OPINION
Before RIVES, Circuit Judge, and THOMAS and JOHNSON, District Judges.I. History of the Litigation
A) The protracted history of this suit is capsulized in Sims v. Amos, M.D.Ala. 1972, 336 F.Supp. 924, 930-932. In that decision, climaxing over a decade of litigation, this Court:
1) Rejected four separate reapportionment plans submitted by the defendants, because each plan utilized multimember districts, rigidly followed existing county lines, and produced wide variances from the ideal of one-man/one-vote (the per cent deviations ranged from a low of 24.28% in the Attorney General’s Senate Plan # 2 to a high of 52.34% in the Attorney General’s House Plan # 3).
2) Adopted the plaintiffs’ proposed reapportionment plan calling for complete reapportionment of the Alabama Legislature through the use of 105 single-member House districts (2.23% total deviation) and 35 single-member Senate districts (1.39% total deviation).
3) Refused to order mid-term elections and, instead, directed that the reapportionment would become effective for use in the 1974 State elections, primary and general, and for the Legislature as chosen in the 1974 elections.
That decision, entered January 3, 1972, was affirmed without opinion by the Supreme Court on October 24, 1972. Sims v. Amos, 1972, 409 U.S. 942, 93 S.Ct. 290, 34 L.Ed.2d 215.
B) In a February 26, 1973 opinion and order to aid in implementing its January 3, 1972 judgment, this Court took note of the Alabama Legislature’s repeated failure to adopt a constitutional reapportionment plan and ordered the defendants to complete by May 1, 1973 the location and establishment of the legislative district lines under the court-ordered plan.
In an addendum to the February 26 order, this Court expressed the firm view that the recent Supreme Court decision in Mahan, Secretary, Board of Elections of Virginia v. Howell, 1973, 410 U.S. 315, 93 S.Ct. 979, 35 L.Ed.2d 320, neither undermined the validity of the original January 3 decree nor warranted a delay in that decree’s implementation.
Yet, in spite of this Court’s view that the Supreme Court’s Virginia decision did not dictate adoption of a state-initiated reapportionment plan in Alabama, and in spite of the Legislature’s prolonged recalcitrance, this Court culminated its February 26 order by stating that,
“Notwithstanding all that has been said in the foregoing opinion and order and in this addendum, if the Alabama Legislature promptly enacts a proposed alternate reapp'ortionment plan, we will consider that plan with open minds and with the sincere hope that it may, under all proper considerations, be more acceptable than the present court-ordered plan.”
[Emphasis added.]
C) In an order entered April 9, 1973, this Court defined the term “promptly,” as used in the preceding quotation, to mean “on or before May 17,1973.”
D) On May 16, 1973, one day prior to the May 17 deadline, Governor Wallace and Attorney General Baxley filed for the defendants a motion seeking relief .from implementation of the January 3 court-ordered plan and acceptance of the reapportionment plan set forth in Act No. 3, House Bill 2, passed in the 1973 Special Session of the Legislature of Alabama and signed by the Governor on May 15, 1973.
Whether to grant or deny the defendants’ May 16 motion is the sole issue now before this Court.
*218II. Defendants’ Reapportionment Plan
The defendants’ current reapportionment plan was prepared by a University of Alabama task force under the supervision of Professor Henry Moore. The task force began operations under instructions set forth in a letter, dated May 16, 1972, from Representative Hugh Merrill to Don James, a task force member.
Representative Merrill’s letter directed the task force to draw a plan conforming to the following guidelines (promulgated by the Alabama Joint Legislative Committee on Reapportionment) :
1) Formation of 105 single-member House districts and 35 single-member Senate districts (composed of 3 House districts each);
2) formation of districts “based solely on total population as provided by the 1970 Census”;
3) construction of districts as nearly equal in population as possible;
4) preservation of existing political subdivisions, including precincts, wherever possible, with concomitant minimization of crossed county lines.
A) - Cherner Plan
In response to Representative Merrill’s directive, the task force completed in October 1972 the Cherner Plan which, according to Professor Moore, incorporated extremely small population deviations between districts (3.8% for the House, 2.8% for the Senate). [Moore Dep., p. 73.] However, due to the death of Representative Cherner and the prevalent opinion that the Cherner Plan had little chance of passing the Senate, that plan was shelved indefinitely. [Moore Dep., p. 76.]
