Billing Code: 3110-01
Federal Register 7/9/97, Part II. Pages 36873-36946
OFFICE OF MANAGEMENT AND BUDGET

Recommendations from the Interagency Committee for the Review of the Racial and Ethnic Standards to the Office of Management and Budget Concerning Changes to the Standards for the Classification of Federal Data on Race and Ethnicity

AGENCY: Executive Office of the President, Office of Management and Budget (OMB), Office of Information and Regulatory Affairs

ACTION: Notice and Request for Comments

SUMMARY: OMB requests comments on the recommendations that it has received from the Interagency Committee for the Review of the Racial and Ethnic Standards (Interagency Committee) for changes to OMB's Statistical Policy Directive No. 15, Race and Ethnic Standards for Federal Statistics and Administrative Reporting (See Appendix 1 for the text of the standards in Directive No. 15, originally issued in 1977). The Interagency Committee's report and recommendations, which are published in Appendix 2 in their entirety, are the result of a four-year, comprehensive review of the current standards.

DATES: To ensure consideration during the final decision making process, written comments must be provided to OMB no later than [INSERT DATE 60 DAYS FROM DATE OF PUBLICATION].

ADDRESSES: Written comments on the recommendations may be addressed to Katherine K. Wallman, Chief Statistician, Office of Information and Regulatory Affairs, Office of Management and Budget, NEOB, Room 10201, 725 17th Street, N.W., Washington, D.C. 20503.

Comments may also be submitted by facsimile to 202-395-7245, or by electronic mail to OMBDIR15@A1.EOP.GOV (please note that "1" in "A1" is the number one and not the letter "l"). Be sure to include your name and complete postal mailing address in the comments sent by electronic mail. If you submit comments by facsimile or electronic mail, please do not also submit them by regular mail.

Electronic availability and addresses: This Federal Register Notice, as well as the June 9, 1994 and the August 28, 1995 Federal Register Notices related to the review, are available electronically from the OMB Homepage on the World Wide Web: /OMB/fedreg/, and in paper copy from the OMB Publications Office, 725, 17th Street, NW, NEOB, Room 2200, Washington, D.C. 20503, telephone: (202) 395-7332, facsimile: (202) 395-6137.

FOR FURTHER INFORMATION CONTACT: Suzann Evinger, Statistical Policy Office, Office of Information and Regulatory Affairs, Office of Management and Budget, NEOB, Room 10201, 725 17th Street, N.W., Washington, D.C. 20503. Telephone: 202-395-3093.

SUPPLEMENTARY INFORMATION:

A. Background

The current standards were developed in cooperation with the Federal agencies to provide consistent and comparable data on race and ethnicity throughout the Federal government for an array of statistical and administrative programs. Development of the data standards stemmed in large measure from new responsibilities to enforce civil rights laws. Data were needed to monitor equal access to housing, education, employment opportunities, etc., for population groups that historically had experienced discrimination and differential treatment because of their race or ethnicity. The categories that were developed represent a political-social construct designed to be used in the collection of data on the race and ethnicity of major broad population groups in this country, and are not anthropologically or scientifically based. The standards are used not only in the decennial census (which provides the "denominator" for many measures), but also in household surveys, on administrative forms (e.g., school registration and mortgage lending applications), and in medical and other research.

The standards provide a minimum set of categories for data on race and ethnicity.

The current standards have four categories for data on race (American Indian or Alaskan Native, Asian or Pacific Islander, Black, and White) and two categories for data on ethnicity ("Hispanic origin" and "Not of Hispanic origin"). The standards als o permit the collection of more detailed information on population groups provided that any additional categories can be aggregated into the minimum standard set of categories. Self-identification is the preferred means of obtaining information about an individual's race and ethnicity, except in instances where observer identification is more practical (e .g., completing a death certificate).

The categories in Directive No. 15 do not identify or designate certain population groups as "minority groups." As the Directive explicitly states, these categories are not to be used for determining the eligibility of population groups for pa rticipation in any Federal programs. Directive No. 15 does not establish criteria or qualifications (such as blood quantum levels) that are to be used in determining a particular individual's racial or ethnic classification. Directive No. 15 does not tell an individual who he or she is, or specify how an individual should classify himself or herself.

B. Review Process

Particularly since the 1990 census, the standards have come under increasing criticism from those who believe that the minimum categories set forth in Directive No. 15 do not reflect the increasing diversity of our Nation's population that has resulted primarily from growth in immigration and in interracial marriages. In response to the criticisms, OMB announced in July 1993 that it would undertake a comprehensive review of the current categories for data on race and ethnicity.

This review has been conducted over the last four years in collaboration with the Interagency Committee for the Review of the Racial and Ethnic Standards, which OMB established in March 1994 to facilitate the participation of Federal agencies in the review.

The members of the Interagency Committee, from more than 30 agencies, represent the many and diverse Federal needs for data on race and ethnicity, including statutory requirements for such data.

The principal objective of the review is to enhance the accuracy of the demographic information collected by the Federal Government. The starting point for the review was the current minimum set of categories for data on race and ethnicity that have provided 20 years of information for a variety of purposes, and the recognition of the importance of being able to maintain this historical continuity. The review process has had two major elements: (1) public comment on the present standards, whi ch helped to identify concerns and provided numerous suggestions for changing the standards; and (2) research and testing related to assessing the possible effects of suggested changes on the quality and usefulness of the resulting data.

Public input, the first element of the review process, was sought through a variety of means: (1) During 1993, Congressman Thomas C. Sawyer, then Chairman of the House Subcommittee on Census, Statistics, and Postal Personnel, held four hearings that included 27 witnesses, focusing particularly on the use of the categories in the 2000 census. (2) At the request of OMB, the National Academy of Sciences' Committee on National Statistics (CNSTAT) conducted a workshop in February 1994 to articulate issues surrounding a review of the categories. The workshop included representatives of Federal agencies, academia, social science research institutions, interest groups, private industry, and a local school district. (A summary of the workshop, Spotlight on Heterogeneity: The Federal Standards for Racial and Ethnic Classification, is available from CNSTAT, 2101 Constitution Avenue, N.W., Washington, D.C. 20418.) (3) On June 9, 1994, OMB published a Federal Register (59 FR 29831-29835) Notice that contained background information on the development of the current standards and requested public comment on: the adequacy of current racial and ethnic categories; the principles that should govern any proposed revisions to the standards; and specific suggestions for change that had been offered by individuals and interested groups over the past sev eral years. In response, OMB received nearly 800 letters. As part of this comment period and to bring the review closer to the public, OMB also heard testimony from 94 witnesses at hearings held during July 1994 in Boston, Denver, San Francisco, and Hon olulu. (4) In an August 28, 1995, Federal Register (60 FR 44674-44693) Notice, OMB provided an interim report on the review process, including a summary of the comments on the June 1994 Federal Register Notice, and offered a final opportunity for comment on the research to be conducted during 1996. (5) OMB staff have also made themselves available to discuss the review process with various interested groups and have made presentations at many meetings.

The second element of the review process involved research and testing of various proposed changes. The categories in OMB's Directive No. 15 are used not only to produce data on the demographic characteristics of the population, but also for ci vil rights enforcement and program administration. Research would enable an objective assessment of the data quality issues associated with various approaches to collecting data on race and ethnicity. For that reason, the Interagency Committee's Research Working Group on Racial and Ethnic Standards, which is co-chaired by the Bureau of the Census and the Bureau of Labor Statistics, reviewed the various criticisms and suggestion s for changing the current categories, and developed a research agenda for some of the more significant issues that had been identified. These issues included collecting and classifying data on persons who identify themselves as "multiracial"; combining race and Hispanic origin in one question or having separate questions on race and Hispanic origin; combining the concepts of race, ethnicity, and ancestry; changing the terminology used for particular categories; and adding new categories to the current minimum se t.

Because the mode of data collection can have an effect on how a person responds, the research agenda addressed the issue of how an individual responds when an interviewer collects the information (in an in-person interview or a telephone intervi ew) versus how an individual responds in a self-administered situation, such as in the decennial census when a form is filled out and mailed back. In addition, cognitive research interviews were conducted with various groups to provide guidance on the wording of the questions and the instructions.

The research agenda included several major national tests during the last two years, the results of which are discussed throughout the Interagency Committee's report:

(1) In May 1995, the Bureau of Labor Statistics (BLS) sponsored a Supplement on Race and Ethnicity to the Current Population Survey (CPS). The findings were made available in a 1996 report, Testing Methods of Collecting Racial and Ethnic Information: Results of the Current Population Survey Supplement on Race and Ethnicity, available from BLS, 2 Massachusetts Avenue, N.E., Room 4915, Postal Square Building, Washington, D.C. 20212, by calling 202-606-7375. The results were also summarized in an October 26, 1995, news release, which is available electronically at http://stats.bls.gov/news.release/ethnic.toc.htm. (2) The Bureau of the Census, as part of its research for the 2000 census, tested alternative approaches to collecting data on race and ethnicity in the March 1996 National Content Survey (NCS). The Census Bureau published the results in a December 1996 report, Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey; highlights of the report are available at http://www.census.gov/population/www/socdemo/96natcontentsurvey.html. (3) In June 1996, the Census Bureau conducted the Race and Ethnic Targeted Test (RAETT), which was designed to permit assessments of effects of possible changes on smaller popul ations not reliably measured in national samples, including American Indians, Alaska Natives, detailed Asian and Pacific Islander groups (such as Chinese and Hawaiians) and detailed Hispanic groups (such as Puerto Ricans and Cubans). The Census Bureau r eleased the results in a May 1997 report, Results of the 1996 Race and Ethnic Targeted Test; highlights of the report are available at http://www/census.gov/population/www/documentation/twps-0018.html. Single copies (paper) of the NCS and RAETT reports may be obtained from the Population Division, U.S. Bureau of the Census, Washington, D.C. 20233; telephone 301-457-2402.

In addition to these three major tests, the National Center for Education Statistics (NCES) and the Office for Civil Rights in the Department of Education jointly conducted a survey of 1,000 public schools to determine how schools collect data o n the race and ethnicity of their students and how the administrative records containing these data are maintained to meet statutory requirements for reporting aggregate information to the Federal Government. NCES published the results in a March 1996 report, Racial and Ethnic Classifications Used by Public Schools. The report is available electronically at http://www.ed.gov/NCES/pubs/98092.html. Single paper copies may be obtained from NCES, 555 New Jersey, NW, Washington, D.C. 20208-5574, or by calling 202-219-1442.

The research agenda also included studies conducted by the National Center for Health Statistics, the Office of the Assistant Secretary for Health, and the Centers for Disease Control and Prevention to evaluate the procedures used and the qualit y of the information in administrative records on race and ethnicity such as that reported on birth certificates and recorded on death certificates. Since these data are used in studies of diseases and of the health and well-being of major population groups, these studies investigated possible impacts of suggested changes on data needed for medical and health research.

C. Overview of Interagency Committee Report

This Federal Register Notice makes available for comment the Interagency Committee's recommendations for how OMB should revise Directive No. 15. These recommendations are elaborated in the Interagency Committee's Report to the Office of Management and Budget on the Review of Statistical Policy Directive No. 15 which is published in its entirety as part of this Notice. The report consists of six chapters. Chapter 1 provides a brief history of Directive No. 15, a summary of the issues considered by the Interagency Committee, a review of the research activities, and a discussion of the criteria used in conducting the evaluation. Chapter 2 discusses a number of general concerns that need to be addressed when considering any changes to the current standards. Chapters 3 through 5 report the results of the research as they bear on the more significant suggestions OMB received for changes to Directive No. 15.

Chapter 6 gives the Interagency's Committee's recommendations concerning the various suggested changes based on a review of public comments and testimony and the research results.

This Notice affords a final opportunity for the public to comment before OMB acts on the recommendations of the Interagency Committee. None of the recommendations has been adopted and no interim decisions have been made concerning them. OMB ca n modify or reject any of the recommendations, and OMB has the option of making no changes. The report and its recommendations are published in this Notice because OMB believes that they are worthy of public discussion and that OMB's decision will benefi t from obtaining the public's views on the recommendations. OMB will announce its decision in mid-October 1997, so that changes, if any, can be incorporated into the questions for the 2000 census "dress rehearsal," which will be conducted in spring 1998.

ISSUES FOR COMMENT: With this Notice, OMB requests comments on the recommendations it has received from the Interagency Committee for the Review of the Racial and Ethnic Standards concerning the revision of Statistical Policy Directive No. 15. These recommendations are contained in Chapter 6 of the Interagency Committee's report.

