National Congress of American Indians
November 2000

 

The Census, American Indians And Alaska Natives

Why An Accurate Census Is Important For American Indians

The U.S. Constitution requires a census to apportion seats in Congress among the 50 States based on population. The census numbers also are used to draw political boundaries within States, for Congress, state legislatures, and local governing bodies such as school boards.

Beyond reapportionment and redistricting, census data are used to allocate about

$200 billion in Federal aid to states, counties, cities, and Indian tribal governments. The information also is needed to assess the well-being of American Indians and other Native people who maintain a unique trust relationship with the Federal government by virtue of treaties and statutes.

Federal programs that distribute aid to Indians and other Native people based in whole or in part on census data include:

Background

The 1990 census was the first to be measurably less accurate than the one before it. Census Bureau evaluations revealed that about 8 million people were missed, while another 4 million were counted twice, resulting in a net national undercount of 4 million people. Including other types of miscounts, such as people counted in the wrong place or included erroneously, the 1990 census contained about 25 million mistakes, or about 10 percent of the count, according to the U.S. General Accounting Office (GAO).

The accuracy of the 1990 census also varied greatly among population subgroups. Continuing a disturbing trend that the Census Bureau first identified with scientific measurements in 1940, people of color and poor people in urban and rural areas were missed at much higher rates than Whites. In 1990, this gap in census coverage between Whites and African Americans, Hispanics, American Indians, and Asian Americans, known as the "differential undercount," was the highest ever recorded. While the net national undercount was 1.6 percent, only 0.7 percent of non-Hispanic Whites were missed, while nearly 5.0 percent of Hispanics and 4.4 percent of African Americans weren't counted.

The undercount of American Indians in 1990 was similarly disproportionate. About 4.5 percent of all Indians (nearly 175,000 people) weren't counted. The census missed 12.2 percent of Indians living on reservations. Like all children, Indian children were missed at a higher rate; 6.2 percent were left out of the census. And young Indian males (age 18 - 29), like their peers in other population subgroups, were more likely to be missed: 6.4 percent weren't counted.

In Pursuit Of A More Accurate Census In Indian Country

In planning for the 2000 census, the Census Bureau established two overarching goals at the direction of Congress: improve accuracy (and in particular, substantially reduce the differential undercount) and contain costs. A heated and often partisan debate over the Bureau's proposed use of statistical sampling methods in Census 2000 eventually made cost containment a far less important goal (for reasons discussed more fully below). But reducing, if not eliminating, the disproportionate undercount of people of color, the poor, and children remained the highest priority for the Census Bureau, and decisions about Census 2000 design, methods, and operations were aimed at achieving that objective.

The Census Bureau used a combination of methods to count Indians living on reservations. The Bureau worked with tribal leaders to determine the best method for each reservation, taking into account the accessibility of homes and the style of addresses used.

Most American households received a census form in the mail by mid-March and were asked to mail it back by mid-April. However, special procedures were used on many Indian reservations, in Alaska Native villages, and in other rural or remote counties where homes did not have city-style addresses (i.e. number and street name). In these areas, census workers visited homes to determine the exact geographic location and to ensure that housing units were not missed.

Many of those homes were counted using a method called "update/leave." In March 2000, census enumerators visited each household and left a questionnaire for the resident(s) to mail back. At the same time, they updated address lists and maps, making any corrections or additions as they went from house to house. In areas where homes were very remote, census workers went from house to house, compiling an address list, spotting the house on a map, and collecting the information requested on the census form. This operation, called "list/enumerate" or "update/enumerate" (depending on the availability of pre-census address information), was conducted in April-May 2000. (The Census Bureau counted Alaska Native villages using the list/enumerate method in January and February because of unique weather and occupational considerations.)

The second major phase of the census, called "non-response follow-up", ran from late April through June. On Indian reservations during this phase, enumerators visited all households that did not mail back a questionnaire (in "mail out/mail back" or "update/leave" areas) or provide information to a census worker on a previous visit (in "list/enumerate" or "update/enumerate" areas).

After the non-response follow-up phase was complete, the Bureau conducted additional programs to improve the accuracy of the census after the initial counting operations. These activities, called "coverage improvement programs," included a recheck of housing units classified as vacant or nonexistent during earlier operations and additional visits to households that failed to provide complete information on the census form. Census workers also visited homes built after the address list was finalized in the summer of 1999.

Also following the non-response follow-up phase, the Bureau conducted a post enumeration survey (PES) in representative census blocks across the country, including on Indian reservations. Here, the Bureau sent its best enumerators out to re-interview 314,000 representative households across the country, starting over completely with a new set of address lists, in order to check the quality of the information collected in the previous counting operations.

