Nos. 94-11614, 94-1631 and 94-1985 In The Supreme Court of The United States OCTOBER TERM, 1995 STATE OF WISCONSIN, PETITIONER v. CITY OF NEW YORK, ET AL. STATE OF OKLAHOMA, PETITIONER v. CITY OF NEW YORK, ET AL. UNITED STATES DEPARTMENT OF COMMERCE, ET AL., PETITIONERS v. CITY OF NEW YORK, ET AL. ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE SECOND CIRCUIT BRIEF FOR THE FEDERAL PETITIONERS DREW S. DAYS, III Solicitor General FRANK W. HUNGER Assistant Attorney General EDWIN S. KNEEDLER Deputy Solicitor General MALCOLM L. STEWART Assistant to the Solicitor General MARK B. STERN MICHAEL S. RAAB Attorneys Department of Justice Washington, D.C. 20530 (202)514-2217 ---------------------------------------- Page Break ---------------------------------------- QUESTION PRESENTED Whether the decision of the Secretary of Commerce not to undertake a statistical adjustment to the 1990 census violated the Constitution. (I) ---------------------------------------- Page Break ---------------------------------------- II PARTIES TO THE PROCEEDINGS Petitioners, defendants below, are: United States Department of Commerce Ronald H. Brown, Esq., As Secretary of the United States Department of Commerce; Everett Ehrlich, As Under Secretary for Economic Affairs of the United States Department of Commerce; Bureau of Census; Harry Scarr, As Acting Director of Bureau of Census;, William J. Clinton, As President of the United States;, Dan Glickman, As Secretary of Ag- riculture; Donna E. Shalala, As Secretary of Health and Human Services; Henry Cisneros, As Secretary of Housing and Urban Development; Robert B. Reich, As Secretary of Labor; Federico Pena, As Secretary of Transportation; Richard W. Riley, As Secretary of Education. The State of Wisconsin and the State of Oklahoma, intervenor-defendants below, are also petitioners. Respondents, plaintiffs below, are: City of New York State of New York; City of Los Angeles; City of Chicago; City of Houston Dade County, Florida; United States Conference of Mayors; National League of Cities; League : of United Latin American Citizens; National Association for the Advancement of Colored People; Marcella Maxwell; Donald H. Elliott; John Mack; Olga Morales; Timothy W. Wright, III; Raymond G. Romero; Antonio Gonzales; Athalie Range; Jerry Alan Wood; Carolyn Sue Lopez; City of Atlanta, Georgia; Maynard Jackson, Individually, and as the Mayor of the' City of Atlanta Florida House of Representative; Florida State Conference; Miguel A. De Grandy; Willye Dennis; Mario Diaz-Balart; Dr. Charles Evans; Rodolfo Garcia, Jr.; Bollowy L. "Bo" Johnson; Alfred J. Lawson, Jr.; Willis Logan, Jr.; Johnnie McMillan; Alzo J. Reddick Peter Rudy Wallace; T.K. Wetherell; State of Texas; City of Phoenix, Arizona State of New ---------------------------------------- Page Break ---------------------------------------- III Jersey State of Florida; City of Cleveland, Ohio; City of Denver, Colorado; City of Inglewood, California City of New Orleans, Louisiana City of Oakland, California; City of Pasadena, California; City of Philadelphia, Penn- sylvania; City of San Antonio, Texas; City of San Francisco, California; Broward County, Florida; State of Arizona City of Baltimore, Maryland; City of Boston, Massachusetts; City of Long Beach, California; City of San Jose, California; Los Angeles County, California; San Bernadino County, California; District of Columbia; Navajo Nation; State of New Mexico; City of Tucson, Arizona; Council of Great City Schools. By virtue of Rule 12.6 of this Court, Donnald K. Anderson, As Clerk of the United States House of Representatives, a defendant below, is also a respondent in this Court. The following were parties to the proceedings in the district court, but were not parties to the proceeding in the court of appeals: People of the State of California ex rel. Daniel E. Lungren, Attorney General; and County of Hudson, New Jersey. ---------------------------------------- Page Break ---------------------------------------- TABLE OF CONTENTS Page Opinions below . . . . 1 Jurisdiction . . . . 2 Constitutional and statutory provisions involved . . . . 2 Statement: A. Constitutional and statutory framework . . . . 2 B. The 1990 census . . . . 4 C. Consideration of an adjustment to the 1990 census . . . . 5 D. The Secretary's decision not to make an adjustment . . . . 13 E. The district court's decision . . . . 19 F. The court of appeals' decision . . . . 21 Summary of argument . . . . 23 Argument: The Secretary's decision not to undertake a statistical adjustment of the 1990 census was fully consistent with the Constitution . . . . 25 A. The Secretary's decision was consistent with the text, history, and purpose of the Constitution . . . . 26 B. The court of appeals erred in holding that the Secretary's decision should be subject to heightened scrutiny and that respondents established a prima facie case of a constitu- tional violation . . . . 39 Conclusion . . . . 51 TABLE OF AUTHORITIES Cases: Anderson v. City of Bessemer City, 470 U.S. 564 (1985) . . . . 34, 35 (V) ---------------------------------------- Page Break ---------------------------------------- VI Cases-Continued: Page Baldrige v. Shapiro, 455 U.S. 345 (1982) . . . . 29, 33 Baltimore Gas & Electric Co. v. Natural Re- sources Defense Council, Inc., 462 U.S. 87 (1983) . . . . 34 Borough of Bethel Park v. Stuns, 449 F.2d 575 (3d Cir. 1971) . . . . 42 Camp v. Pitts, 411 U.S. 138 (1973) . . . . 20 Carey v. Klutznick: 503 F. SUPP. 874 (N.D. 111, 1980) . . . . 42 637 F.2d 834 (2d Cir. 1980) . . . . 42 City of Camden v. Plotkin., 466 F. Supp. 44 (D.N.J. 1978) . . . . 42 City of Detroit v. Franklin, 4 F.3d 1367 (6th Cir. 1993), cert. denied, 114 S. Ct. 1217 (1994) . . . . 4, 23, 30, 35 City of Philadelphia v. Klutznick, 503 F. Supp. 663 (E.D. Pa. 1980) . . . . 42 City of Willacoochee v. Baldrige, 556 F. Supp. 551 (S.D. Ga. 1983) . . . . 42 Cuomo v. Baldrige, 674 F. Supp. 1089 (S.D.N.Y. 1987) . . . . 5 District of Columbia v. United States Dep't of Commerce, 789 F. Supp. 1179 (D.D.C. 1992), appeal voluntarily dismissed, No. 92-5212 (D.C. Cir.) . . . . 42 Federation for American Immigration Reform v. Klutznick, 486 F. Supp. 564 (D.D.C.), appeal dismissed, 447 U.S. 916 (1980) . . . . 42 Florida Power & Light Co. v. Lorion, 470 U.S. 729 (1985) . . . . 19-20 Franklin v. Massachusetts, 112 S. Ct. 2767 (1992) . . . . passim Harlow v. Fitzgerald, 457 U.S. 800 (1982) . . . . 45 International Fabricare Institute v. EPA, 972 F.2d 384 (D.C. Cir. 1992) . . . . 34 Karcher v. Daggett, 462 U.S. 725 (1983) . . . . 21, 31, 39, 40, 41,43,44 Kleppe v. Sierra Club, 427 U.S. 390 (1976) . . . . 34 ---------------------------------------- Page Break ---------------------------------------- VII Cases-Continued Page Marsh v. Oregon Natural Resources Council, 490 U.S. 360 (1989) . . . . 34 Massachusetts v. Mellon, 262 U.S. 447 (1923) . . . . 19 Nebraska v. Wyoming, 115 S. Ct. 1933 (1995) . . . . 19 Quon v. Stans, 309 F. Supp. 604 (N.D. Cal. 1970) . . . . 43 Ridge v. Verity, 715 F. Supp. 1308 (W.D. Pa. 1989) . . . . 42 Thomas Jefferson University v. Shalala, 114 S. Ct. 2381 (1994) . . . . 34 Tucker v. United States Dep't of Commerce, 958 F.2d 1411 (7th Cir.), cert. denied, 113 S. Ct. 407 (1992) . . . . 23, 30, 41 United States Dep't of Commerce v. Montana, 503 U.S. 442 (1992) . . . . 3, 19, 27, 28, 40, 43 United States Dep't of Labor v. Triplett, 494 Us. 715 (1990) . . . . 30 Village of Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U.S. 252 (1977) . . . . 46, 47 Washington v. Davis, 426 U.S. 229 (1976) . . . . 45, 47 West End Neighborhood Corp. v. Stans, 312 F. Supp. 1066 (D.D.C. 1970) . . . . 42 Constitution, statutes and rule: U.S. Const.: Art. I . . . . 19 2, Cl. 3 . . . . 2, 26, 27, 30, 33, 40 Apportionment Clause . . . . 3, 30, 43 Census Clause . . . . 3, 26, 28, 33, 50 8, C1. 18 (Necessary and Proper Clause) . . . . 43 Art. III . . . . 19 Amend. V . . . . 2, 46 Due Process Clause . . . . 46 Amend. XIV . . . . 2 2 . . . . 2, 3, 30 5 . . . . 43 Administrative Procedure Act, 5 U.S.C. 706(2)(A) . . . . 20 2 U.S.C. 2a . . . . 2 2 U.S.C. 2a(a) . . . . 3 ---------------------------------------- Page Break ---------------------------------------- VIII Statutes and rule-Continued: Page 2 U.S.C. 2a(b) . . . . 3 13 U.S.C. 141 . . . . 26 13 U.S.C. 141(a) . . . . 2, 3, 26, 33 13 U.S.C. 141(b) . . . . 2, 3 13 U.S.C. 18 . . . . 26 Fed. R. Civ. P. 52(a) . . . . 34 Miscellaneous: Bureau of the Census, Department of Commerce, Report of the Committee on Adjustment of Post- censal Estimates: Assessment of Accuracy of Adjusted Versus Unadjusted 1990 Census Base for Use in Intercensal Estimates (Aug. 7, 1992) . . . . 4-5 Bureau of the Census, Department of Commerce, State and Metropolitan Area Data Book 1991 (4th ed. 1991) . . . . 17 54 Fed. Reg. 51,002 (1989) . . . . 6 55 Fed. Reg. (1990): p. 9838 . . . . 6 pp. 9839-9842 . . . . 6 p. 9841 . . . . 18 58 Fed. Reg. (1993): p. 70 . . . . 5 pp. 72-73 . . . . 37 27 Weekly Comp. Pres. Dec. 6 (1991) . . . . 4 ---------------------------------------- Page Break ---------------------------------------- IN THE SUPREME COURT OF THE UNITED STATES OCTOBER TERM, 1995 No. 94-1614 STATE OF WISCONSIN, PETITIONER CITY OF NEW YORK, ET AL. No. 94-1631 STATE OF OKLAHOMA, PETITIONER CITY OF NEW YORK, ET AL. No. 94-1985 UNITED STATES DEPARTMENT OF COMMERCE, ET AL., PETITIONERS CITY OF NEW YORK, ET AL. BRIEF FOR THE FEDERAL PETITIONERS OPINIONS BELOW The opinion of the court of appeals (Pet. App. 140) 1. is reported at 34 F.3d 1114. The opinions of the district court (Pet. App. 41-95, 96-120, 121-134) are reported at 822 F. Supp. 906, 739 F. Supp. 761, and 713 F. Supp. 48. The decision of the Secretary of Commerce (Pet. App. 135-415) is published at 56 Fed. Reg. 33,582. ___________________(footnotes) 1 References to "Pet. App." are to the appendix to the petition for a writ of certiorari filed in No. 94-1614. (1) ---------------------------------------- Page Break ---------------------------------------- 2 JURISDICTION The judgment of the court of appeals was entered on August 8, 1994. The petitions for rehearing filed by the States of Oklahoma and Wisconsin were denied on December 12, 1994, and January 4, 1995, respectively. Pet. App. 416-418,419-421. The petition for a writ of certiorari in No. 94-1614 was filed by Wisconsin on April .3, 1995. The petition for a writ of certiorari in No. 94-1631 was filed by Oklahoma on April 4, 1995. On March 27, 1995, Justice Ginsburg extended the federal petitioners' time for filing a petition for a writ of certiorari to and including May 4, 1995. 0n April 25, 1995, Justice Ginsburg further extended the time for filing to and including June 3,1995. The peti- tion for a writ of certiorari in No.94-1985 was filed by the federal petitioners on June 5,1995 (a Monday). The three petitions were granted on September 27, 1995. J.A. 109- 111. The jurisdiction of this Court rests on 28 U.S.C. 1254(l). See also notes 16 and 23, infra. CONSTITUTIONAL AND STATUTORY PROVISIONS INVOLVED Article I, Section 2, Clause 3, of the United States Constitution; the Fifth Amendment and Section 2 of the Fourteenth Amendment to the Constitution; 2 U.S.C. 2a; and 13 U.S.C. 141(a) and(b) are reproduced in relevant part at Pet. App. 422-425. STATEMENT A. Constitutional And Statutory Framework The Constitution requires a decennial census for the purpose of determining the number of Representatives to which each State is entitled. Article I, Section 2, Clause 3 provides that "Representatives * * * shall be apportioned among the several States * * * according to their ---------------------------------------- Page Break ---------------------------------------- 3 respective Numbers" (the Apportionment Clause) and directs that "[t]he 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" (the Census Clause). See also Amend. XIV, 2 ("Repre- sentatives 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."). Pursuant to the Census Clause, Congress has provided in the Census Act that the decennial census shall be conducted by the Secretary of Commerce "in such form and content as he may determine, including the use of sampling procedures and special survey s." 13 U.S.C. 141(a). Congress has, however, prescribed a strict timetable for the census. The population of the United States is to be determined as of April 1 of the census year. The "tabulation of total population by States" for the purpose of apportionment of Representatives is to be completed and reported to the President within nine months after the April 1 census date. 13 U.S.C. 141(b), Within a week of the beginning of the first Session of Congress following the census, the President must transmit to Congress a statement showing the "whole number of persons in each State * * * and the number of Representatives to which each State would be entitled" under the statutorily prescribed "equal proportions" formula for apportioning Representatives. 2 U.S.C. 2a(a); see United States Dep't of Commerce v. Montana, 503 U.S. 442,451-455 (1992). Within 15 days of receiving that statement, the Clerk of the House must "send to the executive of each State a certificate of the number of Representatives to which such State is entitled under this section." 2 U.S.C. 2a(b). Following the 1990 decennial ---------------------------------------- Page Break ---------------------------------------- 4 census, the President transmitted the statement of the apportionment of Representatives to Congress on January 3,1991, see 27 Weekly Comp. Pres. Dec. 6(1991), and the Clerk then sent the required certificates to the Governors of the 50 States. B. The 1990 Census The 1990 decennial census was conducted in accordance with an elaborate, multi-staged procedure designed to locate and identify each housing unit in the country and to enumerate each person in those units by means of a system replete with rechecks and cross-checks. 