with whom Mr. Justice Douglas and Mr. Justice Marshall join, dissenting in No. 71-1476, ante, p. 735, and concurring in part and dissenting in part in No. 72-147.
The Court today upholds statewide legislative apportionment plans for Connecticut and Texas, even though these plans admittedly entail substantial inequalities in the population of the representative districts, and even though the States have made virtually no attempt to justify their failure “to construct districts ... as nearly of equal population as is practicable.” Reynolds v. Sims, 377 U. S. 533, 577 (1964). In reaching this conclusion, the Court sets aside the judgment of the United States District Court for the District of Connecticut holding the Connecticut plan invalid, and the judgment of the United States District Court for the Western Dis*773trict of Texas reaching a similar result as to the Texas plan. In the Texas case, the Court does affirm, however, the District Court’s determination that the use of multi-member districts in Dallas and Bexar Counties had the unconstitutional effect of minimizing the voting strength of racial groups.1 See Whitcomb v. Chavis, 403 U. S. 124, 142-144 (1971); Burns v. Richardson, 384 U. S. 73, 88 (1966); Fortson v. Dorsey, 379 U. S. 433, 439 (1965). With that latter conclusion I am in full agreement, as I also agree with and join Part I of the Court’s opinion in No. 72-147, White v. Regester. But the decision to uphold the state apportionment schemes reflects a substantial and very unfortunate retreat from the principles established in our earlier cases, and I therefore must state my dissenting views.
I
At issue in No. 71-1476, Gaffney v. Cummings, is the 1971 reapportionment plan for election of members of the House of Representatives of Connecticut. The plan was premised on a 151-member House, with each member elected from a single-member district. Since the population of the State was 3,032,217, according to 1970 census data, the ideal would fix the population of each district at 20,081. In fact, the population of many *774districts deviated substantially from the ideal, ranging from a district underrepresented by 3.93% to one overrepresented by 3.9%. The total spread,of deviation — a figure deemed relevant in each of our earlier decisions — ■ was 7.83%. The population of 39 assembly districts deviated from the average by more than 3%. Another 34 districts deviated by more than 2%. The average deviation was just under 2%. To demonstrate that the state plan did not achieve the greatest practicable degree of equality in per-district population, appellees submitted a number of proposed apportionment plans, including one that would have significantly reduced the extent of inequality. The total range of deviation under appellees' plan would have been 2.61%, as compared to 7.83% under the state plan.
The District Court held the state plan invalid on the ground that “the deviations from equality of populations of the . . . House districts are not justified by any sufficient state interest.” 2 341 F. Supp. 139, 148 (Conn. 1972). Instead of adopting one of appellees' plans, the court appointed a Special Master to chart a new plan, and his effort produced a scheme with a total range of deviation of only 1.16%. In overturning the District Court's decision, the Court does not conclude, as it did earlier this Term in Mahan v. Howell, 410 U. S. 315 (1973), that the District Court failed to discern the State's sufficient justification for the deviations. Indeed, in view of appellant's halfhearted attempts to justify *775the deviations at issue here, such a conclusion could hardly be supported. Whereas the Commonwealth of Virginia made a substantial effort to draw district lines in conformity with the boundaries of political subdivisions — an effort that was found sufficient in Mahan v. Howell to validate a plan with total deviation of 16.4%— the evidence in the case before us requires the conclusion that Connecticut’s apportionment plan was drawn in complete disregard of political subdivision lines. The District Court pointed out that “[t]he boundary lines of 47 towns are cut under the Plan so that one or more portions of each of these 47 towns are added to another town or a portion of another town to form an assembly district.” 341 F. Supp., at 142. Moreover, the boundary lines of 29 of these 47 towns were cut more than once, and the plan created “78 segments of towns in the formation of 151 assembly districts.” Ibid.
Although appellant failed to offer cogent reasons in explanation of the substantial variations in district population, the Court nevertheless upholds the state plan. The Court reasons that even in the absence of any explanation for the failure to achieve equality, the showing of a total deviation of almost 8% does not make out a prima facie case of invidious discrimination under the Fourteenth Amendment. Deviations no greater than 8% are, in other words, to be deemed de minimis, and the State need not offer any justification at all for the failure to approximate more closely the ideal of Reynolds v. Sims, supra.
