voters possess the potential to elect representatives in the absence of the challenged structure or practice, they cannot claim to have been injured by that structure or practice.... Thus, if the minority group is spread evenly throughout a multimember district, or if, although geographically compact, the minority group is so small in relation to the surrounding white population that it could not constitute a majority in a single-member district, these minority voters cannot maintain that they would have been able to elect representatives of their choice in the absence of the multimember electoral structure.

478 U.S. at 50 n. 17, 106 S. Ct. at 2766 n. 17 (emphasis added). See Romero v. City of Pomona, 883 F.2d 1418, 1425 (9th Cir.1989) (explaining that the Gingles opinion "repeatedly makes reference to effective voting majorities, rather than raw population totals, as the touchstone for determining geographical compactness") (footnote omitted).

In order to elect a representative or have a meaningful potential to do so, a minority group must be composed of a sufficient number of voters or of those who can readily become voters through the simple step of registering to vote. In order to vote or to register to vote, one must be a citizen. We agree with the Ninth Circuit that because "a section 2 claim will fail unless the plaintiff can establish that the minority group constitutes an effective voting majority in a single-member district," Romero, 883 F.2d at 1426, the proper statistic for deciding whether a minority group is sufficiently large and geographically compact is voting age population as refined by citizenship. Id. at 1424-26. Of course, the requirement that voting age population data be further refined by citizenship data applies only where there is reliable information indicating a significant difference in citizenship rates between the majority and minority populations. As we have previously indicated, such a disparity is unlikely except in areas where the population includes a substantial number of immigrants. We turn now to the question of whether such information exists in this case.

3. The Citizenship Information for Miami Beach

Plaintiffs attempt to justify their expert's decision not to use citizenship information by calling into question the accuracy of the data itself. The Miami Beach citizenship information from the Census Bureau was not obtained by the door-to-door census but is instead based on questions posed to a smaller segment of the population. The citizenship information (as well as some other information) is extrapolated from this smaller sample population. Because the citizenship information is based upon a sample population, it cannot be as precise as the census data, which is based upon the entire population. For example, according to the sample data, there are 42,888 Hispanics in Miami Beach, as compared to 43,342 as shown in the door-to-door census data. Plaintiffs contend that because the sample data underestimates the Hispanic population and simultaneously overestimates the white and black population, mixing the sample data with the census data invites error.

If mixing the two sets of data is an invitation for error, it is an invitation plaintiffs themselves extended. In addressing the totality of the circumstances, plaintiffs offered income, poverty and education statistics showing that Hispanics in Miami Beach were poorer and less educated than their white counterparts. All of those statistics came from sample data. One of the key census tables containing citizenship data, Table 182, was one the district court took judicial notice of at the plaintiffs' request. We will not allow plaintiffs to take inconsistent positions by touting the sample data when it suits their purposes and decrying the validity of that data when it does not.

Even if plaintiffs themselves had not introduced and relied on the sample data, we would nevertheless uphold the district court's consideration of the citizenship statistics, even though those statistics are based on sample data. The use of sample data is a long-standing statistical technique, whose limits are known and measurable. We will not reject the citizenship statistics solely because they are based on sample data without some indication that the sample was tainted in some way. There were no arguments before the district court that the sample was skewed in a statistically significant way due to improper sampling method, small sample size, or sheer random error.4 The plaintiffs contend that the sample's underestimation of the Hispanic population should cause the sample data to be rejected, but plaintiffs fail to show that that underestimation is statistically significant. In fact, there is a close correlation between the door-to-door census data and the sample data. For example, with regard to figures for the Hispanic population (43,342 figure from the census data compared to 42,888 figure from the sample data), the difference amounts to less than one percent. We conclude that the citizenship information for Miami Beach is reasonably accurate and demonstrates a significant disparity between Hispanic citizenship rates and non-Hispanic citizenship rates. In order to obtain an accurate assessment of Hispanic voting strength, this reasonably accurate citizenship information should be taken into account.

The plaintiffs offer two other arguments against using the citizenship information. First, the sample data, including the citizenship information, is available only at the block group level, not at the block level, which is the smallest population group. Plaintiffs contended at oral argument that using the block group data would force Wilson to make several assumptions when drawing the map and result in a less exact map. However, Wilson admitted in his testimony at trial that it was possible to create a district map of Miami Beach using block groups. Accordingly, we reject this argument.

Second, the plaintiffs contend that the district court erred in not accepting Wilson's plan disregarding citizenship, because the defendants presented no expert testimony explaining how or why citizenship should be taken into account. Putting aside for the moment the fact that the district court as factfinder was free to reject Wilson's expert testimony, even if it was uncontradicted, see Gregg v. U.S. Industries, Inc., 887 F.2d 1462, 1469-70 (11th Cir.1989), plaintiffs misconstrue the issue. Whether citizenship should be taken into account for the first Gingles precondition is a question of law. See De Grandy, 512 U.S. at 1007-10, 114 S.Ct. at 2655-56. Wilson's testimony, uncontradicted or not, and any failure of the defendants to put on testimony, have no bearing on the resolution of that legal question.

