Changing Census, Changing America

By Edward Still

Vol. 22, No. 4, 2000 pp. 24-26

Every census is different from the last, but there are some big changes in store with Census 2000. Beginning in early March 2001, the Bureau will publish census data for each state to use in redistricting. What new things can we expect from this census? To begin with, we will have access to census data more easily via computer. The Census Bureau will be posting census data on its website, and a separate site for the American Factfinder, More importantly, the data will be different from past censuses. I want to discuss two changes: the racial data and the sampling controversy.

Reporting One or More Races

For the first time, Census 2000 allowed respondents to identify themselves as a member of more than one race. The census asked, "What is this person's race? Mark one or more races to indicate what this person considers himself/herself to be." The races which the Census Bureau will report are: white; black, African American, or Negro; American Indian or Alaska Native; Asian; Native Hawaiian or other Pacific Islander; and, "Some other race." (In the 1990 Census, the Asian and Native Hawaiian/Pacific Islander groups were combined.)

Because the respondents were allowed to choose more than one race, there are fifty-seven possible combinations of racial groupings-ranging from people who mark two races (ten possible combinations) to people who claim all six racial categories.

All of this stems from a 1997 decision of the Office of Management and Budget to amend Directive 15 which tells all federal statistical agencies--not just the Census Bureau--how to gather information on the race of Americans. Unfortunately, the OMB made the decision to allow multiple checkoffs to the racial question without figuring out how any agency would use or report the data on the

Page 25

small number of people who will actually mark two or more races. (In both a federal health survey in the mid-1990s and the 1998 Census Dress Rehearsal in 3 areas, participants were given the chance to mark multiple races; less than 2 percent marked more than one.)

Since the adoption of Directive 15, the Census Bureau, OMB, and other federal agencies which gather or use racial population statistics have struggled with how to report the racial statistics and how the agencies enforcing civil rights laws would ask employers, schools, etc. to gather and report race statistics on their employees, students, voters, etc. to the federal government.

The idea of sixty-three racial categories instead of five worried many. For instance, at meetings of the Redistricting Task Force of the National Conference of State Legislators, the staff from various states complained that they would have to reprogram their computers to accommodate the larger number of categories, and worried that the data would become meaningless if there were many "cells" (racial combinations) with zero.

In 1998 the Census Bureau proposed to report racial groups as a range of populations-the so-called "broadband method." For instance, consider a population in which 65 percent reported themselves as white, 33 percent as black, and 2 percent as black and white. The broadband method would have reported that the white population was 65 to 67 percent and the black population was 33 to 35 percent. The Voting Rights Project of the Lawyers' Committee was concerned that this level of imprecision would play havoc with its efforts to draw redistricting plans. An ad hoc group within the Leadership Conference on Civil Rights expressed its concern to the U.S. Justice Department's Voting Section, which was then supporting the broadband method. The Census Bureau relented and announced that it would report all sixty-three categories in the redistricting data it will send to the states in March of 2001.

In December 1999, we learned that various agencies were coming to a consensus around an "aggregation" method. Aggregation would combine the sixty-three categories into a smaller number, but no one would be moved from a single-race or multi-race category he or she had chosen. Allocation, on the other hand, would result in people in multi-racial categories being reassigned to single-race groups on the basis of a rule-not necessarily what the person would have chosen if told to of choose only one race. In effect, the census would have asked one question, but modified the answers to fit another (unasked) question.

After more comments by the Leadership Conference, the Administration announced the decision to go with both allocation and aggregation. (For a copy of OMB Bulletin 00-02, go to the OMB website at this address: While the Census Bureau will publish data for all sixty-three categories, states, localities, school boards, and employers will be required to recombine the data into ten or more groups whenever they have to report any racial data about their particular populations. For instance, a school board reporting on the racial breakdown of its school population and teaching staff will aggregate the racial data into the smaller number of categories. The school board would report on the following racial groups:

  • Five single race groups;
  • Four most common bi-racial groups: American Indian-White, Asian-White, Black-White, American Indian-Black;
  • Any other group larger than 1 percent in the particular area;
  • Balance of all other multi-racial groups.

The minimum number of categories would be ten, but could be more if there are one or more multi-racial groups with 1 percent or more in the area for which the report is prepared.