2. The Census Bureau employed approximately 500,000 tem- porary workers, including 300,000 interviewers, in its efforts to obtain an accurate count of the population. Tr. 1743. Although those processes resulted in a very high rate of success, it is undisputed that many persons were not counted. The Secretary of Commerce originally es- timated that the census enumeration had failed to count 2.1% of the population. Pet. App. 158. It was subsequently determined, however, that as a result of certain errors (including a computer processing error), that estimate significantly overstated the undercount; the Census Bureau represented at the trial in this case that the undercount was 1.6% of the population. Id. at 58 n.12; see Tr. 1786-1788; DX 21; see also Bureau of the Census, ___________________(footnotes) 2 That process included the development of comprehensive maps of each block in the United States, the identification of housing units on each block, test censuses and post-enumeration survey (PES) tests, the actual enumeration (including multiple contacts with each household that did not return a census questionnaire), and pre- and post-census review by state and local governments. See generally Pet. App. 319- 332; City of Detroit v. Franklin, 4 F.3d 1367, 1376 (6th Cr. 1993) ("the Census Bureau's efforts to conduct as accurate a census count as possible are extraordinary"), cert. denied, 114 S. Ct. 1217 (1994). ---------------------------------------- Page Break ---------------------------------------- 5 Department of Commerce, Report of the Committee on Adjustment of Postcensal Estimates: Assessment of Accuracy of Adjusted Versus Unadjusted 1990 Census Base for Use in Intercensal Estimates 15 (Aug. 7, 1992) (CAPE Report); 58 Fed. Reg. 70 (1993). That 1.6% undercount is approximately the same as the undercount in the 1980 census. Pet. App. 58 n.12. It is undisputed that the undercount varied among population subgroups. The decision of the Secretary at issue in this case estimated that "Blacks appear to have been undercounted in the 1990 census by 4.8%, Hispanics by 5.2%, Asian-Pacific Islanders by 3.l%, and American Indians by 5.0%." Pet. App. 138-139. A differential under- count also exists among various other subgroups. For example, men are missed at a higher rate than women, the young are missed at a higher rate than the elderly, and renters are missed at a higher rate than homeowners. DX 1, at 457-458; Pet. App. 158. C. Consideration Of An Adjustment To The 1990 Census 1. The problem of the overall undercount and differential undercounts among demographic groups led the Department of Commerce to consider a statistical adjustment to the 1980 decennial census. Ultimately, how- ever, no adjustment was undertaken. That decision was challenged in a number of lawsuits, but none was successful, The action brought by the State of New York was not resolved until 1987. Cuomo v. Baldrige, 674 F. Supp. 1089, 1104 (S.D.N.Y. 1987). In the same year that litigation: regarding the 1980 census finally concluded, the Department of Commerce announced that no statistical adjustment would be made for the 1990 census. In the following year, respondents filed this suit, seeking to compel an adjustment to the 1990 ---------------------------------------- Page Break ---------------------------------------- 6 census for all purposes for which the census is used, including the apportionment of Representatives among the States and the distribution of federal funds. Defendants included the President, the Secretary of Commerce, the Census Bureau and its Director, and the Clerk of the House of Representatives. 3. In July 1989, prior to a hearing on respondents' request for a preliminary injunction, the parties entered into a stipulation under which defendants agreed to reconsider the adjustment question and to reach a decision by July 15, 1991. J.A. 61-67. Defendants also agreed that the Census Bureau would conduct a post-enumeration survey (PES) of at least 150,000 households in order to generate data on which a statistical adjustment could be based. J.A. 62. In accordance with the stipulation, the Secretary published proposed guidelines for public comment and ultimately adopted eight guidelines that he would consider in deciding whether to order an adjustment of the 1990 census totals. See 54 Fed. Reg. 51,002 (1989); 55 Fed. Reg. 9838,9839-9842 (1990); Pet. App. 113-115. 4. ___________________(footnotes) 3 On February 8, 1989, counsel for the Clerk of the House of Representatives entered into a stipulation providing that the Clerk "neither objects nor consents to the relief requested by the plaintiff in their complaint and takes no position with respect to the questions involved in this litigation." Stipulation Regarding Participation of Defendant Donnald K. Anderson at, 1-2. The Clerk agreed to be bound by any judgment entered in the case and accordingly was "relieved of any obligation to answer the complaint or otherwise participate in this litigation." Id. at 2. We do not believe that injunctive relief was properly sought against the President. See Franklin v. Mass- achusetts, 112 S. Ct. 2767,2775-2776 (1992); id. at 2776-2777 (opinion of O'Connor, J.); id. at 2788-2790 (Scalia, J., concurring in part and concurring in the judgment). 4 Those guidelines are reprinted in the district court's opinion of June 7, 1990, in which it considered respondents' contention that the guidelines were inconsistent with the Secretary's obligations under the ---------------------------------------- Page Break ---------------------------------------- 7 2. In July 1990, the Census Bureau conducted the PES, which involved an attempt to identify all individuals living in a selected sample of more than 168,000 housing units across the United States in order to determine whether those persons had been properly counted and located in the census. The Census Bureau placed persons in the PES in one of 1392 categories, called "post-strata," defined by five variables: age, sex, race and Hispanic/non-Hispanic ethnicity, geographic subdivision, 5. and home tenure (owner vs. renter). See Pet. App. 343 (giving examples of post-strata). The Bureau then estimated the rate at which individuals within each post-stratum had been accurately counted in the census. Data obtained from the PES were used to evaluate the accuracy of the original census and, through the use of a complex series of statistical and mathematical processes, to generate adjusted population estimates. See id. at 332-346. a. In constructing the post-strata, the Census Bureau was required to make numerous choices and assumptions concerning how best to divide the population into groups that would have similar "capture probabilities" (i.e., probabilities of being counted in the census). Reasonable alternatives to the 1392 post-strata would have produced different population estimates. To take one example, the PES divided the nation into geographic groups cor- responding to the Census Bureau's census divisions (see Pet. App. 373) and assumed that the likelihood of being captured in the census was the same for persons within ___________________(footnotes) stipulation. See Pet. App. 113-115. The district court rejected that contention. Id. at 116-117. 5 To define the geographic subdivision for a particular indi- vidual, the Bureau considered both the area of the country (e.g., New England, Middle Atlantic) and the type of urban or rural area in which that individual resided. See Pet. App. 373-375. ---------------------------------------- Page Break ---------------------------------------- 8 each division who shared similar demographic char- acteristics. At the trial in this case, respondents' experts acknowledged that geographic groupings other than those used by the Census Bureau in conducting the PES might have been reasonable, and that those alternative groupings would have affected the outcome of the adjustment process. See Tr. 1440-1441 (testimony of Professor John W. Tukey); see also Tr. 1682-1683 (testimony of Steven Fienberg). Defendants' witness, Dr. Kenneth Wachter, also testified that alternative geographic groupings would have been equally reasonable, if not preferable, and would have produced different adjustment results. Tr. 2133-2142. b. The Bureau compared the PES data with the census data to determine whether identified persons in the PES sample had been erroneously missed or counted by the census. That process was known as "matching." On some occasions, however, it was not possible to determine whether a particular person found in the PES had been recorded by the census, because incomplete census and PES forms made it impossible to determine whether PES records matched census records for some addresses, Pet. App. 169. In those instances, mathematical models were used to supply the missing data and then to "impute" a match status. Id. at 170. Experts disagreed over the level of error resulting from the imputation process. The Secretary's administrative decision declining to make an adjustment observed that "[t]he imputation scheme used for the PES is based on a series of assumptions that are mostly guesswork." Ibid. c. By comparing the results of the PES with the re- sults of the census, the Bureau developed an "adjustment factor" for each of the 1392 post-strata that estimated the percentage of the members of that post-stratum that were ---------------------------------------- Page Break ---------------------------------------- 9 undercounted or overcounted in the census, 6. as well as a "variance" for each of the adjustment factors. 7. The Bureau determined, however, that the initial or "raw" adjustment factor should not be used in making an adjustment because of the high level of "sampling error" (a type of error inherent in any attempt to extrapolate for the whole population based on a sample). See Pet. App. 345. Although the overall PES was an unusually large sample survey, the average post-stratum in the PES consisted of only about 300 people (377,000 persons in the PES divided by 1392 post-strata). See ibid.; DX 64, at 7 Tr. 2369-2370. In each post-stratum, moreover, only about 25 persons had been erroneously missed and about 15 persons erroneously counted in the original census. DX 64, at 7 Tr. 2369-2370. Any errors in those small numbers would produce significant changes in estimated undercounts or overcounts for the different post-strata. See DX 64, at 7 Tr. 2369-2370. To deal with the problem of sampling error, the Bureau employed a "smoothing" procedure. In essence, smoothing sought to compensate for the relatively small number of persons in each post-stratum by combining data from many different post-strata, weighted on the basis of variables such as race or age. Pet. App. 219-220,222. That attempt to correct for sampling error, however, introduced important new questions. The post-strata had been developed in the first place because they were believed to ___________________(footnotes) 6 The Secretary explained that "if the [estimated population based on the census and PES] for a particular post-stratum was 1,050,000 and the census count was 1,000,000, then the adjustment fac- tor was 1.05." Pet. App. 344. 7 The variance of an adjustment factor is an estimate of its random variation, see Pet. App. 223, calculated by squaring the standard error, see Tr. 2367. ---------------------------------------- Page Break ---------------------------------------- 10 represent groupings of persons with similar capture probabilities. See id. at 204-205. In combining post-strata in the smoothing process, the Bureau was required to make new assumptions as to which variables mattered and how those variables interacted to produce the undercount. The Secretary's final decision discussed the smoothing process in considerable detail. See Pet. App. 219-226. One step of the smoothing process involved the calculation of a "flattened" estimate of the adjustment factor for each post-stratum. The "flattened" adjustment factor was derived by considering data from that post-stratum in conjunction with data from many other post-strata, through use of a statistical technique known as a linear regression. See id. at 222-223. To determine the "smoothed" adjustment factor, the Bureau calculated a "weighted" average of the raw and flattened adjustment factors, based upon the variances for the raw adjustment factors. If the variance for a particular adjustment factor was high, the smoothed factor would be closer to the flattened factor than to the raw factor; if the variance was low, the reverse would be true. Id. at 223. 8. The smoothing procedure was subject to an additional complication. Due to the small size of each of the post- strata, the variances themselves (which were estimated from the sample data) were subject to substantial random variation. See Pet. App. 223. The Bureau therefore determined that the variances should also be smoothed, a process known as- "pre-smoothing." Ibid.; see also Administrative Record (A. R.) 492 ("The Census Bureau * * * replaced the variances by a flatter, regression-based ___________________(footnotes) 8 As the Secretary explained, "[t]he smaller the random variation in a poststratum the more the smoothed factor relies on the observed data and the less it relies on the regression estimate." Pet. App. 223. ---------------------------------------- Page Break ---------------------------------------- 11 set of approximations to them."). The smoothed variances were then used to smooth the adjustment factors (i. e., to determine the weighted average of the raw and flattened adjustment factors for each post-stratum). See Pet. App. 223; see also id. at 377-415 (table of adjustment factors by post-stratum). The effects of the smoothing procedure were signi- ficant. Dr. Ericksen and Dr. Tukey, who were members of the Special Advisory Panel (a group established pursuant to the stipulation, see J.A. 63-65), observed that the decision to pre-smooth" the variances resulted in a substantial increase in the estimate of the net undercount, and in a 70% increase in the estimate of the differential undercount for minority residents. PX 142 (July 11, 1991, letter to Secretary Mosbacher) at 2. Dr. Wachter, who also was a member of the Special Advisory Panel, observed that "[t]he effect of deciding to use pre-smoothed rather than unsmoothed variances is to raise many of the adjustment factors by several percentage points and raise some by more than six percentage points. * * * These are huge changes for a decision of detail." A.R. 492; see Pet. App. 223-224. Indeed, a single highly technical decision to exclude certain data from the pre-smoothing model affected the process in such a manner as to cause a shift of a Representative as compared to the adjustment that would have occurred if those data had been included. Pet, App. 220. 