The complete report is included in this Notice because Chapters 1 through 5 provide both a context and the bases for the Interagency Committee's recommendations outlined in Chapter 6. As an aid in evaluating the recommendations, readers may wis h to refer to the set of general principles (see Chapter 1) that were developed at the beginning of the Directive No. 15 review to govern the process -- a process that has attempted to balance statistical issues, needs for data, social concerns, and the personal dimensions of racial and ethnic identification. The committee recognized that these principles may in some cases represent competing goals for the standard. For example, having categories that are comprehensive in the coverage of our Nation's diverse population (Principle 4) and that would facilitate self-identification (Principle 2) may not be operationally feasible in terms of the burden that would be placed upon respondents and the public and private costs that would be associated with implementation (Principle 8). The following are just a few examples of questions that might be considered in assessing the recommendations using the general principles:

• Do the recommendations provide categories for classifying data on race and ethnicity that are: generally understood and accepted by the public (Principle 3); comprehensive in coverage (Principle 4); and useful for statistical analysis, and for Federal statutory and programmatic requirements (Principles 5 and 6)?
• Are the recommendations based on sound methodological research (Principle 9)?
• Do the recommendations take into account continuity of historical data series (Principle 10)?

As reflected in the general principles, the goal has been to produce a standard that would result in consistent, publicly accepted data on race and ethnicity which will meet the needs of the Federal Government and the public, while recognizing the diversity of the population and respecting the individual's dignity. We would appreciate receiving your views and comments on any aspects of the Interagency Committee's recommendations, as well as on the extent to which the recommendations were successful in meeting the goals of the governing principles.

_______________________
Sally Katzen
Administrator, Office of Information and Regulatory Affairs

This Directive provides standard classifications for record keeping, collection, and presentation of data on race and ethnicity in Federal program administrative reporting and statistical activities. These classifications should not be interpreted as being scientific or anthropological in nature, nor should they be viewed as determinants of eligibility for participation in any Federal program. They have been developed in response to needs expressed by both the executive branch and the Congress to provide for the collection and use of compatible, nonduplicated, exchangeable racial and ethnic data by Federal agencies.

1. Definitions

The basic racial and ethnic categories for Federal statistics and program administrative reporting are defined as follows:

• a. American Indian or Alaskan Native. A person having origins in any of the original peoples of North America, and who maintains cultural identification through tribal affiliation or community recognition.

  b. Asian or Pacific Islander. A person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands. This area includes, for example, China, India, Japan, Korea, the Philippine Islands, and Samoa.

  c. Black. A person having origins in any of the black racial groups of Africa.

  d. Hispanic. A person of Mexican, Puerto Rican, Cuban, Central or South American or other Spanish culture or origin, regardless of race.

  e. White. A person having origins in any of the original peoples of Europe, North Africa, or the Middle East.

2. Utilization for Record keeping and Reporting

To provide flexibility, it is preferable to collect data on race and ethnicity separately. If separate race and ethnic categories are used, the minimum designations are:

• a. Race:
  -- American Indian or Alaskan Native
  --Asian or Pacific Islander
  --Black
  --White

  b. Ethnicity:
  --Hispanic origin
  --Not of Hispanic origin

When race and ethnicity are collected separately, the number of White and Black persons who are Hispanic must be identifiable, and capable of being reported in that category.

If a combined format is used to collect racial and ethnic data, the minimum acceptable categories are:

• --American Indian or Alaskan Native
  --Asian or Pacific Islander
  --Black, not of Hispanic origin
  --Hispanic
  --White, not of Hispanic origin.

The category which most closely reflects the individual's recognition in his community should be used for purposes of reporting on persons who are of mixed racial and/or ethnic origins.

In no case should the provisions of this Directive be construed to limit the collection of data to the categories described above. However, any reporting required which uses more detail shall be organized in such a way that the additional categories can be aggregated into these basic racial/ethnic categories.

The minimum standard collection categories shall be utilized for reporting as follows:

• a. Civil rights compliance reporting. The categories specified above will be used by all agencies in either the separate or combined format for civil rights compliance reporting and equal employment reporting for both the public and private sectors and for all levels of government. Any variation requiring less detailed data or data which cannot be aggregated into the basic categories will have to be specifically approved by the Office of Management and Budget (OMB) for executive agencies. More detailed reporting which can be aggregated to the basic categories may be used at the agencies' discretion.

  b. General program administrative and grant reporting. Whenever an agency subject to this Directive issues new or revised administrative reporting or record keeping requirements which include racial or ethnic data, the agency will use the race/ethnic categories described above. A variance can be specifically requested from OMB, but such a variance will be granted only if the agency can demonstrate that it is not reasonable for the primary reporter to determine the racial or ethnic background in terms of the specified categories, and that such determination is not critical to the administration of the program in question, or if the specific program is directed to only one or a limited number of race/ethnic groups, e.g., Indian tribal activities.

  c. Statistical reporting. The categories described in this Directive will be used at a minimum for federally sponsored statistical data collection where race and/or ethnicity is required, except when: the collection involves a sample of such size that the data on the smaller categories would be unreliable, or when the collection effort focuses on a specific racial or ethnic group. A repetitive survey shall be deemed to have an adequate sample size if the racial and ethnic data can be reliably aggregated on a biennial basis. Any other variation will have to be specifically authorized by OMB through the reports clearance process. In those cases where the data collection is not subject to the reports clearance process, a direct request for a variance should be made to OMB.

3. Effective Date

The provisions of this Directive are effective immediately for all new and revised record keeping or reporting requirements containing racial and/or ethnic information. All existing record keeping or reporting requirements shall be made consistent with this Directive at the time they are submitted for extension, or not later than January 1, 1980.

4. Presentation of Race/Ethnic Data

Displays of racial and ethnic compliance and statistical data will use the category designations listed above. The designation "nonwhite" is not acceptable for use in the presentation of Federal Government data. It is not to be used in any publication of compliance or statistical data or in the text of any compliance or statistical report.

In cases where the above designations are considered inappropriate for presentation of statistical data on particular programs or for particular regional areas, the sponsoring agency may use:

• (1) The designations "Black and Other Races" or "All Other Races," as collective descriptions of minority races when the most summary distinction between the majority and minority races is appropriate;

  (2) The designations "White," "Black,"and "All Other Races" when the distinction among the majority race, the principal minority race and other races is appropriate; or

  (3) The designation of a particular minority race or races, and the inclusion of "Whites" with "All Other Races," if such a collective description is appropriate.

In displaying detailed information which represents a combination of race and ethnicity, the description of the data being displayed must clearly indicate that both bases of classification are being used.

When the primary focus of a statistical report is on two or more specific identifiable groups in the population, one or more of which is racial or ethnic, it is acceptable to display data for each of the particular groups separately and to describe data relating to the remainder of the population by an appropriate collective description.

May 28, 1997
MEMORANDUM FOR KATHERINE K. WALLMAN
Chief Statistician
Office of Management and Budget

FROM: Interagency Committee for the Review of the Racial and Ethnic Standards

SUBJECT: Transmittal of Report and Recommendations on the Review of Directive No. 15

We are pleased to transmit to you the attached report that provides the recommendations of the Interagency Committee for the Review of the Racial and Ethnic Standards for modifying OMB's Statistical Policy Directive No. 15, Race and Ethnic Standards for Federal Statistics and Administrative Reporting. These recommendations, which are outlined in Chapter 6 of the report, represent our best technical and professional advice for how these data standards could better reflect the increasing racial and ethnic diversity of our Nation's population, while maintaining historical continuity.

Our recommendations for Directive No. 15 are the product of a three-year review process that is briefly described in Chapter 1 of the report. During that time, we developed and carried out a research program to evaluate various proposals for revising the standards. Chapter 2 discusses some general concerns relevant to consideration of any changes in the standards. Chapters 3 through 5 report on the extensive research efforts, including three national tests, that have been conducted to test alternative approaches for questions to collect data on race and ethnicity. The Interagency Committee's recommendations, presented in Chapter 6, are based on our evaluation of the research results and consideration of related public comments and testimony.

We hope that the Office of Management and Budget will find this report with its accompanying recommendations informative and helpful in making its decision on what changes to adopt, if any, in the Federal standards for reporting data on race and ethnicity.

Attachment

CHAPTER 1. Introduction

1.1 Overview
1.2 History of Directive No. 15
1.3 Concerns About the Current Standards
1.4 Principles for the Review Process
1.5 Overview of Research Activities
1.6 Evaluation of Research Results

CHAPTER 2. Issues of General Concern

2.1 Overview
2.2 Satisfying Statutory and Program Needs
2.3 Voting Rights Issues
2.4 Data Continuity Concerns
2.6 Financial Costs

CHAPTER 3. Reporting More Than One Race

3.1 Background
3.2 Current Practice
3.3 Overview of Research on Reporting More Than One Race
3.3.1 Surveys to Explore Options
3.3.2 Cognitive Research to Guide Survey Design
3.4 Evaluating Research on Options for Reporting More Than One Race
3.4.1 Data Comparability
3.4.2 Should a multiracial category be listed among the response options to the question on race?
3.4.3 If a multiracial category is listed, should a "follow-up" format be used, in which individuals who select the category are asked to specify their racial identities?
3.4.4 Should a multiple-response format be used, in which the respondent is instructed to "mark one or more races"?
3.4.5 Should a multiple response format be used in which the respondent is instructed to "mark all that apply" on the race question?
3.4.6 Are there other options for reporting more than one race by respondents?3.5 Trends With Respect to Reporting Multiple Races
3.5.1 Trends Contributing to Reporting of Multiple Races
3.5.1.1 Increases in Interracial Marriages and Households and Births to Parents of Different Races
3.5.1.2 State Requirements for Multiracial Reporting
3.5.2 Public Sentiment
3.6 Measurement Concerns and Opportunities Related to Reporting More Than One Race
3.6.1 Meeting Legislative and Program Needs
3.6.2 Defining and Using the Term "Multiracial"
3.6.2.1 Definition of "Multiracial"
3.6.2.2 Using a Stand-Alone "Multiracial or Biracial" Category or Including a Follow-up Question
3.6.3 Using a "Mark One or More" or a "Mark All That Apply" Instruction in the Race Question
3.6.4 Issues Related to Primary and Secondary Data Collection
3.7 Some Implications of Allowing the Reporting of More Than One Race
3.7.1 Possible Effects on Reporting by Particular Population Groups
3.7.2 Tabulation of Multiple Responses
3.7.3 Monetary Costs and Resource Burdens

CHAPTER 4. A Combined Race and Hispanic Origin Question

4.1 Background
4.2 Concepts of Race and Ethnicity
4.3 Self-Identification
4.4 Some Alternative Formats for Questions
4.5 Research on Data Quality
4.5.1 Reporting in the "Other Race" Category by Hispanics
4.5.2 Item Nonresponse in the Race Question
4.5.3 Item Nonresponse in the Hispanic Origin Question
4.5.4 Reporting Inconsistency
4.6 Measures to Correct Misreporting in the Race Question and the Hispanic Origin Question
4.7 The Effects of Combining the Race Question and the Hispanic Origin Question into a Single Question
4.7.1 Results From the May 1995 CPS Supplement on Race and Ethnic Origin 4.7.2 Results From the Race and Ethnic Targeted Test
4.7.2.1 Reporting of Hispanic Origin
4.7.2.2 Reporting of Multiple Races
4.7.2.3 Summary of Findings
4.8 Public Sentiment
4.9 Additional Cost Concerns

CHAPTER 5. Other Possible Changes

5.1 Background
5.2 Specific Suggestions
5.3 Evaluation of the Possible Effects of Suggested Changes
5.3.1 Changes related to American Indians and Alaska Natives
5.3.1.1 Should the term "American Indian" or "Native American" be used?
5.3.1.2 Should the term "Alaska Native" or "Eskimo and Aleut" be used?
5.3.1.3 Should a distinction be made between federally recognized and nonfederally recognized tribes?
5.3.1.4 What is the best way to elicit tribal affiliation?
5.3.1.5 Should the definition of the "American Indian or Alaskan Native" category be changed to include Indians indigenous to Central America and South America?
5.3.2 Changes related to Asian and Pacific Islanders
5.3.2.1 Should the "Asian or Pacific Islander" category be split into two categories? If yes, how should this be done?
5.3.2.2 Should specific groups be listed under the Asian or Pacific Islander category?
5.3.2.3 Should the term "Guamanian" or "Chamorro" be used?
5.3.3 Changes related to Hawaiians
5.3.3.1 Should the term "Native Hawaiian" or "Hawaiian" be used?
5.3.3.2 Should Hawaiians continue to be included in the "Asian or Pacific Islander" category; be reclassified and included in the "American Indian or Alaskan Native" category; or be established as a separate, new category?
5.3.4 Other terminology issues
5.3.4.1 Should the term "Black" or "African American" be used?
5.3.4.2 Should the term "Hispanic" or "Latino" be used?
5.3.4.3 Should more than one term be used for Black or for Hispanic?
5.3.5 Other new category issues
5.3.5.1 Should an Arab or Middle Eastern category be created and, if so, how should it be defined?
5.3.5.2 Should a Cape Verdean category be created?

