THE TWENTY-SECOND DECENNIAL CENSUS

     Neither the Enumeration Clause of the Constitution nor the Census Act precludes the Bureau of the Census from statistically adjusting "headcounts" in the decennial census for the year 2000 or conducting the non-response follow-up on a sample basis.

     The provision in the Census Act prohibiting sampling for purposes of apportionment of the House of Representatives does not preclude reliance upon statistical adjustments that would improve the accuracy of "headcount" data.

October 7, 1994

MEMORANDUM FOR THE SOLICITOR GENERAL

      You have asked, on behalf of the Department of Commerce, for our advice on the questions whether the use of statistically adjusted census figures would be consistent with the Constitution, U.S. Const. art. I, 2, cl. 3, and with the Census Act, 13 U.S.C. 1-307. The questions arise because the traditional method of taking the census fails to count a significant portion of the population, and in particular disproportionately undercounts identifiable racial and ethnic minorities. In light of these problems, the Department of Commerce is considering the use of statistical adjustments in the twenty-second decennial census (for the year 2000) before the final count is completed in order to improve the accuracy of that census. The Department of Commerce is also considering the use of sampling to conduct the follow-up on households that did not respond to its initial mailing of questionnaires. Accordingly, it desires to know whether such procedures would be lawful. We conclude that both of the proposed changes in conducting the census would be lawful. (*)

I.

      The Constitution "provides the basis for the decennial censuses, but does not specify the details of their administration." Seventeenth Decennial Census, 41 Op. Att'y Gen. 31, 32 (1949). Instead, the Constitution vests in Congress the power to conduct an "actual Enumeration . . . in such Manner as they shall by Law direct." U.S. Const. art. I, 2, cl. 3. Congress's power has in turn been vested in the Bureau of the Census (the "Bureau"), a component of the Department of Commerce. See 13 U.S.C. 2.

     The primary purpose of the decennial census (1) is to provide the basis for Congress's apportionment of seats in the House of Representatives among the States. (2) The census also serves several other legally significant objectives. Historically, the decennial census has been "an enumeration not only of free persons in the States but of free persons in the Territories, and not only an enumeration of persons but the collection of statistics respecting age, sex, and production." Legal Tender Cases, 79 U.S. (12 Wall.) 457, 536 (1870). "The census today serves an important function in the allocation of federal grants to states based on population. In addition, the census also provides important data for Congress and ultimately for the private sector." Baldrige v. Shapiro, 455 U.S. 345, 353 (1982);see generally Note,Demography and Distrust: Constitutional Issues of the Federal Census, 94 Harv. L. Rev. 841, 844-45 (1981).

     The traditional method for conducting the decennial census "is a headcount rather than an estimation based on sampling." Tucker v. United States Dep't of Commerce, 958 F.2d 1411, 1412 (7th Cir.),cert. denied, 506 U.S. 953 (1992). (3) The term "headcount" is somewhat misleading, however. "The census . . . is not a headcount in which each and every person residing in the United States on a given date is counted by the Census Bureau. Rather, it is a survey of the population that through the responses of one member of each household attempts to enumerate the entire population." >Carey v. Klutznick, 508 F. Supp. 420, 426 (S.D.N.Y. 1980),rev'd, 653 F.2d 732 (2d Cir. 1981), cert. denied, 455 U.S. 999 (1982).

     In the 1990 census, the Bureau's tabulation had four phases. First, relying on lists compiled by commercial sources and its own fieldwork, the Bureau derived a mailing list of as many households as it could locate. Second was the "mail out/mail back" phase, in which the Bureau mailed out questionnaires to each household on its list, and requested their return by April 1, 1990. (The return rate was 63%.) The third phase was a follow-up in which the Bureau sent out another round of mailings. The fourth phase comprised efforts by census enumerators, in person, to contact non-responding households (or other reliable sources) to obtain the needed information. Following that, the Bureau undertook "coverage improvement programs" designed to reach non-respondents in other ways, including rechecks of all vacant or uninhabitable housing units, recanvassing of selected blocks, an advertising campaign, checks of parolees and probationers, and a local government review. See City of New York v. United States Dep't of Commerce, 34 F.3d 1114 (2d Cir. 1994),rev'd, 517 U.S. 1 (1996). (4)

      Like earlier censuses, the 1990 census concededly did not count the entire population of the United States. (5) Given the inherent difficulties of census-taking and the existence of financial and time constraints, some degree of inaccuracy in the census count is perhaps inevitable. The Bureau itself believes that "every census has necessarily involved an undercount,"Young v. Klutznick, 497 F. Supp. 1318, 1327 (E.D. Mich. 1980),rev'd, 652 F.2d 617 (6th Cir. 1981),cert. denied, 455 U.S. 939 (1982), and the courts agree that "a perfectly accurate count of upwards of 250 million people" is simply not "feasible." City of Detroit v.Franklin, 4 F.3d at 1377. (6) Far more troubling than the bare existence of an undercount is the fact that the 1990 census perpetuated a pattern, the existence of which has been recognized since 1940, ofdifferentially undercounting African Americans. (7) The 1990 census also differentially undercounted Hispanics: the estimated undercount for that group was 5.2%, as against an estimated undercount of 2.1% for the population at large. (8) The Bureau "specifically acknowledge[d] an undercount in the 1990 census ranging from 1.7 percent of whites to 5.2 percent of Hispanics." (9)

