Could the Maryland General Assembly use statistically adjusted census numbers instead of the official 1990 head count to redraw legislative districts?

Plain-English summary

After the 1990 census, the U.S. Census Bureau ran a follow-up Post-Enumeration Survey and found Baltimore City had been undercounted by about 4.9%, or 36,000 people, with a statewide undercount of 1.8%. Even so, the U.S. Secretary of Commerce decided not to statistically adjust the official 1990 head count, in part because his own analysis found that, for Maryland and Baltimore specifically, the adjusted figures would actually be less accurate than the raw census numbers. A state senator asked the Attorney General whether the General Assembly was legally required to use the official census count for legislative redistricting affecting Baltimore City, and if not, under what conditions it could instead use a higher, adjusted population figure.

The opinion concluded that the General Assembly was not legally required to use head-count census data for redistricting, since neither the U.S. Constitution's Equal Protection Clause nor Article III, §5 of the Maryland Constitution locked the legislature into using only the federal census. But it found that a substantial burden fell on anyone proposing to substitute other data for the census: courts had required "a high degree of accuracy," systematic statewide application rather than an ad hoc, single-city fix, and "clear, cogent and convincing evidence" that the census figures were wrong and the substitute figures were right. Applying that standard, the opinion found Maryland's position weak on the facts, since the Commerce Secretary himself had identified Maryland and Baltimore as places where the adjusted survey data were likely less accurate than the actual census count, and since the Census Bureau had not released the detailed adjustment data needed to support any substitute figure in the first place. The opinion's bottom line: the General Assembly could use adjusted census figures instead of head-count figures, but only if it had clear and convincing evidence the adjusted figures were superior, and only if it applied them statewide rather than singling out Baltimore City.

Currency note

This opinion was issued in 1991, addressing the 1990 census cycle and that era's federal case law on census-data substitution in redistricting. Subsequent statutory amendments, court decisions, or later AG opinions may have changed the analysis, particularly given later U.S. Supreme Court redistricting decisions and subsequent decennial census cycles. Treat this page as historical context, not current legal advice. Verify the current legal standard for using non-census population data in redistricting, and the current text of Article III, §5 of the Maryland Constitution, before relying on any specific rule described here.

Common questions

Did Maryland have to use the official 1990 census count when drawing legislative districts?
According to this 1991 opinion, no. Neither the federal Equal Protection Clause nor Article III, §5 of the Maryland Constitution required the General Assembly to use only head-count census data for redistricting.

Could Maryland just use a higher, adjusted population number for Baltimore City to offset the census undercount?
Not easily, according to the opinion. It said adjusted figures could be used only with clear and convincing evidence of their superior accuracy, applied the same way statewide, not singled out for one city. The opinion found Maryland's case for doing so weak, since the U.S. Secretary of Commerce had specifically identified Maryland and Baltimore as places where the adjustment would likely make the data less, not more, accurate.

Why couldn't the state just rely on the Post-Enumeration Survey's undercount findings?
Because, per the opinion, the Census Bureau had not released the detailed survey data needed to support a substitute population figure for legislative redistricting, and the Commerce Secretary's own published statement expressed doubt about the survey's accuracy at the state and local level, even while acknowledging a national undercount.

Background and statutory framework

Following the 1990 decennial census, the Census Bureau's Post-Enumeration Survey found a 4.9% (36,000-person) undercount in Baltimore City and a 1.8% (87,532-person) statewide undercount, though the only released data were aggregate figures for the state and large subdivisions, with nothing showing how the undercount was distributed within a given jurisdiction. On July 15, 1991, Secretary of Commerce Robert A. Mosbacher decided not to statistically adjust the 1990 count, explaining that the Post-Enumeration Survey itself "missed important segments of the population" and that, while adjustment might improve accuracy nationally, "the experts disagree with respect to the accuracy of the shares measured from an adjustment" at the state and local level. Significantly, Secretary Mosbacher's own analysis identified Baltimore as one of the large metropolitan areas where the census count appeared more accurate than the adjusted figures, and the opinion understood Maryland generally to be among the states an adjustment would make less accurate. The Commerce Department had also declined, citing confidentiality requirements, to release the underlying adjustment data.