B) Defendants’ Current Plan
Following the decision, Mahan v. Howell, supra, in February of 1973, Representative Merrill contacted Professor Moore and asked the task force to draw a new plan which, in Moore’s words, “would not increase the deviation unduly, but would uncross a number of the county lines that were left crossed in the Merrill plan [Cherner Plan].” [Moore Dep., p. 77.] Prodded on cross-examination as to the precise content of Representative Merrill’s directive, Professor Moore replied in the following colloquy:
“Q. Were you instructed, at that time, Professor Moore, that you could go as high as 16 plus per cent, the percentage of differential, which the Supreme Court approved in the Virginia case, Mahan versus Howell ?
“A. We were told not to exceed that.
“Q. Were you—
“A. (Interrupting) But to stay as low as we possibly could.
“Q. All right, sir. Were you told to shoot for 16 per cent ?
“A. No. We were told to stay under. it, was the only admonition we had and to keep the percentage as low as possible.”
[Moore Dep., p. 83.]
Thus, defendants’ current reapportionment plan evolved under a hybrid directive composed of Representative Merrill’s original letter to the task force and his subsequent instructions regarding the Virginia decision.
Defendants’ current reapportionment plan, embodied in Act No. 3 of the 1973 Special Session, assuming its statistical accuracy, produces a total population variance between House districts of 13.-24% and between Senate districts of 13.43%.
Its salient characteristics can be briefly listed:
1) 105 single-member House districts and 35 single-member Senate districts;
2) maintenance of all existing precinct lines;
3) split of 33 out of Alabama’s 67 counties, 29 between districts, 4 internally.
The only reasonably accurate 1970 population information available to the *219task force was that contained in the 1970 decennial census. The 1970 census enumeration is subdivided into four kinds of geographical units — census county divisions, enumeration districts, census tracts and block groups. The smallest identifiable population unit is the enumeration district (ED). Because ED boundaries do not uniformly coincide with precinct boundaries and because the task force interpreted its instructions as requiring complete fidelity to precinct autonomy, a method had to be devised for allocating population to the separate portions of an ED split by a precinct line. In his testimony, Professor Moore described that method as follows :
“This was done by the process of counting the houses on each side of a . precinct line which divided an enumeration district .... We divided the number of people by the number of houses to get the average population per household. We counted the houses on each side of the blue line [precinct line] . . . , multiplied the respective numbers by 3.1, or 3.6, or whatever happened to be the number that came out in dividing the population by houses, and put that-— those respective numbers of people on each side.”
[Moore Dep., pp. 38-39.]
In urban areas where ED’s were split by precinct lines and population was too congested to facilitate house symbol depiction on highway maps, the task force had to concoct another scheme for estimating the population of each portion of a severed ED. Professor Moore’s graduate assistant, Mr. McDowall, gave a revealing, although far from lucid, description of the system used to estimate urban population:
“A. [Professor Moore] (Interrupting) You can’t count houses in all cases. How did we do it when we didn’t count houses?
“MR. MeDOWALL: In the block group data I believe we probably estimated land area.
“Q. [Mr. Dees, attorney for plaintiffs] Now, let me take off from there. You say that in block group data you estimated land area ?
“A. Uh hum.
“Q. In other words, if there was a city of high population density, you took a certain number of square inches of land, and then you divide, drew a line through it and if thirty per cent of the land was on one side and seventy on the other, you apportioned the population accordingly ?
“MR. MeDOWALL: Probably.”
[Moore Dep., p. 146.]
The Moore plan did not emerge intact from the Legislature. The House Committee added several amendments affecting the allocation of significant numbers of people in Pike, Coffee, Chambers and Tallapoosa Counties [Moore Dep., p. 90], and the Senate re-drew the Senate district lines in both the Jefferson/Shelby County area and the northeast corner of the State, departing from the basic district lines established by the Moore task force. [Moore Dep., p. 95.]
III. Denial of Defendants’ Motion
Although we of course adhere to the constitutional principle that reapportionment is primarily a matter for legislative determination, Reynolds v. Sims, 1963, 377 U.S. 533, 84 S.Ct. 1362, 12 L.Ed.2d 506, and in spite of the fact that there are fundamental differences between congressional districting under Article I and state legislative reapportionment governed by the Fourteenth Amendment, Gaffney v. Cummings, 1973, 412 U.S. 735, 93 S.Ct. 2321, 37 L.Ed.2d 298, a careful study of defendants’ proposed reapportionment plan leads us to the conclusion that the motion to substitute that plan for the court-ordered plan must be denied.