      Despite that acknowledgement, the Secretary of Commerce declined in 1991 to adjust the 1990 census figures to correct for the undercounts. (10) The Secretary's decision not to make the adjustment has been the subject of litigation in three circuits, with conflicting results. CompareTucker (plaintiffs had no judicially enforceable rights) andCity of Detroit (same)withCity of New York (remanding with instruction that refusal to adjust could not be upheld unless shown to be necessary to a legitimate governmental interest).

     The Bureau is currently considering whether to adjust the "raw count" of the next decennial census for the year 2000. Sampling was used in connection with the 1990 census to carry out the "Post-Enumeration Survey" (the "PES") that measured the undercount for that year. SeeCity of New York, 34 F.3d at 1117; David A. Freedman,Adjusting the Census of 1990, 34 Jurimetrics J. 99, 102-03 (1993). In that census, the Bureau tested the accuracy of the count by a PES of some 174,000 households and then matching the questionnaires for households in the PES against the same households in the census (including both mail-backs and non-response follow-ups). The matching process provided the Bureau with data to develop adjusting factors, or "multipliers," to capture the estimated under- or overcount for some 1,392 demographic subgroups. The application of the multipliers to the enumeration data for the subgroups produced the conclusion that 1.6% of the total population had not been counted in the census. For the 2000 census, the Bureau is considering the use of a sample-based adjustment as in 1990, except that it would complete the adjustment before its deadline for reporting State totals to the President.

     The Bureau is also considering whether to conduct the non-response follow-up on a sample basis, rather than sending enumerators to each non-responding household. Specifically, it is proposing to contact, by telephone or in person, between 25% and 50% of the households that failed to return the census questionnaire. The Bureau would then extrapolate from the results of this sample to estimate the whole non-respondent population. The Bureau believes that the use of this procedure would save it between $300 and $600 million. At the same time, it advises us that the procedure would also produce greater accuracy than was achieved in the 1990 census.

      In the past, the Bureau took the position that it would be legally precluded from adjusting the census for apportionment purposes. SeeCensus Undercount Adjustment: Basis for Decision, 45 Fed. Reg. 69,366, 69,371-73 (1980). This claim was based on both constitutional and statutory grounds. First, the Bureau has argued that

interpretation of the phrase "actual enumeration" in Article 1, Section 2, Clause 3 must begin with the words themselves, and that the terms "census" and "enumeration" mean nothing more or less than a headcount. [It] say[s] that the use of the modifier "actual" with the word "enumeration" can only reinforce the conclusion that the framers of the Constitution intended a headcount, and nothing but a headcount. [It] further rel[ies] upon the fact that, with the exception of the 1970 census when imputations were performed which added approximately 4.9 million people, the census has been, since 1790, an actual headcount and nothing more.

Young v. Klutznick, 497 F. Supp. at 1332. The Bureau has also argued in the past that "even if the Constitution does not prohibit an adjustment for apportionment of Representatives, Congress has by statute prohibited such an adjustment." Id. at 1334. We consider these issues in turn.

II.

The Enumeration Clause of the Constitution reads in relevant part as follows:

Representatives . . . shall be apportioned among the several States . . . according to their respective Numbers . . . . The actual Enumeration shall be made within three Years after the first Meeting of the Congress of the United States, and within every subsequent Term of ten Years, in such Manner as they shall by Law direct.

U.S. Const. art. I, 2, cl. 3;see also U.S. Const. amend. XIV, 2 ("Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State . . . .").

      The Enumeration Clause was one facet of the "Great Compromise" at the Constitutional Convention, which provided for equal representation of the States in a Senate, and representation of "the People of the several States" in a House of Representatives. U.S. Const. art. I, 2, cl. 1;see generallyWesberry v. Sanders, 376 U.S. 1, 10-16 (1964);Demography and Distrust, 94 Harv. L. Rev. at 846. Because the Framers "intended that in allocating Congressmen the number assigned to each State should be determined solely by the number of the State's inhabitants . . . [t]he Constitution embodied Edmund Randolph's proposal for a periodic census to ensure 'fair representation of the people'" Wesberry, 376 U.S. at 13-14 (citations omitted).