However, the U.S. government would analyze data for civil rights monitoring and enforcement using an allocation method in which the five single-race categories would be left as they are; multi-racial groups which have some white ancestry (e.g., white-Asian) would be allocated to the minority race and multi-racial groups who are minority-minority would be treated in several ways depending on the context and enforcement need. If the question is how all minority groups are being treated, the analyst will probably run multiple analyses with successive allocations of multi-racial groups to their respective single-race groups. For instance, an analysis of educational opportunities will necessarily have to look at the treatment of a variety of racial groups. If the federal government is responding to a discrimination complaint from, say, an Asian, it would probably reallocate all Asian multi-racial groups as Asian for purposes of its analysis. In this way, the analyst could determine whether there is a bias against "pure" Asians as well as "mixed-race" Asians.

How much impact will all of this have? Nationally, probably not much. Fewer than 2 percent of all Americans have marked multiple racial boxes in past surveys that allowed such responses. And, of course, the numbers will vary from place to place. In the 1998 Dress Rehearsal of Census 2000, Sacramento County, California, had 5.4 percent multi-racial respondents while the eleven counties around Columbia, South Carolina, had 0.8 percent.

For several reasons, we can expect the numbers of persons checking multiple racial boxes to increase in the future. First, with increased social acceptance of interra-

Page 26

cial marriages, there will be more children with parents of different races. I expect that these children will be the most likely group to define themselves as bi-racial. The number of mixed race marriages has been growing, but the overwhelming number of marriages are still same-race.

Second, there could be a trend of blacks to claim multiple racial heritages although there is a strong countervailing argument in the African-American community against this. Some African-American leaders have argued in private meetings that African Americans should claim only one racial group. Their argument is that nearly all African Americans have some mixed racial heritage, and yet discrimination occurs against them because of their African ancestry without any benefit being accrued from their European or American Indian ancestry.

Census Sampling Controversy

In January 1999, the Supreme Court decided appeals in two suits which had been filed by the U.S. House of Representatives and by a group of private plaintiffs. These suits had been filed to prevent the Census Bureau from using statistical sampling as part of Census 2000.

The Supreme Court decision only blocked sampling for the count to be used for the determination of the number of House members to be selected from each state. In the far more important area of redistricting congressional and legislative districts within each state--that is actually drawing the lines for the districts--the Supreme Court has held that the law requires the Census Bureau to use sampling and make that corrected data available to the states and localities for use in drawing their own redistricting plans.

Everyone in the United States has an interest in being equally represented in Congress, in their state legislatures, and in their local governments. Those districts can only be drawn equally and accurately if the census data is accurate.

In a 1998 report to Congress, the Bureau of the Census explained its reasons for using sampling. Among the lessons it had learned from the 1990 Census was that some groups were counted less effectively than others ("undercounted"), including children, renters (particularly in rural areas), and racial and ethnic minorities. The undercount rate for African Americans was six times greater than that for non-Hispanic whites. The undercount rate for Hispanics was seven times greater, and the rate for American Indians more than seventeen times greater than that for non-Hispanic whites. This greater undercount rate is called "the differential undercount."

The Bureau reported that the methods it had employed in previous censuses could not sufficiently remedy the undercount, and especially the differential undercount of minorities, that had led to a decrease in accuracy between the 1980 and 1990 censuses. To correct this differential undercount and achieve an overall more accurate count, the Bureau adopted statistical sampling.

A differential undercount adversely affects the voting rights of minorities in specific ways at the district level. If there is a differential undercount of minorities relative to whites, then blacks, Latinos, and other minorities are in more danger than whites of being placed in districts that are unconstitutionally malapportioned.

Also, if there is a differential undercount of blacks or Hispanics relative to whites, creation of majority-minority districts where necessary to comply with Section 2 of the Voting Rights Act will become more difficult.

For these reasons, the Lawyers' Committee for Civil Rights Under the Law and other civil rights organizations are working to insure that accurate census numbers are used in each state's redistricting.

Several states (Colorado, Kansas, Arizona, Alaska, and Virginia) have passed legislation forbidding the use of sampled census data in 2001 redistricting. Since Arizona, Alaska, and Virginia are covered by Section 5 of the Voting Rights Act, they must obtain preclearance from either the Justice Department or the U.S. District Court in Washington D.C. before they utilize any new election standard, practice, or procedure.

Virginia filed suit in the District Court for the District of Columbia in April for preclearance of its anti-sampling law. The Lawyers' Committee, representing the Virginia State Conference of the NAACP and other voters, moved to intervene in the suit. Virginia quickly moved for a summary judgment (a finding that there are no material facts in dispute and the law supports its position). At this writing, we are awaiting a decision by the three-judge panel on the motion.

While Census 2000 will probably be more accurate than the 1990 Census, it will also be marked by more controversy. There will be fights over whether to use adjusted figures for redistricting. There will be fights over the proper interpretation of the racial statistics. As usual, more data means more interpretations-with everyone wanting to interpret the data for their own purposes.