The Bureau is comparing the PES results with the information gathered from the 314,000 survey households during the direct counting operations. This matching process allows the Bureau to measure how many people were missed, counted twice, or otherwise miscounted the first time. The PES and any resulting corrections to the first set of numbers counts are designed to eliminate the persistent, disproportionate undercount of American Indians, other people of color, and the urban and rural poor. The Bureau will use statistical estimation to project those findings onto demographically and geographically similar census blocks across the country. It will use that information to adjust the first set of numbers to account for people missed or eliminate double-counts. Virtually all experts who have studied the census process agree that the Census Bureau cannot reduce this 'differential undercount' without using sampling and statistical methods to supplement the direct counting effort.

Preliminary reports indicate that census operations generally went well on Indian reservations. The pace of operations ran well ahead of similar operations in 1990, when list/enumerate and coverage improvement efforts dragged through July and into August on some Indian reservations.

Most Indian reservations were counted using the "update/enumerate" method, a new hybrid of update/leave and list/enumerate. Initial update/enumerate operations were finished by the end of May 2000, while coverage improvement activities were completed in July in most areas. The Bureau has almost completed the 'matching' process for the sample quality-check survey and is starting its statistical estimation work to measure undercounting and overcounting. Most of the 520 local census offices established for Census 2000 are now closed; the remaining offices will close by the end of October. The three temporary data capture and processing centers (Pomona, CA; Phoenix; and Baltimore) will stop operating by December 1, 2000, and the Census Bureau will dismantle the equipment and other infrastructure by next spring.

The recruitment and hiring of Indians to help conduct the count on reservations was reasonably successful, although there were some problem areas. In some parts of Minnesota, the Bureau brought in enumerators from other local census offices to complete the update/enumerate and coverage improvement work. This transfer of temporary employees who had completed their initial assignments was not unique to Indian reservations, however; in some urban neighborhoods, the Bureau successfully reassigned more productive census takers from other locations to improve productivity in areas where the door-to-door counting operations were lagging behind schedule.

The census counts compiled through mailed questionnaires, door-to-door visits, and coverage improvement programs, will be used to apportion the 435 seats in the U.S. House of Representatives among the 50 states, as required by the Constitution. By law, the Census Bureau must report the total population of each state to the President by December 31, 2000, for congressional apportionment purposes. More detailed population counts for counties, cities, tribal lands, census tracts, and census blocks must be transmitted to the states by April 1, 2001, for redistricting purposes, and will include data on race, Hispanic origin, and voting age (over or under age 18). (The redistricting data will be available to the public on the Internet and on CD-ROM within a day or two of transmittal to the state legislatures.)

Race and Tribal Affiliation

As in previous censuses, the 2000 census questionnaire asked respondents to identify their race. For the first time, however, respondents were given the choice to check off more than one race. The racial categories set by the Office of Management and Budget (OMB), which oversees the collection of statistics by Federal agencies, are: American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White. The census asks a separate question on whether respondents are Hispanic or Latino. People who identify themselves as American Indian are asked to write-in the tribe in which they are enrolled or their principal tribe.

The race categories and the multiple response option for Census 2000 are the product of a lengthy review and revision of the former OMB Statistical Policy Directive 15, which had been in place since the late 1970s. The new "standards" (as they are called), put in place on October 30, 1997, must be adopted by all federal agencies no later than January 1, 2003.

Following the issuance of the "Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity," OMB established a working group to draft guidelines for reporting multiple responses to the race question in the census and other federal surveys. In February 1999, OMB issued "Draft Provisional Guidance on the Implementation of the 1997 Standards for the Collection of Federal Data on Race and Ethnicity." The report set forth several approaches for tabulating multiple race responses. There are 63 possible combinations of race (including single race responses) from the data that will be collected in Census 2000. All 63 race combinations can be further classified by ethnicity: "Hispanic or Latino" or "Not Hispanic or Latino".

The Census Bureau will not decide how to tabulate race responses to the census for the purpose of administering various government programs or setting policies. The one exception is the determination of how to tabulate the data that the Bureau, by law, must transmit to the state legislatures by April 1, 2001, for use in the political redistricting process. After consultations with the U.S. Department of Justice and other stakeholders, the Census Bureau decided to provide data for 63 race categories, combined with the separate Hispanic origin responses, in the block-level data transmitted to the States for redistricting.

In a March 9, 2000 memo, OMB issued further guidance on tabulating race data under its revised standards, to federal agencies charged with monitoring and enforcing civil rights laws. For the purpose of reporting data to federal agencies, the OMB guidelines call for aggregating the information by the five single race categories; four double race combinations projected to be chosen most frequently; other multiple race combinations that comprise more than one percent of the population in the relevant jurisdiction (as determined by the responsible agency); and a balance category for all other responses. The four, frequently reported double race combinations are American Indian/Alaska Native and White; Asian and White; Black/African American and White; and American Indian/Alaska Native and Black/African American.