The opinion found that federal courts had long treated the decennial census as the "best population data available," Kirkpatrick v. Preisler, 394 U.S. 526, 528 (1969); Karcher v. Daggett, 462 U.S. 725, 738 (1983), and presumed its accuracy, Latino Political Action v. City of Boston, 568 F. Supp. 1012, 1018 (D. Mass. 1983), but states were not constitutionally required to use only census data for redistricting. Federal courts had addressed the question in cases involving mid-decade redistricting, Wells v. Rockefeller, 273 F. Supp. 984, aff'd 389 U.S. 421 (1967); exclusion of certain population classes from the count, Burns v. Richardson, 384 U.S. 73 (1966), and the Maryland Court of Appeals' own decision in DuBois v. City of College Park, 286 Md. 677, 410 A.2d 577 (1980); and unsuccessful state attempts to justify population deviations based on alleged census inaccuracies, Karcher, 462 U.S. at 738; Kirkpatrick, 394 U.S. at 535, with one federal court upholding reliance on the state census headcount, Black Political Task Force v. Connolly, 679 F. Supp. 109 (D. Mass. 1988). The U.S. Supreme Court had held the Equal Protection Clause "does not require the states to use total population figures derived from the federal census as the standard" for measuring population equivalency, Burns v. Richardson, 384 U.S. at 91, and the opinion read Article III, §5 of the Maryland Constitution, which begins the redistricting process "following each decennial census," as reflecting an expectation that census data would ordinarily be used, without precluding the use of other data.

Even so, the opinion found a substantial burden fell on any proponent of substitute data: non-census figures needed "a high degree of accuracy," Kirkpatrick, 394 U.S. at 535, had to be "thoroughly documented and applied throughout the State in a systematic, not an ad hoc, manner," id., and federal courts required "clear, cogent and convincing evidence" that the census figures were invalid and the substitute figures valid, Dixon v. Hassler, 412 F. Supp. 1036, 1040 (W.D. Tenn.), aff'd sub nom. Republican Party v. Dixon, 429 U.S. 934 (1976); McNeil v. Springfield Park Dist., 851 F.2d 937, 946 (1989). The opinion concluded that any use of the Post-Enumeration Survey figures had to satisfy this standard, and that Maryland's showing was weak given the Commerce Secretary's own conclusion that adjustment would make Maryland's data less, not more, accurate, and given the incomplete state of the released survey data. It also noted that any substitute figures would have to be applied statewide, not just in Baltimore City, Kirkpatrick, 394 U.S. at 535; DuBois, 286 Md. at 689 n. 6, and that the data released to date would not support statewide use in any event.

Citations and references

Statutes:

• Article III, §5 of the Maryland Constitution, requiring the redistricting process to begin following each decennial census

Cases:

• Kirkpatrick v. Preisler, 394 U.S. 526, 528 (1969), U.S. Supreme Court decision on the census as the "best population data available" and the accuracy standard for substitute data
• Karcher v. Daggett, 462 U.S. 725, 738 (1983), U.S. Supreme Court decision reinforcing the census's presumed reliability in redistricting
• Latino Political Action v. City of Boston, 568 F. Supp. 1012, 1018 (D. Mass. 1983), federal district court decision on the presumption of census accuracy
• Wells v. Rockefeller, 273 F. Supp. 984, aff'd 389 U.S. 421 (1967), federal case on mid-decade redistricting and census data
• Burns v. Richardson, 384 U.S. 73 (1966), U.S. Supreme Court decision holding the Equal Protection Clause does not require exclusive use of federal census figures
• DuBois v. City of College Park, 286 Md. 677, 410 A.2d 577 (1980), Maryland Court of Appeals decision on state exclusion of population classes from census-based redistricting figures
• Black Political Task Force v. Connolly, 679 F. Supp. 109 (D. Mass. 1988), federal district court decision upholding use of a state's own census headcount
• Dixon v. Hassler, 412 F. Supp. 1036, 1040 (W.D. Tenn.), aff'd sub nom. Republican Party v. Dixon, 429 U.S. 934 (1976), federal case setting the "clear, cogent and convincing evidence" standard for substitute population data
• McNeil v. Springfield Park Dist., 851 F.2d 937, 946 (1989), federal appellate decision applying the same evidentiary standard

Source

• Landing page: https://oag.maryland.gov/resources-info/Pages/attorney-general%E2%80%99s-opinions.aspx
• Original PDF: https://oag.maryland.gov/resources-info/Documents/pdfs/Opinions/1991/Volume76_1991.pdf (this opinion appears at printed pages 209-213 of the bound annual volume; Maryland's site does not publish a standalone PDF of this opinion)

Original opinion text

Best-effort transcription from a scanned PDF. Minor errors may remain — the linked PDF is authoritative.

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

                         August 12, 1991

The Honorable John A. Pica, Jr.
Maryland Senate

  You have requested our opinion about the legality of using certain

statistically adjusted census data, as distinct from head-count census
data, for legislative redistricting affecting Baltimore City. You point out
the United States Secretary of Commerce has acknowledged that the
census missed a significant number of people, particularly among
minority groups. Nevertheless, the Secretary decided not to allow a
statistical adjustment of the head-count data based on a survey conducted
after the census, called the Post-Enumeration Survey.