9. The Secretary in his final decision expressed concern that "the statistical artifice of variance ___________________(footnotes) 9 The Bureau excluded from the pre-smoothing model variances from 28 post-strata on the extreme ends of the estimated adjustment factors, the so-called "outliers." Pet. App. 220. Respondents' expert acknowledged at trial that this exclusion was not required by any generally accepted statistical model. Tr. 851-854. ---------------------------------------- Page Break ---------------------------------------- 12 smoothing [wa]s making substantial differences in adjustment factors." Id. at 224. d. Once final ("smoothed") adjustment factors were computed, the number of people in each post-stratum in a block was multiplied by the appropriate adjustment factor. Thus, if the adjustment factor for a post-stratum was 1.10, and a particular block was found in the enumeration to have 10 persons within that post-stratum, an adjustment would have required defendants to multiply that number by 1.10. Consequently, the block total would be adjusted upward by one person (10x 1.1 = 11). This process would be repeated for every block in which members of that post- stratum had been counted in the census. Adjusted counts for larger areas-States, cities, counties, and congres- sional districts-were obtained by adding up the adjusted counts for the blocks within them. See Pet. App. 204-205, 346. e. The adjustment process relied on the assumption that the undercount rate is constant throughout a post- stratum. For example, a post-stratum for white single male renters, aged 20-29, living in New York City "assumes" that the capture probability for a 29-year-old white male law firm associate living in a rented con- dominium in Manhattan is the same as `-the capture probability for a 20-year-old white male laborer who rents a room in Queens. If that "homogeneity" assumption were incorrect, the accuracy of adjusted counts for States, cities, counties, and congressional districts would be open to question. See generally Pet. App. 204-205. The im- portance of the homogeneity assumption was highlighted by Dr. Wachter, who explained that "[t]he homogeneity assumption, in a sense, is the whole ballgame." Tr. 2114. The validity of the homogeneity assumption was there- fore the subject of considerable concern and debate, and the Census Bureau expended substantial effort to address ---------------------------------------- Page Break ---------------------------------------- 13 it. Pet. App. 205-208. In a study known as "[P]roject P12" (id. at 205), the Bureau analyzed factors often correlated with undercounts and found, as the Secretary stated, that "these factors showed significant heterogeneity by state within post-stratum for well over 80% of the post-stratum groups." Id. at 207 (emphasis added). Dr. Wachter testi- fied at trial that the Secretary's conclusions regarding the homogeneity assumption were supported by the de- tailed analysis of the Census Bureau in its Project P12 report, as well as by Dr. Wachter's independent analysis of Bureau data. Tr. 2154-2158. D. The Secretary's Decision Not To Make An Adjustment 1. In making the final determination whether to adjust the 1990 census figures based on the PES, the Secretary of Commerce considered the views of a large and diverse array of persons both inside and outside the Department. He received advice from senior officials in the Economics and Statistics Administration (which includes the Census Bureau) and other senior advisors, as well as individual recommendations from each of the eight members of the Special Advisory Panel. Pet. App. 137, 139-14; see also id. at 258-319 (Secretary's summary and analysis of Special Advisory Panel recommendations). The Secretary also had access to a total of 21 studies that were performed in an effort to evaluate the PES, id. at 168, and he solicited and received extensive public comments on the question whether an adjustment should be made, id. at 137. 2. On July 15, 1991, the date established -by the stipulation as the deadline for the adjustment decision, 10. ___________________(footnotes) 10 Guideline Six stated that, "[i]f sufficient data and analysis of the data are not available in time to publish adjusted counts by July 15, 1991, a determination will be made not to adjust the 1990 census." Pet. App. 241. In his final decision, the Secretary noted that "because of the court imposed deadline for the decision, the analyses of ---------------------------------------- Page Break ---------------------------------------- 14 the Secretary issued his determination that the census headcount would not be statistically adjusted. Pet. App. 135-415. At the outset, the Secretary explained that the crucial question was not whether an adjustment would improve accuracy in terms of absolute numbers at the national level, but whether adjusted numbers would provide an improved account of the way the population is distributed among the States and their subdivisions. See id. at 141, 146-147, 161, 184. He noted that this focus on "distributive" rather than "numeric" accuracy follows from the constitutional purpose of the census, which is to apportion Representatives among the States, and from the other uses of the census, including intrastate electoral redistricting and the distribution of federal funds. See id. at 141, 184, 20(-201. The Secretary also stated that, in accordance with the guidelines he had adopted to imple- ment the parties' stipulation, the unadjusted figures would be regarded as the most accurate unless the adjusted numbers were shown to be more accurate. Id. at 184. 11. ___________________(footnotes) the data are far from complete." Id. at 243. He expressed "particular] concern[ ] about problems in data quality and analysis that were revealed, or occurred, in the final weeks before the decision," Ibid.; see also, e.g., id. at 144, 225 n.108, 244-246. The Secretary concluded, however, that "[notwithstanding my concerns about the effect the July 15, 1991, deadline bad on research efforts, * * * sufficient data exist to permit me to decide whether to adjust the census." Id. at 247. 11 Guideline one stated that "[t]he Census shall be considered the most accurate count of the population of the United States, at the national, State and local level, unless an adjusted count is shown to be more accurate." Pet. App. 151. The Secretary explained in his July 15, 1991, decision that "[t]he true population counts cannot be observed. However, classical statistics provides a standard way of approaching the required inference. In accordance with Guideline One, we take as a working (null) hypothesis that the actual enumerations in fact better characterize the true population. The adjusted counts are an alternative measure and the question is whether the available evidence ---------------------------------------- Page Break ---------------------------------------- 15 The Secretary concluded that a statistical adjustment would be likely to increase the overall numeric accuracy of the census at the national level. Pet. App. 184-185,200. He further concluded, however, that the adjusted figures had not been shown to increase distributive accuracy, and that in fact "the evidence provided by the Census Bureau tends to support the superior distributive accuracy of the actual enumeration." Id. at 185. In support of that deter- mination, the Secretary analyzed a variety of statistical disputes regarding the accuracy of the proposed adjust- ment, explaining that the adjustment process consisted of a series of statistical assumptions, each of which was susceptible to error. He also noted that the difficulty of his decision was increased by the "diversity of opinion among [his] advisors." Id. at 140. "The Special Advisory Panel split evenly as to whether there was convincing evidence that the adjusted counts were more accurate. There was also disagreement among the professionals in the Commerce Department, which includes the Economics and Statistics Administration and the Census Bureau." Ibid. 3. The Secretary analyzed at length the efforts made to estimate the relative accuracy of the adjusted and unadjusted numbers. Pet. App. 185-200. That inquiry focused largely on the Census Bureau's "loss function model," which the Bureau had developed to measure the "loss" from using the census as compared to the "loss" from using an adjusted count. 12. The Secretary concluded ___________________(footnotes) permits us to reject the hypothesis that the census better describes the true population." Id. at 184. 12 The loss is the difference between each count and the "true population." Pet. App. 18$-189. The "true population" is, of course, unknown. Ibid. Indeed, the purpose of an adjustment is to arrive at an estimate of the "true population." To deal with that problem, the Bureau attempted to create a hypothetical "true population" by tak- ---------------------------------------- Page Break ---------------------------------------- 16 that the loss function analysis involved so many uncertainties that its utility was questionable. Id. at 188- 192. In the Secretary's view, however, the analysis tended to show that the adjusted figures were less accurate than the unadjusted count at every level from the state level down. In this regard, the Secretary noted that the Bureau's initial loss function analysis had significantly overstated the distributive accuracy of the adjusted figures at the state and local levels, where the census is actually used for apportionment and funding. See Pet. App. 141. The Secretary pointed out that the original loss function analysis appeared seriously to underestimate the var- iances, or estimates of sampling error, in the adjusted data. Id. at 180-191. The Undercount Steering Committee (a panel of nine senior Census Bureau officials, see J.A. 71) had reported that the actual variances were between 1.7 and 3 times as great as the estimates used in conducting the loss function analysis. Pet. App. 190-191. 13. The Secretary concluded that if the variances were increased by a factor of 2, a figure at the low end of the Bureau's range, "the proportional shares of about 28 or 29 states would be worsened by an adjustment in terms of distributive accuracy." Id. at 191; see id. at 74. The record further showed that, even before that variance correction, an adjustment was estimated to worsen the distributive accuracy of 11 of the 23 metropolitan areas in ___________________(footnotes) ing the proposed adjusted figures and attempting to correct them for various "biases" or errors. See id. at 186-187. The census and the adjusted counts were then compared to the estimated "true population." 13 The Undercount Steering Committee expressed that conclu- sion in an Addendum issued on June 27, 1991-less than three weeks before the Secretary was required by the stipulation to announce his decision for or against adjustment. See Pet. App. 190. ---------------------------------------- Page Break ---------------------------------------- 17 cities with 500,000 or more inhabitants, including the City of New York. Id. at 191. 14. 4. The Secretary also emphasized that even very small changes in the assumptions utilized in developing the adjustment methodology would shift Representatives from State to State. For example, "[o]ne expert found that among five reasonable alternative methods of calculating adjustments, none of the resulting apportionments of the House were the same, and eleven different states either lost or gained a seat in at least one of the five methods." Pet. App. 142 see also id. at 218. The Secretary noted as well the gravity of departing from the 200-year practice of relying upon unadjusted figures to apportion Repre- sentatives among the States, Id. at 138; see also id. at 248. Finally, the Secretary pointed out that "[decisions that may be nearly equally defensible from a technical ___________________(footnotes) 14 Moreover, although the adjustment was in large Part proposed to remedy the disproportionate undercount of minorities, the Secretary explained that the driving force of the proposed adjustment may not have been the undercount at all. He noted that "[o]ne would hope that the predominant determinant of the adjustment would be the number of people missed in the census;' so that "areas with high miss rates get high adjustments." Pet. App. 173. His evaluation revealed, however, that the areas with the three highest omission rates in fact had very different adjustment rates. To a large extent, it turned out, the adjust- ment outcome reflected attempts to correct for overcounting in certain areas and certain respects. Ibid.. The Secretary's decision also suggested the lack of a firm basis for finding a direct correlation between a State's percentage of minority residents and how it fares under an adjustment, noting that the undercount rate for Montana was higher than that of New York. See id. at 194. Indeed, despite its relatively large proportion of minorities (see Bureau of the Census, Department of Commerce, State and Metropolitan Area Data Book 1991, at XIV-XV (4th ed. 1991)), New York's share of the population would actually decrease as a result of the proposed adjustment. See A.R., App. 10, Table 5. ---------------------------------------- Page Break ---------------------------------------- 18 standpoint may have very different outcomes which can he known in advance of the decisions: thereby creating the danger of actual or perceived "manipulation of the census for partisan gain." Id. at 236-237. 