[ Footnote * ] Page I Together with No. 94-1631, Oklahoma v. City of New York et al., and No. 94-1985, Department of Commerce et al. v. City of New York et al., also on certiorari to the same court.
The Constitution's Census Clause vests Congress with the responsibility to conduct an "actual Enumeration" of the American public every 10 years, with the primary purpose of providing a basis for apportioning congressional representation among the States. That responsibility has been delegated to the Secretary of Commerce, who determined that an "actual Enumeration" would best be achieved in the 1990 census by not using a postenumeration survey (PES) statistical adjustment designed to correct an undercount in the initial enumeration. In this action brought by several of the respondents and others, the District Court concluded that the Secretary's decision not to statistically adjust the census violated neither the Constitution nor federal law. In reversing and remanding, the Court of Appeals looked to a line of precedent involving judicial review of intrastate districting decisions, see Wesberry v. Sanders, 376 U.S. 1 , and its progeny, and held, inter alia, that a heightened standard of review was required here because the Secretary's decision impacted the fundamental right to have one's vote counted and had a disproportionate impact upon certain identifiable minority racial groups.
Because it was reasonable to conclude that an "actual Enumeration" could best be achieved in the 1990 census without the PES-based statistical adjustment, the Secretary's decision Page II not to use that adjustment was well within the constitutional bounds of discretion over the conduct of the census that is provided to the Federal Government. Pp. 11-23.
REHNQUIST, C.J., delivered the opinion for a unanimous Court. [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 1]
CHIEF JUSTICE REHNQUIST delivered the opinion of the Court.
In conducting the 1990 United States Census, the Secretary of Commerce decided not to use a particular statistical adjustment that had been designed to correct an undercount in the initial enumeration. The Court of Appeals for the Second Circuit held that the Secretary's decision was subject to heightened scrutiny because of its effect on the right of individual respondents to have their vote counted equally. We hold that the Secretary's decision was not subject to heightened scrutiny, and that it conformed to applicable constitutional and statutory provisions.
The Constitution requires an "actual Enumeration" of [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 2] the population every 10 years and vests Congress with the authority to conduct that census "in such Manner as they shall by Law direct." 1 Art. I, 2, cl. 3. Through the Census Act, 13 U.S.C. 1 et seq., Congress has delegated to the Secretary of the Department of Commerce the responsibility to take "a decennial census of the population . . . in such form and content as he may determine. . . ." 141(a). The Secretary is assisted in the performance of that responsibility by the Bureau of the Census and its head, the Director of the Census. See 2; 21 ("[The] Director shall perform such duties as may be imposed upon him by law, regulations, or orders of the Secretary").
The Constitution provides that the results of the census shall be used to apportion the Members of the House of Representatives among the States. See Art. I, 2, cl. 3 ("Representatives . . . shall be apportioned among the several States . . . according to their respective Numbers. . ."); Amdt. XIV, 2 ("Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State. . ."). Because the Constitution provides that the number of Representatives apportioned to each State determines in part the allocation to each State of votes for the election of the President, the decennial census also affects the allocation of members of the electoral college. See Art. II, 1, cl. 2 ("Each State shall appoint . . . a Number of Electors, equal to the whole Number of Senators and Representatives to which the State may be entitled in the Congress. . ."). Today, census data also have important consequences not delineated in the Constitution: The [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 3] Federal Government considers census data in dispensing funds through federal programs to the States, and the States use the results in drawing intrastate political districts.
There have been 20 decennial censuses in the history of the United States. Although each was designed with the goal of accomplishing an "actual Enumeration" of the population, no census is recognized as having been wholly successful in achieving that goal. 2 Cf. Karcher v. Daggett, 462 U.S. 725, 732 (1983) (recognizing that "census data are not perfect," and that "population counts for particular localities are outdated long before they are completed"); Gaffney v. Cummings, 412 U.S. 735, 745 (1973) (census data "are inherently less than absolutely accurate"). Despite consistent efforts to improve the quality of the count, errors persist. Persons who should have been counted are not counted at all or are counted at the wrong location; persons who should not have been counted (whether because they died before or were born after the decennial census date, because they were not a citizen of the country, or because they did not exist) are counted; and persons who should have been counted only once are counted twice. It is thought that these errors have resulted in a net "undercount" of the actual American population in every decennial census. In 1970, for instance, the Census Bureau concluded that the census results were 2.7% lower than the actual population. 3 Brief for Respondents 12. [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 4]
The undercount is not thought to be spread consistently across the population: Some segments of the population are "undercounted" to a greater degree than are others, resulting in a phenomenon termed the "differential undercount." Since at least 1940, the Census Bureau has thought that the undercount affects some racial and ethnic minority groups to a greater extent than it does whites. In 1940, for example, when the undercount for the entire population was 5.4%, the undercount for blacks was estimated at 8.4%, (and the undercount for whites at 5.0%). Brief for Respondents 12. The problem of the differential undercount has persisted even as the census has come to provide a more numerically accurate count of the population. In the 1980 census, for example, the overall undercount was estimated at 1.2%, and the undercount of blacks was estimated at 4.9%. Brief for Respondents 12.