      Before the first decennial census in 1790, no modern Nation had undertaken a census (although all the States of the United States, with some exceptions in the South, had done so). See Hyman Alterman,Counting People: The Census in History 164 (1969). Thus, when the Framers were apportioning seats in the first House of Representatives, their decisions were the outcome of "conjecture and political compromise: [they] apparently assigned some of the smaller States a number of Representatives not justified by the size of their populations." Memorandum for Wendell L. Wilkie II, General Counsel, Department of Commerce, from Stuart M. Gerson, Assistant Attorney General, Civil Division at 4 (July 9, 1991) (the "Gerson Memorandum"). (11) The Constitution's reference to an "actual Enumeration" must be explained by reference to the Framers' ignorance of the exact size of the population and its distribution among the States: "[w]hen the Constitution speaks of actual enumeration, it speaks of that as opposed toestimates." Young v. Klutznick, 497 F. Supp. at 1332 (emphasis added). Accord Memorandum for Alice Daniel, Assistant Attorney General, Civil Division, from John M. Harmon, Assistant Attorney General, Office of Legal Counsel,Re: Pending Litigation Concerning Statistical Adjustment of 1980 Decennial Census Population Data at 2 (Sept. 25, 1980) (the "Harmon Memorandum") ("the phrase ['actual Enumeration'] was chosen because an accurate population count was essential once the Convention decided, in the Great Compromise, that representation in the House would be apportioned on the basis of population.").

      The proposal for a periodic enumeration of the population originated, as noted above, with Edmund Randolph, as an incident to the Great Compromise. On July 10, Randolph moved a proposal calling for Congress "to cause a census, and estimate to be taken within one year after its first meeting; and every [left blank] years thereafter -- and that the Legisl[ature] arrange the Representation accordingly." James Madison,Notes of Debates in the Federal Convention of 1787, at 265 (Adrienne Koch ed., 1966) (bracketed material added). George Mason spoke in favor of the motion on the next day, declaring that "[h]e did not object to the conjectural ratio which was to prevail in the outset; but considered a Revision from time to time according to some permanent & precise standard as essential to [the] fair representation required in the [first] branch." Id. at 266. Later in the debate, Randolph repeated Mason's point that "the ratio fix[ed] for the [first] meeting [of Congress] was a mere conjecture." Id. at 267. On August 21, Madison repeated that "[t]he last apportionment of Cong[ress], on which the number of Representatives was founded, was conjectural and meant only as a temporary rule till a Census should be established." Id. at 497. Madison also explained inThe Federalist that the provision in Article I, Section 2, Clause 3 of the Constitution for a House of Representatives that would consist of sixty-five members in the First Congress was merely "a temporary regulation," to be revised when the findings of the census of 1790 became known. The Federalist No. 55, at 343 (James Madison) (Clinton Rossiter ed., 1961). (12)

     These discussions make it clear that, in requiring an "actual" enumeration, the Framers meant a set of figures that was not a matter of conjecture and compromise, such as the figures they had themselves provisionally assumed. An "actual" enumeration would instead be based, as George Mason put it, on "some permanent & precise standard." There is no indication that the Framers insisted that Congress adopt a "headcount" as the sole method for carrying out the enumeration, even if later refinements in the metric of populations would produce more accurate measures.

     Furthermore, the Framers left it to Congress to conduct the enumeration "in such Manner as they shall by Law direct." U.S. Const. art. I, 2, cl. 3. That explicit delegation implies that the Framers were willing to allow for innovation in the choice of measuring techniques; and, not surprisingly, "the Census Bureau's unbroken historical practice really has been to use modern knowledge and scientific techniques to get further and further away from simple headcounting." Young v. Klutznick, 497 F. Supp. at 1333. (13) "The result, and not the method, is the important lesson of the historical experience." Harmon Memorandum at 2.

     In addition, Article I, Section 2, Clause 3 of the Constitution was amended by section 2 of the Fourteenth Amendment. Section 2 declares that "Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State, excluding Indians not taxed." U.S. Const. amend. XIV, 2. Further, section 5 confers on Congress the "power to enforce, by appropriate legislation, the provisions of this article." Id. 5. Congress's powers under section 5 have been "equated . . . with the broad powers expressed in the Necessary and Proper Clause, U.S. Const. art. I, 8, cl. 18. 'Correctly viewed, 5 is a positive grant of legislative power authorizing Congress to exercise its discretion in determining whether and what legislation is needed to secure the guarantees of the Fourteenth Amendment.'" Fullilove v. Klutznick, 448 U.S. 448, 476 (1980) (plurality opinion) (quotingKatzenbach v. Morgan, 384 U.S. 641, 651 (1966)). It follows that Congress has broad power to determine how to carry out the apportionment called for by section 2, and to conduct the enumeration on which that apportionment is based. See Massachusetts v. Mosbacher, 785 F. Supp. 230, 253 (D. Mass.) (three-judge court) ("the exercise of Section 5 powers here in defining the methodology for reapportionment falls squarely within the settled recognition of the competence of Congress as a legislative fact finder"),rev'd sub nom.Franklin v. Massachusetts, 505 U.S. 788 (1992). It would be strange indeed to suppose that Congress -- or its delegate, the Bureau -- lacked the power to authorize a statistical adjustment that would correct the persistent and acknowledged undercounting of African Americans in that enumeration, particularly in view of the fact that the Fourteenth Amendment was primarily intended for the protection of that class. SeeStrauder v. West Virginia, 100 U.S. 303, 306 (1880).