CHAPTER 6. Recommendations and Major Findings

6.1 Summary of Recommendations and Major Findings
6.1.1 Recommendations concerning reporting more than one race
6.1.1.1 Findings concerning a method for reporting more than one race
6.1.1.2 Findings concerning different formats for reporting more than one race
6.1.2 Recommendations concerning a combined race and Hispanic ethnicity question
6.1.2.1 Findings concerning whether race and Hispanic origin should be combined into a single question
6.1.2.2 Findings concerning different formats if race and Hispanic origin are combined in a single question
6.1.3 Recommendations concerning the retention of both reporting formats
6.1.4 Recommendation concerning the ordering of the Hispanic origin and race questions
6.1.5 Recommendation concerning adding Cape Verdean as an ethnic category
6.1.6 Recommendation concerning the addition of an Arab or Middle Eastern ethnic category
6.1.7 Recommendation concerning the addition of any other categories to the minimum set
6.1.8 Recommendation concerning changing the term "American Indian" to "Native American"
6.1.9 Recommendation concerning changing the term "Hawaiian" to "Native Hawaiian"
6.1.10 Recommendation concerning the classification of Hawaiians
6.1.11 Recommendations concerning the use of Alaskan Native instead of Eskimo and Aleut
6.1.12 Recommendations concerning the classification of South and Central American Indians
6.1.13 Recommendations concerning the term or terms to be used for the name of the Black category
6.1.14 Recommendations concerning the term or terms to be used for Hispanic
6.2 Comparison of the Current Standards with the Recommended Standards
6.2.1 The Current Standards in Directive No. 15
6.2.2 Recommended Standards
6.3 Recommendations for Further Research

CHAPTER 1. Introduction

1.1 Overview

This report evaluates a variety of proposals for modifying the Office of Management and Budget's (OMB) Statistical Policy Directive No. 15, "Race and Ethnic Standards for Federal Statistics and Administrative Reporting." The Directive sets forth a minimum set of categories for collecting and presenting data on race and Hispanic origin. This basic set of categories has served as the guideline for Federal Government data collections since it was issued in May 1977. The report presented here, including its recommendations, is the culmination of three years of research undertaken by Federal agencies to evaluate the possible impact of suggested changes on the quality and cost of the resulting data. It is the work of the Interagency Committee for the Review of the Racial and Ethnic Standards and its Research Working Group on Racial and Ethnic Standards. OMB established the Interagency Committee in 1994 to evaluate various proposed changes and provide recommendations. The committee created the Research Working Group to develop and carry out a research agenda for evaluating the proposals.

The report consists of six chapters. This first chapter provides a brief history of Directive No. 15, a summary of the issues considered by the Interagency Committee, a review of the research activities over the past three years, and a discussion of the criteria used in conducting the evaluation. Chapter 2 discusses several general concerns that need to be addressed when considering any changes to the current standards. Chapters 3 through 5 report the research results as they bear on the more significant suggestions for changes to Directive No. 15. These suggestions include, but are not limited to, permitting respondents to report multiple racial backgrounds, a single question on race and ethnicity that would include Hispanic as a category, expanding the minimum set of categories to include other specific ethnic or racial groups, and adding to, or replacing the names of categories used to identify specific racial or ethnic groups. Chapter 6 presents the committee's recommendations on various suggested changes based on its evaluation of the research results and consideration of related public comments and testimony.

1.2 History of Directive No. 15

The United States Government has long collected statistics on race and ethnicity. Such data have been used to monitor changes in the social, demographic, health, and economic characteristics of various groups in our population. Federal data collections, through censuses, surveys, and administrative records, have provided an historical record of the Nation's population diversity and its changing social attitudes, health status, and policy concerns.

Since the 1960s, data on race and ethnicity have been used extensively in monitoring and enforcing civil rights laws covering areas such as education, employment, housing and mortgage lending, health care, voting rights, and the administration of justice. These legislatively based priorities created the need among Federal agencies for compatible, nonduplicative data for population groups that historically had suffered discrimination on the basis of their race or ethnicity. In response, OMB issued, in 1977, the current set of categories for use in the collection and presentation of data on race and ethnicity. The categories also implemented the requirements of Public Law 94-311 of June 16, 1976, which called for the collection, analysis, and publication of economic and social statistics on persons of Spanish origin or descent.

The current standard provides that, if racial and ethnic data are collected separately, the minimum racial categories are:

• --American Indian or Alaskan Native. A person having origins in any of the original peoples of North America, and who maintains cultural identification through tribal affiliation or community recognition.

  --Asian or Pacific Islander. A person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands. This area includes, for example, China, India, Japan, Korea, the Philippine Islands, and Samoa.

  --Black. A person having origins in any of the black racial groups of Africa.

  --White. A person having origins in any of the original peoples of Europe, North Africa, or the Middle East.

For ethnicity, the categories are:

• --Hispanic origin. A person of Mexican, Puerto Rican, Cuban, Central or South American or other Spanish culture or origin, regardless of race.

  -- Not of Hispanic origin. A person not of any Spanish culture or origin. When a combined format is used, the minimum categories are: (1) American Indian or Alaskan Native; (2) Asian or Pacific Islander; (3) Black, not of Hispanic origin; (4) Hispanic; and (5) White, not of Hispanic origin.

The current categories originated in the work of the Federal Interagency Committee on Education (FICE) whose membership represented some 30 Federal agencies. In June 1974, FICE created an Ad Hoc Committee on Racial and Ethnic Definitions, whose 25 members came from Federal agencies with major responsibilities for the collection or use of data on race and ethnicity. This ad hoc committee was charged with developing terms and definitions for a broad range of data on race and ethnicity to be collected by Federal agencies on a compatible and nonduplicative basis. The committee sought to ensure that the categories could be aggregated, disaggregated, or otherwise combined so that the data developed by one agency could be used in conjunction with the data developed by another agency. The committee also suggested that the basic categories could be subdivided into more detailed ethnic subgroups to meet users' needs, but that to maintain comparability, data from one major category should never be combined with data from any other category.

In the spring of 1975, FICE completed its work on a draft set of categories. An agreement was reached among OMB, the General Accounting Office (GAO), the Department of Health, Education, and Welfare's (HEW) Office for Civil Rights, and the Equal Employment Opportunity Commission (EEOC) to adopt these categories for a trial period of at least one year. This trial was undertaken to test the new categories and definitions and to determine what problems, if any, would be encountered in their implementation.

At the end of the test period, OMB and GAO convened an Ad Hoc Committee on Racial/Ethnic Categories to review the experience of the agencies that had implemented the standard categories and definitions and to discuss any potential problems that might be encountered in extending the use of the categories to all Federal agencies. The Committee met in August 1976 and included representatives of OMB; GAO; the Departments of Justice, Labor, HEW, and Housing and Urban Development; the Bureau of the Census; and the EEOC. Based upon the discussion in that meeting, OMB prepared minor revisions to the FICE definitions and circulated the proposed final draft for agency comment. These revised categories and definitions became effective in September 1976 for all compliance record keeping and reporting required by the Federal agencies represented on the Ad Hoc Committee.

Based upon this interagency agreement, OMB drafted for agency comment a proposed revision of the "race and color designations in Federal statistics" contained in its circular on Standards and Guidelines for Federal Statistics. Some agencies published the draft revision for public comment. Following receipt of comments and incorporation of suggested modifications, OMB, on May 12, 1977, promulgated the racial and ethnic categories now set forth in Directive No. 15. Thus, for the first time, standard categories and definitions were to be used by all Federal agencies in both the collection and the presentation of data on race and ethnicity. The categories and definitions were developed primarily on the basis of geography; therefore, they were not to be interpreted as being scientific or anthropological in nature. The racial and ethnic categories in the Directive reflected, in particular, agency needs for data for use in monitoring and enforcing civil rights laws.

Although the standards given in Directive No. 15 have not been revised since 1977, OMB did publish in the January 20, 1988, Federal Register a draft Statistical Policy Circular soliciting public comment on a comprehensive revision of existing Statistical Policy Directives. Among the proposed changes was a revision of Directive No. 15 that would have added an "Other" racial category and required classification by self-identification. This proposal was supported by many multiracial and multiethnic groups and some educational institutions, but it drew strong opposition from large corporations and Federal agencies such as the Civil Rights Division of the Department of Justice, the Department of Health and Human Services, the EEOC, and the Office of Personnel Management (OPM). Critics asserted that the present system provided adequate data, that any changes would disrupt historical continuity, and that the proposed changes would be expensive and potentially divisive. Some members of minority communities interpreted the proposal as an attempt to provoke internal dissension within their communities and to reduce the official counts of their populations. Because it was evident from all of these comments that this proposal would not be widely accepted, no changes were made to Directive No. 15.

1.3 Concerns About the Current Standards

The population of the United States has become increasingly diverse during the 20 years that the current standards have been in effect. During the 1980s, immigration to the United States from Mexico, Central and South America, the Caribbean, and Asia reached historic proportions. The 1990 census data show that the population of the United States is more racially and ethnically diverse than ever. Furthermore, as a result of the growth in interracial marriages, there is an increase in the number of persons born who are of mixed race or ethnicity. In recent years, Directive No. 15 has been criticized for not sufficiently reflecting this growing diversity.

In addition, there have been a number of other concerns expressed. For example:

• --The categories and their definitions have been criticized as failing to be comprehensive and scientific.

  --Some have suggested that the geographic orientation of the definitions for the various racial and ethnic categories is not sufficiently definitive. They believe that there is no readily apparent organizing principle for making such distinctions and that definitions for the categories should be eliminated.

  --Others maintain that the identification of an individual's racial and ethnic "category" often is a subjective determination, rather than one that is objective and factual. Thus, they believe that it may no longer be appropriate to consider the categories as a "statistical standard."

  --There is disagreement over the use of self-identification versus observer identification.

  --Some critics have said that the two formats permitted by Directive No. 15 are not compatible. They argue that, when using the two separate questions, race and Hispanic origin can be kept analytically distinct, but in the combined race/ethnicity format, they cannot. While many find the combined format particularly suitable for observer identification, the use of this format does not provide information on the race of those selecting it. As a result, the combined format makes it impossible to distribute persons of Hispanic origin by race and, therefore, may reduce the utility of counts in the four racial categories by excluding from them persons who would otherwise tend to be included.

  --Certain critics have requested an open-ended question to solicit information on race and ethnicity that would combine the concepts of race, ethnicity, and ancestry.

  --The importance of maintaining comparability over time also has also been questioned, given that the categories have changed in the decennial censuses over the decades.

  --Some have said that the collection categories should allow for capturing greater diversity, but that the categories used to present data should be aggregations of the more detailed categories.

  --Others assert that the collection of data on race and ethnicity should be eliminated because it perpetuates racism and the fragmentation of society.

The following are some of the suggestions for changes to the current categories that OMB received during the current review process:

• --Add a "multiracial" category to the list of racial designations so that respondents would not be forced to deny part of their heritage by having to choose a single category.

  --Add an "other" category for individuals of multiracial heritage and for those who want the option of specifically stating a unique identification.

  --Change the name of the "Black" category to "African American."

  --Change the name of the "American Indian or Alaskan Native" category to "Native American."

  --Since race and ethnicity are not distinct concepts, include Hispanic as a racial category, rather than as a separate ethnic category.

  --Add a "Middle Eastern" or "Arab" ethnic category.

  --Add a "Cape Verdean" ethnic category.

  --Make "Native Hawaiians" a separate category or include "Native Hawaiians" in the American Indian or Alaskan Native category, rather than retain "Native Hawaiians" in the Asian or Pacific Islander category.

  --Change the name of the "Hispanic" category to "Latino."

During 1993, Thomas C. Sawyer, then Chairman of the House of Representatives' Subcommittee on Census, Statistics, and Postal Personnel, held four hearings on the measurement of race and ethnicity in the decennial census. In testimony on July 29, 1993, OMB announced that it would undertake a comprehensive review of the categories, including an analysis of the possible effects of any proposed changes to the categories on the quality and utility of the resulting data that are used for a multiplicity of purposes.