The Census Bureau has recognized the undercount and the differential undercount as significant problems, and in the past has devoted substantial effort toward achieving their reduction. Most recently, in its preparations for the 1990 census, the Bureau initiated an extensive inquiry into various means of overcoming the impact of the undercount and the differential undercount. As part of this effort, the Bureau created two task forces: the Undercount Steering Committee (USC), responsible for planning undercount research and policy development; and the Undercount Research Staff [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 5] (URS), which conducted research into various methods of improving the accuracy of the census. In addition, the Bureau consulted with state and local governments and various outside experts and organizations.
Largely as a result of these efforts, the Bureau adopted a wide variety of measures designed to reduce the rate of error in the 1990 enumeration, including an extensive advertising campaign, a more easily completed census questionnaire, and increased use of automation, which among other things facilitated the development of accurate maps and geographic files for the 1990 census. Pet. App. 321a-322a. 4 The Bureau also implemented a number of improvements specifically targeted at eliminating the differential undercount; these included advertising campaigns developed by and directed at traditionally undercounted populations and expanded questionnaire assistance operations for non-English speaking residents. Id., at 321a-322a.
In preparing for the 1990 census, the Bureau and the task forces also looked into the possibility of using large-scale statistical adjustment to compensate for the undercount and differential undercount. Although the Bureau had previously considered that possibility (most recently in 1980), it always had decided instead to rely upon more traditional methodology and the results of the enumeration. See Cuomo v. Baldrige, 674 F. Supp. 1089 (SDNY 1987) (noting that Bureau rejected large-scale statistical adjustment of 1980 census). In 1985, preliminary investigations by the URS suggested that the most promising method of statistical adjustment was the "capture-recapture" or "dual system estimation" approach.
The particular variations of the "dual system estimation" [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 6] considered by the Bureau are not important for purposes of this opinion, but an example may serve to make the "dual system estimation" more understandable. Imagine that one wanted to use DSE in order to determine the number of pumpkins in a large pumpkin patch. First, one would choose a particular section of the patch as the representative subset to which the "recapture" phase will be applied. Let us assume here that it is a section exactly 1/10 the size of the entire patch that is selected. Then, at the next step - the "capture" stage - one would conduct a fairly quick count of the entire patch, making sure to record both the number of pumpkins counted in the entire patch and the number of pumpkins counted in the selected section. Let us imagine that this stage results in a count of 10,000 pumpkins for the entire patch and 1,000 pumpkins for the selected section. Next, at the "recapture" stage, one would perform an exacting count of the number of pumpkins in the selected section. Let us assume that we now count 1,100 pumpkins in that section. By comparing the results of the "capture" phase and the results of the "recapture" phase for the selected section, it is possible to estimate that approximately 100 pumpkins actually in the patch were missed for every 1,000 counted at the "capture" phase. Extrapolating this data to the count for the entire patch, one would conclude that the actual number of pumpkins in the patch is 11,000.
In the context of the census, the initial enumeration of the entire population (the "capture") would be followed by the post-enumeration survey (PES) (the "recapture") of certain representative geographical areas. The Bureau would then compare the results of the PES to the results of the initial enumeration for those areas targeted by the PES, in order to determine a rate of error in those areas for the initial enumeration, (i.e., the rate at which the initial enumeration undercounted people [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 7] in those areas). That rate of error would be extrapolated to the entire population, and thus would be used to statistically adjust the results of the initial enumeration.