     Finally, constitutional plaintiffs injured by the decision to use adjusted census data for apportionment might argue that so sharp a departure from the Bureau's longstanding practices was unjustified. (14) SeeSenate of California, 968 F.2d at 978 ("the method by which the Secretary is to do the count . . . is generally expected to be a head count");see alsoSeventeenth Decennial Census, 41 Op. Att'y Gen. at 34 (if the Director "has consistently followed the practice in question over a long period of time, and it has not been challenged in the Congress or elsewhere . . . his interpretation ought not to be disturbed except for very weighty reasons"). (15) It could be contended that the use of unadjusted "headcounts" almost invariably since the first census of 1790 represents a practical construction of the Enumeration Clause which the Executive, at least absent weighty reasons, may not reverse. See, e.g.,Smiley v. Holm, 285 U.S. 355, 369 (1932) ("long and continuous interpretation in the course of official action under the law may aid in removing doubts as to its meaning. This is especially true in the case of constitutional provisions governing the exercise of political rights . . . ."); The Pocket Veto Case, 279 U.S. 655, 688-90 (1929). We believe, however, that the change in the Bureau's policy would be upheld against an attack of this nature if there were adequate proof that statistical adjustments would be feasible and would generate more accurate counts of both the total population and of minorities.

      Thus, inFranklin v. Massachusetts, the Court upheld the Bureau's changed policy of allocating overseas government personnel to the several states for residence purposes for the 1990 census. The Court stated that

the Secretary of Commerce made a judgment, consonant with, though not dictated by, the text and history of the Constitution, that many federal employees temporarily stationed overseas had retained their ties to the States and could and should be counted toward their States' representation in Congress . . . . The Secretary's judgment does not hamper the underlying constitutional goal of equal representation, but, assuming that employees temporarily stationed abroad have indeed retained their ties to their home States, actually promotes equality.

505 U.S. at 806.

      In the present case, the validity of the policy change would turn largely on the evidentiary showing that the use of statistical adjustments will produce a more accurate count of the population than the bare "headcount" data alone. It appears to us that the factual predicate for the change to adjusted figures is adequate. As the Second Circuit pointed out, the district court inCity of New York found "that the PES-indicated statistical adjustment was feasible; that for most purposes and for most of the population that adjustment would result in a more accurate count than the original census; and that the adjustment would lessen the disproportionate undercounting of minorities." City of New York, 34 F.3d at 1129. Assuming that similar findings would hold true for the next decennial census, then we see no reason why the Bureau, in the exercise of its expertise and discretion, may not alter its past practice and adjust the census figures it obtains through a "headcount." (16)

      Accordingly, we conclude that the Constitution does not preclude the Bureau from employing technically and administratively feasible adjustment techniques to correct undercounting in the next decennial census.

III.

      The Census Act includes two provisions authorizing the use of statistical methods, including "sampling," in conducting its statutory responsibilities. The first statute, 13 U.S.C. 141(a), states that

[t]he Secretary shall, in the year 1980 and every 10 years thereafter, take a decennial census of population as of the first day of April of such year, which date shall be known as the "decennial census date", in such form and content as he may determine, including the use of sampling procedures and special surveys.

      The second statute, 13 U.S.C. 195, authorizes, indeed mandates, the use of sampling, but with a limitation relating to apportionment

[e]xcept for the determination of population for purposes of apportionment of Representatives in Congress among the several States, the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as "sampling" in carrying out the provisions of this title.

      In the past, the Bureau has taken the position that 195 prohibits statistical adjustment of census data for purposes of apportionment. The difficulty centered on 195's prohibition on the use of "sampling" in determining the size of the population for purposes of apportionment. Since the scope of 195's exception is not plain from the language of the statute, we turn to the legislative history of that section.

      Congress enacted 195 in 1957, but in a form that authorized, rather than required, the use of sampling; a 1976 amendment transformed the Secretary's authorization into the conditional mandate of the current statute. (17) The enacting Congress of 1957 considered 195 to be merely a change "of an administrative nature" that was "needed for the timely and efficient performance of the biggest jobs the Bureau of the Census has ever undertaken." S. Rep. No. 85-698, at 2 (1957),reprinted in 1957 U.S.C.C.A.N. 1706, 1707. The proviso gave the Bureau the "authority to use sampling in connection with censuses except for the determination of the population for apportionment purposes." Id. at 3, reprinted in 1957 U.S.C.C.A.N. at 1708.