The Texas reapportionment case, No. 72-147, White v. Regester, presents a similar situation, except that the range of deviation in district population is greater and the State’s justifications are, if anything, more meager. An ideal district in Texas, which chooses the 150 members of the State House of Representatives from 79 single-member and 11 multimember districts, is 74,645. As *776defined in the State’s 1970 plan, a substantial number of districts departed significantly from the ideal. The total range of deviation was at least 9.9%, and arguably almost 30%, depending on the mode of calculation.3 The District Court pointed out that
“[i]n all of the evidence presented in this case, the State has not attempted to explain in terms of rational State policy its failure to create districts equal in population as nearly as practicable, nor has the State sought to justify a single deviation from precise mathematical equality. The lengthy depositions of the members of the legislative redistricting board and of the staff members who did the actual drawing of the legislative district lines are devoid of any meaningful indications of the standards used.” 343 F. Supp. 704, 714 (WD Tex. 1972).
As the District Court’s opinion makes clear, the variations surely cannot be defended as a necessary byproduct of a state effort to avoid fragmentation of political subdivisions. Nevertheless, the Court today sets aside the District Court’s decision, reasoning, as in the Connecticut case, that a showing of as much as 9.9% total deviation still does not establish a prima facie case under the Equal Protection Clause of the Fourteenth Amendment. Since the Court expresses no misgivings about our recent decision in Abate v. Mundt, 403 U. S. 182 (1971), where we held that a total deviation of 11.9% must be *777justified by the State, one can reasonably surmise that a line has been drawn at 10% — deviations in excess of that amount are apparently acceptable only on a showing of justification by the State; deviations less than that amount require no justification whatsoever.
II
The proposition that certain deviations from equality of district population are so small as to lack constitutional significance, while repeatedly urged on this Court by States that failed to achieve precise equality, has never before commanded a majority of the Court.4 Indeed, in Kirkpatrick v. Preisler, 394 U. S. 526, 530 (1969), we expressly rejected the argument
“that there is a fixed numerical or percentage population variance small enough to be considered de minimis and to satisfy without question the 'as nearly as practicable’ standard. The whole thrust of the 'as nearly as practicable’ approach is inconsistent with adoption of fixed numerical standards which excuse population variances without regard to the circumstances of each particular case.”
The Court reasons, however, that Kirkpatrick v. Preisler, *778supra, a case that concerned the division of Missouri into congressional districts, has no application to the apportionment of seats in a state legislature. In my dissenting opinion in Mahan v. Howell, supra, I pointed out that the language, reasoning, and background of the Kirkpatrick decision all command the conclusion that our holding there is applicable to state legislative apportionment no less than to congressional districting. In fact, this Court specifically recognized as much in the context of a challenge to an Arizona apportionment scheme in Ely v. Klahr, 403 U. S. 108 (1971). Describing the opinion of the District Court whose judgment was under review, we noted that the court below had “properly concluded that this plan was invalid under Kirkpatrick v. Preisler, 394 U. S. 526 (1969), and Wells v. Rockefeller, 394 U. S. 542 (1969), since the legislature had operated on the notion that a 16% deviation was de minimis and consequently made no effort to achieve greater equality.” 403 U. S., at 111. Yet it is precisely such a notion that the Court today approves.5
Moreover, even if Kirkpatrick should be deemed inapplicable to the apportionment of state legislative districts, the reasoning that gave rise to our rejection of a *779de minimis approach is fully applicable to the case before us. We pointed out there that the “as nearly as practicable” standard — the standard that controls legislative apportionment as well as congressional districting, Reynolds v. Sims, supra, at 577 — demands that “the State make a good-faith effort to achieve precise mathematical equality. . . . Equal representation for equal numbers of people is a principle designed to prevent debasement of voting power and diminution of access to elected representatives. Toleration of even small deviations detracts from these purposes.” 394 U. S., at 530-531. Kirkpatrick recognized that “to consider a certain range of variances de minimis would encourage legislators to strive for that range rather than for equality as nearly as practicable.” 394 U. S., at 531.