As a first step, OMB asked the Committee on National Statistics (CNSTAT) of the National Academy of Sciences to convene a workshop to provide an informed discussion of the issues surrounding a review of the categories. The workshop, held on February 17-18, 1994, included representatives of Federal agencies, academia, social science research institutions, interest groups, private industry, and a local school district.

1.4 Principles for the Review Process

In March 1994, OMB established and held the first meeting of the Interagency Committee for the Review of the Racial and Ethnic Standards, whose members from more than 30 agencies represent the many and diverse Federal needs for data on race and ethnicity, including statutory requirements for such data. Given the range of suggestions and criticisms concerning Directive No. 15, OMB sought in constituting the committee to have all agency stakeholders participate in this comprehensive review of the standards. Agencies represented on the Interagency Committee included:

Department of Agriculture

National Agricultural Statistics Service
Economic Research Service

Department of Commerce

Bureau of the Census

Department of Defense

Defense Manpower Data Center
Office of the Secretary

Department of Education

National Center for Education Statistics
Office for Civil Rights

Department of Health and Human Services

Administration for Native Americans
Agency for Health Care Policy and Research
Centers for Disease Control and Prevention
Indian Health Service
National Center for Health Statistics
National Institutes of Health
Office for Civil Rights
Office of Minority Health
Office of Refugee Resettlement

Department of Housing and Urban Development

Department of the Interior

Bureau of Indian Affairs

Department of Justice

Bureau of Justice Statistics
Civil Rights Division
Immigration and Naturalization Service

Department of Labor

Bureau of Labor Statistics
Office of Federal Contract Compliance Programs

Department of Transportation

Bureau of Transportation Statistics

Department of Veterans Affairs
Equal Opportunity Employment Commission
Federal Reserve Board
National Science Foundation
Office of Personnel Management
Small Business Administration
U.S. Commission on Civil Rights
Office of Management and Budget, ex officio

The Interagency Committee developed a set of general principles to govern the review process. This process was designed not only to evaluate suggestions received from the public but also to balance statistical issues, data needs, social concerns, and the personal dimensions of racial and ethnic identification. These principles were as follows:

1. The racial and ethnic categories set forth in the standards should not be interpreted as being primarily biological or genetic in reference. Race and ethnicity may be thought of in terms of social and cultural characteristics as well as ancestry.

2. Respect for individual dignity should guide the processes and methods for collecting data on race and ethnicity; ideally, respondent self-identification should be facilitated to the greatest extent possible, recognizing that in some data collection systems observer identification is more practical.

3. To the extent practicable, the concepts and terminology should reflect clear and generally understood definitions that can achieve broad public acceptance. To assure they are reliable, meaningful, and understood by respondents and observers, the racial and ethnic categories set forth in the standard should be developed using appropriate scientific methodologies, including the social sciences.

4. The racial and ethnic categories should be comprehensive in coverage and produce compatible, nonduplicative, exchangeable data across Federal agencies.

5. Foremost consideration should be given to data aggregations by race and ethnicity that are useful for statistical analysis and program administration and assessment, bearing in mind that the standards are not intended to be used to establish eligibility for participation in any federal program.

6. The standards should be developed to meet, at a minimum, Federal legislative and programmatic requirements. Consideration should also be given to needs at the State and local government levels, including American Indian tribal and Alaska Native village governments, as well as to general societal needs for these data.

7. The categories should set forth a minimum standard; additional categories should be permitted provided they can be aggregated to the standard categories. The number of standard categories should be kept to a manageable size, determined by statistical concerns and data needs.

8. A revised set of categories should be operationally feasible in terms of burden placed upon respondents; public and private costs to implement the revisions should be a factor in the decision.

9. Any changes in the categories should be based on sound methodological research and should include evaluations of the impact of any changes not only on the usefulness of the resulting data but also on the comparability of any new categories with the existing ones.

10. Any revision to the categories should provide for a crosswalk at the time of adoption between the old and the new categories so that historical data series can be statistically adjusted and comparisons can be made.

11. Because of the many and varied needs and strong interdependence of Federal agencies for racial and ethnic data, any changes to the existing categories should be the product of an interagency collaborative effort.

12. Time will be allowed to phase in any new categories. Agencies will not be required to update historical records.

13. The new directive should be applicable throughout the U.S. Federal statistical system. The standard or standards must be usable for the decennial census, current surveys, and administrative records, including those using observer identification.

The committee recognized that these principles may in some cases represent competing goals for the standards. By applying these principles to the review process, the committee hoped to produce a standard that would result in consistent, publicly accepted data on race and ethnicity that would meet the needs of the Federal Government and the public while, at the same time, recognizing the diversity of the population and respecting the individual's dignity.

OMB invited comment on the principles when they were published in a June 9, 1994, Federal Register Notice. That Notice also contained background information on the development of Directive No. 15; the revision proposed but not made in 1988; the 1993 congressional hearings; and the CNSTAT workshop. OMB requested public comment on the adequacy of the current categories, as well as on the suggested changes it had received over the years. As part of the public comment period, OMB also held hearings in Boston, Denver, San Francisco, and Honolulu during July 1994. OMB received nearly 800 letters in response to the 1994 Federal Register Notice and heard testimony of 94 witnesses during the four public hearings. A wide array of interested parties provided comments, including individuals, data users, and data providers from within and outside the Federal Government.

1.5 Overview of Research Activities

The Interagency Committee created a Research Working Group to outline an agenda for researching and testing key concerns. The Research Working Group, in August 1995, issued the "Research Agenda for the Review of the Racial and Ethnic Categories in Directive No. 15," based on an examination of the information in the June 1994 Federal Register Notice, the public comments it engendered, and previous research. This agenda identified five central research issues together with a number of questions associated with these issues. Some of the questions cut across several of the central issues, and others were unique to a particular issue. In developing the research agenda, the Research Working Group gave equal weight to the conceptual and the operational questions that must be answered before any changes to Directive No. 15 can be considered. The five central issues were:

• (1) Reporting of multiple races. What are the possible effects of including a multiple race response option or a multiracial category in data collections that ask individuals to identify their race and ethnicity?

  (2) Combining questions on race and Hispanic origin. Should a combined race/Hispanic origin question be used instead of separate questions on race and Hispanic origin?

  (3) Concepts of race, ethnicity, and ancestry. Should the concepts of race, ethnicity, and ancestry be combined and include, for example, a follow-up, open-ended question with no fixed categories? How well does the public understand these three concepts?

  (4) Terminology. Should any of the current terminology for the racial and ethnic categories be replaced or modified?

  (5) New classifications. Should new racial or ethnic categories be developed for specific population groups and be added to the minimum basic set of categories?

The most important conceptual questions surrounding these issues were (1) who are the stakeholders, (2) how are various terms used and understood, (3) what is the respondent's view of the task of self-identification, (4) what would be the effects of any changes on population counts and historical trends, and (5) what would be the effects of any changes on the quality and usefulness of the resulting data? The most important operational questions were (1) how would the changes affect data collection procedures, (2) what differences might there be between collection and reporting categories, (3) how could continuity be maintained, (4) how should any changes be implemented, and (5) how might cognitive research assist in implementing any changes? In addition to recommending research that should be done, the Research Working Group both encouraged and supported a number of more specific research projects carried out by the individual agencies.

The first national test related to the central issues was the May 1995 Supplement on Race and Ethnicity to the Current Population Survey (CPS), which had a sample of approximately 60,000 households and more than 100,000 persons. The supplement, sponsored by the Bureau of Labor Statistics and conducted by the Bureau of the Census, tested the effects of: (1) adding a multiracial category to the list of races, and (2) including "Hispanic" as a category on the race question. Respondents also were asked about their preferences for terms to describe themselves (e.g., African-American or Black and Latino or Hispanic). Originally, questions concerning the respondent's understanding of the concepts of race, ethnicity, and ancestry were to be included, but extensive cognitive testing prior to creating the survey instrument indicated that these types of questions were confusing and difficult to administer in a large-scale survey. Additional analysis of open-ended responses by cognitive researchers provided possible explanations for the inconsistencies in some respondents' answers to the race and ethnicity questions.

As a part of the research on the subject content for the 2000 census, the Bureau of the Census tested alternative versions of questions on race and Hispanic origin in the March 1996 National Content Survey (NCS). This test was designed to provide information on how members of approximately 90,000 households identify their race and ethnicity in a self-reporting context, in contrast to the CPS Supplement which was administered by interviewers either in person or by telephone. Some NCS panels, comprising about 18,000 households, tested the effects of adding a multiracial category to the race question, placing the Hispanic origin question immediately before the race question, and combining both of these changes. The NCS sample was not designed to detect possible effects of different treatments on relatively small population groups, such as American Indians and Alaskan Natives, detailed Asian and Pacific Islander groups (such as Chinese and Hawaiians), or detailed Hispanic origin groups (such as Puerto Ricans and Cubans). Moreover, because the results were based on the responses from households in the national sample that mailed back questionnaires, the results do not represent the entire national population.

In contrast to the NCS, the Race and Ethnic Targeted Test (RAETT) was designed by the Bureau of the Census to provide findings for smaller population groups. Conducted in June 1996, the RAETT sample included approximately 112,000 urban and rural households. The sample was taken from geographic areas of the country with concentrations of different racial and ethnic populations including American Indians, Alaskan Natives, Asians, Pacific Islanders, Hispanics, Blacks, and White ethnic groups. This design permits assessments of the effects of changes on relatively small populations not reliably measured in national samples. The RAETT tested and evaluated the effects of adding a "multiracial or biracial" category; having instructions in the race question to "mark one or more" or to "mark all that apply; placing the Hispanic origin item before the race item; combining race, Hispanic origin, and ancestry in a single, two-part question; using a combined "Indian (Amer.) or Alaska Native" category; and using a "Native Hawaiian" or "Hawaiian" category.

In the spring of 1995, the National Center for Education Statistics and the Office for Civil Rights in the Department of Education conducted a survey of a thousand public schools. This survey obtained information on how schools currently collect data on students' race and ethnicity, how administrative records containing data on race and ethnicity are maintained and reported, what state laws mandate or require of school systems with respect to collecting data on race and ethnicity, and current issues in schools regarding categories for reporting data on race and ethnicity.

The Centers for Disease Control and Prevention held a Workshop on the Use of Race and Ethnicity in Public Health Surveillance. The workshop had three objectives: (1) to describe the current measures of race and ethnicity and their use in public health surveillance, (2) to assess the use of data on race and ethnicity in surveillance for planning, operation, and evaluation of public health programs, and (3) to propose better use of existing measures for race and ethnicity or to identify alternative measures. The limitations inherent in the current concepts, measures, and uses of race and ethnicity in public health surveillance were identified, and recommendations were made regarding their improvement.

The National Center for Health Statistics and the Office of Public Health and Science sponsored interviews with 763 multiracial and Hispanic women who had a baby during the preceding three years. The purpose of the study was to determine the effects of different question formats on reporting of race on birth certificates. The standard open-ended race question was compared with two experimental versions: (1) an open-ended race question that included the term "multiracial" as one of several examples, and (2) a "mark all that apply" format. When possible, results were compared with the race the respondent recorded on the youngest child's birth certificate.

A literature search on work related to racial classification in the health field (using MEDLINE) was conducted by the Department of Health and Human Services (HHS). An inventory of HHS minority health data bases that provides information on the data available and on the data collection problems that have been encountered was developed.

A focus group was conducted with state and local government members of the Association of Public Data Users. The participants were asked about possible effects of various suggested changes on their organizations. An expert on redistricting and reapportionment was interviewed concerning the effects these same changes might have on reapportionment and redistricting following the 2000 census. A survey of a small number of businesses and professional associations that rely on Federal statistics also was undertaken to ascertain views about the time and costs involved if various changes were made.

1.6 Evaluation of Research Results

Although some of the issues surrounding the proposed revisions may ultimately be settled through policy discussion and the criteria used may at times be subjective, there is an important place in the discussion for empirically grounded research. Thus, this evaluation, while considering such subjective information as stakeholder positions and respondent burden, focuses on the following objective criteria:

• (1) ease of adhering to the principle of self-identification;
  (2) consistency and quality of measurement across time with respect to various subgroups;
  (3) magnitude of changes to current time series;
  (4) ability to provide categories that are meaningful for policy purposes;
  (5) ability to develop implementable reporting standards for all data providers;
  (6) ease of using the measures in different data collection settings;
  (7) ease of creating data editing and adjustment procedures; and
  (8) costs associated with changing or not changing the standards.

To facilitate the use of research results to evaluate alternatives and develop recommendations, the Research Working Group has acted as a clearinghouse for data gathering activities. As such, the Research Working Group has monitored various projects and overseen the consolidation of results in a form intended to be useful for policy makers.