The URS thought that the PES also held some promise for correcting the differential undercount. The PES would be conducted through the use of a system called post-stratification. Thus, each person counted through the PES would be placed into one, and only one, of over 1,000 post-strata defined by five categories: geography; age; sex; status of housing unit (rent v. own); and race (including Hispanic versus non-Hispanic origin). 5 By comparing the post-stratified PES data to the results of the initial enumeration, the Bureau would be able to estimate not only an overall undercount rate, but also an undercount rate for each post-strata. Hence, the statistical adjustment of the census could reflect differences in the undercount rate for each post-strata.
Through the mid-1980's, the Bureau conducted a series of field tests and statistical studies designed to measure the utility of the PES as a tool for adjusting the census. The Director of the Bureau decided to adopt a PES-based adjustment, and in June 1987, he informed his superiors in the Department of Commerce of that decision. The Secretary of Commerce disagreed with the Director's decision to adjust, however, and in October 1987, the Department of Commerce announced that the 1990 Census would not be statistically adjusted. [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 8]
In November 1988, several plaintiffs (including a number of the respondents in this case) brought suit in the United States District Court for the Eastern District of New York, arguing that the Secretary's decision against statistical adjustment of the 1990 census was unconstitutional and contrary to federal law. The parties entered into an interim stipulation providing, inter alia, that the Secretary would reconsider the possibility of a statistical adjustment.
In July 1991, the Secretary issued his decision not to use the PES to adjust the 1990 census. Pet. App. 135a-415a. The Secretary began by noting that large-scale statistical adjustment of the census through the PES would "abandon a two hundred year tradition of how we actually count people." Id., at 138a. Before taking a "step of that magnitude," he held, it was necessary to be "certain that it would make the census better and the distribution of the population more accurate." Ibid. Emphasizing that the primary purpose of the census was to apportion political representation among the States, the Secretary concluded that "the primary criterion for accuracy should be distributive accuracy - that is, getting most nearly correct the proportions of people in different areas." Id., at 146a-147a.
After reviewing the recommendations of his advisors and the voluminous statistical research that had been compiled, the Secretary concluded that although numerical accuracy (at the national level) might be improved through statistical adjustment, he could not be confident that the distributive accuracy of the census - particularly at the state and local level - would be improved by a PES-based adjustment. 6 Id., at 140a-141a, 200a-201a. [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 9] In particular, the Secretary noted, the adjusted figures became increasingly unreliable as one focused upon smaller and smaller political subdivisions. Id., at 142a.
The Secretary stated that his decision not to adjust was buttressed by a concern that adjustment of the 1990 census might present significant problems in the future. Id., at 143a. Because small changes in adjustment methodology would have a large impact upon apportionment - an impact that could be determined before a particular methodology was chosen - the Secretary found that statistical adjustment of the 1990 census might open the door to political tampering in the future. The Secretary also noted that statistical adjustment might diminish the incentive for state and local political leaders to assist in the conduct of the initial enumeration. See id., at 143a-144a. In conclusion, the Secretary stated that the Bureau would continue its research into the possibility of statistical adjustment of future censuses, and would maintain its efforts to improve the accuracy and inclusiveness of the initial enumeration. Id., at 145a.
The plaintiffs returned to court. The District Court concluded that the Secretary's decision violated neither the Constitution nor federal law. See New York v. United States Dept. of Commerce, 822 F. Supp. 906 (EDNY 1993).
Respondents appealed, arguing that the District Court had adopted the wrong standard of review for their constitutional claim, 7 and the Court of Appeals for the [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 10] Second Circuit reversed by a divided vote. 34 F.3d 1114 (1994); Pet. App. 1a-40a. The majority looked to a line of precedent involving judicial review of intrastate districting decisions, see, e.g., Karcher v. Daggett, 462 U.S. 725 (1983); Wesberry v. Sanders, 376 U.S. 1 (1964), and found that a heightened standard of review was required here both because the Secretary's decision impacted a fundamental right, viz., the right to have one's vote counted, and because the decision had a disproportionate impact upon certain identifiable minority racial groups. Pet. App. 33a. The court then held that the plaintiffs had shown that the Secretary had failed to make a good-faith effort to achieve equal districts as nearly as possible, id., at 38a, and therefore that the defendants must bear the burden of proving that population deviations were necessary to achieve some legitimate state goal, id., at 39a-40a. The court remanded for an inquiry into whether the Secretary could show that the decision not to adjust was essential for the achievement of a legitimate governmental objective. Id., at 40a.