      What Congress originally meant by "sampling" is not clear. In testimony in support of the 1957 legislation, Robert W. Burgess, the Director of the Bureau of the Census, explained that

[t]he use of sampling procedures would be authorized by the proposed new section 195. It has generally been held that the term "census" implies a complete enumeration. Experience has shown that some of the information which is desired in connection with a census could be secured efficiently through a sample survey which is conducted concurrently with the complete enumeration of other items; that in some instances a portion of the universe to be included might be efficiently covered on a sample rather than a complete enumeration basis and that under some circumstances a sample enumeration or a sample census might be substituted for a full census to the advantage of the Government. This section, in combination with section 193, would give recognition to these facts and provide the necessary authority to the Secretary to permit the use of sampling when he believes that it would be advantageous to do so.

Amendment of Title 13, United States Code, Relating to Census: Hearings on H.R. 7911 Before the House Committee on Post Office and Civil Service, 85th Cong. 7-8 (1957).

      The Director's testimony suggests that in enacting 195, Congress intended that the Bureau conduct a "complete enumeration" or a "full census" when determining the size of the population for apportionment purposes, but that the Bureau could use "sampling" in other contexts, where a "sample enumeration" or a "sample census" might be used "to the advantage of the Government." Read in the light of the testimony, the statute's preclusion of "sampling" need not have meant that statistical adjustment of census figures was forbidden: Congress may well have intended only that the decennial census not be a "sample census." Moreover, a "complete enumeration" or "full census" may affirmatively require statistical adjustments of "headcount" data to be made.

      Our Office has previously argued that the 1957 legislative history should not be understood to preclude statistical adjustment. Citing the testimony quoted above, we argued that "[s]ampling refers to a representative portion of the whole . . . while adjustment refers to additions to the whole, here the headcount. As we read the Census Act, there is no statutory prohibition of statistical adjustment." Harmon Memorandum at 3 (citation omitted). The Congressional Research Service (the "CRS"), however, reviewed the same testimony and drew a contrary inference:

it appears that when Section 195 was originally enacted, the Department of Commerce took the position that an actual enumeration was required for all decennial census purposes. Section 195 was enacted in order to relieve this restriction for purposes other than apportionment by sanctioning the use of sampling when appropriate. There was no need to mention other forms of estimating population since this section was making an exception to the general requirement of an actual enumeration only for sampling. Therefore, one may conclude that Section 195 was not intended to sanction the use of methods of estimating population other than "sampling," and did not intend to permit the use of this method for purposes of apportionment.

Congressional Research Service, Library of Congress,Legal Considerations in Census Bureau Use of Statistical Projection Techniques to Include Uncounted Individuals For Purposes of Congressional Reapportionment (Mar. 27, 1980), (report prepared for Congressional use),reprinted inProblems with the 1980 Census Count: Joint Hearing Before the Subcomm. on Commerce, Consumer, and Monetary Affairs of the House Comm. on Government Operations, and the Subcomm. on Census and Population of the House Comm. on Post Office and Civil Service, 96th Cong. 190 (1980) (the "Joint Hearing").

      The 1976 legislation amending the Census Act, Act of Oct. 17, 1976, Pub. L. No. 94-521, 90 Stat. 2459, 2464, was primarily concerned with the establishment of mid-decade censuses. In carrying forward (and amending) 195, we believe that Congress meant that while reliance on samplingalone might be appropriate or desirable for mid-decade censuses, it should not be the exclusive procedure for tabulating the population in decennial censuses. (18) So understood, the 1976 re-enactment does not bar the statistical adjustment of the decennial census if such adjustments would improve their accuracy.

      This interpretation of the 1976 legislative history is not uncontroverted. See Gerson Memorandum at 11 ("Congress' amendment of Section 195 in 1976 is similarly open to two alternative interpretations."). The CRS, noting that both the Comptroller General and the Bureau had advised Congress in 1976 of ongoing developments in estimating or allocating populations other than sampling, argued that "it would be logically inconsistent for Congress to prohibit sampling for purposes of reapportionment, but at the same time to permit the use of other techniques whose reliability had not yet been determined." Joint Hearing at 188. Based on its review of the legislative history, CRS concluded that "the use of demographic estimates for purposes of apportionment of Representatives among the States . . . is prohibited by Section 195 of Title 13." Id. at 192. (19)

      In our judgment, the better view is that the Census Act does not preclude the Bureau from engaging in statistical adjustments of the next set of decennial census figures. SeeFranklin v. Massachusetts, 505 U.S. at 820 (Stevens, J., joined by Blackmun, Kennedy and Souter, JJ., concurring in part and concurring in judgment) (Census Act "embodies a duty to conduct a census that is accurate and that fairly accounts for the crucial representational rights that depend on the census and the apportionment"). A non-preclusive reading gives due weight to the fact that, when it re-enacted 195 in 1976, Congress was primarily concerned with institutingmid-decade censuses. Its prohibition on "sampling" in decennial censuses appears to have meant only that while a procedure relying on "sampling" alone might be the most cost-effective means to discover the information sought in a mid-decade census, the Bureau should not rely on "sampling" as itsexclusive method of tabulating population figures in the decennial census. The use of sampling techniques in the mid-decade census is "probably a pragmatic necessity in that instance, given the vast mobilization of people and resources needed to conduct an even somewhat accurate head count." Senate of California, 968 F.2d at 978. Despite the additional costs entailed, however, Congress did not wish the decennial census to consist of "a mere statistical manipulation through the use of sampling and other techniques." Id. Nothing in amended 195proscribed the use of sampling or other statistical devices in connection with the decennial "headcount," however, if such adjustments would result in a more accurate tabulation.

