Chapter 2 - The Census

The information produced by the year 2000 census, the 22^nd decennial count, will determine the apportionment of the 435 seats in the U.S. House of Representatives among the 50 states and prompt the drawing of new boundaries for congressional, state and local election districts. This chapter reviews some of the practical and legal issues that will affect the 2000 census. It discusses the mandate for the decennial census found in the U.S. Constitution and statute law. Case law and issues that have created controversy about the conduct of the census also are covered. It includes an outline of how the 2000 census will be conducted and reported. Finally, it touches on the use of statistical bases other than the decennial census in redistricting.

Two major issues that dominated census developments during the 1990s are covered below: changes in the census questions and reports to reflect multi-racial population categories, and the possible use of sampling techniques to refine census population counts and improve accuracy.

The decennial census of 1790 began an unbroken series of nationwide population counts called for by Article I, Section 2, Clause 3 of the U.S. Constitution:

Representatives ... shall be apportioned among the several states ... according to their respective Numbers which shall be determined by addition to the whole Number of Free Persons ... [and] three-fifths of all other persons ... . The actual Enumeration shall be made within three Years after the first Meeting of the Congress of the United States, and within every subsequent term of ten Years, in such manner as they shall by law direct.

Section 2 of the Fourteenth Amendment to the U.S. Constitution eliminated the “three-fifths” requirement for counting slaves and provides that: “Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State ... .”

Congress has delegated responsibility for taking the census to the Department of Commerce and its Bureau of the Census in Title 13 of the United States Code. The law directs the secretary of commerce to “take a decennial census of population as of the first day of April . . . the ‘decennial census date’” in 1980, 1990, 2000, etc. The secretary must complete the census and report the total population, by states, to the president by December 31 of the census year. The purpose of the report is “the apportionment of Representatives in Congress among the several States” as required by Article I, Section 2, of the U.S. Constitution.

When the Congress convenes in 2001, the president must transmit to that body a statement of the apportionment of the 435 representatives’ seats among the states. The number of representatives allocated to each state is based on the census results and determined by the “method of equal proportions.” Each state is guaranteed at least one representative, and the remaining 385 seats are apportioned among the states by assigning priority values to each seat.

Title 13, as amended by Public Law 94-171 (1975), also requires the secretary of commerce to report census results no later than April 1, 2001, to the bodies or officials charged with state legislative redistricting and to the governors. These reports contain the population data for various geographical areas within the state, including the smallest areas, the “census blocks.” The April 1 reports provide the basis for state and local level decennial redistricting efforts as well as for the redrawing of congressional districts within each state.

Figure 1 explains the three-phase process by which the Census Bureau plans to gather and distribute this redistricting data by the April 1 deadline.

Controversy over the use of statistical sampling in determining the census count presents the principal federal legislative issue affecting the 2000 census and its use in reapportionment and redistricting.

There are two references to sampling in Title 13 that appear to be in conflict. Section 141 (a) directs the secretary of commerce to take the decennial census “in such form and content as he may determine, including the use of sampling procedures and special surveys.” Section 195, however, directs the secretary to use sampling methods in fulfilling his duties under Title 13, “[e]xcept for the determination of population for purposes of apportionment of Representatives in Congress among the several States.”

The controversy over sampling led to the passage of the Decennial Census Improvement Act of 1991, which called for studies by the National Academy of Sciences on ways to achieve the most accurate census, including the possible use of sampling techniques.

Based on these studies, the Bureau of the Census developed its plan for the 2000 census. The plan includes sampling procedures.

In late 1997, Congress declared that “the use of statistical sampling or statistical adjustment in conjunction with an actual enumeration to carry out the census with respect to any segment of the population poses the risk of an inaccurate, invalid, and unconstitutional census.” It authorized “any person aggrieved by the use of any statistical method in violation of the Constitution or any … [other] law … in connection with the 2000 or any later decennial census, to determine the population for purposes of the apportionment or redistricting of members of Congress” to obtain declaratory, injunctive or other appropriate relief against the use of the method. Congress also specified that “the number of persons enumerated without using statistical methods must be publicly available for all levels of census geography which are being released by the Bureau of the Census for ... the data contained in the 2000 decennial census Public Law 94-171 data file released for use in redistricting” as well as other specified data.

The speaker of the House of Representatives brought suit under this 1997 authorization. The case is discussed below.

The undercount. The census is not, and cannot be, 100 percent accurate. According to the Census Bureau, the 1990 census was the first census less accurate than its predecessor, even though it was better designed and executed. The undercount rate of 1.6 percent was 50 percent greater than the 1980 undercount. The undercount is not uniform and the census misses a disproportionate number of racial and ethnic minorities. Undercount figures for 1990 reported by the bureau include 4.4 percent of African Americans, 5 percent of Hispanics, 12.2 percent of American Indians living on reservations, and 0.7 percent of non-Hispanic whites.