A) Under section 5 of the Voting Rights Act, 42 U.S.C. § 1973c, certain states of the Deep South, including Alabama are required to submit any legislative reapportionment plan either to the United States Attorney General for pos*220sible objections, or to the United States District Court for the District of Columbia for a declaratory judgment that the proffered plan does not have the purpose and will not have the effect of denying or abridging the right to vote on account of race or color.
The implementing regulation, 28 C.F.R. § 5119, which sets out the standards the Attorney General is to employ in deciding whether or not to object to a state submission, places the burden of proof on the state to establish the nondiscriminatory effect of the proposed plan.1
In the instant case, defendants’ proposed plan need not be submitted to the Attorney General for scrutiny, since a decree of this Court would not fall within the ambit of section 5 of the Voting Rights Act. Connor v. Johnson, 1971, 402 U.S. 690, 91 S.Ct. 1760, 29 L.Ed.2d 268; Conner v. Board of Supervisors of Oktibbeha Co., Miss., N.D. Miss.1971, 334 F.Supp. 280. Nevertheless, the defendants retain the affirmative burden of proving that their plan is color blind or racially nondiscriminatory. Georgia v. United States, 1973, 411 U.S. 526, 93 S.Ct. 1702, 36 L.Ed.2d 472.2
Plaintiffs vigorously attack the alleged racial neutrality of defendants’ plan. They claim that the method of allocating population between the separate portions of divided ED’s diluted Negro representation by failing to take into account the disparity between the average size of Negro and white households. In addition, they assert that population estimates based on the assumption, employed by the Moore task force, that Negroes live in evenly distributed patterns creates a built-in discriminatory factor.
Directing attention to the revised Senate district lines in the Jefferson/Shelby County area and emphasizing that the legislators who drew those lines possessed complete racial data, plaintiffs raise a more serious allegation — that certain portions of the defendants’ plan were constructed with racially discriminatory intent.
While the evidence falls short of affirmatively establishing any intended invidious discrimination in the defendants’ plan, plaintiffs’ arguments amply demonstrate that defendants have failed to bear their burden of proving that the proposed plan is racially nondiscriminatory.
B) In Ely v. Klahr, 1971, 403 U.S. 108, 91 S.Ct. 1803, 29 L.Ed.2d 352, the Supreme Court affirmed a district court decision ordering use of a legislative reapportionment plan for the 1970 elections in Arizona. However, the Supreme Court carefully explained that the legislature’s plan was adopted by the district court as the least unsatisfactory alternative, only for use in the 1970 elections, the last elections before the 1970 census data would become available. Moreover, the Supreme Court catalogued the deficiencies in the legislative plan, listing as an initial defect the fact that “the population formula used did not ‘truly represent the population within [the] precincts in either 1960 or 1968,’ and thus ‘the figures produced . are not truly population figures.’ [Meltzer v. Crescent Leaseholds Inc.] 313 F.Supp. [142], at 152.” Ely v. Klahr, supra, at 112, 91 S.Ct. at 1806 (footnote omitted).
The Arizona Legislature chose to use the local political precinct as the founda*221tion for each district, but because no population data was available for precincts the legislature Was forced to devise a method for estimating precinct population through use of county population figures. The formula utilized was succinctly described by the Klahr Court at footnote 4, quoting the district court:
“4. 'The population factor in each of the election precincts comprising part of a legislative district was obtained by instructing the computer to take the 1968 voter registration for the precinct and divide it by the 1968 voter registration for the county in which the precinct was located, thereby obtaining the percentage of registered voters of the county residing within the precinct. The computer was then directed to multiply that percentage figure by the 1960 census for the county in which the precinct was located, thereby obtaining the population factor for the precinct.’ 313 F.Supp., at 151-152.”
403 U.S. at 112, 91 S.Ct. at 1806.
In his concurrence to the majority opinion in Klahr, Justice Douglas pinpointed the characteristic which made the estimation formula constitutionally inaccurate:
“If all segments of society were equally likely to register to vote, then the Arizona method of computing population would be unobjectionable. But all members of a community are not equally likely to register. For example, only two counties out of eight with Spanish surname populations in excess of 15% showed a voter registration equal to the statewide average. Not only are the poor, the blacks, the Chícanos, and the Indians less likely to register in the first place, they are also likely to have a higher rate of illiteracy among their members. Arizona law at the time of the decision below required a literacy test for voter registration. Ariz.Bev.Stat.Ann. §§ 16-101(A) (4), 16-101(A) (5).”