Plaintiffs argue in the alternative that, assuming citizenship rates should be taken into account, Plan 7-C still allows for a majority Hispanic district. However, that argument is based upon application of citizenship rates to raw population figures, not to voting age population. As we have explained, both voting age and citizenship are to be considered in determining whether the minority group has the potential to elect its preferred representatives. When both citizenship and voting age are taken into account, plaintiffs cannot make the requisite showing that Hispanics will constitute a majority in a single-member district.

Plaintiffs point to one other piece of evidence in their effort to show that the district court erred in deciding that the first Gingles precondition had not been met. Wilson apparently created another map of Miami Beach on which he drew a single district with a 66% majority Hispanic VAP. Neither the map nor the chart giving the racial breakdown of the single district were introduced into evidence. However, Wilson testified about the single district, and we will consider this evidence to the extent he testified about it. In any event, Wilson stopped short after this first district and never drew the other six districts. There is no information as to what the six remaining districts might look like, where they would be located, or what their racial and ethnic makeup might be. Nor is there information about how many people reside in the district or how the single district compares in population size to the remaining six districts. Thus, this is not really a plan, but only a part of one.

As this Court has interpreted the first Gingles precondition, "plaintiffs in vote dilution cases must demonstrate that the challenged system suppressed minority voting strength in comparison to some alternative, feasible benchmark system." Nipper v. Smith, 39 F.3d 1494, 1531 (11th Cir.1994) (en banc) (citing Holder v. Hall, 512 U.S. 874, 114 S. Ct. 2581, 129 L. Ed. 2d 687 (1994)), cert. denied, 514 U.S. 1083, 115 S. Ct. 1795, 131 L. Ed. 2d 723 (1995); see also SCLC v. Sessions, 56 F.3d 1281, 1289 (11th Cir.1995) (en banc) ("[P]laintiffs must show that an appropriate remedy can be fashioned" as part of first Gingles precondition), cert. denied, --- U.S. ----, 116 S. Ct. 704, 133 L. Ed. 2d 660 (1996). A district court cannot implement an incomplete plan, containing only a single district, with the rest of the map left blank. Because this plan does not provide an "alternative, feasible benchmark system," plaintiffs' testimony regarding this incomplete plan fails to meet the first Gingles precondition.

To summarize, we conclude that the first Gingles precondition requires the consideration of citizenship information to determine whether a minority has the potential to elect its preferred representative in a single-member district, when that information is reasonably accurate and demonstrates a significant difference between minority and majority citizenship rates. When citizenship rates are applied to plaintiffs' Plan 7-C, the plan fails to create any districts where Hispanics constitute a majority of potential voters. We agree with the district court that plaintiffs have failed to establish the first Gingles precondition. Because we hold that plaintiffs' § 2 vote dilution claim fails with the first Gingles precondition, we do not address the remainder of the district court's findings of fact and conclusions of law concerning the claim.

Plaintiff Diaz appeals from the district court's refusal to grant his motion for voluntary dismissal pursuant to Federal Rule of Civil Procedure 41. Diaz contends that his motion to dismiss was a Rule 41(a)(1)(ii) stipulation. That rule provides the following:

[A]n action may be dismissed by the plaintiff without order of court ... by filing a stipulation of dismissal signed by all parties who have appeared in the action.

Fed.R.Civ.P. 41(a)(1)(ii). Diaz's motion fails to meet the requirement of the rule in two ways. First, the motion was not made in writing but was made orally in open court. Second, even if an oral motion could satisfy the rule, there is no indication in the record that anyone besides Diaz agreed to the motion. The rule is clearly stated: a voluntary dismissal by stipulation is applicable only if all the parties sign off on it. Diaz was not entitled to a voluntary dismissal by stipulation.

Rule 41(a)(2) provides the other avenue for a voluntary dismissal. It states:

[A]n action shall not be dismissed at the plaintiff's instance save upon order of the court and upon such terms and conditions as the court deems proper.

Fed.R.Civ.P. 41(a)(2). If Diaz's motion to dismiss is considered as a Rule 41(a)(2) motion, it requires leave of the court and is subject to the court's discretion. We review the district court's decision on a Rule 41(a)(2) dismissal only for abuse of discretion. See Fisher v. Puerto Rico Marine Management, Inc., 940 F.2d 1502, 1503 (11th Cir.1991). We can find nothing in the record that indicates the district court's refusal to grant his motion was an abuse of discretion.IV. CONCLUSION

For the reasons set forth above, we AFFIRM the judgment of the district court.