5. If Secretary Mosbacher had authorized the particular adjustment proposed for his consideration in July 1991, 15. and those adjusted figures had been made the basis for a reapportionment of Representatives among the States, the result would have been a loss of one Representative each for Wisconsin and Pennsylvania and a gain of one Representative each for California and Arizona. See Pet. App. 17, 250-251. Subsequently, however, the Census Bureau determined that its initial estimation of the net undercount had been substantially overstated. Seepage 4, supra. The Bureau then released corrected adjusted figures for each State. See CAPE Report, Apt. 4. We have been informed by the Department of Commerce that if the method of equal proportions (see page 3, supra) is applied to the corrected adjusted census figures, Wisconsin would lose a Representative and California would gain one (as compared to the current ___________________(footnotes) 15 Guideline Three required that the adjusted figures be derived through "pre-specified" procedures. See Pet. App. 213-214, 215-217. As the Secretary explained in announcing the guidelines, "[t]his guideline specified] that a set of procedures for generating proposed adjusted counts [would] be determined in advance of receiving the 1990 post- enumeration-survey estimates." 55 Fed. Reg. 9841 (1990). The Secretary's final decision made clear that the pre-specification requirement served as a safeguard against actual or perceived political manipulation, by requiring that choices regarding the details of the adjustment methodology would be made at a time when the effects of those choices upon the various States and localities could not yet be foreseen. See Pet. App. 217; see also id. at 83. As a result, although the Secretary recognized the existence of a variety of defensible ad- justment methodologies, he was ultimately presented with a choice between the unadjusted census and a single set of adjusted figures. ---------------------------------------- Page Break ---------------------------------------- 19 apportionment), but the number of Representatives allotted to Arizona and Pennsylvania would remain unchanged. 16. E. The District Court's Decision Following the Secretary's decision, and over the government's objection, the district court permitted extensive discovery and set the case for trial. 17. The trial ___________________(footnotes) 16 Although California alone would gain a Representative from the proposed adjustment (as since corrected), California did not appeal from the judgment of the district court sustaining the Secretary's decision not to make an adjustment. See C.A. App. A387-A390 (notice of appeal). Several municipalities and individual voters in California did appeal. We have serious doubts, however, that municipalities have a right (either on their own behalf or in a parens patriae capacity) to challenge the apportionment of Representatives among the States, especially when the State itself is a party. Cf. Massachusetts v. Mellon, 262 U.S. 447, 485-486 (1923). Similarly, although we may assume that individual residents could be accorded Article III standing to challenge the apportionment of Representatives among the States, there is a question whether they are proper parties to do so, or whether instead the State itself is the proper party to challenge the allocation of political power among the States under the structure prescribed by Article I of the Constitution. Compare Nebraska v. Wyoming, 115 S. Ct. 1933, 1944-1945 (1995). In Montana and Franklin v. Massachusetts, 112 S. Ct. 2767 (1992), for example, the State itself was a plaintiff (along with individual voters). See Montana, 503 U.S. at 458 (referring to "[t]he controversy between Montana and the Government"). More- over, even if individual residents may initiate an action to challenge the apportionment of Representatives among the States, there is a further question whether they may continue such a challenge when their State, representing the interests of all its residents, has chosen to forgo an appeal, and thereby to accept the judgment of the district court that sustained the apportionment. 17 The government argued in the district court, and continues to believe, that judicial review ought to have been confined to the administrative record that was before the Secretary at the time of his decision against adjustment. See, e.g., Florida Power & Light Co. v. ---------------------------------------- Page Break ---------------------------------------- 20 lasted 13 days, involving numerous witnesses- and thousands of pages of exhibits. At the conclusion of the trial, the court granted judgment for the governmental defendants, using the "arbitrary and capricious" standard of the Administrative Procedure Act (APA), 5 U.S.C. 706(2)(A). See Pet. App. 65-66. 18. The court concluded that the Secretary's consideration of each of the guidelines adopted pursuant to the stipulation was reasonable. In particular, it found that the "Secretary's decision to focus on distributive, rather than numeric, accuracy was consonant with the constitutional goal of assuring the most accurate census practicable, given the census's function as a standard by which to distribute political representation and economic benefits." Id. at 77-78. The district court also explored several of the underlying statistical controversies. Although the court stated in passing that it probably would have ordered an adjustment if it had been called upon to decide the issue de novo (id. at 89), it found no basis to set aside the Secretary's de termination. The court stated, in particular, that the Secretary "was neither arbitrary nor capricious" in ___________________(footnotes) Lorion, 470 U.S. 729, 743-744 (1985); Camp v. Pitts, 411 U.S. 138, 142 (1973) (per curiam): compare Franklin, 112 S. Ct. at 2771-27732777- 2779; id. at 2786 & n.22 (Stevens, J., concurring in part and concurring in the judgment). In our view, however, the evidence at trial con- firmed the reasonableness of the Secretary's decision. 18 At a prior stage of the litigation, the district court had con- cluded that "Article I, 2 requires that the census be as accurate as practicable." Pet. App. 109 (brackets and internal quotation marks omitted). The court adhered to that conclusion in its final ruling, while making clear that its application of that standard would reflect def- erence to the Secretary's resolution of disputed technical and policy questions. See id. at 68 (test is "whether the Secretary's decision was arbitrary and capricious in light of the requirement that the decision provide the most accurate census practicable"). ---------------------------------------- Page Break ---------------------------------------- 21 concluding that the superior distributive accuracy of the adjusted figures had not been established. Id. at 77. F. The Court Of Appeals' Decision On appeal to the Second Circuit, respondents did not contest the district court's determination that the Secretary 's decision against adjustment was not arbitrary or capricious. They argued instead that the district court should have analyzed the adjustment issue de novo because the outcome of the decision affected the allocation of Representatives and thus raised constitutional concerns. The court of appeals rejected the argument for de novo review. Pet. App. 23, 34. It nevertheless reversed the judgment of the district court, on a rationale not advanced by respondents. It held that "both 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 pro- tected, and the unadjusted census undercount dispro- portionately disadvantages certain identifiable minority groups." Id. at 33. The court noted that in "one person- one vote" challenges to intrastate districting plans, a court must first determine "whether the population differences among districts could have been reduced or eliminated altogether by a good-faith effort to draw districts of equal population." Id. at 36-37 (quoting Karcher v. Daggett, 462 U.S. 725, 730 (1983)), In that setting, if a court determines that a good-faith effort has not been made, "the State must bear the burden of prov- ing that each significant variance between districts was necessary to achieve some legitimate goal." Id. at 37 (quoting Karcher, 462 U.S. at 731). Applying that analysis here, the court of appeals concluded that "plaintiffs carried their burden of proving ---------------------------------------- Page Break ---------------------------------------- 22 that the Secretary's refusal to adjust the census in accordance with the PES did not reflect an effort to achieve equality as nearly as practicable." Pet. App. 38. The court reached that conclusion by focusing on "the Secretary's acknowledgment" that use of the adjusted population estimates likely would increase the overall numeric accuracy of the census at the national level and would reduce the disproportionate undercount of minori- ties. Ibid. The court stated that the Secretary "gave other factors priority over achievement of greater accuracy" at the national level by, inter alia, "valu[ing] `distributive accuracy' over numerical accuracy." Ibid. It also expressed the view that the Secretary's deter- mination that an adjustment should not be made unless it would result in greater distributive accuracy implied that he did not make the requisite good-faith effort in light of the improved count of the total national population and the improved absolute count of minorities under the adjusted figures. Ibid.; see also id, at 38-39. 19. The court of appeals concluded that the Secretary's decision against adjust- ment was "subject to scrutiny not under an arbitrary-and- capricious standard of review but rather under the more traditional standard applicable to an equal protection claim that a fundamental right has been denied on the basis of race or ethnicity." Id. at 3940. It therefore vacated the judgment of the district court and remanded for further proceedings to determine whether the Secretary's deci- ___________________(footnotes) 19 The court of appeals also noted the Secretary's concern that an adjustment that did not clearly provide a more accurate account of the distribution of the population would introduce fears of political mani- pulation. The court stated that that concern reflected a be- lief that "eliminating the possibility of manipulation of statisti- cal surveys in the future was more important than using the admittedly unmanipulated 1990 PES to achieve a more, accurate overall count ." Pet. App. 38. ---------------------------------------- Page Break ---------------------------------------- 23 sion "(a) furthers a governmental objective that is legiti- mate, and (b) is essential for the achievement of that objective." Id. at 40. Judge Timbers dissented. He expressed agreement with the decision and reasoning of the district court and noted that the majority's holding was in conflict with the decisions of two other courts of appeals. Pet. App. 40 (citing City of Detroit v. Franklin, 4 F.3d 1367 (6th Cir. 1993), cert. denied, 114 S. Ct. 1217 (1994); Tucker v. United States Dep't of Commerce, 958 F.2d 1411 (7th Cir.), cert. denied, 113 S. Ct. 407 (1992)). SUMMARY OF ARGUMENT A. In light of the Census Bureau's extraordinary efforts to conduct an accurate enumeration, and its success in counting 98.49% of the population, respondents bear a heavy burden in contending that the Constitution mandated a statistical adjustment to the 1990 census figures. In fact, the Secretary's decision against adjust- ment was fully consistent with applicable constitutional requirements. The Secretary's emphasis on distributive rather than numeric accuracy, and his determination that the unadjusted figures would be considered the most accurate absent a contrary showing, were "consistent with the constitutional language and the constitutional goal of equal representation." Franklin v. Massachusetts, 112 S. Ct. 2767, 2777 (1992). The focus on distributive accuracy follows necessarily from the constitutions] purpose of the census, which is to determine the apportion- ment of Representatives among the States. And it was surely permissible for the Secretary to require proof that the adjusted population estimates were more accurate before abandoning the 200-year practice of using unadjusted figures. ---------------------------------------- Page Break ---------------------------------------- 24 Based on an exhaustive analysis, the Secretary con- cluded that the available evidence "tends to support the superior distributive accuracy of the actual unumer- ation." Pet. App. 185. Any review of the Secretary's analysis of disputed technical issues must be conducted under a highly deferential standard. The adjustment process involved a host of complex assumptions that were susceptible to error, and the use of plausible alternative assumptions would have affected the apportionment of seats in the House of Representatives. Absent un- equivocal evidence that an adjustment would have improved the distributive accuracy of the census, there is no constitutional basis upon which a court could set aside the Secretary's determination that no adjustment was warranted. B. The