The dissenting judge stated that he would have affirmed based upon the decision of the District Court. See ibid. He also noted that the majority's decision created a conflict with two other decisions of the Courts of Appeals. See Detroit v. Franklin, 4 F.3d 1367 (CA6 1993), and Tucker v. United States Dept. of Commerce, 958 F.2d 1411 (CA7 1992).
Wisconsin, Oklahoma, and the United States each filed a petition for certiorari. We granted those petitions, and consolidated them for argument. 515 U.S. ___ (1995). We now reverse.
In recent years, we have twice considered constitutional challenges to the conduct of the census. In Department of Commerce v. Montana, 503 U.S. 442 (1992), the [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 11] State of Montana, several state officials, and Montana's Members of Congress brought suit against the Federal Government, challenging as unconstitutional the method used to determine the number of Representatives to which each State is entitled. A majority of a three-judge District Court looked to the principle of equal representation for equal numbers of people that was applied to intrastate districting in Wesberry v. Sanders, supra, and held it applicable to congressional apportionment of seats among the States. Noting a significant variance between the population of Montana's single district and the population of the "ideal district," the court found that Congress' chosen method of apportionment violated the principle of Wesberry, and therefore voided the federal statute providing the method of apportionment.
In an unanimous decision, this Court reversed. We began by revisiting Wesberry, a case in which the Court held unconstitutional wide disparities in the population of congressional districts drawn by the State of Georgia. Montana, supra, at 459-460. We recognized that the principle of Wesberry - "`equal representation for equal numbers of people'" - had evolved though a line of cases into a strictly enforced requirement that a State "`make a good-faith effort to achieve precise mathematical equality'" among the populations of congressional districts. See Montana, supra, at 460, quoting Kirkpatrick v. Preisler, 394 U.S. 526, 530 -531 (1969) (disparities between congressional districts in Missouri held unconstitutional); see also Karcher v. Daggett, supra, (1% disparity between population of New Jersey districts held unconstitutional). Returning to Montana's challenge to Congress' apportionment decision, we noted that the Wesberry line of cases all involved intrastate disparities in the population of voting districts that had resulted from a State's redistricting decisions, whereas Montana had challenged interstate disparities resulting [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 12] from the actions of Congress. Montana, supra, at 460.
We found this difference to be significant beyond the simple fact that Congress was due more deference than the States in this area. Wesberry required a State to make "a good-faith effort to achieve precise mathematical equality" in the size of voting districts. Kirkpatrick, supra, at 530-531. While this standard could be applied easily to intrastate districting because there was no "theoretical incompatibility entailed in minimizing both the absolute and the relative differences" in the sizes of particular voting districts, we observed that it was not so easily applied to interstate districting decisions where there was a direct trade-off between absolute and relative differences in size. Montana, supra, at 461-462. Finding that Montana demanded that we choose between several measures of inequality in order to hold the Wesberry standard applicable to congressional apportionment decisions, we concluded that "[n]either mathematical analysis nor constitutional interpretation provide[d] a conclusive answer" upon which to base that choice. Montana, supra, at 463.
We further found that the Constitution itself, by guaranteeing a minimum of one representative for each State, made it virtually impossible in interstate apportionment to achieve the standard imposed by Wesberry. Montana, supra, at 463. In conclusion, we recognized the historical pedigree of the challenged method of apportionment, and reemphasized that Congress' "good-faith choice of a method of apportionment of Representatives among the several States `according to their respective Numbers' commands far more deference than a state districting decision that is capable of being reviewed under a relatively rigid mathematical standard." Montana, supra, at 464.
In Franklin v. Massachusetts, 505 U.S. 788 (1992), we reiterated our conclusion that the Constitution vests Congress with wide discretion over apportionment [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 13] decisions and the conduct of the census. In Franklin, the State of Massachusetts and two of its registered voters sued the Federal Government, arguing that the method used by the Secretary to count federal employees serving overseas was (among other things) unconstitutional. Restating the standard of review established by Montana, we examined the Secretary's decision in order to determine whether it was "consistent with the constitutional language and the constitutional goal of equal representation." See Franklin, supra, at 804; Montana, supra, at 459. After a review of the historical practice in the area, we found that the plaintiffs had not met their burden of proving that a decision contrary to that made by the Secretary would "make representation . . . more equal." Franklin, 505 U.S., at 806 . Concluding that the Secretary's decision reflected a "judgment, consonant with, though not dictated by, the text and history of the Constitution . . .," we held the Secretary's decision to be well within the constitutional limits on his discretion. Ibid.
