(2000)

85 Op. Att'y Gen. 18

Opinion

Dear President Miller,

You have asked for our opinion on the constitutionality of employing statistically adjusted census data — as opposed to the traditional head-count data — in the preparation of State legislative districts. For the reasons detailed below, it is our opinion that the use of such data would be constitutional.

I BackgroundA. Development of Statistically Adjusted Census Data

For a variety of reasons, the decennial census of the United States has failed to count the entire population. Following extensive litigation over the 1980 census, the Census Bureau developed a method to estimate the undercount in the census and to adjust the final numbers to reflect the actual population more accurately. This method involves a large sample survey, called the "Post-Enumeration Survey," that is conducted in conjunction with the decennial census. The Bureau compares the information gathered during this survey with information obtained in the

head-count and uses the comparison to estimate the number and characteristics of the people omitted from the head-count. See Department of Commerce v. United States House of Representatives, 525 U.S. 316, 119 S.Ct. 765, 769 (1999).

The Census Bureau has made various decisions about the disclosure and use of the statistically adjusted figures for the 1990 and 2000 enumerations. Those decisions have been challenged in the courts with varying results.

B. 1990 Census

The Census Bureau initially planned to use statistically adjusted figures for the 1990 census. However, the Secretary of Commerce overruled that decision and the Commerce Department planned to rely on the head-count alone.¹ Then, in response to a law suit, the Department agreed to have the Census Bureau perform the necessary post-enumeration survey and to consider whether to use the adjusted figures. See City of NewYork v. Department of Commerce, 739 F. Supp. 761 (E.D.N Y 1990). After the survey was done, the Secretary of Commerce determined not to use adjusted figures for the official census numbers and also not to release the adjusted figures, thus rendering them unavailable to those states that desired to use them. Both of those decisions were later upheld by the courts.Wisconsin v. City of New York, 517 U.S. 1 (1996) (decision not to adjust 1990 census results within the discretion of the Secretary); Senate of the State of California v.Mosbacher, 968 F.2d 974 (9th Cir. 1992) (Secretary not required to release adjusted data).

C. 2000 Census

In preparation for the 2000 census, the Census Bureau announced that it intended to use adjusted figures for the official census numbers, on the grounds that the adjusted figures would be more accurate than the raw head-count. Members of Congress and others brought court actions to challenge this decision. The Supreme Court held that the Census Act requires the use of the actual head-count data in the apportionment of members of Congress among the states and bars the use of adjusted figures for that purpose. The Court did not reach the issue of whether other uses of adjusted figures would be constitutional. Department ofCommerce v. United States House of Representatives,525 U.S. 316, 119 S.Ct. 765 (1999).

Following that decision, the Secretary of Commerce announced that the Census Bureau would be releasing two sets of figures following the 2000 census: (1) the actual head-count, for use in apportioning the members of Congress among the states, and (2) statistically adjusted figures for other purposes, including redistricting and the allocation of federal funds. Statement of U.S. Secretary of Commerce William M. Daley on Plan for Census 2000 (February 24, 1999).

You ask whether there is a constitutional impediment to the State's using the Census Bureau's adjusted figures for redistricting.

II AnalysisA. Standard for Use of Alternate Figures

A 1991 opinion of this Office concluded that use of statistically adjusted census data for State redistricting is constitutionally permissible so long as the data are reasonably have a "high degree of accuracy" accurate and are used consistently throughout the State. 76 Opinions of theAttorney General 209 (1991). That opinion concluded that neither the State Constitution nor the federal constitution requires use of head-count census figures for legislative redistricting,² and that the use of other measures is permissible so long as they are sufficiently accurate to allow the State to comply with the constitutional mandate of one person, one vote. Noting the Secretary of Commerce's refusal to release the 1990 adjusted figures and his statement, with specific reference to Maryland and Baltimore City, that those figures were less accurate than the head-count, the opinion questioned whether the requisite showing could be made that the adjusted figures were superior to the head-count.

To date, there have been no cases assessing the actual use of statistically adjusted census data for State redistricting because no such data have yet been available for that purpose. However, cases arising out of the controversy over the 1990 census suggest that the conclusion of the 1991 opinion of this Office remains valid.