Your specific questions are as follows:

 1. For purposes of legislative redistricting, is the General

Assembly required to use the 1990 U.S. Census count for Baltimore
City "if it can be demonstrated that more reliable data ... justify a higher
population figure"?

 2. If the General Assembly is not required to use the 1990 census

count, under what conditions could it "legally and constitutionally use
the highest population figure" for Baltimore City?

 For the reasons detailed below, we conclude that the General

Assembly is not required by law to use head-count census data as a basis
for legislative redistricting. However, because the federal census is
presumed accurate until proven otherwise, any other basis for
redistricting, including adjustment figures, must be thoroughly
documented, have a high degree of accuracy, and be applied in a
systematic manner throughout the entire State, not just in Baltimore
City. The State's ability to make the requisite showing is called into
serious question by the Commerce Secretary's statement that Post-
Enumeration Survey figures for both the State of Maryland and
Baltimore City are "less accurate" than the decennial census data. A
further problem is that the Census Bureau has declined to release Post-
Enumeration Survey data for Baltimore City and the State of Maryland
in sufficient detail to permit their use for purposes of legislative
redistricting.

                                  I

                            Background

 Following the taking of the 1990 decennial census, the U.S. Census

Bureau conducted a Post-Enumeration Survey. Survey results, released
in June 1991, disclosed that Baltimore City had a census undercount of
4.9%, or 36,000 people. Statewide, the Survey showed an undercount
of 1.8%, or 87,532 people.1 The only undercount data for Maryland
released by the Census Bureau were aggregate figures for the State and
11 subdivisions with populations in excess of 100,000. No figures were
released that would have disclosed how the undercount was distributed
within a particular jurisdiction.

 On July 15, 1991, Secretary of Commerce Robert A. Mosbacher

decided that, despite the results of the Post-Enumeration Survey, the
1990 census count would not be altered by a statistical adjustment. See
Statement of Secretary Robert A. Mosbacher on Adjustment of the 1990
Census (July 15, 1991) ("Mosbacher Statement"). Although conceding
"a diversity of opinion among [his] advisors," Secretary Mosbacher
opted against adjustment because the Post-Enumeration Survey itself
"missed important segments of the population." Mosbacher Statement
at 3. He noted "general agreement that at the national level, the adjusted
counts are better, though independent analysis shows that adjusted
counts, too, suffer from serious flaws." But, he continued, "[b]elow the
national level, ... the experts disagree with respect to the accuracy of the
shares measured from an adjustment. The classical statistical tests of
whether accuracy is improved by an adjustment at state and local levels
show mixed results ...." Mosbacher Statement at 4.

    1
     Because the national undercount rate, according to the Survey, was

2.1%, Maryland's figure was below the national average.

 Significantly, Secretary Mosbacher specifically identified Baltimore

as a city where use of the adjusted data would be less accurate than the
head-count census data:

            Based on the measurements so far completed, the
        Census Bureau estimated that the proportional share of
        about 29 states would be made more accurate and
        about 21 states would be made less accurate by
        adjustment. looking at cities, the census appears
        more accurate in 11 of the 23 metropolitan areas with
        500,000 or more persons: Phoenix, Washington,
        D.C., Jacksonville, Chicago, Baltimore, New York
        City, Memphis, Dallas, El Paso, Houston and San
        Antonio. Many large cities would appear to be less
        accurately treated under an adjustment. While these
        analyses indicate that more people live in jurisdictions
        where the adjusted counts appear more accurate, one
        third of the population lives in areas where the census
        appears more accurate. As the population units get
        smaller, including small and medium sized cities, the
        adjusted figures become increasingly unreliable.
        When the Census Bureau made allowances for
        plausible estimates of factors not yet measured, these
        comparisons shifted toward favoring the accuracy of
        the census enumeration. Using this test, 28 or 29
        states were estimated to be made less accurate if the
        adjustment were to be used.

Mosbacher Statement at 4-5. We understand that Maryland is one of the
states included in Secretary Mosbacher's reference to those states whose
count would be made "less accurate by adjustment."2

Citing census confidentiality requirements, the Commerce

Department has declined to release the adjustment figures compiled in
connection with the Post-Enumeration Survey. Washington Post, July
17, 1991 at A21. However, California Assembly Speaker William L.
Brown, Jr. has filed a freedom of information request for the data. Id.