The 1990 Census–Wisconsin v. City of New York. Litigation before the 1990 census focused on the undercounting of population and the disproportionate effect of the undercount on minority groups. Plaintiffs representing undercounted areas sought statistical adjustments to the actual enumeration based on post-census surveys. The Department of Commerce agreed to conduct the surveys, but declined to report the statistically adjusted counts as the official census. In July 1991, the secretary of commerce announced that there would be no statistical adjustment based on post-enumeration surveys to the apportionment and redistricting data released in early 1991. The secretary cited three rationales for his decision: first, an adjustment might improve numerical accuracy but would lessen the distributive accuracy of the census (the correct proportions of population in the various areas covered by the census); second, the long-standing practice of relying on an actual enumeration; and third, concern that statistical adjustments would be subject to future manipulation.

The court cases involving the 1990 census were finally resolved by the U.S. Supreme Court when it upheld the secretary’s 1991 decision in Wisconsin v. City of New York in 1996.

The Supreme Court ruled that the secretary’s decision was valid and that it bore “a reasonable relationship” to the task required by the Constitution. The Court cited the broad discretion lodged by the Constitution in the government on the conduct of the census and the broad discretion given the secretary under Title 13 to determine the “form and content” of the census.

The Court left open the issue of whether sampling or adjusted census figures were forbidden by the Constitution’s reference to an “actual enumeration.”

The 2000 Census–U.S. House of Representatives v. U.S. Department of Commerce.

In this case, the House of Representatives sought to prohibit the use of sampling techniques under the authorization passed by the Congress in 1997. The Department of Commerce defended the use of sampling techniques to improve the accuracy of the 2000 census, a reversal of its position in the Wisconsin case, and was joined by intervenor-defendants such as the City of Los Angeles group.

On August 24, 1998, a three-judge special panel ruled unanimously in favor of the House of Representatives and permanently enjoined the Department of Commerce from using statistical sampling to determine the population for purposes of apportioning representatives in Congress among the states. The court held that the provisions of Title 13, the Census Act, Sections 141 and 195, prohibited the use of sampling in determining apportionment counts. The court’s order specifically enjoined the department and defendants “from using any form of statistical sampling, including their program for nonresponse follow-up and Integrated Coverage Measurement, to determine the population for purposes of congressional apportionment.” The court did not reach the constitutional issue of whether the requirement for an “actual enumeration” in Article I prohibits the use of sampling techniques.

The defendants appealed the decision. The Supreme Court heard oral arguments on the appeal on November 30, 1998.

A second, similar case was heard by the Supreme Court at the same time. The federal district court in Glavin v. Clinton

reached the same conclusion as the court in the House of Representative case.

As of December 1, 1998, the Bureau of the Census was proceeding to prepare for either a traditional enumeration or an enumeration combined with sampling, pending the outcome of the litigation and further action by the Congress.

The Constitution requires that the apportionment of seats in the House of Representatives be determined by an “actual enumeration” of persons “in each State,” conducted every 10 years.

In the 1990 census, the Census Bureau allocated the overseas employees of the Department of Defense to particular states, based on their “usual residence.” Including these overseas military personnel in the counts of each state caused one congressional district to be shifted from Massachusetts to Washington. Massachusetts sued the secretary of commerce, alleging that the procedure was contrary to the language of the Constitution, was arbitrary and capricious in violation of the Administrative Procedures Act. The Massachusetts federal district court agreed,

but the Supreme Court reversed.

In 1992, in Franklin v. Massachusetts,

the Supreme Court found that the secretary’s decision was not subject to the Administrative Procedures Act, because it was not a “final agency action.” Rather, it formed the basis for the secretary’s report to the president on the population of each state, which the president was free to adopt or revise as he saw fit.

The decision to allocate overseas federal employees to their home state based on their usual residence was consistent with Census Bureau practice since the first census of trying to assign each person to the person’s home state, even though he or she might be temporarily absent on the day of the census. Assuming that the overseas employees have retained ties to their home states, the Court found that the secretary’s method of allocation actually promotes equality of representation.

In compliance with the October 1997 decision by the U.S. Office of Management and Budget, the Census 2000 questions (and other federal surveys) will permit individuals to choose more than one of several race categories (White, African American or Black, American Indian or Alaska Native, Asian, Native Hawaiian and Other Pacific Islander, Other Race). This new standard is in response to requests by individuals who are multi-racial and by parents of multi-racial children who wish to reflect that fact in federal surveys. This change also recognizes the growing number of multi-racial individuals in the United States. The new standard requires federal agencies to include a separate question so people of any race can indicate whether they identify as being of Hispanic/Latino origin.