For civil rights monitoring and enforcement, federal agencies will allocate multiple race responses in several ways: combinations of one minority race and White are allocated to the minority race; combinations that include two or more minority races are allocated to the race that is cited as the basis for discrimination, in the case of an individual complaint of discrimination. In cases that require an assessment of disparate impact or discriminatory patterns (such as the employment record of a particular company), the enforcement agency will review the patterns based on alternative allocations to each of the minority groups reported.

Census Bureau staff are continuing their discussions about how to tabulate responses to the race question for all other Census 2000 data products. Other tabulation methods being considered include the "all inclusive" distribution (multiple-race responses would be assigned to each of the selected single race categories, resulting in totals that exceed 100 percent of respondents), the "single race" distribution (single race responses would be reported independently, and all multiple-race responses would be included in a category called, "two or more races"), and various allocation methods (i.e., allocating multiple race responses to one of the single race categories selected, based on a pre-determined standard, in order to facilitate comparison with data collected under the old policy).

In addition to the redistricting files, the Census Bureau's data products include tabulations of demographic and socio-economic data collected on the census short and long forms. (The long form, sent to one in six households nationwide, collects more detailed information on topics such as housing, educational attainment and labor force participation, computing patterns, veterans status, income, and ancestry.) Data from the census, tabulated for varying levels of geography, will be published on a flow basis from 2001 to 2003. Summary File 1 (SF1), which the Census Bureau will publish starting in June 2001 through the end of that year, will include population counts for detailed race and Hispanic origin categories, as well as American Indian and Alaska Native tribes, at the census block and tract levels. SF1 will be available on the Internet and CD-ROM.

Sampling In The Census

In January 1999, the U.S. Supreme Court ruled in two lawsuits, filed by Republicans in the U.S. House of Representatives and by a conservative legal advocacy organization based in Atlanta, that the Census Act (title 13, United States Code) prohibits the use of sampling methods to derive the population totals used as the basis for congressional apportionment. (The Court did not rule on the constitutionality of those methods.) The opinion in Department of Commerce v. U.S. House of Representatives left open the possibility of using statistically corrected, or adjusted, census numbers for non-apportionment purposes, such as drawing legislative district boundaries (redistricting) and distributing federal funds. The scientific sample survey (described above) is part of the original Census 2000 plan and was not barred by the Supreme Court ruling except to calculate the state population totals used for congressional apportionment.

As the result of the Court's ruling, the Census Bureau will produce two sets of census population numbers: one derived solely from traditional, direct counting methods, and the second using statistical sampling methods to adjust the direct count for people missed or counted twice, as measured by the quality check survey. On October 6, 2000, the Commerce Department issued a final rule (sometimes called a regulation), delegating to the Census Bureau director final authority for the decision on whether to release statistically adjusted census data to the States next spring for redistricting purposes.

The Census Bureau will not decide until early March 2001 if the adjusted numbers are more accurate than the uncorrected numbers collected in Census 2000, but it has issued a preliminary determination that statistically corrected data can be produced within the time frame required by law (i.e. for redistricting purposes) and that those data will be more accurate than the raw census counts, assuming the Bureau's methodology is carried out in a sound manner. A law passed by Congress in late 1997 requires the Census Bureau to publish unadjusted numbers down to the census block level if it decides to adjust the numbers officially.

The Census And Redistricting

In April 1999, Arizona became the first state to enact a law against the use of statistically corrected numbers for drawing congressional and state legislative district boundaries following the 2000 census. Arizona is one of 16 states required to obtain approval from the U.S. Department of Justice (USDOJ) under section 5 of the 1965 Voting Rights Act for any change to state election laws. States (or sub-state areas) covered under section 5 must show that any such changes do not have the purpose or effect of denying or abridging the right to vote of racial, ethnic, or language minorities.

Since then, four other states - Alaska, Colorado, Kansas, and Virginia - have enacted similar laws barring the use of adjusted census data for redistricting. Of those states, Alaska and Virginia also are subject to section 5 pre-clearance requirements before the anti-sampling laws can take effect. None of the three section 5 states that enacted anti-sampling redistricting laws have obtained USDOJ approval. Arizona decided to withdraw its pre-clearance submission last winter (giving no substantive reason). USDOJ informed Alaska that it could not properly evaluate its new law until data from Census 2000 are available in March 2001. And in October 2000, a federal district court (in the District of Columbia) rejected Virginia's lawsuit seeking approval for its anti-sampling law without obtaining USDOJ pre-clearance under section 5 of the Voting Rights Act. The court in that case agreed with USDOJ's argument that it could not fully evaluate the effect of Virginia's law on the voting rights of racial minorities until Census 2000 data are available next year. Virginia is likely to appeal the decision directly to the Supreme Court.

In addition to the five states that enacted prohibitions against using adjusted data for redistricting, one chamber of the legislature in several other states passed bills to accomplish the same goal. State legislatures that have considered such bills or non-binding resolutions in the past 18 months include New Jersey, Michigan, Indiana, Nevada, and Utah.

* * * * *