      Furthermore, in adopting the Census Act, Congress "left the actual administration of a great number of necessary details to the judgment and discretion of the Director of the Census." Seventeenth Decennial Census, 41 Op. Att'y Gen. at 33. Standing alone, 141(a), which authorizes the Director to take the decennial census "in such form and content as he may determine, including the use of sampling procedures and special surveys," would seem to permit statistical adjustments, if in the Director's judgment they would produce greater accuracy. While 195 undoubtedly makes an exception for the use of sampling in apportionment, that exception can be construed narrowly, as befits Congress's otherwise broad delegation of power to the Bureau: the section could be taken to mean that while census figures used for apportionment may not be based on sampling alone, it is permissible to use population samples as one element in a more complex operation by which a prior "headcount" is corrected. Such a reading has in fact generally been adopted by the courts. SeeCarey v. Klutznick, 508 F. Supp. at 415;Young v. Klutznick, 497 F. Supp. at 1334-35;see also Gerson Memorandum at 18 ("the weight of existing caselaw" is "that Section 195 does not preclude statistical adjustment"). (20)

      Moreover, if 195 were read as preclusive, its constitutionality would be highly suspect. Because (as shown above) a non-preclusive reading is a reasonable one, it should be preferred.

      Substantial constitutional issues would arise under both the Enumeration Clause and the Fifth Amendment if 195 were construed to prevent the Bureau from adjusting census data for apportionment. The Enumeration Clause prescribes that Representatives be apportioned to the several States "according to their respective Numbers," and it can be argued that the Clause is violated if Representatives are apportioned on the basis of a census count that is known to be deficient, but that could be rendered more accurate by feasible adjustments. SeeFranklin v. Massachusetts, 505 U.S. at 806 (Bureau's decision to allocate government personnel stationed abroad to State designated as home of record "does not hamper the underlying constitutional goal of equal representation, but . . . actually promotes equality");United States Dep't of Commerce v. Montana, 503 U.S. 442, 461 (1992) (Court "might well find" that requirement that Representatives be apportioned by reference to the populations of the several States "embod[ied] the same principle of equality" as found inWesberry);Carey v. Klutznick, 508 F. Supp. at 414 (language of Enumeration Clause evinces "an intent that apportionment be based on a census that most accurately reflects the true population of each state");cf.Wesberry, 376 U.S. at 13-14.

      Furthermore, "[t]he Fifth Amendment . . . might be thought, by analogy to the decisions invalidating the malapportionment of state legislatures under the equal protection clause, to require the federal government to apportion congressional seats . . . in accordance with an accurate estimate of the number of people in each state." Tucker, 958 F.2d at 1414. See, e.g.,Reynolds v. Sims, 377 U.S. 533, 555 (1964) ("the right of suffrage can be denied by a debasement or dilution of the weight of a citizen's vote just as effectively as by wholly prohibiting the free exercise of the franchise"). Thus, the Second Circuit has found that the Bureau's decision not to adjust the 1990 census figures was constitutionally suspect under the Fifth Amendment:

[B]oth the nature of the right and the nature of the affected classes are factors that traditionally require that the government's action be given heightened scrutiny: the right to have one's vote counted equally is fundamental and constitutionally protected, and the unadjusted census undercount disproportionately disadvantages certain identifiable minority groups. . . . That the goal of precise equality cannot be achieved nationwide . . . does not relieve the federal government of the obligation to make a good-faith effort to achieve voting-power equality "as nearly as is practicable."

City of New York, 34 F.3d at 1128, 1129 (citation omitted).

      We need not here consider whether the Second Circuit's view of the merits is correct; nor need we address the issue whether the question the court decided was litigable. Suffice it to say that there would be substantial constitutional difficulties under both the Enumeration Clause and the Fifth Amendment if 195 were understood to prohibit the Bureau from making practicable statistical adjustments that would result in a more accurate tally than the traditional headcount. Section 195 should be construed, if "'fairly possible,'" to avoid those difficulties. See, e.g.,Ashwander v. TVA, 297 U.S. 288, 348 (1936) (Brandeis, J., concurring) (citation omitted). Because a constitutionally unproblematic reading is justified (and has, in fact, been adopted by most courts), it should be adopted.