Subsequent to the 1991 opinion, a number of courts addressed the authority of states to make use of alternate data for state redistricting. For example, in City of Detroit v.Franklin, 4 F.3d 1367, 1373-74 (6th Cir. 1993), cert.denied, 510 U.S. 1176 (1994), the Sixth Circuit concluded that neither the Constitution nor the Supreme Court's decisions concerning state redistricting require the states to use only unadjusted census figures, but simply the "best census data available" in their attempts to create districts of substantially equal population. Similarly, in Senate of the State ofCalifornia v. Mosbacher, 968 F.2d 974, 979 (9th Cir. 1992) the Ninth Circuit held that "[i]f the State knows that the census data is underrepresentative, it can, and should, utilize noncensus data in addition to the official count in its redistricting process. It is the State's responsibility, not the Secretary's, to satisfy the mandates of the Voting Rights Act." (citations omitted).

In two redistricting cases concerning the population base on which redistricting must be based, the Fourth Circuit — the circuit that includes Maryland — has "not [foreclosed] the possibility of basing apportionment on a measure other than total census population, provided that such an alternative resulted in only minor deviations from the census population count." Dalyv. Hunt, 93 F.3d 1212, 1225 (4th Cir. 1996) (state not required to use voting age population, as opposed to total population, as apportionment base); Ellis v. Mayor and CityCouncil of Baltimore, 352 F.2d 123, 129 (4th Cir. 1965) (invalidating apportionment based on population of registered voters). See also Tucker v. United States Department ofCommerce, 958 F.2d 1411, 1418 (7th Cir. 1992) ("Not only because population shifts over a ten year period, . . . but also because population and voting population are not the same, states are not required to use census figures for the apportionment of their legislatures.").

Similarly, the Attorney General of Nevada has concluded that non-census population data can constitutionally be used for state redistricting if those data have a very high degree of accuracy. 1998 Nev. Op. Atty. Gen. No. 15, 1998 WL 309048 (May 14, 1998).

B. Whether Statistically Adjusted Census Data Satisfy theStandard

In our view, there is likely to be sufficient evidence to support the use of the Census Bureau's adjusted figures for state redistricting purposes. Commentators have suggested that use of head-count figures adjusted by the Census Bureau "would be the most unobjectionable source" for alternate population data.See Issacharoff and Lichtman, The Census Undercountand Minority Representation: The Constitutional Obligation of theStates to Guarantee Equal Representation, 13 Rev. Litig. 1, 22-23 (1993). There also appears to be broad support among experts in statistics that adjusted census numbers would yield a more accurate enumeration. For example, in recounting the history of the debate over the use of statistically adjusted figures for the census, the Supreme Court noted that two of the three review panels of the National Academy of Sciences concluded that "differential undercount cannot be reduced to acceptable levels at acceptable cost without the use of [statistical sampling]"and that all three panels recommended the use of statistical sampling in the 2000 census. Department of Commerce v. United StatesHouse of Representatives, 119 S.Ct. 765 (1999). The National Research Council, Commission on Behavioral and Social Sciences and Education, and the American Statistical Association have endorsed the accuracy of adjusted figures.³

Despite the decision of the Secretary of Commerce not to use the adjusted figures in the 1990 census, there was significant support for their use at that time. The Census Bureau's Undercount Steering Committee voted 7 to 2 in favor of adjustment, and the Census Director recommended official use of adjusted numbers. See Issacharoff and Lichtman,supra, at 14. Unlike the situation in 1990, the use of adjusted figures from the 2000 census has the support of both the Commerce Department and the Census Bureau, which will be releasing the figures for official use. Furthermore, the Census Bureauhas concluded that a "dress rehearsal" it conducted for the 2000 census shows the need to use statistical methods to reduce the undercount. Statement of U.S. Secretary of Commerce William M. Daley on Plan for Census 2000 (February 24, 1999). In the past, the courts have accorded substantial deference to the decisions of the Secretary and Bureau concerning the disclosure and use of adjusted figures.⁴

C. Whether the State Must Use Statistically AdjustedFigures

Some commentators have suggested that the use of the adjusted data may be required. Note, Vote Dilution and theCensus Undercount: A State-by-State Remedy, 94 Mich. L.Rev. 1098 (1996). We disagree. In our view, the courts will not invalidate redistricting that is based on the official census head-count. Kirkpatrick v. Preisler, 394 U.S. 526, 528 (1969) (characterizing official census count as "best population data available"); Latino Political Action v. City ofBoston, 568 F. Supp. 1012, 1018 (D.Mass. 1983) (census data presumed to be accurate).