Although not purporting to quarrel with the principle that precise mathematical equality is the constitutionally mandated goal of reapportionment, the Court today establishes a wide margin of tolerable error, and thereby undermines the effort to effectuate the principle. For it is clear that the state legislatures and the state and federal courts have viewed Kirkpatrick as controlling on the issue of legislative apportionment, and the outgrowth of that assumption has been a truly extraordinary record of compliance with the constitutional mandate. Appellees in No. 71-1476 make the point forcefully by comparing the extent of inequality in the population of legislative districts prior to 1969, the year of our decision in Kirkpatrick, with the extent of inequality in subsequent years.6 Prior to 1969, the range of variances in population of state senatorial districts exceeded 15% in 44 of the 50 States. Three States had *780reduced the total variance to between 10% and 15%; two had cut the variance to between 5% and 10%; only one had reduced the variance below 5%. The record of apportionment of state House districts was even less encouraging. Variances in excess of 15% characterized all but two of the States, and only one of these had brought the total variance under 10%. The improvement in the post-1969 years could not have been more dramatic. The table provided by appellees, set out in full in the margin,7 reveals that in almost one-half of the States the total variance in population of senatorial districts was within 5% to zero. Of the 45 States as to which information was available, 32 had reduced the total variance below 10% and only eight had failed to bring the total variance below 15%. With regard to House districts the improvement is similar. On the basis of information concerning 42 States, it appears that 20 had achieved a total variance of less than 5%, and only 14 retained districts with a total variance of more than 15% from the constitutional ideal.
To appreciate the significance of this encouraging development, it is important to understand that the demand for precise mathematical equality rests neither on *781a scholastic obsession with abstract numbers nor a rigid insensitivity to the political realities of the reapportionment process. Our paramount concern has remained an individual and personal right — the right to an equal vote. “While the result of a court decision in a state legislative apportionment controversy may be to require the restructuring of the geographical distribution of seats in a state legislature, the judicial focus must be concentrated upon ascertaining whether there has been any discrimination against certain of the State’s citizens which constitutes an impermissible impairment of their constitutionally protected right to vote.” Reynolds v. Sims, supra, at 561. We have demanded equality in district population precisely to insure that the weight of a person’s vote will not depend on the district in which he lives. The conclusion that a State may, without any articulated justification, deliberately weight some persons’ votes more heavily than others, seems to me fundamentally at odds with the purpose and rationale of our reapportionment decisions. Regrettably, today’s decisions are likely to jeopardize the very substantial gains that have been made during the last four years.
Moreover, if any approach ascribes too much importance to abstract numbers and too little to the realities of malapportionment, it is not Kirkpatrick’s demand for precise equality in district population, but rather the Court’s own de minimis approach. By establishing an arbitrary cutoff point expressed in terms of total percentage variance from the constitutional ideal, the Court fails to recognize that percentage figures tend to hide the total number of persons affected by unequal weighting of votes. In the Texas casé, for example, the District Court pointed out that
“the total deviations for Dallas and Bexar Counties, respectively, amount to about 16,000 people and 5,500 people, for a total of around 21,500 people. *782The percentage deviation figures are only a shorthand method of expressing the ‘loss,’ dilution, or disproportionate weighting of votes. Just as the Court in Reynolds concluded that legislators represent people, not trees or cows, so we would emphasize that legislators represent people, not percentages of people.” 343 F. Supp., at 713 n. 5.
Finally, it is no answer to suggest that precise mathematical equality is an unsatisfactory goal in view of the inevitable inaccuracies of the census data on which the plans are based. That argument, which we implicitly rejected in Kirkpatrick v. Preisler, supra,8 mixes two distinct questions. In the first place, a state apportionment plan must be grounded on the most accurate available data, and the unreliability of the data may itself necessitate the invalidation of the plan. But once the data are established, the State’s constitutional obligation is to achieve the highest practicable degree of equality with reference to the information at hand. In my view, the District Courts properly concluded that neither Texas nor Connecticut had satisfied this obligation. I would therefore affirm both judgments.