district court held that the Secretary "was neither arbitrary nor capricious" in concluding that the superior accuracy of the adjusted figures had not been demonstrated, Pet. App. 77, and respondents did not challenge that holding on appeal. The court of appeals concluded, however, that the Secretary's adjustment decision was subject to review under a heightened stand- ard adapted from this Court's intrastate redistricting de- cisions. The court reasoned, in particular, that "plaintiffs `carried their burden of proving that the Secretary's refusal to adjust the census in accordance with the PES did not reflect an effort to achieve equality as nearly as practicable" Id. at 38. That statement is without basis. The Secretary's acknowledgment that an adjustment would have increased numeric accuracy does not evidences lack of effort to apportion Representatives accurately among the States, since only distributive accuracy is relevant to the apportionment process. Nor did the district court find, implicitly or explicitly, that a good-faith effort had not been made. ---------------------------------------- Page Break ---------------------------------------- 25 The court of appeals also erred in suggesting that equal protection principles required heightened scrutiny of the Secretary's decision because the census dispropor- tionately undercounted racial minorities. It is well established that an equal protection violation based upon racial discrimination requires proof of a discriminatory purpose. Respondents have not alleged, and neither the court of appeals nor the district court found, that either the Secretary or other federal officials acted for the purpose of reducing the electoral power of (or otherwise disadvantaging) minority residents. Accordingly, review of the Secretary's decision under a heightened standard is not warranted. ARGUMENT THE SECRETARY'S DECISION NOT TO UNDER- TAKE A STATISTICAL ADJUSTMENT OF THE 1990 CENSUS WAS FULLY CONSISTENT WITH THE CONSTITUTION In considering whether to make a statistical adjustment of the population figures resulting from the 1990 census, the Secretary of Commerce received sharply conflicting recommendations from independent advisors and from subordinates within his own Department. The question for this Court is not whether, on balance, the Secretary made the appropriate determination as a matter of statistics or policy, but whether his decision not to make an adjustment was within the range of constitutionally permissible options. As we demonstrate below, it was. Resolution of the technical and policy disputes bearing on the adjustment determination is substantially en- trusted to the Secretary's expert judgment and discretion, and judicial review must reflect appropriate deference to his decision. Such deference is mandated by generally applicable principles of administrative law. It is required ---------------------------------------- Page Break ---------------------------------------- 26 as well by the broad delegation of authority expressed in the Constitution itself, which provides that the census shall be conducted "in such Manner as [Congress] shall by Law direct," Art. I, 2, Cl. 3, and in the Census Act, which authorizes the Secretary of Commerce to conduct the decennial census "in such form and content as he may determine," 13 U.S.C. 141(a). And it is required by the compelling interest in the finality of the President's apportionment of Representatives among the States in January 1991, which was based on the unadjusted census data. Secretary Mosbacher's July 15, 1991, decision against adjustment of the 1990 census figures was reasonable and consistent with applicable constitutional standards. We agree with the court of appeals that the Constitution and the Census Act do not bar the use of statistical sampling in conducting the decennial census. See Pet. App. 23-2313 U.S.C. 141 (decennial census), 181 (intercensal population data). The remainder of the court of appeals' analysis, however, reflects a fundamental misunderstanding both of the bases for the Secretary's decision and of the governing constitutional principles. A. The Secretary's Decision Was Consistent With The Text, History, And Purpose Of The Constitution 1. The Census Clause of the Constitution provides that the "actual Enumeration" of the population shall be conducted "in such Manner as [Congress] shall by Law direct." Art. I, 2, Cl. 3. That text affords Congress great discretion to choose from among the broad range of possible ways to conduct the census the particular "Manner" it believes to be the most appropriate. Congress in the Census Act has delegated to the Secretary broad authority to assess and weigh myriad factors, such as feasibility, efficiency, resource allocation, public partici- ---------------------------------------- Page Break -------------------------------------- 27 pation, state and local support, various standards of accuracy, and the degree of confidence in the results. In particular, nothing in the Census Clause prescribes whether (or to what extent) Congress or the Secretary should provide for face-to-face interviews, mailed questionnaires, follow-up visits, or statistical sampling, so long as the process is reasonably calculated to produce state population totals having a sufficient degree of accuracy to constitute an "actual Enumeration" on which to base an apportionment of Representatives among the several States "according to their respective Numbers," Art. I, 2, Cl. 3. Compare United States Dep't of Commerce v. Montana, 503 U.S. 442, 463 (1992) ("The polestar of equal representation does not provide sufficient guidance to allow us to discern a single constitutionally permissible course."). There can be no doubt that the manner in which the Census Bureau conducted the 1990 decennial census satisfied that fundamental standard. The Census Bureau first sought to identify every housing unit in the country and then mailed or delivered a questionnaire to every unit. See Pet. App. 323-324. "The Census Bureau followed up every housing unit for which a questionnaire was not returned," id. at 327 its nonresponse follow-up policy required Bureau enumerators "to make up to six attempts to contact a household member and complete a census questionnaire," id. at 328. The Bureau also developed special procedures directed at individuals-such as residents of group quarters, transients, military per- sonnel, homeless persons found at street and shelter locations, and parolees and probationers-who the Census Bureau believed were particularly likely to be missed. See id. at 326, 330-331. The Bureau's `Were You Counted?" campaign provided individuals who believed that they had been missed with an additional opportunity for ---------------------------------------- Page Break ---------------------------------------- 28 inclusion in the census. Id. at 330. Finally, units of local government were permitted to challenge the housing unit or group quarters count for any block. Id. at 332. 20. The Census Bureau has concluded that that elaborate process counted 98.4% of the nation's population, which was roughly equal to the percentage counted in 1980 and exceeded the percentage counted in every census before 1980. See CAPE Report 4. Especially when measured against historical practice and experience, the "headcount" in 1990 was well within the range of accuracy necessary to fulfill Congress's constitutional responsibility to make an actual enumeration of the population. Compare Franklin v. Massachusetts, 112 S. Ct. 2767, 2777-2778 (1992) (noting importance of historical experience); id. at 2785 (Stevens, J., concurring in part and concurring in the judgment) ("The `usual residence' policy that has guided the census since 1790 provides a" further standard by which to evaluate the Secretary's exercise of discretion."); Montana, 503 U.S. at 465-466. Respondents accordingly assume a heavy burden in attempting to demonstrate that the Census Clause nevertheless imposed a judicially enforceable duty on the Secretary to revise that consti- tutionally acceptable enumeration-e g., by means of a statistical adjustment. Respondents have wholly failed to carry that burden. 2. The Secretary's decision not to undertake an adjustment of the 1990 census figures rested primarily on three determinations. First, the Secretary concluded that distributive rather than total numeric accuracy should be of paramount importance in his decision whether to make a ___________________(footnotes) 20 Each of the country's 51 largest cities challenged at least some blocks, and eight cities challenged more than 2000 blocks. Pet. App. 332. ---------------------------------------- Page Break ---------------------------------------- 29 statistical adjustment, Second, the Secretary determined that, in making his decision, the unadjusted census figures would "be considered the most accurate count of the population of the United States, at the national, State, and local level, unless an adjusted count is shown to be more accurate." Pet. App. 151. Finally the Secretary con- cluded that the adjusted figures had not been shown to improve distributive accuracy and that adjustment was therefore not warranted. 21. The first two determinations are matters of policy that a court may review to the extent of determining whether they are "consistent with the constitutional language and the constitutional goal of equal representation." Franklin, 112 S. Ct. at 2777. The third is an exercise of technical judgment that is properly reviewable, if at all, only under a deferential "rational basis" or "arbitrary and capricious" standard. 22. Each of the determinations at issue easily withstands judicial scrutiny. ___________________(footnotes) 21 Pursuant to the stipulation and guidelines, see page 6, supra, the Secretary's 1991 decision also discussed in detail several other factors bearing on the decision not to adjust. We believe that the Secretary's consideration of those factors was "consonant with, though not dictated by, the text and history of the Constitution." Franklin, 11.2 S. Ct. at 2778. However, disputes concerning the propriety of those aspects of the Secretary's decision are not central to this ease for present purposes. Although the Secretary believed that such factors as encouraging participation in future censuses and the fear of actual or perceived political manipulation lent further support to the decision against adjustment, he stated unequivocally that the adjusted figures had not been shown to improve the accuracy of the census in the relevant (i.e., distributive) sense. Moreover, such factors as concern about actual or perceived political manipulation are not irrelevant to the overall, long-term accuracy of the census Cf. Baldridge v. Shapiro, 455 U.S. 345, 353-354 & n.8, 361 & n.17 (1982). 22 Two courts of appeals have concluded, in rejecting con- stitutional challenges to the Secretary's decision against net adjustment, ---------------------------------------- Page Break ---------------------------------------- 30 a. The constitutional purpose of the census is to determine the apportionment of Representatives among the States, U.S. Const. Art. I, 2, Cl. 3, based on their "respective numbers, counting the whole number of persons in each State." Amend. XIV, 2. 23. Even a ___________________(footnotes) that the Apportionment Clause "does not authorize law suits founded on disagreement with the Census Bureau's statistical methodology" because such a lawsuit "invokes no judicially administrable stan- dards." Tucker v. United States Dep't of Commerce, 958 F.2d 1411, 1418 (7th Cir.), cert. denied, 113 S. Ct. 407 (1992); accord City of Detroit v. Franklin, 4 F.3d 1367,1375-1378 (6th Cir. 1993), cert. denied, 114 S. Ct. 1217 (1994). We shall assume for present purposes, however, that the Secretary's decision is reviewable under a "rational basis" or "arbitrary and capricious" standard. Cf. Franklin, 112 S. Ct. at 2786 n.22(Stevens, J., concurring in part and concurring in the judgment). 23 As explained above, see pages 18-19, supra, use of the parti- cular adjustment proposed for Secretary Mosbacher's consideration would have resulted in a loss of one Representative each for Wisconsin and Pennsylvania and a gain of one each for California and Arizona, while use of the "corrected" adjusted figures would have resulted in a transfer of one Representative from Wisconsin to California. The large majority of the plaintiffs in this case, however, are States other than California and Arizona, or municipalities or residents outside of California and Arizona. Although many of those plaintiffs may have suffered injury-in-fact due to the impact of the Secretary's decision upon the distribution of federal funds, cf. City of Detroit, 4 F.3d at l374. 