      Accordingly, 195 does not preclude reliance upon technically feasible statistical adjustments to improve the accuracy of "headcount" data, and specifically to correct the differential undercounting of minority group populations. It also does not prohibit the Bureau from conducting the non-response follow-up on a sample basis, rather than sending enumerators to every non-responding household, where the use of the former technique would improve accuracy while substantially lowering administrative costs.

Conclusion

      Neither the Constitution nor the Census Act precludes the Bureau from making the proposed statistical adjustments of "headcount" data in the decennial census for the year 2000.

WALTER DELLINGER
Assistant Attorney General
Office of Legal Counsel

*  Editor's Note: Subsequent to the date of this opinion, the Supreme Court held that the Census Act prohibits the proposed uses of statistical sampling in calculating population for congressional apportionment purposes. See Department of Commerce v. United States House of Representatives, 119 S.Ct. 765, 779 (1999). The Court did not reach the constitutional question. Id.

1 There is also a mid-decade census. See 13 U.S.C. 141(d).

2 The apportionment of Representatives among the States in turn affects the allocation of Electoral College votes to the States. See U.S. Const. art. II, 1, cl. 2.

3 The first statute authorizing a census, "An Act providing for the enumeration of the Inhabitants of the United States" (Mar. 1, 1790), declared that "the marshals of the several districts of the United States" were "authorized and required to cause the number of the inhabitants within their respective districts to be taken," omitting Indians not taxed. 4National State Papers of the United States, 1789-1817, at 1 (Eileen Daney Carzo ed., 1985). It further placed on "each and every person more than sixteen years of age" the obligation to provide the census-taker "a true account, if required, to the best of his or her knowledge, of all and every person belonging to [the respondent's] family." Id. at 3.

4 The Bureau's efforts to obtain as accurate a count as possible have been found to be "extraordinary. According to one court, the 1990 census is said to be one of the best ever taken in this country because despite our large population, approximately 98 percent of the population was counted." City of Detroit v. Franklin, 4 F.3d 1367, 1376 (6th Cir. 1993),cert. denied, 510 U.S. 1176 (1994).

5 The first census in 1790 counted over 3,890,000 people, but fell short of the expected 4,000,000 figure. George Washington thought it "certain" that "ourreal numbers will exceed, greatly, the official returns of them," and Thomas Jefferson considered the uncounted population "very great." SeeBaldrige v. Shapiro, 455 U.S. at 353 n.8.

6 See alsoKarcher v. Daggett, 462 U.S. 725, 732 (1983) ("the census data are not perfect, and the well-known restlessness of the American people means that population counts for particular localities are outdated long before they are completed");id. at 772 (White, J., dissenting) ("the census . . . cannot be perfect");Gaffney v. Cummings, 412 U.S. 735, 745 (1973) (decennial census figures "may be as accurate as such immense undertakings can be, but they are inherently less than absolutely accurate.").

7 In the 1990 census, "Blacks were undercounted by 4.8%, Hispanics by 5.2%, Asian-Pacific Islanders by 3.1%, American Indians by 5.0%, and non-Blacks by 1.7%." Senate of California v. Mosbacher, 968 F.2d 974, 975 (9th Cir. 1992). "In 1940, 10.3 percent of blacks were missed, compared to 5.1 percent of whites, a gap of 5.2 percentage points. In 1980, 6.2 percent of blacks were missed, compared to 1.3 percent of whites, for a similar disparity of 4.9 percentage points." Samuel Issacharoff & Allan J. Lichtman,The Census Undercount and Minority Representation: The Constitutional Obligation of the States to Guarantee Equal Representation, 13 Rev. Litig. 1, 8 (1993). See also Gaffney v. Cummings, 412 U.S. at 745 n.10.

8 See Stephen E. Fienberg,The New York City Census Adjustment Trial: Witness for the Plaintiffs, 34 Jurimetrics J. 65, 70-71 (1993).

9 Tucker, 958 F.2d at 1413;see generallyDecision of the Secretary of Commerce on Whether a Statistical Adjustment of the 1990 Census of Population and Housing Should Be Made for Coverage Deficiencies Resulting in an Overcount or Undercount of the Population, 56 Fed. Reg. 33,582 (1991).

10 The Secretary's reasoning, as recapitulated by the Seventh Circuit, was that

while adjustment by the best method available would increase the census totals, it would not significantly alter the apportionment of seats in the House of Representatives among the states, in part because there is overcounting as well as undercounting. After the dust settled, Illinois's representation would be unchanged, although California and Arizona would pick up a few seats at the expense of Pennsylvania and Wisconsin. Federal grant allocations might not be much affected either . . . . Moreover, any attempt to make a statistical adjustment to the mechanical headcount would, by injecting judgmental factors -- and ones of considerable technical complexity to boot, . . . -- open the census process to charges of political manipulation. And while a statistical adjustment for the undercount would undoubtedly improve the accuracy of the nationwide census total, there is no consensus among statisticians and demographers that it would make the state and district census totals -- the level at which the adjustment would actually affect representation and funding -- more accurate.