In Fortson v. Dorsey, 379 U. S. 433 (1965), we held that a multimember district is not per se unconstitutional under the Equal Protection Clause, even though we had previously recognized certain inherently undesirable features of the device. See Lucas v. Colorado General Assembly, 377 U. S. 713, 731 n. 21 (1964). We have concluded, however, that the use of the device is, in fact, unconstitutional, where it operates to “ 'minimize or cancel out the voting strength of racial or political elements of the voting population,’ ” Burns v. Richardson, 384 U. S. 73, 88 (1966), quoting from Fortson v. Dorsey, supra, at 439. Today’s decision is the first in which we have sustained an attack on the use of multimember districts. Cf. Whitcomb v. Chavis, 403 U. S. 124, 144 (1971).
With regard to the senatorial districts, the 1971 plan produced a total variance of 1.81%. Although appellees did not specifically challenge the apportionment of senatorial districts, the District Court properly concluded that its finding of unconstitutional deviation in one house required invalidation of the entire apportionment plan. Maryland Committee for Fair Representation v. Tawes, 377 U. S. 656, 673 (1964); Lucas v. Colorado General Assembly, supra, at 735. Burns v. Richardson, supra, at 83.
The District Court pointed out that “the State’s method of computing deviations in the multi-member districts may distort the actual percentage deviations in those eleven districts. . . . Since we have concluded that the 9.9% total deviation is not the result of a good faith attempt to achieve population equality as nearly as practicable, it is unnecessary for us to resolve this complex computational conflict.” 343 F. Supp. 704, 713 n. 5. A similar conflict existed in Mahan v. Howell, 410 U. S. 315 (1973), as I pointed out in my dissenting opinion, id., at 333, and there too the Court declined to indicate any awareness of the dispute.
There is a statement, to be sure, in Swann v. Adams, 385 U. S. 440, 444 (1967), that “\_d\e minimis deviations are unavoidable,” but that statement must be viewed in context. By way of clarification, the Court immediately added that “the Reynolds opinion limited the allowable deviations to those minor variations which ‘are based on legitimate considerations incident to the effectuation of a rational state policy.' 377 U. S. 533, 579.” Ibid. Similarly, the Court noted, quoting from Roman v. Sincock, 377 U. S. 695, 710 (1964), that “the Constitution permits ‘such minor deviations only as may occur in recognizing certain factors that are free from any taint of arbitrariness or discrimination.’ ” 385 U. S., at 444. Swann v. Adams does not, in my view, suggest any support for the proposition that deviations as great as 10% are tolerable in the absence of any justification or explanation by the State.
By contrast, in Mahan v. Howell, supra, the Court expressly reaffirmed the holding of Reynolds v. Sims, 377 U. S. 533 (1964), that “some deviations from the equal-population principle are constitutionally permissible” “[s]o long as the divergences from a strict population standard are based on legitimate considerations incident to the effectuation of a rational state policy.” Id., at 579, quoted in Mahan v. Howell, supra, at 325 (emphasis added). In my view, the Court incorrectly concluded in Mahan v. Howell that Virginia had justified the population variations at issue there. Nevertheless, the Court did follow the line of analysis prescribed in our earlier decisions — requiring the State to justify every deviation from precise equality. The approach of Mahan is, therefore, directly at odds with'the approach adopted today. See also, e. g., Abate v. Mundt, 403 U. S. 182, 185 (1971); Kilgarlin v. Hill, 386 U. S. 120, 122 (1967); Swann v. Adams, supra, at 443-446.
Appellees’ figures are compiled from a table entitled Apportionment of Legislatures, in 17 Council of State Governments, the Book of the States: 1968-1969, pp. 66-67 (1968), and from Council of State Governments, Reapportionment in the Seventies (1973).
Deviations After 1970
See 394 U. S., at 538-540 (1969) (Fortas, J., concurring); Wells v. Rockefeller, 394 U. S. 542, 554 (1969) (White, J., dissenting).