1375, there is a substantial question whether plaintiffs outside Arizona and California may contest the allocation of Representatives among the States or invoke a heightened standard of review based upon the purported effect of the Secretary's decision upon the right to vote. See United States Dep't of Labor v. Triplett, 494 U.S. 715, 720 (1990) ("Ordinarily, of course, a litigant must assert his own legal rights and interests, and cannot rest his claim to relief on the legal rights or interests of third parties.") (internal quotation marks omitted); cf. Franklin, 112 S. Ct. at 2776 (plurality opinion) (Massachusetts lacked standing to challenge accuracy of census data absent allegation that it would have been allocated an additional Representative if different data had been used). Nor does the Administrative Procedure Act fur- ---------------------------------------- Page Break ---------------------------------------- 31 dramatic undercounting of the total population would not be inconsistent with the goal of fair apportionment if each State's share was accurately determined. Cf. Karcher v. Daggett, 462 U.S. 725, 736 (1983) ("the existence of a one- percent undercount would be irrelevant to population deviations among districts if the undercount were distri- buted evenly among districts"). Conversely, a precise determination of the nation's total population as a result of a statistical adjustment would not improve the apportionment process if the States' respective shares were inaccurately estimated under the adjustment. Thus, the Secretary's focus on distributive rather than numeric accuracy was "consistent with the constitutional lan- guage and the constitutional goal of equal representation." Franklin, 112 S. Ct. at 2777 accord Pet. App. 77-78 (district court's decision). Respondents acknowledge (Br. in Opp. 39) that "distributive accuracy is the object of adjustment." See also Resp. C.A. Reply Br. 14 n.4 (stating that respondents "agree (as do, for that matter, all experts on either side of the adjustment question) that it is distributive accuracy that is of paramount importance for the constitutional and legal purposes for which the census is conducted"). b. The Secretary's determination to resolve the issue of an adjustment to the 1990 census by regarding the unadjusted figures as the most accurate unless a contrary showing was made is also a judgment "consonant with, though not dictated by, the text and history of the Constitution." Franklin, 112 S. Ct. at 2778. Because unadjusted headcounts had been used for 200 years, the Constitution surely did not bar the Secretary from adopting a working hypothesis that unadjusted figures ___________________(footnotes) nish a statutory basis for challenging the decennial census. Franklin, 112 S. Ct. at 2773-2776. ---------------------------------------- Page Break ---------------------------------------- 32 derived from the 1990 census would be treated as the most accurate unless alternative numbers were shown to be better. See pages .28-29, supra. Nor could "the constitutional goal of equal representation" among the States, Franklin, 112 S. Ct. at 2777, plausibly bethought to require the Secretary to use adjusted figures that he had not found to be more accurate than the unadjusted numbers in achieving that goal. In any event, the Secretary did not simply determine that proponents of adjustment had failed to prove the greater distributive accuracy of adjusted figures; he affirmatively concluded that ''the evidence provided by the Census Bureau tends to support the superior distributive accuracy of the actual enumeration." Pet. App. 185. 24. c. As Census Bureau Director Bryant observed in her recommendation to the Secretary in favor of an ___________________(footnotes) 24 The court of appeals suggested in passing that the Secretary regarded the adjusted and unadjusted numbers equally accurate, see Pet. App. 38, and stated that the adjusted figures would improve accuracy for the areas in which up to two-thirds of the population resided, see id. at 39. Those characterizations of the record reflect a basic misreading of the Secretary's decision and the evidence at trial. The two-thirds figure was an original estimate provided by the Census Bureau using its `loss function model." The Secretary's decision explained in considerable detail that the loss function model had failed to take into account variances that the Undercount Steering Commit- tee subsequently reported to be significantly understated. See id. at 185-192. As the district court explained, "when the Secretary em- ployed a statistical variance toward the low end of the acceptable range envisioned by the [Undercount Steering Committee], he found that the proportional shares of 28 or 29 states would be worsened by adjustment," Id. at 74. As noted above, subsequent research by the Census Bureau and further evidence presented at trial demonstrated that the original loss function analysis substantially overstated the accuracy of the adjusted data. See page 16, supra. The court of appeals' reliance on data that the Secretary had specifically corrected underscores the error of its analysis. ---------------------------------------- Page Break ---------------------------------------- 33 adjustment, "[adjustment is an issue about which reason- able men and women and the best statisticians and demographers can disagree." J.A. 73. The Secretary engaged in exhaustive analysis of the pertinent evidence, however, and his resolution of disputed technical questions can be reviewed only under an extremely deferential standard. See page 29 and note 22, supra. There is, in particular, no basis for respondents' contention in the court of appeals that the central statistical issue in this case-whether use of the adjusted figures would have improved the distributive accuracy of the census-should have been resolved by the district court de novo on the ground that the case involves a constitutional claim. De novo review of the expert agency's technical judg- ments would be especially inappropriate in the present context, where the pertinent constitutional and statutory provisions vest broad discretion in the political Branches. See Baldrige v. Shapiro, 455 U.S. 345, 361 (1982). The Census Clause provides that the census shall be conducted "in such Manner as [Congress] shall by Law direct." Art. I, 2, Cl. 3. The Census Act similarly contains a broad delegation of authority to the Executive Branch, directing the Secretary to conduct the decennial census "in such form and content as he may determine." 13 U.S.C. 141(a). The governing constitutional and statutory provisions therefore counsel the courts to exercise particular caution in reviewing the Secretary's discretionary deter- minations regarding the conduct of the census. Moreover, although issues pertaining to the accuracy of the census implicate constitutional values, resolution of the statistical dispute in this case does not involve the interpretation of constitutional text. Rather, it involves an exercise of technical expertise that lies uniquely within the competence of the Executive Branch officials to ---------------------------------------- Page Break ---------------------------------------- 34 whom Congress has entrusted the extraordinarily complex task of conducting the decennial census. As this Court has repeatedly recognized, where "analysis of the relevant documents requires a high level of technical expertise, [courts] must defer to the informed discretion of the responsible federal agencies." Marsh v. Oregon Natural Resources Council, 490 U.S. 360, 377-378 (1989) (internal quotation marks omitted); see also Thomas Jefferson University v. Shalala, 114 S. Ct. 2381, 2387 (1994); Baltimore Gas & Electric Co. v. Natural Resources Defense Council, Inc., 462 U.S. 87, 103 (1983); Kleppe v. Sierra Club, 427 U.S. 390, 412 (1976); International Fabricare Institute v. EPA, 972 F.2d 384, 400 (D.C. Cir. 1992) ("As we are not scientists and must defer to the Agency's judgments on matters within its technical competence, our task is to assure that they be reasoned, not that they be right."). That respondents' attack on the Secretary's decision is framed as a constitutional challenge does not detract from the respect for insti- tutional competence that underlies the rule of deference to an expert agency's technical judgments. Finally, de novo resolution by the district courts of the question whether an adjustment would increase distri- butive accuracy would create the possibility of incon- sistent factual findings by courts in different jurisdictions -each reviewable on appeal only under the "clearly erroneous" standard of Federal Rule of Civil Procedure 52(a). See Anderson v. City of Bessemer City, 470 U.S. 564, 573-576 (1985). Where a variety of reasonable alternatives were presented for the Secretary's consideration, 25. no definitive resolution of the conflict ___________________(footnotes) 25 As explained above, see note 15, supra, the Secretary was ultimately con fronted, due to the pre-specification requirement, with a choice between the unadjusted census and a single set of adjusted ---------------------------------------- Page Break ---------------------------------------- 35 would be possible under ordinary principles of appellate review. And because the constitutional purpose of the census is to determine the manner in which a fixed number of Representatives should be apportioned among the States, the Secretary could not utilize adjusted figures in some regions of the country and unadjusted figures in others. See City of Detroit v. Franklin, 4F. 3d 1367, 1378 (6th Cir. 1993) ("Plaintiffs' request for a statistical adjustment for Michigan alone cannot be considered in isolation. It would be impractical for this Court to order ([and] the Secretary to do) such an adjustment without considering the consequences for the other states and other cities."), cert. denied, 114 S. Ct. 1217 (1994). 26. ___________________(footnotes) figures. The district court framed the question before it as whether the Secretary had made a permissible choice between those two alterna- tives, and respondents have not contended that some other set of numbers should have been used. If respondents (or some other plaintiffs) had made such an argument, however, and if the role of the district court were to resolve pertinent statistical disputes de novo, there would be no harrier to courts in different jurisdictions determin- ing that each of several adjustment methodologies would have maxi- mized distributive accuracy. 26 Nor would it be a satisfactory solution for conflicting district court determinations regarding the choice of census figures that would maximize distributive accuracy to be reviewed de novo by the courts of appeals, and ultimately by this Court TO the institutional limitations applicable to all federal courts would be added the disadvantages associated with requiring appellate judges to function as triers of fact. See Anderson, 470 U.S. at 574-575. Those disadvantages are parti- cularly severe where resolution of factual disputes requires credibility determinations. Id. at 575. Where (as here) expert testimony is pre- sented on each side of the adjustment question, a court's attempt to determine de novo whether adjustment would enhance distributive accuracy would necessarily involve an assessment of the credibility of the competing experts. ---------------------------------------- Page Break ---------------------------------------- 36 d. The district court held that the Secretary "was neither arbitrary nor capricious" in concluding that the superior distributive accuracy of the adjusted figures had not been demonstrated. Pet. App. 77. That holding was not contested on appeal, and it was clearly correct. As explained above, the formulation and evaluation of the proposed adjustment involved a series of complex pro- cesses and assumptions in a highly technical area at the frontier of generally accepted statistical science. In the Secretary's view, however, the analysis tended to show that the adjusted figures were less accurate than the unadjusted count at every level from the state level down. See pages 15-17, supra. There were, moreover, substantial uncertainties in the record that weighed against a determination that the particular adjustment under consideration was distinctly meritorious and superior to the unadjusted figures. Thus, the Secretary noted that "[o]ne expert found that among five reasonable alternative methods of calculating adjust- ments, none of the resulting apportionments of the House were the same, and eleven different states either lost or gained a seat in at least one of the five methods." Pet. App. 142; see also id. at 218. 27. The evidence at trial confirmed that the use of reasonable alternative assumptions in the ___________________(footnotes) 27 The Secretary was referring to Dr. Wachter's study, which considered five alternative adjustment calculations. See Pet. App. 142, 218, 303-304. Dr. Wachter compared adjustments using the smoothed adjustment estimates with the raw estimates. Id. at 218. Recognizing the problems inherent in the imputation of persons who could not be matched with census data, Dr. Wachter also substituted two alterna- tive assumptions for dealing with unmatched persons. Ibid. Dr. Wachter's fifth estimate employed adjustment factors formulated on the basis of data within each State rather than the multi-State aggre- gations used in the post-strata, Ibid. Each alternative approach produced a different apportionment result. Id. at 142, 218,304. ---------------------------------------- Page Break ---------------------------------------- 37 adjustment process could have altered the apportionment of Representatives. For example, both plaintiffs' and defendants' experts recognized that alternative geo- graphic groupings could have been used in formulating post-strata and that those alternatives would have produced different adjusted population figures. See pages 7-8, supra. Moreover, experts on both sides recognized that the decision to "pre-smooth" the variances, and then to "smooth" the adjustment factors, significantly affected the adjustment outcome. See pages 9-11, supra. 