CHAPTER 2. Issues of General Concern

2.1 Overview

This section provides a discussion of several general concerns that the Research Working Group considered during its review of Directive No. 15. They are: (1) statutory and programmatic needs of the Federal agencies for data on race and ethnicity, (2) voting rights issues, (3) data continuity concerns, and (4) financial costs of making changes to the Directive. These concerns merit general consideration because they must be confronted to some degree when dealing with any of the proposed changes. The relationship of specific suggested changes to these concerns will be addressed in later chapters.

2.2 Satisfying Statutory and Program Needs

Federal agencies that collect data on race and ethnicity include, but are not limited to, the Bureau of the Census, the Bureau of Labor Statistics, the Centers for Disease Control and Prevention, the National Center for Health Statistics, and the National Center for Education Statistics. Agencies use data on race and ethnicity for administrating Federal programs, for enforcing the civil rights laws, and for analyses of social, economic, and health trends for population groups.

A principal driving force in the 1970s for the development of the current standards was the need for data on race and ethnicity to enforce the civil rights laws. Some of the agencies that use these data for monitoring and enforcing civil rights laws include the Equal Employment Opportunity Commission (EEOC), the U.S. Commission on Civil Rights, the Civil Rights Division of the Department of Justice, the Office of Federal Contract Compliance Programs in the Department of Labor, the Office for Civil Rights in the Department of Education, and the Office for Civil Rights in the Department of Health and Human Services. State and local governments, educational institutions, and private sector employers use the categories when providing data on race and ethnicity to meet Federal reporting requirements.

Reliable and consistent information is important for enforcing Federal laws. In recent U.S. Supreme Court decisions involving education, employment, and voting rights, the Court has interpreted the Fourteenth Amendment to the United States Constitution to require that governmental decision-making based on racial classifications be subjected to "strict scrutiny" to determine whether it is "narrowly tailored" to meet "compelling State interests." Changes in Directive No. 15 could affect the ability of agencies to carry out the court's mandate. If, for instance, allowing individuals to identify with more than one race would make it more difficult to identify the members and characteristics of a particular racial or ethnic group (such as American Indians and Alaska Natives, or Asians and Pacific Islanders), then determining whether a "compelling State interest" exists with regard to such persons--and whether the government's action is narrowly enough tailored to meet that interest--could become correspondingly more difficult.

Generally, the statutes that require collection of data on race and/or ethnicity do not specify the exact categories that Federal agencies must use. Most of these laws simply require that data on race and ethnicity be collected. The following examples illustrate statutory requirements that specify the exact categories particular agencies must use:

• The Federal Affirmative Employment Program of the U.S. Equal Employment Opportunity Commission is required by 29 CFR 1607.4B. to use the minimum OMB Directive No. 15 categories except in Hawaii (where detailed Asian or Pacific Islander subgroups are to be collected) and Puerto Rico (Hispanic and non-Hispanic)

• Federal agencies are required by the Office of Personnel Management's Federal Personnel Manual 292-I (Book III, pp. 106-107, 296-233 and 298-302) to collect the minimum racial and ethnic categories and eleven national origin categories (Asian Indian, Chinese, Filipino, Guamanian, Hawaiian, Japanese, Korean, Samoan, Vietnamese, all other Asian or Pacific Islanders, and not Hispanic in Puerto Rico) for the Central Personnel Data Files.

• Legislation covering collection of data on race by the Bureau of Indian Affairs has varying definitions of Indian depending on the program (Indian Reorganization Act of 1934, 25 U.S.C. 479 and 25 CFR part 5).

• Contract Compliance Programs of the Employment Standards Administration are required by 41CFR chapter 60 (EEO) to collect data on race and ethnicity for workforce analysis using the categories "Blacks, Spanish-surnamed Americans, American Indians, and Orientals" (41 CFR 60-2.11).

• Data on race and ethnicity from employee selection tests and procedures are to be collected using the categories "Blacks (Negroes), American Indians (including Alaskan Natives), Asians (including Pacific Islanders), Hispanic (including persons of Mexican, Puerto Rican, Cuban, Central or South American, or other Spanish origin or culture regardless of race), Whites (Caucasians) other than Hispanic, and totals" (41 CFR 60-3.4B.).

• The Center for Minority Veterans of the Department of Veterans Affairs is required by Sec. 509, Public Law 103-446 and 38 U.S.C. 317 to use the categories Asian American, Black, Hispanic, Native American (including American Indian, Alaskan Native, and Native Hawaiian), and Pacific-Islander American.

2.3 Voting Rights Issues

Concerns have been raised that changes to the current categories for data on race and ethnicity may affect the usefulness of the data for congressional reapportionment, legislative redistricting, and enforcement of the Voting Rights Act.

Following each decennial census, congressional reapportionment--the redistribution of the 435 seats in the U.S. House of Representatives among the 50 States--is calculated using the population totals for each state and the formula of "equal proportions" adopted by the Congress in 1941 (United States Code, Title 2, Section 2a). Redistricting is the process of redrawing the boundaries of congressional, state, and local legislative districts in accordance with the Fourteenth Amendment's "one-person/one-vote" principle and the standard of population equality as set forth in Wesberry v. Sanders, Reynolds v. Sims, and subsequent court decisions. Changes to Directive No. 15 would be expected to affect congressional reapportionment and one-person/one-vote compliance in redistricting only to the extent that such changes affect the overall response to the decennial census.

Charges of minority vote dilution--the claim that the redistricting plan or at-large election system minimizes or cancels out the voting strength of a minority group--under Section 2 of the Voting Rights Act (which applies nationwide) are usually determined by reference to decennial census data on race and ethnicity. In addition, compliance with Section 5 of the Voting Rights Act--which requires Federal preclearance for new voting practices and procedures in certain states--also is generally determined by reference to decennial census data on race and ethnicity. Changes to Directive No. 15 could have implications for the effective implementation of the Voting Rights Act.

Decennial census data are used to determine the count and distribution of the voter-eligible minority population. Proof that it is possible to draw a district with a voter-eligible minority population in the majority is usually needed to establish a vote dilution claim under Section 2 of the Voting Rights Act. Changes to the current categories that alter the counts of voter-eligible minorities could affect the ability of such groups to mount successful vote dilution claims. The Attorney General's preclearance determinations pursuant to Section 5 of the Voting Rights Act--whether to grant or deny Section 5 preclearance--are often affected by the size and distribution of the minority population.

In addition, data on race and ethnicity from the decennial census frequently are used as independent variables in statistical procedures that estimate group voting behavior, particularly when counts of registered voters by race or ethnicity are not available. These estimates of group voting behavior are essential to vote dilution claims under Section 2 of the Voting Rights Act, as well as to the analysis of many types of voting changes under Section 5 of the Voting Rights Act.

2.4 Data Continuity Concerns

If changes are made to the Federal standards for collecting data on race and ethnicity, it will be critically important to data users to understand the impact of those changes vis-a-vis the categories they have been using for the past 20 years. The acceptance of new ways of reporting race and ethnicity may require supporting information so that users can assess the magnitude of changes to current time series. To that end, alternative methods of tabulating multiple responses on race into the current minimum set of categories must be investigated further.

2.5 Financial Costs

If OMB were to revise the categories for data on race and ethnicity by modifying Directive No. 15, a sizeable number of Federal agencies and others would have to change data collection forms, computer programs, interviewers' and coders' manuals, and other related materials for their data systems. Although Directive No. 15 is a standard for use by Federal agencies, many State and local agencies and private sector entities also follow the Federal standards for collection, record keeping, and presentation data on race and ethnicity. On the other hand, there will be other costs incurred if changes are not made to the current categories, and these costs are also discussed in this section.

If a decision were made either to use separate questions exclusively, or to use a combined format always, or to use a "mark one or more" reporting option for race, or to add a "multiracial" category, there would also be costs for redesigning data editing, coding, and processing systems to accommodate the changes.

Other costs would be associated with changing data base management, retrieval and aggregation programs, and historical table formats. Data base management systems might have to be significantly expanded to provide data comparability with historical series. Procedures might have to be developed for editing multiple responses to achieve this comparability. Staff would have to be trained in the new procedures resulting from any change to the current categories. Since the estimated transition time for changing EEOC data bases would be 2-3 years, data for these years could be severely hampered for enforcement purposes. This would likely result in additional costs for protracted processing of grievances.

The Health Resources and Services Administration (HRSA) of the Department of Health and Human Services has noted that substantial changes for 23 categorical grant programs would be required for competing and noncompeting grant application materials, data entry and report programs, and the preference/priority databases. Alterations in the current collection categories for data on race and ethnicity would require restructuring of the definitions and data collection tools designed to report cross-cutting outcome measures for Title VII and VIII Health Professions and Nursing education and training programs.

During informal discussions, company representatives offered a few examples of the potential impact on private sector employers if changes to the categories were to be made. The costs of making changes to forms is considered to be minimal. Changes in the data systems would be more expensive than changes in the forms, since this effort would be very labor intensive. In addition, if there were new categories, employees might have to be resurveyed in order to update the information on race and ethnicity.

Any changes from the current collection mechanism would entail major program changes for the 700 institutions participating in the seven student assistance campus-based loan and scholarship programs. Review and revision of records for eligibility and fiscal accounting data would be required, including manual review of data, computer programming changes, and changes to the scope of work for contract services. In addition, the Student Financial Aid Guideline and the User Manual for the Electronic Reporting System would require review and revision. Moreover, changes in definitions would require that schools reconcile past and current submissions of data for compatibility to enable HRSA to make appropriate awards to participating institutions.

The Administration on Children and Families (ACF) of the Department of Health and Human Services considers the overall effect of change to the racial and ethnic categories to be marginal. ACF collects data on race and ethnicity for several internal data systems (e.g., foster care, personnel, grant-related information). However, in relation to the total cost of maintenance of these internal data systems, possible changes in the classification of data on race and ethnicity are likely to have only marginal effects. Alterations to racial and ethnic categories used for data systems maintained by private contractors for ACF (e.g., Head Start, Child Abuse and Neglect, Developmentally Disabled, Native American) would not likely cause excessive burden to the data collection effort.

In addition, ACF has data systems that are legislatively mandated and involve data collections by states (such as temporary assistance to needy families, child support enforcement). If the alterations to existing systems are profound, states might be resistant to change or they might seek Federal funds to defray costs of updating state data systems, particularly to meet Federal reporting requirements.

While financial costs would be incurred if changes are made to Directive No. 15, there are other types of costs associated with not making changes. Problems that exist with use of the current Directive will not be resolved. These continuing problems include lack of standardization for classifying data on race and ethnicity across state and Federal agencies; less than optimal participation in Federal surveys (especially item nonresponse); misidentification of individuals and groups in surveys; inaccurate counts and rates; inaccurate research; inaccurate program design, targeting and monitoring; and possibly misallocation of funds. There will continue to be inconsistency even within the same Federal agency if Hispanic origin data continues to be collected using either the combined format or two separate questions. It is not uncommon for the denominator of a rate for Hispanics to be based on data collected using separate questions on race and ethnicity while the numerator is based on data collected using the combined format.

Chapter 3. Reporting More Than One Race

3.1 Background

This chapter addresses issues related to whether or not the Federal standards for data on race and ethnicity should provide an option that permits the reporting of more than one race. The chapter discusses different approaches that have been studied by Federal agencies to provide such an option. It presents findings of the research conducted by Federal agencies on the alternative approaches and identifies potential implications of providing or not providing a response option for reporting more than one race. Following a review of the current standards and an overview of the research conducted, the chapter addresses the following questions:

• Should a multiracial category be listed among the response options to the question on race? (section 3.4.2)

• If a multiracial category is listed, should a "follow-up" format be used, in which individuals who select "multiracial" are asked in a follow-up question to specify their racial identities? (section 3.4.3)

• Should a multiple-response format be used in which the respondent is instructed to "mark one or more races"? (section 3.4.4)

• Should a multiple-response format be used in which the respondent is instructed to "mark all that apply" on the race question? (section 3.4.5)

• Are there other options for reporting of more than one race by respondents? (section 3.4.6)

Sections 3.5 through 3.7 discuss some of the trends, concerns, and potential implications related to adding (or not adding) an option for reporting more than one race to the Federal standard for collecting and reporting racial categories, including the effects on such areas as legal and program needs, measurement issues, and data production.

3.2 Current Practice

Directive No. 15 provides a minimum set of racial and ethnic categories--four categories for data on race (White, Black, American Indian or Alaskan Native, and Asian or Pacific Islander) and two categories for data on ethnicity (Hispanic origin and not of Hispanic origin). The current standard permits Federal agencies to use more detailed categories for collecting data on population groups, so long as the data collection is organized in a way that makes it possible for the agencies to aggregate the more detailed designations into the Directive No. 15 categories.