In its decision in this case, the Court of Appeals found that a standard more strict than that established in Montana and Franklin should apply to the Secretary's decision not to statistically adjust the census. The court looked to equal protection principles distilled from the same line of state redistricting cases relied upon by the plaintiffs in Montana, and found that both the nature of the right asserted by respondents - the right to have one's vote counted equally - and the nature of the affected classes - "certain identifiable minority groups" - required that the Secretary's decision be given heightened scrutiny. Pet. App. 33a. The court drew from the District Court's decision "implicit" findings: that the census did not achieve equality of voting power as nearly as practicable; "that for most purposes and for most of the population [the PES-based] adjustment would result in a more accurate count than the original [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 14] census; and that the adjustment would lessen the disproportionate undercounting of minorities." Id., at 34a.
The court recognized two significant differences between the intrastate districting cases and the instant case: first, that this case involves the federal rather than a state government; and second, that constitutional requirements make it impossible to achieve precise equality in voting power nationwide. Id., at 34a-35a. But it found these differences nondeterminative, deciding that no deference was owed to the Executive Branch on a question of law, and that the "impossibility of achieving precise mathematical equality is no excuse for [the Federal Government] not making [the] mandated good-faith effort." Id., at 35a. The court found that the respondents here had established a prima facie violation of the Wesberry standard both by showing that the PES-based adjustment would increase numerical accuracy, and by virtue of the fact that "the differential undercount in the 1990 enumeration was plainly foreseeable and foreseen." Pet. App. 38a-39a. The court held that the Secretary's decision would have to be vacated as unconstitutional unless on remand he could show that the decision not to adjust "(a) furthers a governmental objective that is legitimate, and (b) is essential for the achievement of that objective." Id., at 40a.
We think that the Court of Appeals erred in holding the "one person-one vote" standard of Wesberry and its progeny applicable to the case at hand. For several reasons, the "good-faith effort to achieve population equality" required of a State conducting intrastate redistricting does not translate into a requirement that the Federal Government conduct a census that is as accurate as possible. First, we think that the Court of Appeals understated the significance of the two differences that it recognized between state redistricting cases [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 15] and the instant case. The court failed to recognize that the Secretary's decision was made pursuant to Congress' direct delegation of its broad authority over the census. See Art. I, 2, cl. 3 (Congress may conduct the census "in such Manner as they shall by Law direct.") The court also undervalued the significance of the fact that the Constitution makes it impossible to achieve population equality among interstate districts. As we have noted before, the Constitution provides that "[t]he number of Representatives shall not exceed one for every 30,000 persons; each State shall have at least one Representative; and district boundaries may not cross state lines." Montana, 503 U.S., at 447 -448.
While a court can easily determine whether a State has made the requisite "good-faith effort" toward population equality through the application of a simple mathematical formula, we see no way in which a court can apply the Wesberry standard to the Federal Government's decisions regarding the conduct of the census. The Court of Appeals found that Wesberry required the Secretary to conduct a census that "would achieve population equality," which it understood to mean a census that was as accurate as possible. Pet. App. 34a. But in so doing, the court implicitly found that the Constitution prohibited the Secretary from preferring distributive accuracy to numerical accuracy, and that numerical accuracy - which the court found to be improved by a PES-based adjustment - was constitutionally preferable to distributive accuracy. See id., at 39a (". . . the Secretary did not make the required effort to achieve numerical accuracy as nearly as practicable, . . . the burden thus shifted to the Secretary to justify his decision not to adjust. . . .") As in Montana, where we could see no constitutional basis upon which to choose between absolute equality and relative equality, so here can we see no ground for preferring numerical [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 16] accuracy to distributive accuracy, or for preferring gross accuracy to some particular measure of accuracy. The Constitution itself provides no real instruction on this point, and extrapolation from our intrastate districting cases is equally unhelpful. Quite simply, "[t]he polestar of equal representation does not provide sufficient guidance to allow us to discern a single constitutionally permissible course." Montana, supra, at 463.