Among the reasons given by the Secretary of Commerce in 1991 for not releasing the adjusted Census figures were that the statistical procedures used in making adjustments were "unstable" and could produce different results and different mistakes, that political tampering could result from using statistics to achieve desired results, and that future censuses could be affected in a way that would discourage full and active participation by the state and localities. See Senate of the State ofCalifornia, 968 F.2d at 975. Moreover, some contend that while adjusted figures may correct an overall undercount, they will not eliminate local undercounts or compensate for a differential undercount of minorities.⁵ Whether the traditional head-count is less vulnerable to these problems is subject to debate;⁶ nevertheless, these considerations could support the use of the traditional head-count data, even in the face of evidence that the adjusted numbers compensate for the undercount.

Because there is no clear answer to the question of which of the two sets of numbers to be issued by the Census Bureau for official use is the more accurate, the choice of data to use for State redistricting is an issue on which the courts are likely to defer to the State. See Daly v. Hunt,93 F.3d 1212, 1225 (4th Cir. 1996) (whether to favor representational or electoral equality in redistricting should be left to the state);see also Issacharoff and Lichtman, The CensusUndercount and Minority Representation: The ConstitutionalObligation of the States to Guarantee Equal Representation, 13 Rev. Litig. 1, 22 (Fall 1993).

III Conclusion In our view, the State may choose to use the statistically adjusted census data published by the Census Bureau for purposes of State redistricting, or the unadjusted head-count figures, so long as one set of data is chosen and used consistently for all districts.

Very truly yours, J. Joseph Curran, Jr. Attorney General

Kathryn M. Rowe Assistant Attorney General

Robert N. McDonald Chief Counsel Opinions Advice

Elections — Reapportionment and Redistricting — General Assembly — Use of Adjusted Census Data

¹ The Census Act assigns responsibility for the decennial census to the Secretary of Commerce. 13 U.S.C. § 141(a). The Census Bureau is an agency within the Department of Commerce.13 U.S.C. § 2.

² The opinion noted that Article III, § 5 of the State Constitution evinces an expectation that census data would ordinarily be used for redistricting. 76 Opinions of theAttorney General at 212.

³ See Letter of Janet Norwood, Chair, Panel to Review the 2000 Census, National Research Council, Commission on Behavioral and Social Sciences and Education, to Kenneth Prewitt, Director, U.S. Bureau of the Census (May 3, 1999); Brief of American Statistical Association as Amicus Curiae inDepartment of Commerce v. United States House ofRepresentatives, 119 S.Ct. 765 (1999),incorporating Association's Blue Ribbon Panel Report.

⁴ See Wisconsin v. City of New York,517 U.S. 1 (1996) (deferring to Secretary's decision not to use the figures in the 1990 census); Franklin v. Massachusetts,505 U.S. 788 (1992) (deferring to Bureau decision on allocation among the states of federal employees working overseas); seealso Note, The Census Adjustment Cases: The Hunt for theWily Trout, 37 Jurimetrics 35 (Fall 1996).

⁵ See Report of Congressional Census Monitoring Board (September 30, 1999).

⁶ Compare Note, The Census Adjustment Cases: TheHunt for the Wily Trout, 37 Jurimetrics 35, 48 (1996) ("A preference for not adjusting . . . recognizes that while there is only one way not to adjust census count, there are many statistical models and assumptions that can be used to make adjustments. In the long run, a system that favors the status quo avoids arguments that the enumeration has been manipulated to achieve partisan ends.") with Note, 113 Harv. L.Rev. 349, 357 n. 60 (1999) ("From a statistician's standpoint, the traditional headcount is itself no more than one category of statistical sample. . . . There are many ways of performing headcounts by `traditional' methods, as the 1970 census's revolutionary use of `self enumeration by mail' illustrates. . . . [C]ritics of the `traditional headcount' see it as involving a variety of arbitrary and politically significant decisions, such as determinations of how to assign states of residency to college students and military personnel.").