For persons who identify with more than one race, Directive No. 15 indicates that the single racial category which most closely reflects the individual's recognition in his or her community should be used. Directive No. 15 does not provide for identifying two or more races.

3.3 Overview of Research on Reporting More Than One Race

To assist OMB in deciding whether or not the Federal standard should provide for reporting more than one race, Federal agencies have conducted several major surveys to test the possible effects on data quality of various options. Major objectives of the research and testing programs carried out in 1995 and 1996 have included:

• Analysis of the growth, characteristics, and self-identification patterns of persons in interracial marriages and households;

• Cognitive research to develop alternative race questions with a category called "multiracial" or response options such as "mark one or more" or "mark all that apply;"

• Empirical research on how reporting more than one race is likely to affect current racial distributions in self-administered censuses and surveys (compared, for example, with interviewer and telephone surveys); and

• Research on whether most respondents who self-identified as multiracial will specify more than one race.

3.3.1 Surveys to Explore Options

The Current Population Survey, conducted jointly by the Bureau of Labor Statistics (BLS) and the Bureau of the Census, included a Supplement on Race and Ethnicity in May 1995 (the CPS Supplement). The CPS Supplement was designed to test the effect of asking questions about race and Hispanic ethnicity, with and without a multiracial response option. As part of its research and testing program for Census 2000, the Bureau of the Census conducted two additional studies--the National Content Survey (also known as the 1996 census survey or the Census 2000 survey) and the Race and Ethnic Targeted Test (the RAETT)--to explore the implications of using different formats for questions on respondents' racial identification and reporting of Hispanic origin.

3.3.2 Cognitive Research to Guide Survey Design

The agencies conducted extensive cognitive research to pretest the racial and ethnic categories and the sequencing of the questions on race and Hispanic origin in the survey instruments. An interagency team conducted cognitive research on several versions of the CPS Supplement questionnaire designed for face-to-face and telephone interviews. The race question included a multiracial category, with a follow-up question for reporting the races with which the respondent identified. The questionnaire was tested with a range of racial and ethnic groups in various regions of the United States, and respondents from all groups were able to report that the term "multiracial" meant more than one race. (McKay and de la Puente, 1995)

The Bureau of the Census conducted cognitive research on two different options for reporting more than one race on the race item in a mail survey form. The options consisted of including (1) a "multiracial" category in the race question, and (2) an instruction to mark one or more of the racial categories provided in the race question.

The cognitive research guided the placement of a separate multiracial category in the race item, determined the appropriate number of write-in lines to the multiracial-response box, identified the appropriate terminology for soliciting response from persons of mixed racial parentage (without providing a definition of "multiracial" for this population), and guided the development of the instructions allowing respondents to choose more than one box. Because the cognitive research revealed that some respondents believed the term "multiracial" meant more than two races, the wording "multiracial or biracial" was used in the NCS and the RAETT to convey to respondents that the category is to be used by those who identify with two or more racial groups. (Gerber and de la Puente, 1996)

The cognitive research also was used to develop a "mark one or more" instruction, indicating that respondents could mark more than one racial category as applicable. The initial cognitive work, which offered respondents the choice of marking one racial category or marking more than one racial category, asked those selecting more than one group to specify the race with which they most identified.

Cognitive interviews tested several versions of this question. A number of problems were identified in these interviews. First, some respondents could not absorb or understand the complex instructions that were necessary. Second, the formatting (which was subject to space limitations) made it difficult for some respondents to read and absorb the question fully. Third, respondents who expected a "multiracial" category were disappointed that this response option was not provided. And finally, some respondents were not comfortable with being asked to designate a single race, when they did not want to discount any part of their racial heritage. The question that was ultimately used asked respondents merely to mark the boxes, without also asking them to designate the race with which they most identified. (Gerber and de la Puente, 1996)

Respondents for the cognitive research were recruited on the basis of interracial parentage or ancestry. In testing the use of multiracial reporting options in both the interview and self-administered mail modes, researchers found that many of the respondents recruited based on known multiracial status did not choose to report as multiracial. Reasons they gave for not selecting the multiracial category included: identification with the racial and cultural group of one parent; acceptance of the racial identity perceived to be conferred by their community; and a lack of identification with a "multiracial" group encompassing members of different racial ancestries. (McKay and de la Puente, 1995; Gerber and de la Puente, 1996)

3.4 Evaluating Research on Options for Reporting
More Than One Race

The sections that follow present results from the CPS Supplement, the National Content Survey, and the RAETT as they bear on the alternative approaches outlined at the beginning of this chapter (See section 3.1). Brief descriptions of these surveys follow.

The Current Population Survey is a monthly national sample survey of approximately 60,000 households; it routinely collects information on the race and ethnic origin of household members using the current Directive No. 15 categories. The May 1995 CPS Supplement collected additional racial and ethnic data on the households under four different panel conditions:

• Panel 1 Separate race and Hispanic-origin questions, with no "multiracial" category
  Panel 2 Separate race and Hispanic-origin questions, with "multiracial" category
  Panel 3 Combined race and Hispanic-origin question, with no "multiracial" category
  Panel 4 Combined race and Hispanic-origin question, with "multiracial" category

The CPS Supplement had a response rate of 82.9 percent.

The National Content Survey (NCS), conducted from March through June 1996, was a mail survey of 94,500 households drawn from 1990 decennial census "mail back areas" representing about 95 percent of the country. The NCS included thirteen panels, four of which were designed to evaluate the effects of adding a " multiracial or biracial" category and reversing the sequence of the questions on race and Hispanic origin. It is less representative of American Indians and Alaska Natives, given that about 25 percent of those populations live outside "mail back areas."

The NCS panels were as follows:

• Panel 1 - Separate race and Hispanic origin questions--no "multiracial or biracial" category; race first sequence
  Panel 2 - Separate race and Hispanic origin questions--with "multiracial or biracial" category; race first sequence
  Panel 3 - Separate race and Hispanic origin questions--no "multiracial or biracial" category; Hispanic origin first sequence
  Panel 4 - Separate race and Hispanic origin questions--with "multiracial or biracial" category; Hispanic origin first sequence

Each of the four questionnaires was mailed to a panel of about 6,000 households. The response rate for the four panels was 72 percent; the results are thus based on approximately 18,000 households. Computer-assisted telephone reinterviews were conducted with each household that had completed and returned the NCS form. Because the NCS sample excluded households outside 1990 census mailback areas, and some households did not return a questionnaire, results from the NCS cannot be generalized to the entire national population.

The RAETT, conducted by the Bureau of the Census in the summer of 1996, was the principal vehicle for testing and evaluating several important proposed changes for the race question. The RAETT targeted 112,000 households in areas that have, relative to the Nation as a whole, high concentrations of households in any one of six specified racial or ethnic groups: White ethnic (whether European, Canadian, or American), Black, American Indian, Alaska Native, Asian or Pacific Islander, and Hispanic origin. A total of 58,911 questionnaires were returned, yielding an overall response rate of 53 percent. The RAETT included questions designed to test the effects of a "multiracial or biracial" category as well as "mark one or more" and "mark all that apply" approaches to reporting more than one race, and a combined question on race and Hispanic origin, using eight different panels or versions of the questionnaire. The RAETT panels were as follows:

• Panel A - Separate race and Hispanic origin questions--no "multiracial or biracial" category; Hispanic origin first sequence

  Panel B - Separate race and Hispanic origin questions with "multiracial or biracial" category with write-ins; Hispanic origin first sequence

  Panel C - Separate race and Hispanic origin questions with " mark one or more races" instruction; Hispanic origin first sequence

  Panel D - Separate race and Hispanic origin questions with a "multiracial or biracial" category with write-ins; race first sequence

  Panel E - Combined race, Hispanic origin, and ancestry question with a "multiracial or biracial" category

  Panel F - Combined race, Hispanic origin, and ancestry with "mark one or more boxes" instruction

  Panel G - Separate race and Hispanic origin questions with "multiracial or biracial" category with write-ins; Hispanic origin first sequence; tested terminology and alphabetization of categories

  Panel H - Separate race and Hispanic origin questions with "mark all that apply" instruction; Hispanic origin first sequence

Each of these surveys provides important information about options for collecting and classifying data on race and ethnicity, but each also has its limitations. The CPS Supplement is nationally representative and data were gathered for over 80 percent of the sample, but it could not provide reliable information for smaller groups in the population. The NCS is close to being nationally representative and its use of a mail out/mail back questionnaire is particularly relevant for designing the 2000 census, but the response rate was only 72 percent, and it too could not provide reliable information for smaller groups.

The RAETT design provides a good test of the possible effects of suggested new racial categories because it focuses on populations for which the national surveys often do not provide sufficiently large samples. However, even with a 100 percent response to the RAETT, results could be generalized only to the population in the census tracts in each targeted sample frame. The actual response rate averaged 53 percent, and the response rates in some targeted samples were as low as 34 percent. The sample design of RAETT also does not permit results for different targeted samples to be combined.

3.4.1 Data Comparability

A key concern of some Federal agencies, reflected in the principles that have guided the review of the current standards, has been the comparability of data from any new categories with information produced under the existing categories. In its report on the RAETT, the Bureau of the Census presented -- for purposes of illustration -- different approaches for tabulating the data, using the information provided in the write-in entries to the "multiracial or biracial" category and in multiple responses to the race question. Some of these classification approaches provide examples of procedures that could be developed and used by the agencies as "bridges" between the current and any new classification. The three illustrative approaches were termed the single-race approach, the all-inclusive approach, and the historical series approach. They may be characterized as follows:

Single-race approach. Responses indicating only one racial category would be assigned to that category. Responses from individuals who reported multiple races would be classified into a separate "multiple race" category. This method provides a lower bound for the number who identify with a given category. The results from this approach are readily available from standard tabulations.

All-inclusive approach. Responses are classified into each racial category specified using the minimum set of categories in Directive No. 15. With a single race/ethnicity question using the combined format in Directive No. 15, the all-inclusive Hispanic proportion would be most comparable to the proportion reporting Hispanic when there are separate questions, one for race and one for ethnicity.

The sum of the percentages reported for the four separate racial categories would exceed 100 percent, because multiple race responses would be counted in each reported racial category. In spite of this disadvantage, the all-inclusive approach would provide information on the total number of times the racial category had been selected.

Historical series approach. Unlike the single race or the all-inclusive approach, the historical series approach can take on many variations, just one of which was used in the RAETT illustrative tabulations. The intent of this approach is to classify data into categories that resemble those that have been used historically to enforce current civil rights laws. An individual's response (or responses) is classified into one and only one category, in a set of mutually exclusive and exhaustive categories that add up to 100 percent. For example, in the report on the RAETT, which tested a " multiracial or biracial" category with a write-in to specify races as well as other options for reporting more than one race, the historical series approach classified into the Asian or Pacific Islander category responses of (1) only the Asian or Pacific Islander category, (2) the Asian or Pacific Islander category and also White, (3) the Asian or Pacific Islander category and Other Race, and (4) the Asian or Pacific Islander category and the multiracial category, with no specification of additional races. The "multiracial" or "other" category in the historical series was a residual category which consisted of responses to the "multiracial" category that did not specify any races; and responses of two race categories other than "White" or "Some Other Race." A more complete description of the historical series approach is provided in the RAETT report.

Under the historical series approach, the percentages allocated to each of the major categories were comparable to the data collected without a multiple race reporting option (Panel A of the RAETT), except for the Alaska Native targeted sample. The discrepancy in this group may be due to the fact that this particular targeted sample suffered from both a small sample size and from an extremely low response rate (34 percent).

3.4.2 Should a multiracial category be listed among the response options to the question on race?

The CPS Supplement on Race and Ethnicity, the National Content Survey, and the Race and Ethnic Targeted Test all allowed testing of the effects of adding a multiracial category to the list of races. The CPS Supplement used the term "multiracial" to identify the category, and the NCS and the RAETT used the term "multiracial or biracial."

CPS Supplement. In the CPS Supplement, the race question on Panels 2 and 4 included a " multiracial" category; results were very similar -- a little more than 1.5 percent identified as multiracial in each panel.

Table 3.1 shows that the multiracial response option drew respondents primarily from the American Indian, Eskimo, and Aleut population, and from those who reported in the "Something Else" category. Without a multiracial response category, about 1 percent reported as American Indian, Eskimo, and Aleut. With a multiracial category, about 0.75 percent reported in the American Indian, Eskimo, and Aleut category only.