28. The uncertainties discussed at length in the Secre- tary's decision were underscored by the trial testimony of Dr. Robert Fay, a Senior Mathematical Statistician at the Census Bureau and one of the members of the Undercount Steering Committee who had initially supported adjustment. Dr. Fay testified that there had been a great deal of uncertainty at the time of the July 15, 1991, decision surrounding the original estimated variances used in the loss function model. Because of that uncertainty, Dr. Fay, on behalf of the Undercount Steering Committee, attempted to estimate the size of the variances based on judgment, not calculations. Tr. 1906. ___________________(footnotes) 28 Dr. David Freedman, defendants' smoothing expert, testified that the dramatic effects of smoothing militated against an adjustment decision because the assumptions in the smoothing model had not been validated. Tr. 2385-2386. Similarly, Dr. Wachter stated that the substantial effects of pre-smoothing were especially troubling because "[p]re-smoothing the variances that go into smoothing the adjustment factors is at three removes from the data. It incorporates little, if any, further empirical information. It depends entirely on a set of further assumptions." A.R. 495; see Pet. App. 224. Indeed, because of the serious concerns with the smoothing model, the Census Bureau aban- doned the smoothing process altogether in devising a proposal for a possible statistical adjustment for use in the intercensal estimates program. See 58 Fed. Reg. 72-73 (1993). ---------------------------------------- Page Break ---------------------------------------- 38 On that basis, Dr. Fay had "roughly guessed" that there appeared to be enough of a "margin of safety" to support a recommendation in favor of adjustment. Tr. 1907. After the July 15, 1991, decision, Dr. Fay conducted additional research and determined that the original estimates failed to account for all the sources of variance in the smoothing model. When those previously un- accounted-for variances were taken into account, Dr. Fay determined that the initial variance estimates were understated by approximately a factor of 2. Tr. 1914 1917. 28. That result, Dr. Fay explained, represents a significant increase in the uncertainty of the PES estimates. Tr. 1920. As a result of this work, Dr. Fay no longer advocated the adjustment recommended by the Committee and sought by respondents. Tr. 1920-1921. Insure, the adjustment process and the loss function analysis designed to evaluate that process were based on technical assumptions and hypotheses that were subject to substantial uncertainty. The record demonstrated that had several of those assumptions been altered in a manner that would itself have been reasonable, the apportionment of Representatives would have been different from the one resulting from the particular adjustment submitted to the Secretary. For that reason, and in the absence of evidence demonstrating that even that particular adjustment proposal would have improved the distributive accuracy of the census, the Secretary's decision not to undertake the adjustment was clearly permitted under the Constitution. That is especially so in light of the need for the Secretary to make a decision by July 15, 1991, the deadline to which respondents stipulated in the district court and which ___________________(footnotes) 28 Two of plaintiffs' experts, Dr. Welter and Dr. Tukey, tes- tified that such results would have caused them to review their recommendations. Tr. 726-727, 1420. ---------------------------------------- Page Break ---------------------------------------- 39 they did not challenge in the court of appeals. As that deadline approached, the Secretary concluded that the state of the administrative record and the various expert evaluations of it were not yet sufficiently settled to permit him to make a decision in favor of adjustment with the degree of confidence he believed was appropriate for such a significant departure from historical practice. See Pet. App. 140-141,144. B. The Court Of Appeals Erred In Holding That The Secretary's Decision Should Be Subject To Heightened Scrutiny And That Respondents Established A Prima Facie Case Of A Constitutional Violation The court of appeals analogized the Secretary's 1991 decision to a State's decision to create congressional or state legislative districts of unequal size. Under this Court's precedents, parties challenging a State's dis- tricting decisions first bear the burden of proving that differences between the populations of individual districts within a State are "not the product of a good-faith effort to achieve population equality." Karcher, 462 U.S. at 740; see id. at 730-731 (discussing burden of proof). If the plaintiffs are able to make that showing, "the burden shift[s] to the State to prove that the population deviations in its plan were necessary to achieve some legitimate state objective." Id. at 740. In the present case, the court of appeals concluded that "plaintiffs amply showed that the Secretary did not make the required effort to achieve numerical accuracy as nearly as practicable," and that "the burden thus shifted to the Secretary to justify his decision not to adjust the census in a way that the [district] court found would for most purposes be more accurate and would lessen the disproportionate counting of minorities." Pet. App. 39. ---------------------------------------- Page Break ---------------------------------------- 40 The court of appeals thus treated the Secretary's decision as suspect on the basis of (a) the Secretary's acknow- ledgment that an adjustment would probably improve total numeric accuracy at the national level, (b) a purported district court finding that the adjusted figures "would for most purposes be more accurate," ibid., and (c) the uncontested fact that the undercount of racial minorities was greater than the undercount of nonminority residents. The court of appeals' attempt to apply the Karcher analysis in this quite different setting does not withstand scrutiny. 30. ___________________(footnotes) 30 Any effort to apply a Karcher-type analysis to the Executive Branch's conduct of the census and certification of state population figures would have to take account of certain significant differences between the census and apportionment process and a State's districting decisions. First, an attempt to ensure that congressional districts in different States are precisely equal in population "is constrained by three requirements. The 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. Thus, notwithstanding the federal government's efforts to achieve population equality "as nearly se is practicable: Karcher, 462 U.S. at 730, disparities between districts in different States are likely to be of a magnitude that would be unacceptable in the context of state districting policy. Second, because of the technical complexity of the enumeration process and the multitude of discretionary determinations that it entails, a reviewing court should be especially cognizant of the difference between the question whether the agency has in fact maximized population equality and the question whether it has made a good-faith effort to do so. See pages 44-45, infra. Finally, the sweeping constitutional grant of authority to Congress to conduct the census "in such manner as they shall by Law direct," Art. I, 2, Cl. 3, suggests that a broader range of legitimate governmental objectives may be available to justify use of census figures that do not maximize population equality than would be true of decisions made in the state districting context. ---------------------------------------- Page Break ---------------------------------------- 41 1. As an initial matter, the extraordinary effort expended by the Census Bureau in conducting the 1990 decennial census, and the success of the Bureau in counting 98.4% of the nation's population, are sufficient in themselves to refute any suggestion that the headcount totals respondents challenge were "not the product of a good-faith effort to achieve population equality." Karcher, 462 U. S. at 740. As we have explained above (see pages 27- 28, supra.), the headcount was well within the range of accuracy necessary to fulfill Congress's constitutional responsibility to make an actual enumeration of the population for apportionment purposes. For that reason alone, even if we assume arguendo that the Karcher framework should reapplied by analogy here, respondents have failed to carry their threshold burden under that framework. 2. In the state redistricting cases, where a given population is to be divided into a fixed number of districts, courts can readily "decree equality." Tucker v. United States Dep't of Commerce, 958 F.2d 1411, 1418 (7th Cir.), cert. denied, 113 S. Ct. 407 (1992). In that setting, if significant differences between the sizes of individual districts are conceded to exist, a court can confidently conclude, not simply that the State has failed to achieve population equality, but that it has failed to make a good- faith effort to do so. This case, by contrast, involves the antecedent task of determining the population, a complex inquiry circum- scribed by limitations in resources and statistical science. That task involves numerous technical and policy de- cisions as to which reasonable alternatives are available. 31. ___________________(footnotes) . 31 Each choice among such alternatives would in turn invite litigation under the court of appeals' decision. In addition to the sev- eral lawsuits seeking to compel a statistical adjustment to the census, ---------------------------------------- Page Break ---------------------------------------- 42 ___________________(footnotes) numerous suits have been filed challenging a host of other census- related decisions. See, e.g., Franklin v. Massachusetts, supra (chal- lenging Census Bureau's method of allocating certain overseas federal employees to particular States); National Law Center on Homelessness and Poverty v. Brown, appeal pending, No. 94-5312 (D.C. Cir.) (argued Oct. 6, 1995) (challenging Census Bureau's procedures for locating and enumerating homeless persons) Carey v. Klutznick, 637 F.2d 834 (2d Cir. 1980) (seeking, in addition to a statistical adjustment to 1980 census, an order requiring the Census Bureau to process certain forms and to compare Census Bureau records of New York City residents with a computerized list of 1.2 million persons in New York City eligible for Medicaid); Borough of Bethel Park v. Stans, 449 F.2d 575 (3d Cir. 1971) (challenging manner in which Census Bureau determined residence of college students, persons confined to institutions, federal personnel and their dependents located outside the United States, and other citizens studying or working outside the United States who are not federal employees); District of Columbia v. United States Dep't of Commerce, 789 F. Supp. 1179 (D.D.C. 1992) (challenging Census Bureau's decision to count as residents of Virginia inmates of Lorton Correctional Facility, which is located in Virginia, but operated by the District of Columbia), appeal voluntarily dismissed, No. 92-5212 (D.C. Cir.) Ridge v. Verity, 715 F. Supp. 1308 (W.D. Pa. 1989) (challenging inclusion of illegal aliens in 1990 census); City of Willacoochee v. Baldrige, 556 F. Supp. 551 (S.D. Ga. 1983) (challenging accuracy of population count of Georgia city); Carey v. Klutznick, 503 F. Supp. 874 (N.D. III. 1980) (suit to require Census Bureau to keep its offices in Cook County, Illinois, open and to continue the 1980 census to ensure that all housing units in Cook County were canvassed); City of Philadelphia v. Klutznick, 503 F. Supp. 663 (E.D. Pa. 1980) (challenging, among other things, Census Bureau's alleged hiring of unskilled enumerators and its failure to properly conduct local review); Federation for American Immigration Reform v. Klutznick, 486 F. Supp. 564 (D.D.C.) (three-judge court) (challenging inclusion of illegal aliens in 1980 census), appeal dismissed, 447 U.S. 916 (1980); City of Camden v. Plotkin, 466 F. Supp. 44 (D.N.J. 1978) (challenging manner in which Census Bureau conducted its 1976 "Pretest Census"); West End Neighborhood Corp. v. Stans, 312 F. Supp. 1066 (D.D.C. 1970) (challenging Census Bureau's decision to replace door-to-door method of ---------------------------------------- Page Break ---------------------------------------- 43 Karcher itself recognized the distinction between the determination of population by the decennial census, in which an undercount is to be expected, and the apportion- ment of seats in a state legislature based on the census totals, in which mathematical exactitude is achievable. See 462 U.S. at 731, 735-738. A reviewing court's belief that alternative measures would more likely than not have increased the distributive accuracy of the census scarcely suggests that the Secretary has failed to make a good- faith effort to achieve a fair apportionment. The Consti- tution allows for varying approaches to the conduct of the census and varying measures of accuracy and equity. In light of the Constitution's plenary grants of authority to Congress in the Apportionment Clause, the Necessary and Proper Clause, and Section 5 of the Fourteenth Amend- ment, the Secretary's "apparently good-faith choice of a method" for conducting the census (pursuant to broad discretionary authority conferred on him by Congress) "commands far more deference than a state districting decision." Montana, 503 U.S. at 464. 3. Even under the court of appeals' analysis, a neces- sary predicate for concluding that the Secretary had failed to make "a good-faith effort to achieve population equality" in this setting, Karcher, 462 U.S. at 740, would be that the Secretary had determined that the proposed adjustment would increase the accuracy of the States' "respective Numbers" (i.e., distributive accuracy), but had never- theless declined to make an adjustment. Only then could the burden shift to the government to show that the decision against adjustment was "necessary to achieve some legitimate [federal] objective." Ibid. The court of appeals simply erred, however, in inferring the lack of a ___________________(footnotes) enumeration with mail-out/mail-back procedure); Quon v. Stans, 309 F. Supp. 604 (N.D. Cal. 1970) (same). ---------------------------------------- Page Break ---------------------------------------- 44 good-faith effort to achieve population equality from the Secretary's acknowledgment that an adjustment would likely increase numeric accuracy at the national level. As discussed above (see pages 30-31, supra), the Secretary's focus on distributive rather than numeric accuracy was fully justified. The court of appeals did not explain how the goal of equal representation for equal numbers of people in the several States could be better served by focusing on the total numeric accuracy of the census figures at the national level. As the district court recognized, "[t]he Secretary's decision to focus on distri- butive, rather than numeric, accuracy was consonant with the constitutional goal of assuring the most accurate census practicable, given the census's function as a standard by which to distribute political representation and economic benefits." Pet. App. 77-78. See also Br. in Opp. 39 (acknowledging that "distributive accuracy is the object of adjustment"). 