403 U.S. at 118, 91 S.Ct. at 1809 (footnote omitted).
The method used by the Moore task force to estimate the population of the distinct parts of ED’s split by precinct lines is remarkably similar to the impermissible formula used by the Arizona Legislature in Klahr3 The most important viable distinction between that case and this one is the fact that defendants’ plan in the present case contains far more inaccuracies.
The plan proffered in Klahr contained an admitted maximum deviation of only 1.8%, whereas defendants’ present plan with a facial 13;% plus deviation approaches the limits of tolerable inexactness without regard to the accuracy of the estimation method used.
More importantly, the Moore task force did not estimate precinct population by computer analysis and projection of existing reliable data, but, instead, followed two practices of dubious propriety :
1) House counts from old highway maps (many pre-dating the 1960 census) were used as the basis for rural population estimates.4
2) Visual comparison of map depicted land area was the sole foundation for allocation of population in split urban blocks.
We recognize that census figures can never represent an exact population count,5 and that maintenance of existing political subdivisions is a legitimate *222state goal.6 Yet, legislators represent people, not houses,7 and
“Whatever the means of accomplishment, the overriding objective must be substantial . equality of population among the various districts, so that the vote of any citizen is approximately equal in weight to that of any other citizen in the State.”
Reynolds v. Sims, 1964, 377 U.S. 533, 579, 84 S.Ct. 1362, 1390, 12 L.Ed.2d 506.
The facial inaccuracy of defendants’ proposed reapportionment plan (13% plus), coupled with the imprecise calculations made in split precincts, convinces us that the figures ultimately obtained by the Moore task force are not truly population figures and fail to satisfy the overriding constitutional objective of substantial equality.
C) Recent decisions of the Supreme Court in the evolving area of legislative reapportionment suggest utilization of a rule of thumb three-stage test to determine the constitutionality of the proposed plan.
If the districts of largest and smallest population are shown to differ by less than 10%, the plaintiffs’ threshold requirement of demonstrating invidious discrimination is not satisfied, and the state need not justify the discrepancy. White v. Regester, 1973, 412 U.S. 755, 93 S.Ct. 2332, 37 L.Ed.2d 314.8
If the deviation exceeds the de minimis limit set in White, but falls short of 16.4%, then the state is required to justify the population difference by demonstrating that the deviation is based on a rational state policy. Mahan v. Howell, supra9 When the maximum deviation surpasses the upper limit justifiable by a rational state policy, the statute becomes intolerable under the one-man/one-vote doctrine.
Defendants’ current proposed plan contains a maximum deviation, assuming the plan’s accuracy, of 13.4%. Hence, that deviation must be justified by a rational state policy. Defendants contend that maintenance of the integrity of political subdivision lines constitutes a rational justification notwithstanding the statistical inaccuracy of the Legislature’s plan. While we agree that maintenance of political boundaries is a rational justification for population disparities, Mahan v. Howell, supra, the defendants’ plan, which joins an admitted 13.4;% variance with highly questionable estimation procedures, has, in our opinion, the effect of sacrificing “substantial equality to justifiable deviations.”
Although substantial political reasons exist for preserving county lines where possible, Davis v. Mann, 1964, 377 U.S. 678, 686, 84 S.Ct. 1441, 12 L.Ed.2d 609, the grounds for total fidelity to precinct autonomy are, in our view, less significant. If defendants’ plan coupled reliance on existing precinct lines with an accurate, verifiable method for estimating the population of contiguous portions of split ED’s, our decision might be different.
In Mahan v. Howell, supra, the Supreme Court reaffirmed the holding in Reynolds v. Sims, supra, that, “the Equal Protection Clause requires that a State make an honest and good faith effort to construct districts, in both houses of its legislature, as nearly of equal population as is practicable.” Mahan, *223supra. In spite of the remarkable improvement of the current state plan over prior ones, we are unable confidently to hold that that plan is the result of a “good faith effort to construct districts . as nearly of equal population as practicable.”