The proportions reporting in the White category, in the Black category, and in the Asian or Pacific Islander category were not affected by the introduction of the multiracial option in the CPS Supplement.

Panel 1 - separate Hispanic question, no multiracial category
Panel 2 - separate Hispanic question, multiracial category
Panel 3 - no separate Hispanic question, no multiracial category
Panel 4 - no separate Hispanic question, multiracial category.

From Tucker et al., 1996.

National Content Survey. In the NCS, the race question included a multiracial category (using the term "multiracial or biracial") in two of four panels. The percent of respondents identifying themselves as multiracial on the NCS was 1.2 percent on the panel with the race question first (Panel 2), and 1.1 percent on the panel with the Hispanic-origin question first (Panel 4). Thus, as in the CPS, less than 2 percent of the total population chose the multiracial category on the NCS. Hispanics on the NCS were more likely than the total population to identify as multiracial (6.7 percent in Panel 2 and 10.0 percent in Panel 4.)

The addition of a multiracial category had no statistically significant effect on the percentage of persons who reported as White, as Black, as American Indian, or as Asian or Pacific Islander regardless of whether the race or the Hispanic-origin question was asked first. However, the relatively small sample size in the NCS might not detect changes that were substantively important for small populations.

For example, although not statistically significant, the declines in the proportion reporting in the Asian or Pacific Islander category, from 4.0 percent to 2.7 percent in panels where the race question came first, and from 3.4 percent to 2.8 percent when the Hispanic-origin question was asked first, suggested that further analyses should be undertaken. An analysis of the Asian or Pacific Islander write-in responses for those who reported in the multiracial category revealed that if these write-in responses had been reported solely as Asian or Pacific Islander, the proportion of the population in that category would have increased to about 3 percent. These findings, however, cannot be used to draw a firm conclusion about the effects of adding a multiracial category on reporting as Asian and Pacific Islander because the sample sizes were too small.

Adding a multiracial category significantly decreased reporting in the "Other race" category when race was asked first, from 3.3 percent to 1.7 percent. Reporting as "Other race" decreased only 0.3 percent with a multiracial category when the Hispanic-origin question was asked first.

Race and Ethnic Targeted Sample. The RAETT used a total of eight panels, Panels A through H (with A as the control panel). Three of the panels specifically tested the effects of reporting more than one race. In Panel B, the RAETT tested the effects of including a "multiracial or biracial" category. In Panel C, it tested the effects of instructing respondents to "mark one or more" in response to the race question; and in Panel H, it tested the effects of instructing respondents to "mark all that apply" in response to the race question. The results are discussed in succeeding sections of this chapter.

To determine the effects of including a multiracial category, responses to Panel B are compared with responses to Panel A. The findings indicate that the availability of the option to report as "multiracial or biracial" had the most substantial effect in the Asian and Pacific Islander and in the Alaska Native targeted samples. In the other targeted samples, use of the multiracial category had no significant effect on how race was reported. The percentages using the multiracial category in each of the other targeted samples were under 1.0 percent for the White ethnic and the Black targeted samples, 2.33 percent for the Hispanic targeted sample, and 3.67 percent for the American Indian targeted sample. (See Table 3.2.)

From Bureau of the Census, 1997.

In the Asian and Pacific Islander targeted sample, 7.58 percent in Panel B selected the multiracial category, and another 3.06 percent marked more than one race, even though they were instructed to mark only one. The corresponding percentages in the Alaska Native targeted sample were 7.07 percent and 6.32 percent.

The RAETT results show that, if there were the addition of a new category (e.g., multiracial), the proportion reporting in at least one of the current categories may be reduced. In the Asian and Pacific Islander targeted sample, about 2 percent fewer reported in the White (only) category in Panel B, and about 4.5 percent fewer reported in the Asian and Pacific Islander (only) category. Within the Asian and Pacific Islander category, the Hawaiian and the Asian Indian categories had the largest drops in reporting from Panel A to Panel B. However, the response rate for the Asian and Pacific Islander targeted sample was only 55 percent, and the possible impact of nonresponse bias on these comparisons is not known without further research. (See Table 3.3.)

* An asterisk indicates that the difference is statistically significant at the 90-percent confidence interval assuming there is no bias due to a low response rate.

From Bureau of the Census, 1997, Table 1-4R, p. D-6.

In the Alaska Native targeted sample, the response rate was only 34 percent, leading again to the possibility of nonresponse bias and the need for further research. This, and the fact that the percent reporting White (only) increased by about 4.5 percent with the addition of a multiracial category, suggests that the group reporting in Panel A was different in some way from the group reporting in Panel B. In this targeted sample, the multiracial category drew primarily from the American Indian and Alaska Native category. (See Table 3.4).

* An asterisk indicates that the difference is statistically significant at the 90-percent confidence interval as long as there is no bias due to a low response rate. From Bureau of the Census, 1997.

3.4.3 If a multiracial category is listed, should a "follow-up" format be used, in which individuals who select the category are asked to specify their racial identities?

All three of the major research surveys--the CPS Supplement, the NCS, and the RAETT-- used a two-part question to evaluate the effects of a follow-up question on reporting by different racial groups.

CPS Supplement. The responses on the CPS Supplement to the follow-up question for individuals who identified themselves as multiracial are shown in Table 3.5.

From Tucker, et al., 1996.

With the exception of respondents who named only one race, the "American Indian + one other race" group had the highest frequency in both panels, followed by "Asian/Pacific Islander + one race" on Panel 4. All but a small percentage of the Hispanics who used the multiracial category reported only an Hispanic ethnic group. (McKay, Stinson, de la Puente, and Kojetin, 1996)

More than 60 percent of multiracial responses on Panel 2 and close to 20 percent of multiracial responses on Panel 4 did not provide two or more different races. Respondents who reported only a single race, or reported ethnicities as races, were designated as "unconfirmed multiracials." With the addition of an Hispanic category, there was a 90 percent decline among Hispanic "unconfirmed multiracials" between Panels 2 and 4. There was also a 60 percent decline in such entries for non-Hispanics between Panels 2 and 4, which is not readily explained by the presence of the Hispanic category on Panel 4. (See Table 3.6.)

The decline in "unconfirmed multiracials" among Hispanics in Panel 4 may reflect the effect of the combined race and Hispanic origin question on Hispanic reporting. In the case of non-Hispanics, the decline might result from the absence of the influence of a preceding Hispanic origin question.

From McKay, Stinson, de la Puente, and Kojetin, 1996.

Researchers were able to compare the racial identification of CPS respondents on the CPS control card, which represents the current time series, with their racial identification on the CPS Supplement. Table 3.7 displays the results.

From Tucker et al., 1996.

Note: The percentage distribution of the other races for "American Indian, Eskimo, Aleut" respondents in the CPS Supplement was as follows: Panel 1: White, 17.89; Black, 0.64; Asian or Pacific Islander, 0.70; Panel 2: White, 22.10; Black, 10.17; Asian or Pacific Islander, 1.95; Panel 3: White, 14.24; Black, 0.63; Asian or Pacific Islander, 3.50; Hispanic, 4.57; Panel 4: White, 14.44; Black, 2.52; Asian or Pacific Islander, 1.98; Hispanic, 7.43.

As reported above, only the percent of people identifying as American Indian, Eskimo, or Aleut was significantly smaller when a multiracial category was used. However, the largest movement from the American Indian, Eskimo, or Aleut category is always to the White category. (See Note to Table 3.7.) Only 4.24 percent of this group used the multiracial category on Panel 2. On Panel 4, 7.94 percent of those identifying with this group on the CPS Supplement selected multiracial while 7.43 percent chose Hispanic. In sum, a large number of individuals of mixed American Indian and White ancestry changed their racial identification on the CPS Supplement but not necessarily to the multiracial category. This change had a noticeable effect on the American Indian, Eskimo, and Aleut population counts without noticeably affecting counts of the White population.

Researchers analyzed the distribution of CPS Supplement respondents choosing the multiracial category by State to consider whether State legislative requirements for a multiracial category on State records influenced the frequency with which this category was chosen. At the time of the study, Georgia was the only State with a law requiring a multiracial category; six other States (Florida, Illinois, Indiana, Michigan, North Carolina, and Ohio) were in the process of framing legislation requiring a multiracial category. The highest percentage of CPS respondents choosing the multiracial category for these States was 1.5 percent. Among other States, the five with the highest percentage of respondents choosing the multiracial category were: Hawaii, Nevada, Washington, Tennessee, and Alaska. Of these, Hawaii was the highest, with 11.6 percent; the others had percentages between 3.0 percent and 4.7 percent.

The CPS Supplement data were also analyzed to consider the effect of having parents of different races on the reporting of the racial identity of children. Of the CPS households, less than 1 percent involved married partners of different races with children under the age of 16 in the household. About 13 percent of these households involved an Asian/Pacific Islander mother and White father; about 11 percent, a White mother and Black father; about 9 percent, a White mother and multiracial father; about 8 percent, an Hispanic mother and White father; and about 8 percent, a multiracial mother and White father. Almost 32 percent of the children in these households identified as "multiracial."

National Content Survey. In the National Content Survey (NCS), virtually all persons (98 percent) who marked the multiracial category in the panels that included this category provided a write-in response. More than half of these write-in responses (55 percent) identified two or more different races, and about a third showed a racial category and a Hispanic-origin group. The remainder of the write-in responses indicated only one of the racial categories specified in Directive No. 15.

The vast majority (more than 80 percent) of the write-in responses to the multiracial category included White. (This result is consistent with research on interracial and inter-ethnic marriages and households, which usually involve one White spouse (92 percent) or White parent (86 percent).) About 30 percent of the write-in responses included the Asian or Pacific Islander category, about 25 percent involved the Black category, and about 7 percent involved the American Indian category. If the Asian and Pacific Islander write-ins to the multiracial category had been tabulated solely as Asian and Pacific Islander, the proportion of the population in that category would have increased to about 3 percent, still smaller than the 4 percent who selected Asian and Pacific Islander in Panel 1, without a multiracial category.

Race and Ethnic Targeted Test. Information from the write-ins for panels B, D, E, F, and G in the RAETT was tabulated in accordance with the "historical series" and the "all inclusive" approaches described in section 3.4.1. The results are useful in assessing the extent to which write-ins can be used to provide the bridges to the distributions provided by the current classifications. These results are described in other parts of this report.

3.4.4 Should a multiple-response format be used, in which the respondent is instructed to "mark one or more races"?

Another option for collecting data is to allow respondents to select more than one race. Some suggest that this approach has the advantage of preserving detailed data about racial identification that might not be captured with a single multiracial response category, even with write-in lines. This section discusses one instruction that respondents might be given; the next section discusses an alternative instruction. Only the RAETT tested these alternative approaches.

Race and Ethnic Targeted Test--Panels A and B. In the RAETT, some respondents marked more than one box on Panels A and B, despite the instruction on both panels to "mark ONE box..." (Panel B included a "multiracial" category; Panel A did not.) Reporting multiple races on Panel A was especially high in the Alaska Native targeted sample (5.16 percent). This percentage nearly approached the percentage who selected the multiracial category on Panel B in this targeted sample (7.07 percent). Multiple responses on Panel A were also substantial (3.76 percent) in the Asian and Pacific Islander targeted sample. (By comparison, it is estimated that 0.5 percent of respondents to the 1990 census selected more than one race when asked to select only one.)

In the targeted samples of the RAETT, the lowest frequency of marking multiple races on panels with instructions to "mark ONE box" was 0.7 percent in the Black targeted sample. In the Asian and Pacific Islander targeted sample, persons who were born in the United States were far more likely to report multiple races than the foreign-born.

In addition, respondents in all of the targeted samples marked one or more boxes even for the panel that included a multiracial category. That finding suggests that marking multiple races may have a different meaning to some respondents than identifying in a category labeled "multiracial."

Race and Ethnic Targeted Test--Panel C. In the RAETT, Panel C instructed respondents to "mark one or more" races. The percentages in each of the targeted samples that provided multiple responses were under 2 percent for the White ethnic targeted sample and the Black targeted sample, 3.57 percent for the Hispanic targeted sample, 4.22 percent for the American Indian, and 10.03 percent for the Asian and Pacific Islander targeted sample. Approximately the same percentage marked only the Asian and Pacific Islander category in Panel C as selected only that category in Panel A. (The Alaska Native targeted sample did not receive the option to mark one or more.) (See Table 3.8.)

(NA) Not available. From Bureau of the Census, 1997.

3.4.5 Should a multiple response format be used in which the respondent is instructed to "mark all that apply" on the race question?

Respondents evidently interpreted the instruction to "mark all that apply" somewhat differently than the instruction to "mark one or more."