In Montana, we held that Congress' "apparently good-faith choice of a method of apportionment of Representatives among the several States `according to their respective numbers'" was not subject to strict scrutiny under Wesberry. Montana, 503 U.S., at 464 . With that conclusion in mind, it is difficult to see why or how Wesberry would apply to the Federal Government's conduct of the census - a context even further removed from intrastate districting than is congressional apportionment. Congress' conduct of the census, even more than its decision concerning apportionment, "commands far more deference than a state districting decision that is capable of being reviewed under a relatively rigid mathematical standard." 8 Montana, supra, at 464.
Rather than the standard adopted by the Court of Appeals, we think that it is the standard established by this Court in Montana and Franklin that applies to the Secretary's decision not to adjust. The text of the Constitution vests Congress with virtually unlimited [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 17] discretion in conducting the decennial "actual Enumeration," 9 see Art. I, 2, cl. 3, and notwithstanding the plethora of lawsuits that inevitably accompany each decennial census, 10 there is no basis for thinking that Congress' discretion is more limited than the text of the Constitution provides. See also Baldrige v. Shapiro, 455 U.S. 345, 361 (1982) (noting broad scope of Congress' discretion over census). Through the Census Act, Congress has delegated its broad authority over the census to the Secretary. 11 See 13 U.S.C. 141(a) (Secretary shall take "a decennial census of [the] population . . . in such form and content as he may determine. . . .") Hence, so long as the Secretary's conduct of the census is "consistent with the constitutional language and the constitutional goal of equal representation," [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 18] Franklin, 505 U.S., at 804 , it is within the limits of the Constitution. In light of the Constitution's broad grant of authority to Congress, the Secretary's decision not to adjust need bear only a reasonable relationship to the accomplishment of an actual enumeration of the population, keeping in mind the constitutional purpose of the census.
In 1990, the Census Bureau made an extraordinary effort to conduct an accurate enumeration, and was successful in counting 98.4% of the population. See 58 Fed. Reg. 70 (1993); Brief for Federal Petitioners 28. The Secretary then had to consider whether to adjust the census using statistical data derived from the PES. He based his decision not to adjust the census upon three determinations. First, he held that in light of the constitutional purpose of the census, its distributive accuracy was more important than its numerical accuracy. Second, he determined that the unadjusted census data would be considered the most distributively accurate absent a showing to the contrary. And finally, after reviewing the results of the PES in light of extensive research and the recommendations of his advisors, the Secretary found that the PES-based adjustment would not improve distributive accuracy. Each of these three determinations is well within the bounds of the Secretary's constitutional discretion.
As we have already seen, supra, at ___, the Secretary's decision to focus on distributive accuracy is not inconsistent with the Constitution. Indeed, a preference for distributive accuracy (even at the expense of some numerical accuracy) would seem to follow from the constitutional purpose of the census, viz., to determine the apportionment of the Representatives among the States. Respondents do not dispute this point. See Brief for Respondents 54 ("Distributive accuracy is an appropriate criterion for judging census accuracy because it calls attention to a concern with the uses to which [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 19] census data are put"). Rather, they challenge the Secretary's first determination by arguing that he improperly "regarded evidence of superior numeric accuracy as `not relevant' to the determination of distributive accuracy." Id., at 39 (quoting Pet. App. 201a); see also Brief for Respondents 51-54. In support of this argument, respondents note that an enumeration that results in increased numerical accuracy will also result in increased distributive accuracy.
We think that respondents rest too much upon the statement by the Secretary to which they refer. When quoted in full, the statement reads: "[w]hile the preponderance of the evidence leads me to believe that the total population at the national level falls between the census counts and the adjusted figures, that conclusion is not relevant to the determination of distributive accuracy." Pet App. 201a. In his decision, the Secretary found numerical accuracy (in addition to distributive accuracy) to be relevant to his decision whether to adjust. See id., at 157a. Even if the Secretary had chosen to subordinate numerical accuracy, we are not sure why the fact that distributive and numerical accuracy correlate closely in an improved enumeration would require the Secretary to conclude that they correlate also for a PES-based statistical adjustment.
Turning to the Secretary's second determination, we previously have noted, and respondents do not dispute, the importance of historical practice in this area. See Franklin, 505 U.S., at 803 -806 (noting importance of historical experience in conducting the census); cf. Montana, 503 U.S., at 465 ("To the extent that the potentially divisive and complex issues associated with apportionment can be narrowed by the adoption of both procedural and substantive rules that are consistently applied year after year, the public is well served. . . .") Nevertheless, respondents challenge the Secretary's second determination by arguing that his understanding [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 20] of historical practice is flawed. According to respondents, the Secretary assumed that the census traditionally was conducted via a simple "headcount," thereby ignoring the fact that statistical adjustment had been used in both the 1970 and 1980 census. See Brief for Respondents 4-5.