Tucker, 958 F.2d at 1413 (citations omitted);see alsoCity of New York, 34 F.3d at 1122-23;Senate of California, 968 F.2d at 975.

11 See also Hyman Alterman,Counting People at 187-88 ("The Convention had available to it estimates of the white and slave populations in the various states. Mainly on the basis of these estimates the Convention decided how many representatives each state should have until the first census was taken.").

12 U.S. Const. art. I, 2, cl. 3 provided that "until such enumeration shall be made," the States were to have predetermined numbers of Representatives: three for New Hampshire, eight for Massachusetts, one for Rhode Island, five for Connecticut, six for New York, four for New Jersey, eight for Pennsylvania, one for Delaware, six for Maryland, ten for Virginia, five for North Carolina, five for South Carolina and three for Georgia, for a total of sixty-five.

13 "Instead of headcounting people, [the Bureau] uses the mail-out form and the mail-out/mail-back format to enumerate most persons today." Id. See alsoCity of Detroit, 4 F.3d at 1377 ("[t]he Census Bureau has not undertaken a door-to-door campaign since the 1960 census and plaintiffs have presented no evidence indicating that such an effort would lead to any more accurate results").

14 The Court has held that "[c]onstitutional challenges to apportionment are justiciable." Franklin v. Massachusetts, 505 U.S. at 801. Whether constitutional plaintiffs "have standing to challenge the accuracy of the data" tabulated by the Bureau, and "whether the injury is redressable by the relief sought,"id. at 802, are of course separate issues. We shall assume here that those conditions might be met. The availability of review under the Administrative Procedure Act (the "APA") of the use of adjusted data for reapportionment seems doubtful after Franklin, however. The APA permits review only of certain "final" agency actions under 5 U.S.C. 704. In this case, as inFranklin, it would appear that "the final action complained of is that of the President, and the President is not an agency within the meaning of the Act." 505 U.S. at 796. We note thatFranklin's ruling on the APA represented the view of a bare majority of five Justices (including Justice White), and might not be extended by the present Court.

15 For analogous reasons, if APA review were available, a change in policy to allow statistical adjustments might be attacked as arbitrary, capricious or abusive of discretion under 5 U.S.C. 706(2)(A). SeeMotor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 42-45 (1983) (presumption in favor of settled agency practice). We believe that the proposed policy change would survive review under that standard.

16 Moreover, in light of the Bureau's position that the use of a sample-based follow-up for enumerating non-respondent households would improve the accuracy of the final count while at the same time saving the Bureau upwards of $300 million, we can see no constitutional objection to the introduction of that procedure.

17 As enacted in 1957, the statute had stated that "[e]xcept for the determination of population for apportionment purposes, the Secretary may, where he deems it appropriate, authorize the use of the statistical method known as 'sampling' in carrying out the provisions of this title." Pub. L. No. 85-207, 14, 71 Stat. 481, 484 (1957).

18 The Senate Report stated that the section of the 1976 legislation that modified 13 U.S.C. 195 "differs from present language which grants the Secretary discretion to use sampling when it is considered appropriate. The section as amended strengthens congressional intent that, whenever possible, sampling shall be used." S. Rep. No. 94-1256, at 6 (1976),reprinted in 1976 U.S.C.C.A.N. 5463, 5468.

19 One further aspect of the 1976 legislative history should be noted. In the 1970 decennial census, the Bureau used "sampling" to add to the national total the figure of almost five million people believed missing from the headcount. The Bureau estimated that it had not contacted some 10.2 million people, or about 5% of the population. Of this 10.2 million not actually counted, 4.9 million were included in the official count by "imputation" and allocated among the States for apportionment of House seats. Young v. Klutznick, 497 F. Supp. at 1321;see also Gerson Memorandum at 15 ("In effect, a portion of the population was not tabulated directly in 1970. Instead, the Bureau obtained an estimate of its size from the results of statistical sampling and added that estimate to the total population count."). The district court inYoung inferred that when Congress amended 195 in 1976, it was "well aware" of the Bureau's adjustment of the 1970 census data and impliedly consented to that practice. 497 F. Supp. at 1334-35. The court cited no direct evidence, however, that Congress was aware of, and approved, the 1970 census adjustment. See Gerson Memorandum at 15. Moreover, as the Bureau argued,seeYoung, 497 F. Supp. at 1334, the re-enactment of 195 (with essentially minor changes from 1957) could be interpreted as a ratification of the Bureau's more traditional practice of using only a headcount.

20 But see Jeffrey S. Cramptor, Comment,Lies, Damn Lies and Statistics: Dispelling Some Myths Surrounding the United States Census, 1990 Det. C.L. Rev. 71 (criticizing case law); Gerson Memorandum at 18 ("[w]e can foresee a court deciding that Section 195, on its face, prohibits statistical adjustment").