The result of this change will provide more information as part of the Census 2000 Public Law 94-171 redistricting data reports. The prototype data from the 1998 Dress Rehearsal Census will include both “single race” totals for those who choose only one race and “all-inclusive” totals for those who mark more than one of the basic race groups.

Apart from sampling issues, efforts to improve the census involve the distribution by mail of the forms and the basic address list for the country. After the 1990 census, local and tribal government officials expressed concern about the completeness of the 1990 census and their belief that they had address information available that would make the census address list more accurate. In 1994, Congress passed the Census Address List Improvement Act

to deal with this concern. This law directed that the Census Bureau provide an opportunity for every government to review the census address list for accuracy and completeness before it is used to deliver Census 2000 questionnaires. That same law also required that this review process adhere to the confidentiality provision under which the Census Bureau operates as specified in Title 13, U.S. Code.

Census 2000 will be conducted by mail-out/mail-back procedures in most areas. The initial step in building the census mailing list of residential addresses (master address list or MAF) in areas with city-style street addresses was to combine addresses from the U.S. Postal Service’s delivery sequence file with the addresses on the Census Bureau’s 1990 census address file.

In areas that have predominately non-city style addresses (e.g., rural route and box number), the Census Bureau creates its master address list by hiring local staff to travel each such area, locate each housing unit, record its physical location, and “spot” the location on a census map. This process allows the bureau to incorporate these “listed” addresses and map information into the master address file.

The Census Bureau has invited local and tribal governments to review census maps and compare address information they maintain to the Census Bureau’s address list and to make additions, corrections or deletions to the Census Bureau’s address list and identify missing units. This program, Local Update of Census Addresses (LUCA), will take place between spring 1998 and spring 1999 on a flow basis for urban and rural areas.

In areas that have predominantly city-style addresses, bureau field staff, in 1999, will canvass every road and street where someone may live to add addresses for dwellings that are not present in the master address file and to update address information that is not correct.

Figure 2 explains how the Census Bureau decides where each person should be counted.

Between fall 1997 and spring 1998, the Census Bureau opened 12 regional census centers (RCCs) as temporary sites near the 12 permanent regional offices, to provide additional space to manage the large Census 2000 work load. These RCCs support the address list development activities and the reviews by tribal and local governments.

In late spring 1998, the Census Bureau selected three sites–Baltimore County, Md., Pomona, Calif., and Phoenix, Ariz., for three of its four data capture centers. The fourth center is located in the Census Bureau’s permanent processing facility in Jeffersonville, Ind. These data capture centers are where the completed census forms will be received and scanned into the computer. They will house imaging equipment, data keying devices and other systems for converting responses to census questions into computer-readable form.

In fall 1999, the Census Bureau plans to open more than 480 temporary local census offices to manage and control the many operations to collect data from households that do not mail back the census forms. These offices will recruit and train a peak work force of approximately 260,000.

In March 2000 advance letters announcing the census, questionnaire packages and reminder cards will be delivered by the U.S. Postal Service in areas that have city-style addresses. In areas that do not have city-style addresses, temporary field workers will leave questionnaires at each dwelling. Census Day is April 1, 2000, and respondents will be urged in publicity and by instructions on the questionnaire to mail the form back by that date.

The Census Bureau will provide additional ways to be counted by advertising how one can provide their answers to 1-800 census operators, or by obtaining an unaddressed census form at walk-in community assistance centers, tribal offices, etc. Census forms will be translated and available in other languages, such as Spanish, Japanese and others.

As discussed above, the Supreme Court will decide in its 1998-1999 term whether the Census Bureau may employ sampling techniques in conducting Census 2000. This section describes the bureau’s plan, which has not yet been approved, to reach households that do not respond to the census.

The Census Bureau will rely mainly on mail responses to the census questionnaires. Based on experience in focus groups and the response rates in other surveys, the Census Bureau fears that, despite its best efforts, 34 million occupied households will not respond by mail in 2000. For those that do not respond by mail, the bureau has proposed that a sample be selected for nonresponse followup.

As part of the nonresponse followup, the Bureau also has proposed to use sampling to conduct a followup of housing units the U.S. Postal Service identifies as “vacant, undeliverable-as-addressed” (UAA). The Census Bureau designed both sampling operations to provide a less expensive and more efficient way of collecting information on vacant and nonresponding housing units than multiple return visits to all nonresponding households.