We need not tarry long with this argument. The Secretary reasonably recognized that a PES-based statistical adjustment would be a significant change from the traditional method of conducting the census. The statistical adjustments in 1970 and 1980 to which respondents refer were of an entirely different type than the adjustment considered here, and they took place on a dramatically smaller scale. See Cuomo v. Baldrige, 674 F. Supp., at 1107 (rejecting argument that Secretary had to conduct PES-like statistical adjustment of 1980 census and finding that "none of [the] adjustments in 1970 were even remotely similar to the types of wholesale adjustments presently suggested. . . .") Moreover, the PES-based adjustment would have been the first time in history that the States' apportionment would have been based upon counts in other States. See Pet. App. 251a-252a. Here, the Secretary's understanding of the traditional method of conducting the census was well-founded, as was his establishment of a rebuttable presumption that the traditional method was the most accurate.
The Secretary ultimately determined that the available evidence "tends to support the superior distributive accuracy of the actual enumeration," id., at 185a, and it is this determination at which respondents direct the brunt of their attack. Respondents contend that the Secretary's review of the evidence is due no deference from this Court. They argue that the Secretary's decision is not the sort of "highly technical" administrative decision which normally commands judicial deference, and that regardless of its technical complexity, the [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 21] Secretary's review of the evidence presents a constitutional issue that deserves no deference. Respondents contend that the Secretary's review of the evidence is of dubious validity because the Secretary is admittedly "not a statistician," id., at 139a, and because his conclusion is at odds with that of the Director of the Census. According to respondents, we should carefully comb the Secretary's decision in order to review his conclusions de novo.
Respondents' argument fundamentally misapprehends the basis for our deference to the Secretary's determination that the adjusted census results do not provide a more distributively accurate count of the population. Our deference arises not from the highly technical nature of his decision, but rather from the wide discretion bestowed by the Constitution upon Congress, and by Congress upon the Secretary. Regardless of the Secretary's statistical expertise, it is he to whom Congress has delegated its constitutional authority over the census. For that same reason, the mere fact that the Secretary's decision overruled the views of some of his subordinates is by itself of no moment in any judicial review of his decision.
Turning finally to review the Secretary's conclusion that the PES adjustment would not improve distributive accuracy, we need note only that the Secretary's conclusion is supported by the reasoning of some of his advisors, and was therefore a reasonable choice in an area where technical experts disagree. Cf. Tucker v. United States Dept of Commerce, 958 F.2d, at 1418 (Plaintiffs seeking PES-based statistical adjustment "are asking [courts] to take sides in a dispute among statisticians, demographers, and census officials concerning the desirability of making a statistical adjustment to the census headcount"). The Under Secretary of Commerce for Economic Affairs and the Administrator of the Economics and Statistics Administration both voted [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 22] against adjustment. Pet. App. 59a, 140a. Moreover, even those who recommended in favor of adjustment recognized that their conclusion was not compelled by the evidence: The Director of the Census Bureau, upon whose recommendation the respondents heavily rely, stated in her report to the Secretary that "[a]djustment is an issue about which reasonable men and women and the best statisticians and demographers can disagree." App. 73. And one of the principal statisticians at the Bureau, Dr. Robert E. Fay, "`told the Secretary that . . . reasonable statisticians could differ'" on the question of adjustment. Pet. App. 91a. Therefore, and because we find the Secretary's two prior determinations as well to be entirely reasonable, we conclude that his decision not to adjust the 1990 census was "consonant with, . . . the text and history of the Constitution. . . ." Franklin, 505 U.S., at 806 .
The Constitution confers upon Congress the responsibility to conduct an "actual Enumeration" of the American public every 10 years, with the primary purpose of providing a basis for apportioning political representation among the States. Here, the Secretary of Commerce, to whom Congress has delegated its constitutional authority over the census, determined that in light of the constitutional purpose of the census, an "actual Enumeration" would best be achieved without the PES-based statistical adjustment of the results of the initial enumeration. We find that conclusion entirely reasonable. Therefore we hold that the Secretary's decision was well within the constitutional bounds of discretion over the conduct of the census provided to the Federal Government. The judgment of the Court of Appeals is
[ Footnote 2 ] Indeed, even the first census did not escape criticism. Thomas Jefferson, who oversaw the conduct of that census in 1790 as Secretary of State, was confident that it had significantly undercounted the young Nation's population. See C. Wright, History and Growth of the United States Census 16-17 (1900).