4. The district court did not purport to issue formal Endings regarding the relative accuracy of the adjusted and unadjusted census figures. The court's treatment of the issue was somewhat enigmatic. On the one hand, the court stated (without significant explanation) that it was "satisfied that for most purposes the PES resulted in a more accurate-or to be statistically fashionable, a less inaccurate-count than the original census: Pet. App. 59, and it further observed that "were this Court called upon to decide this issue de novo, I would probably have ordered the adjustment," id. at 89. On the other hand, the court stated that "plaintiffs' failure to illustrate affirmatively the superior accuracy of the adjusted counts either (1) at any level mentioned in Guideline One, or (2) for any reasonable definition of accuracy, is sufficient to support a finding that Guideline One favors use of the original census counts." Id. at 78. ---------------------------------------- Page Break ---------------------------------------- 45 In any event, the district court clearly did not find that the Secretary had failed to make a good-faith effort to maximize distributive accuracy. A determination that the Secretary's stated conclusions were arbitrary or un- reasonable conceivably could have suggested the lack of such an effort, cf. Harlow v. Fitzgerald, 457 U.S. 800, 815- 819 (1982) (official's entitlement to "good faith" immunity is determined by reference to the objective reasonableness of the challenged conduct); but the district court held that the Secretary was "neither arbitrary nor capricious" in concluding that the superior distributive accuracy of the adjusted numbers had not been established. Pet. App. 77. There is consequently no basis for the court of appeals' assertion (id. at 38) that "the findings of the district court * * * plainly show that plaintiffs carried their burden of proving that the Secretary's refusal to adjust the census in accordance with the PES did not reflect an effort to achieve equality as nearly as practicable." 5. The court of appeals also referred repeatedly to the disproportionate undercount of racial minorities (see, e.g., Pet. App. 33, 38, 39, 40), invoking that differential to support the proposition that the Secretary's decision not to adjust should be reviewed "under the more traditional standard applicable to an equal protection claim that a fundamental tight has been denied on the basis of race or ethnicity." Id. at 39-40. The court of appeals was simply wrong, however, in its belief that "[e]qual protection analysis requires that heightened scrutiny be given to the Secretary's decision to adhere to an acknowledged under- count that concededly impacts minority groups more severely than nonminority groups." Id. at 34. To the contrary, establishment of an equal protection violation based upon racial discrimination requires proof of a discriminatory purpose. Washington v. Davis, 426 U.S. 229, 239-245 (1976). The intentional undercounting of ---------------------------------------- Page Break ---------------------------------------- 46 racial minorities in the conduct of the census would be a matter of grave concern to the integrity of our democratic process. The courts would be open to consider such claims under the analytical framework followed in Village of Arlington Heights v. Metropolitan Housing Dev. Corp., 429 U.S. 252 (1977), and its progeny, and to afford relief where appropriate, see Franklin, 112 S. Ct. at 2776-2777 (opinion of O'Connor, J.). Respondents have not contended, however, and neither the court of appeals nor the district court found, that either the Secretary or other federal officials acted for the purpose of reducing the electoral power of (or otherwise disadvantaging) minority residents. Indeed, in their briefs in the court of appeals, respondents did not even cite the Fifth Amendment's Due Process Clause, on which an equal protection claim against the federal government must be based. In any event, the evidentiary consid- erations identified in Arlington Heights do not suggest a basis for a claim of racial discrimination in violation of the Fifth Amendment. The Secretary's decision against adjustment was fully in keeping with historical practice, a fact that suggests a lack of discriminatory intent. Compare Arlington Heights, 429 U. S. at 267 & n.16 (abrupt departures from prior policy in ways that disadvantage minorities may constitute evidence of discriminatory motivation); id. at 270 (defendant "has applied [its] policy too consistently for [the Court] to infer discriminatory purpose from its application in this case"). 32. Secretary Mosbacher's exten- ___________________(footnotes) 32 The Court in Arlington Heights also observed that "[d]epart- ures from the normal procedural sequence also might afford evidence that improper purposes are playing a role." 429 U.S. at 267. In the present case, the Secretary departed from customary procedures only by devoting enhanced consideration to the possibility of a statistical adjustment and explaining his ultimate decision in greater detail. ---------------------------------------- Page Break ---------------------------------------- 47 sive explanation published in the Federal Register sets out a host of race-neutral reasons for his decision not to adjust. Compare Arlington Heights, 429 U.S. at 268 (''The legislative or administrative history may be highly relevant, especially where there are contemporary state- ments by members of the decisionmaking body, minutes of its meetings, or reports."). The government's extra- ordinary efforts to enumerate minority residents 33. and to identify persons who initially failed to return question- naires (see pages 27-28, supra) also belie any inference that Census officials pursued a deliberate policy of under- counting minority residents. A disparate impact upon racial minorities, though not itself sufficient to establish a constitutional violation, may sometimes furnish evidence of discriminatory purpose. See Arlington Heights, 429 U.S. at 265-266; Davis, 426 U.S. at 242. In the present case, however, the ultimate effect of the undercount upon minority residents is both indirect and of uncertain magnitude. Those affected by ___________________(footnotes) 33 In her recommendation to Secretary Mosbacher, Census Bureau Director Bryant stated (J.A. 76): For the 1990 census, the Census Bureau mounted the most ex- tensive effort ever to enumerate Blacks and other minorities. This included the hiring of 280 community outreach workers who worked in communities two years before census taking; involve- ment of 56,000 community organizations-mostly minority but also city and state Complete Count Committee outstanding cooperation from Black and Spanish language media in running public service announcements and programs about the importance of the census; and the hiring of follow-up enumerators from minority populations, bilingual and multi-language enumerators, and residents of public housing projects and American Indian reservations to enumerate persons in their neighborhoods. Despite this effort, the undercount differential was not reduced below its historical level. ---------------------------------------- Page Break ---------------------------------------- 48 the undercount, it should be emphasized, are not the uncounted individuals themselves; an individual derives no concrete benefit from inclusion in the census, nor does he or she suffer any concrete injury from being left uncounted. Rather, those affected by the undercount include all residents of geographic areas that are assigned less than their true share of the population due to errors in the enumeration process. More precisely for purposes of this case, which involves a constitutional challenge based on the impact of the unadjusted census on the apportionment of Representatives among the States, those affected by the undercount include all residents (minority and non-minority, men and women, renters and home- owners) of a State that was assigned less than its "true" share of the total population of the nation and that, as a result, was assigned less than its "true" share of the total number of Representatives. Respondents assert that "[b]ecause members of mi- nority groups are not evenly distributed geographically, the consequence of the differential undercount is that areas in which members of minority groups are concentrated are systematically deprived of political representation and funding they would otherwise have received." Br. in Opp. 9. 34. Whatever the intuitive plausi- bility of that theory, however, respondents made no comprehensive effort at trial to demonstrate whether and to what extent States with large minority populations ___________________(footnotes) 34 As noted above, see page 5, supra, the PES also indicated that men, the young, and renters were undercounted at a higher rate than women, older residents, and homeowners. Assuming that mem- bers of those groups "are not evenly distributed geographically: it would be equally plausible to hypothesize that areas in which those demographic groups are concentrated would be "systematically de- prived of political representation and funding they would otherwise have received." ---------------------------------------- Page Break ---------------------------------------- 49 would as a rule have been credited with a higher share of the country's population if the proposed adjustment had been made, let alone to demonstrate that their shares of the population would have been more accurately estimated by use of adjusted figures. See note 14, supra (noting that New York State's share of the nation's population would have been reduced by an adjustment). But in addition, respondents have not contended, and the courts below did not find, that the indirect and uncertain impact of the undercount upon the electoral power of minority residents furnished persuasive evidence of discriminatory intent. Absent a showing of intentional discrimination, the differential undercount of minorities at the national level provides no basis for subjecting the Secretary's decision against adjustment to a heightened standard of review. 35. Respondents also argue that if use of an adjustment would alleviate the racially differential undercount at the national level, then, " ceteris paribus, adjustment would improve the distributive accuracy of the census." Br. in Opp. 43. The flaw in that argument lies in its premise ("ceteris paribus") that other things would remain equal -that the proposed statistical adjustment would lessen the racial disparity without introducing new errors or uncertainties into the enumeration. After extensive con- sideration of the views of experts on both sides of the adjustment question, the Secretary reached a different conclusion. He noted that he was "deeply troubled by this problem of differential participant ion and undercount of minorities," but concluded that "an adjustment does not address this phenomenon without adversely affecting the ___________________(footnotes) 35 Of course, the differential undercount of minority groups (and other demographic groups) may properly be taken into account by Congress or the Secretary in deciding whether to make a statistical adjustment as a matter of policy. ---------------------------------------- Page Break ---------------------------------------- 50 integrity of the census." Pet. App. 139. That determi- nation was based on a reasoned analysis of the pertinent data and of the conflicting recommendations of the Secretary's advisors. Accordingly, even if we assume that a Karcher-type analysis applies to the Secretary's de- cision not to make a statistical adjustment to a census headcount that otherwise fulfills the purposes of the Census Clause, the court of appeals erred in concluding that respondents had carried their threshold burden of showing that the Secretary had failed to make a good-faith effort to maximize population equality. 6. With the issues of Karcher and racial discrimination thus put to one side, the only remaining issue is whether the Secretary acted within the range of discretion afforded by the Constitution in declining to make a statistical adjustment. The district court correctly concluded that the Secretary's determination was not arbitrary and capricious. Because respondents have not challenged that conclusion, the judgment of the district court should be affirmed. Now, halfway through the decade governed by the 1990 census, it is time to bring this challenge to that census to a close. ---------------------------------------- Page Break ---------------------------------------- 51 CONCLUSION The judgment of the court of appeals should be reversed, and the case should be remanded with directions to enter judgment for petitioners. Respectfully submitted. DREW S. DAYS, III Solicitor General FRANK W. HUNGER Assistant Attorney General EDWIN S. KNEEDLER Deputy Solicitor General MALCOLM L. STEWART Assistant to the Solicitor General MARK B. STERN MICHAEL S. RAAB Attorneys NOVEMBER 1995