Direct sampling of nonresponding housing units will be conducted to achieve a final response rate of 90 percent or better in each census tract. A systematic sample of housing units will be selected to ensure that the sample is evenly distributed across all nonresponding addresses in the census tract. The sample will be selected immediately after the cut-off for mail returns, using a variable sampling rate that depends on the initial mail-back response to the census.

To obtain information from 90 percent of the housing units in a census tract, tracts with lower initial mail-back response rates will have a larger proportion of housing units sampled. Table 1shows sampling rates for various initial response rates:

Initial Response Rate	Sampling Rate
60%	3-in-4
70%	2-in-3
80%	1-in-2
85% or higher	1-in-3

The initial response rate used to set the sampling rate for each census tract is shown in table 2:

As of December 1998, approval of the bureau’s plan for the integrated coverage measurement (ICM ) quality check also was pending the Supreme Court’s ruling. The ICM is an intensive telephone or face-to-face interview of about 750,000 housing units in selected blocks across the country. It is designed to enable the Census Bureau to correct the census count, obtained from the procedures described above, for the differential undercount of some segments of the population and the overcount of the other segments. ICM results will be combined with results of the initial phase of Census 2000 to yield census population counts that are reported for use in redistricting.

Most jurisdictions rely on decennial census data to redistrict, and most redistrict on a 10-year cycle keyed to the release of those data. Federal courts have upheld the use of alternative population bases for redistricting if the alternative data base is used uniformly and if the results are comparable to those produced by a census population-based plan.

In Burns v. Richardson, the Supreme Court upheld Hawaii’s legislative redistricting, which was based on the number of registered voters.

Given Hawaii’s special military and tourist populations, the Court allowed the use of an alternative population base after finding that the results were not substantially different from results of a redistricting based on total citizen population. The Court, however, cautioned that the use of registered or actual voter bases could be susceptible to manipulation to maintain underrepresentation. Again, the Supreme Court cautioned in Ely v. Klahr that a new plan for Arizona legislative districts could use registered voter data only if the result would be a “distribution of legislators not substantially different from that which would have resulted from the use of a permissible population basis.”

One court has specifically ruled that a plan based on registered voters is unconstitutional.

In Kirkpatrick v. Preisler,

the Supreme Court implied that eligible voter population could be a valid factor to use in redistricting if identified properly and applied uniformly to an entire redistricting plan. In that case, the Court disapproved a Missouri congressional plan because the data had not been applied in a uniform manner. One divided court has rejected the proposition that the voting age population, rather than the total population, should be the principal basis for redistricting.

Some states have conducted their own censuses, and courts have upheld use of those state population counts in redistricting.

However, the 2000 census is sure to shape redistricting efforts across the country in the next few years.

It is too early to write a conclusion to this chapter on the Census 2000. The best advice to those who are planning for redistricting in the year 2001 is to stay tuned and in touch with the Census Bureau and NCSL as developments proceed.

There will be new developments in 1999. Under the auspices of the Office of Management and Budget, a panel of representatives from statistical agencies is developing guidelines to be issued in early 1999 that will be used in tabulating and publishing these new racial and ethnic data. Because of its need to conduct the 1998 Dress Rehearsal Census, the Census Bureau will be the first agency to produce data using this new standard. In the first quarter of 1999, the Census Bureau plans to issue a “test” version of the Public Law 94-171 Redistricting Data, including 20 tabulations that encompass population totals, and both “single-race” tallies (for those who selected only one race category) and “all-inclusive” tallies (for those who picked two or more races). These data will be cross-tabulated by voting age and by Hispanic/Non-Hispanic origin. Those planning for redistricting should review the test data. States covered by Section 5 of the Voting Rights Act will want to be alert to comments by the Department of Justice on the information it will require in preclearance submissions for redistricting plans.

The sampling issue remains undecided now and may or may not be resolved by the Supreme Court decision in the pending House of Representatives case. Those planning for redistricting will want to follow developments in the Court and Congress through 1999 up to the release of the Public Law 94-171 redistricting data. One issue of interest is whether sampling may be used to produce census reports used for fund distributions or redistricting even if it is not used to apportion the House of Representatives among the states.

This page is maintained by the Redistricting Task Force for the National Conference of State Legislatures
Update: 10/31/03 (psw)
Comments: peter.wattson@senate.leg.state.mn.us

			Respondents + UAA vacants
	Initial Response Rate =
			Total Addresses
Where:
Respondents =		The number of self responding housing units in the census tract,
UAA vacant =		The number of undeliverable as addressed vacant units identified by the U.S. Postal Service in the census tract, and
Total Addresses =		The number of addresses mailed or delivered a questionnaire.