[ Footnote 3 ] One might wonder how the Census Bureau is able to determine whether there is an undercount and its size. Specifically: against what standard are the census results measured? After all, if the actual population of the United States is known, then the conduct of the census [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 4] would seem wholly redundant.
For the most part, we are told, the size of the error in a particular census is determined by comparing the census results not with some definite and established measure of the population, but rather with estimates of the population developed from demographic data. See Pet. App. in 94-1614, pp. 158a-168a, 366a-369a. A similar procedure traditionally has been used to determine the size and make-up of the differential undercount, see infra.
[ Footnote 4 ] All references to Pet. App. are to the appendix to the petition for certiorari in No. 94-1614 unless otherwise noted.
[ Footnote 5 ] Examples of post-strata actually used include: female blacks between the ages of 20 and 29 who owned a home in either Detroit or Chicago; non-black non-Hispanic females, age 45-64, living in owned or rented housing in a non-metropolitan area with a population of 10,000 or more in Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, or Wyoming; and male Asians or Pacific Islanders, age 65 or above, renting a home in either the Los Angeles-Long Beach area or another central city in a metropolitan area in Alaska, California, Hawaii, Oregon, or Washington.
[ Footnote 6 ] The distinction between distributive and numerical accuracy becomes clear with an example. Imagine that the Bureau somehow were able to determine definitely that the census had failed to count exactly 10 [ WISCONSIN v. CITY OF NEW YORK, ___ U.S. ___ (1996) , 9] million people nationwide. If those 10 million "persons" were added to the Nation's total population, and all 10 million were allocated to one particular State, then the numerical accuracy of the census would be improved, but the distributive accuracy would almost certainly be significantly impaired.
[ Footnote 7 ] Respondents did not appeal the District Court's treatment of their statutory claims.
[ Footnote 8 ] Nor do we think that strict scrutiny applies here for some other reason. Strict scrutiny of a classification affecting a protected class is properly invoked only where a plaintiff can show intentional discrimination by the Government. Washington v. Davis, 426 U.S. 229, 239 -245 (1976). Respondents here have not argued that the Secretary's decision not to adjust was based upon an intent to discriminate on the basis of race. Indeed, in light of the Government's extraordinary efforts to include traditionally undercounted minorities in the 1990 census, see Pet. App. 78a, 321a-322a, we think that respondents here would have had a tough row to hoe had they set out to prove intentional discrimination by the Secretary.
[ Footnote 9 ] We do not decide whether the Constitution might prohibit Congress from conducting the type of statistical adjustment considered here. See Brief for Wisconsin 40-42.
[ Footnote 10 ] See, e. g., Franklin v. Massachusetts, 505 U.S. 788, 790 (1992) ("As one season follows another, the decennial census has again generated a number of reapportionment controversies."); National Law Center on Homelessness and Poverty v. Brown, appeal pending, No. 94-5312 (CADC) (argued Oct. 6, 1995) (challenging Census Bureau's procedures for finding and counting homeless persons); Carey v. Klutznick, 637 F.2d 834 (CA2 1980) (seeking order directing Census Bureau to adopt certain processes for counting persons); Borough of Bethel Park v. Stans, 449 F.2d 575 (CA3 1971).
[ Footnote 11 ] We do not here decide the precise bounds of the authority delegated to the Secretary through the Census Act. First, because no party here has suggested that Congress has, in its delegation of authority over the conduct of the census to the Secretary, constrained the Secretary's authority to decide not to adjust the census, we assume here that the Secretary's discretion not to adjust the census is commensurate with that of Congress. See Brief for Petitioner in 94-1614, p. 24 n. 19 (stating that "Congress did not enact any . . . legislation . . . to compel . . . statistical adjustment" of the 1990 Census.) Second, although Oklahoma argues that Congress has constrained the Secretary's discretion to statistically adjust the decennial census, see 13 U.S.C. 195, we need not decide that question in order to resolve this case.