                                    II

                                Analysis

 In the past, the U.S. Supreme Court has characterized the decennial

census as the "best population data available...." Kirkpatrick v. Preisler,
394 U.S. 526, 528 (1969). See also Karcher v. Daggett, 462 U.S. 725,
738 (1983). Further, courts presume the accuracy of this data. See
Latino Political Action v. City of Boston, 568 F. Supp. 1012, 1018 (D.
Mass. 1983).

Nevertheless, states have not been required, as a matter of federal

constitutional law, to use only census data for redistricting purposes.
The issue has typically arisen in cases involving mid- or late-decade
redistricting, see, e.g., Wells v. Rockefeller, 273 F. Supp. 984, aff'd
389 U.S. 421 (1967); state exclusion of certain classes from the
population base, see, e.g., Burns v. Richardson, 384 U.S. 73 (1966),
and DuBois v. City of College Park, 286 Md. 677, 410 A.2d 577
(1980); and unsuccessful state attempts to justify variances from "one-
person, one-vote" requirements on the basis of alleged census
undercounts or inaccuracies, see, e.g., Karcher v. Daggett, 462 U.S. at
738; Kirkpatrick v. Preisler, 394 U.S. at 535. But see Black Political
Task Force v. Connolly, 679 F. Supp. 109 (D. Mass. 1988) (upholding
use of state census headcount). These cases make it clear that the Equal
Protection Clause of the Fourteenth Amendment "does not require the
states to use total population figures derived from the federal census as
the standard by which ... substantial population equivalency is to be
measured." Burns v. Richardson, 384 U.S. at 91.

    2
       Another reason for Secretary Mosbacher's decision was his belief that

the use of adjusted data would disrupt redistricting, which he said "is contentious
and litigious enough without an adjustment." Mosbacher Statement at 6.

 Nor does the Maryland Constitution. Article III, §5 provides that

the legislative redistricting process is to begin "[f]ollowing each
decennial census of the United States ...." See generally 76 Opinions of
the Attorney General 246 (1991).3 Although Article III, §5 evinces an
expectation that census data would ordinarily be used in redistricting, we
do not read the provision to preclude use of other data.

 Nevertheless, a substantial burden is cast on a proponent of the use

of a substitute for federal census data. Non-census data must have "a
high degree of accuracy." Kirkpatrick v. Preisler, 394 U.S. at 535.
They "must be thoroughly documented and applied throughout the State
in a systematic, not an ad hoc, manner." Id. Moreover, lower courts
have said that there must be "clear, cogent and convincing evidence" that
decennial census figures are not valid and that the substitute figures are
valid. Dixon v. Hassler, 412 F. Supp. 1036, 1040 (W.D. Tenn.), aff'd
sub nom. Republican Party v. Dixon, 429 U.S. 934 (1976). See also
McNeil v. Springfield Park Dist., 851 F.2d 937, 946 (1989). In our
opinion, use of the Post-Enumeration Survey figures for State legislative
redistricting must satisfy this legal standard in order to be used as a
supplement or substitute for decennial census data.

    3
      The Governor will submit his plan on the first day of the 1992 Session

of the General Assembly. Unless the General Assembly adopts an alternative
plan by the 45th day of the session, the Governor's plan becomes law.

Although reasonable minds may and do differ about the accuracy of

both the 1990 census and the Post-Enumeration Survey, one thing is
clear: The U.S. Secretary of Commerce has taken the position that,
with respect to Maryland, the census data are more reliable than the
adjusted figures. Of course, the views of the Commerce Secretary are
not conclusive on the point. Nevertheless, they cannot be ignored. If
adjusted figures are used in fashioning the 1992 legislative redistricting
plan, the Secretary's views would surely play a prominent role in any
legal challenge to the plan.

 Two additional factors must be considered in connection with the

use of Post-Enumeration Survey figures for legislative redistricting.
First, the adjusted data must be used statewide, not just in Baltimore
City. Kirkpatrick v. Preisler, 394 U.S. at 535; DuBois v. City of
College Park, 286 Md. at 689 n. 6. Second, all of the necessary data
are not presently available, because the Commerce Department has
refused to release them. The data that have been released to date would
not provide a sufficient base for redistricting purposes.

                               III

                          Conclusion

 In summary, it is our opinion that the General Assembly may use

adjusted census figures instead of head-count census figures for
legislative redistricting if, but only if, the General Assembly does so on
the basis of clear and convincing evidence of the superiority of the
adjusted figures and uses the adjusted figures statewide.

                                           J. Joseph Curran, Jr.
                                           Attorney General

                                           Robert A. Zarnoch
                                           Assistant Attorney General

Jack Schwartz
Chief Counsel
Opinions & Advice