CONCLUSION
Defendants’ proposed reapportionment plan, while commendably superior to previous state plans, comes to us in a posture which compels the defendants to carry the burden of sustaining the plan’s constitutionality. Three separate rationales — (1) failure to satisfy the burden of proving the plan racially nondiscriminatory, (2) reliance upon an inaccurate estimation procedure, as a result of which the population differences may far surpass the maximum deviation of 16.4% which might possibly be justified if based on a rational state policy, and (3) inability to establish a reasonable basis for use of precinct building blocks and consequent failure to justify the admitted population difference by demonstrating that the population deviation is based on rational state policy— unite to convince this Court that the motion to substitute the State’s reapportionment plan for the currently operative court-ordered plan should be denied. Accordingly, such a separate judgment is entered.
The Supreme Court’s landmark decision of Reynolds v. Sims, supra, 377 U. S. 533, 583, 584, 84 S.Ct. 1362, 1393, 12 L.Ed.2d 506, recognized that decennial reapportionment “would clearly, meet the minimal requirements for maintaining a reasonably current scheme of legislative representation.” Sections 198, 199 and 200 of the 1901 Constitution of Alabama are in line with that requirement. Section 200 makes reapportionment “the duty of the legislature at its first session after taking of the decennial census of the United States.” For more than seventy years the Legislature of Alabama has been unable to perform that duty. History shows that the legislatures of the other states have suffered from a similar incapacity.
Since Reynolds v. Sims, supra, the legislatures of some states have devised commissions or like bodies to aid them in the performance of their difficult duty, and with such help have succeeded.10 Perhaps a good faith effort will suggest the creation of some such commission or board.
Certainly the federal courts should not again be called on to reapportion Alabama’s legislature. Indeed, the present court-ordered plan, if it proves politically acceptable to the people of Alabama, as we ardently hope, may simplify the Legislature’s task. Most legislative districts may require no more than addition or subtraction of a few building blocks (ED’s) to bring the reapportionment up-to-date. It is, however, too much to hope that the task can really be compared with mere child’s play. Nonetheless, we rest more confident that the reapportioned Alabama Legislature, perhaps with the aid of like commissions or boards employed by some other states, will be able to and will actually accomplish all future reapportionments of the Alabama Legislature.
. In a declaratory judgment action, the burden of proof is on the party seeking relief. Reliance Life Ins. Co. v. Burgess, 8th Cir. 1940, 112 F.2d 234, 237, 238, cert. denied 311 U.S. 699, 61 S.Ct. 137, 85 L.Ed. 453, rehearing denied 311 U.S. 730, 61 S.Ct. 391, 85 L.Ed. 475 ; 22 Am.Jur.2nd, Evidence § 97.
. Even if the explicit mandate of section 5 of the Voting Rights Act were not applicable to the present case, the history of racial gerrymandering in Alabama would, like the history of de jure segregation, create a presumption that defendants’ plan is discriminatory and impose upon the state the burden of proving that its present plan, unlike past plans, does not dilute minority votes. Cf. Keyes v. School District No. 1, Denver, Colo., 1973, 413 U.S. 189, 93 S.Ct. 2686, 37 L.Ed.2d 548.
. The portion of Klahr upon which we rely is dictum. However, the underlying principle expressed in Klahr — that true population figures must form the bedrock of any constitutional reapportionment plan — has peculiar relevance to the present case because of the unusually inexact estimation procedures employed.
. These house counts ignored foundationless mobile homes and some houses on private roads, while failing to identify abandoned dwellings.
. Gaffney v. Cummings, supra.
. Mahan v. Howell, supra.
. Reynolds v. Sims, 1964, 377 U.S. 533, 563, 84 S.Ct. 1362, 12 L.Ed.2d 506.
. The precise deviation legitimatized by the Supreme Court in White was 9.9%.
. In finding the 16.4% deviation incorporated in the Virginia legislative reapportionment plan acceptable, the Court stated :
“The 16-odd percent maximum deviation which the District Court found to exist in the legislative plan for the reapportionment of the House is substantially less than the percentage deviations which have been found invalid in the previous decisions of this Court. While this percentage may well approach tolerable limits, we do not believe it exceeds them. Virginia has not sacrificed substantial equality to justifiable deviations.”
Mahan, supra.
. See Supreme Court opinions in the Virginia ease, Mahan v. Howell, supra, in the Connecticut ease, Gaffney v. Cummings, supra, and in the Texas case, White v. Regester, supra.