Race and Ethnic Targeted Test--Panel H. The percentages in each of the RAETT targeted samples that provided multiple responses in the "mark all that apply" option were under 2.0 percent for the White ethnic and the Black targeted samples, 2.24 percent for the Hispanic, 4.27 percent for the American Indian, and 11.47 percent for the Asian and Pacific Islander targeted samples. The Alaska Native targeted sample did not receive this option. (See Table 3.9.)

(NA) Not available. From Bureau of the Census, 1997.

In contrast to Panel C, significantly fewer respondents in the Asian and Pacific Islander targeted sample in Panel H, with the "mark all that apply" instruction, selected only the Asian and Pacific Islander category than was the case in Panel A. (See Table 3.10.) If those who marked Asian and Pacific Islander in combination with another category are included with those who marked only Asian and Pacific Islander, the percentages are about the same. The "historical series" approach, described in section 3.4.1 above, also largely eliminated these reductions in reporting. With this tabulation of responses, the percentages reporting as Asian and Pacific Islander on Panel H no longer differed significantly from the percentage on Panel A.

(NA) Not available.

Note: The sample size for Panels A and B is approximately the same, and is approximately twice the sample size of Panels C and H.

From Bureau of the Census, 1997.

3.4.6 Are there other options for reporting more than one race by respondents?

Another option for addressing concerns about reporting multiple races would be to add the category "Other" to the list of races in all Federal data collections. As discussed in Chapter 1 of this report , in 1988, OMB considered a proposal to add "Other" to the list of races. Comments at that time indicated that the proposal was controversial and consensus would not be easily reached. The debate over the "Other" category has continued in the current review of racial and ethnic categories. Some who commented expressed support for the adoption of an "Other" category--if it is open-ended, allowing the identification of biracial and multiracial people and ethnic groups who do not identify with one of the major race groups. Others viewed use of the term as demeaning, or stated that the category was unnecessary or that it was too broad to be of much use. (OMB Federal Register Notice, 1995)

A special exemption from Directive No. 15, granted by OMB, allows the Bureau of the Census to collect data using an "Other race" category, and that category was included in the 1980 and 1990 decennial censuses. In the 1990 Census, more than 250,000 Americans wrote in as their race designation a combination of races or used a term such as "Eurasian" that indicates two or more races.

Under its special exemption, the Bureau of the Census does not assign the "Other race" responses to the Directive No. 15 race categories. The Bureau has, however, developed a Modified Age-Race-Sex (MARS) file that assigns respondents to the standard race categories in order to provide data comparable to vital statistics and other statistical sources. In developing the MARS file, the Bureau of the Census used a complicated set of algorithms. If OMB were to establish a new classification system that provided the "Other race" option, a standard algorithm might be needed across agencies. Alternatively, agencies could simply list "Other race" in tabulations. (National Research Council, 1996)

3.5 Trends With Respect to Reporting Multiple Races

3.5.1 Trends Contributing to Reporting of Multiple Races

As noted earlier in this chapter, a significant number of respondents select more than one race even when asked to select only one. At least two trends may be contributing to this phenomenon.

3.5.1.1 Increases in Interracial Marriages and Households and Births to Parents of Different Races

Some of the impetus for considering an option that allows the reporting of more than one race comes from the increasing number of interracial marriages and births to parents of different races in the past 25 to 30 years. Allowing individuals to report more than one race could provide a more complete report of the Nation's changing society.

Data suggest that individuals from smaller racial population groups are more likely to form interracial unions with individuals from outside their racial population group than are individuals from the White and the Black populations. The White population is such a large proportion of the total United States population, however, that in most interracial marriages one partner is White; similarly, for most children with parents of different races, one parent is White.

• In the 1970 census, there were about 321,000 interracial unions. By 1980, the number had increased to about 1 million; and by 1990 there were about 1.5 million interracial couples. In all but 8 percent of these interracial couples, one spouse (or unmarried partner) was White. In 14 percent of all interracial couples, the non-White spouse was Black; in 22 percent, American Indian and Alaska Native; in 31 percent, Asian and Pacific Islander; and in 25 percent, "Other race" (most of whom were of Hispanic origin).

• Census data indicate that the number of children in interracial families grew from less than one-half million in 1970 to about 2 million in 1990. In 1990, in interracial families with one white partner, for about 34 percent of all children the other parent was American Indian; for 45 percent the other parent was Asian; and for about 20 percent the other parent was Black.

• In 1968, for 2 percent of the births with at least one Black parent, the second parent was reported as White on the birth certificate (8,800). This percentage had increased to 9 percent in 1994 (63,000). Analysis of the change in the numbers of births where one parent is Black and the other is some other race is complicated by the increasing number of births for which the race of the second parent, usually the father, is not given on the birth certificate--40 percent in 1994, compared with 24 percent in 1968. (See Graph 3.1, Births to Minority and White Parents as a Percent of All Births to Minority Parents by Race of Minority Parent: 1968 to 1994.)

• Even with this limitation it can be inferred, from births for which both parents' races are known, that births involving one Black parent and a second parent of another race other than White also are increasing.

• Among births to American Indian and Alaska Native parents, a high percentage of all births involve a second parent of another race. In 1968, 28 percent of all the births with at least one American Indian or Alaska Native parent listed the second parent as White on the birth certificate (6,900); in 1994, it was 45 percent (23,000).

• Among births to Asian or Pacific Islander parents, the percentage of births in which the second parent was listed as White was 28 percent in 1968, about 32 percent between 1971 and 1979, and 26 percent in 1994.

Graph 3.1 Comparison of Births In Which One Parent is White and the Second Parent is American Indian/Alaska Native, Asian or Pacific Islander, or Black as a Percentage of All Births in Which Either Parent Is American Indian/Alaska Native, Asian or Pacific Islander, or Black: 1968 to 1994

3.5.1.2 State Requirements for Multiracial Reporting

Legislative activity at the State level generates further impetus for considering a modification to the Federal standard to provide reporting of more than one race. Advocacy groups for multiracial persons have lobbied many State legislatures for laws to add a multiracial category to all forms and applications used to collect information on race and ethnicity.

Due at least in part to these advocacy efforts, Georgia, Indiana, and Michigan require the use of a stand-alone multiracial category (Georgia since 1994 and Indiana and Michigan since 1995). In these States, the requirement applies to all State forms and applications used to collect data on race and ethnicity, including health department forms. Ohio and Illinois have similarly adopted legislation adding a multiracial category, but these laws affect only school forms that collect data on race and ethnicity. Florida and North Carolina have added a multiracial category (by administrative directives) to school forms that collect information on race and ethnicity.

At least nine other States are considering legislation to add a reporting category of multiracial: California, Massachusetts, New Jersey, New York, Oklahoma, Oregon, Pennsylvania, Texas, and Wisconsin. In Maryland, a bill adding a multiracial category was passed by the legislature in 1995, but was vetoed by the Governor; a task force has been established to review the issue.

State laws enacted thus far specify that if a Federal agency does not accept the multiracial data as a category, then the reporting State agency is to reclassify individuals identified as multiracial to racial or ethnic classifications approved by the Federal agency according to the racial and ethnic distribution of the general population. The term "general population" is not defined in the legislation.

3.5.2 Public Sentiment

Some advocacy groups support adding a category called "multiracial." They represent, for the most part, persons who identify themselves as multiracial, or persons who want to identify their children as multiracial in cases where the parents are of different races. Some are highly critical of an approach that allows for the reporting of only one racial category. This approach, they say, forces children to deny the racial heritage of one parent, thereby adversely affecting self-esteem, sense of family, pride, and psychological well-being. (OMB Federal Register Notice, 1995)

Public comment on how to allow for the reporting of more than one race has ranged from suggestions for a specific category called "multiracial" (without further specification of races) to a preference for identification by listing more than one race (with or without a category called "multiracial"). (OMB Federal Register Notice, 1995)

In some respects, the consequences of adding a multiracial category or of providing an option to report more than one race might be minor. At present, less than 2 percent of the general U. S. population identifies as "multiracial" when the category is included as a response option. Thus, it would be less disturbing to historical data series to add a multiracial category soon, while the size of the population reporting would cause only small changes in data series. A decade or two from now, the multiracial population will be larger and the disturbance to historical series correspondingly greater.

3.6 Measurement Concerns and Opportunities Related
to Reporting More Than One Race

3.6.1 Meeting Legislative and Program Needs

Many Federal agencies use data on race and ethnicity for policy development, program evaluation, and civil rights monitoring and enforcement. A number of these agencies are concerned that adding a new multiracial category, or allowing individuals to report more than one race, could affect the comparability and historical continuity of data series that they rely on to meet their mandates or missions. Some of the concern is related to uncertainty about how the new data (if a new multiracial category were provided) would be reported or how the multiple responses (if respondents were allowed to report more than one race) would be tabulated. For example, in the employment area, representatives of the Equal Employment Opportunity Commission (EEOC) have indicated that adding a multiracial category or using an instruction that permits reporting more than one race could affect the historical comparability of data used for resolving complaints and charges as well as for research, making it difficult particularly to analyze trends.

Other Federal agencies that measure and report on various conditions suggest that the interest in the reporting of multiracial information reflects a growing phenomenon that will have to be addressed sooner or later. In the health field, for example, it is important to collect comprehensive data about the racial heritage of individuals. Studies have indicated that rates of low birth weight, very low birth weight, pre-term delivery, and small-for-gestational-age -- key indicators of children's health status -- were highest when both parents were Black, followed by rates for children with Black mother/White father, White mother/Black father, and both parents White. (Carter-Pokras and LaViest, 1996) In the context of health research, a Federal standard that permitted the reporting of more than one race could better accommodate efforts to identify individuals at high risk for certain medical conditions.

Another example of reporting more than one race is provided by the National Health Interview Survey (NHIS) which since 1982 has been collecting responses on more than one race through the use of a two- part question. The first part allows respondents to select the race or races with which they identify from among those listed on a hand card. Persons who identify more than one race are given a follow-up question which asks them to pick the race that best describes them, and the information from both questions is entered into the person's electronic record. In the surveys that were fielded through 1996, only the first two races circled in the first question and the race that best described the respondent are available for analysis. (The 1997 redesign of the NHIS enables the inclusion of up to five of the races reported in the first question, as well as the race that best describes the respondent.) For persons who reported multiple races, information on the race that best describes them (i.e., that race obtained from the follow-up question) is used to prepare statistics for NHIS publications.

However, an analysis of the data from the first NHIS question asked of multiracial persons (see Table 3.11) revealed the following:

• From 1982-1994, an average of 1.4 percent, nearly 1,500 persons out of a sample of 100,000 per year, reported more than one race in the NHIS. The annual proportion of persons reporting multiple races ranged from 1.2 to 1.8 percent.

• For persons reporting more than one race, the most commonly reported combination was White and Aleut, Eskimo, or American Indian (55 percent).

• About 11.4 percent of respondents who reported more than one race did not select a single race that best represented their background. This group represents 0.2 percent of the total population.

¹ Percentages may not add up to 100.0 percent because of rounding or because persons for whom the first and second reported races are in the same race category are included in the total but not in the multiple-race column.

² The count excludes persons for whom the first and second reported races are in the same race category.

³ The count includes persons who reported a first, second, and/or main race, including those for whom the first and second reported race are in the same race category.

3.6.2 Defining and Using the Term "Multiracial"

A Federal standard adding a "multiracial" category would have to address issues of terminology and definition as well as the issue of whether or not data on specific races would be collected in addition.

3.6.2.1 Definition of "Multiracial"

In the five States that have enacted "multiracial" legislation, the laws call for use of the term "multiracial." (The same is true in several other States where legislation is pending.) Georgia, Indiana, and Michigan have defined "multiracial" as involving parents of different races. In pending legislation, California defines the term "multiracial" as meaning an individual whose biological parents, grandparents, or great-grandparents are of more than one race.

The research findings on the terminology preferred by persons of more than one race are inconclusive. The May 1995 CPS Supplement on Race and Ethnicity indicates that almost the same percentage of multiracial persons preferred the term "multiracial" (28.4 percent) as stated "no preference" (27.8 percent); "Mixed race" was preferred by 16.0 percent, "More than one race" by 6.0 percent, and "Biracial" by 5.7 percent.

Other evidence about terminology comes from a study sponsored by the National Center for Health Statistics involving women whose parents were of different races. The mail and telephone survey interviewed 763 women, some of whom were of mixed racial or Hispanic background, who had had a baby within the preceding three years. Among the respondents, 393 had parents of different races, 149 had one Hispanic parent, and 221 had parents who were either both Hispanic or non-Hispanic and who were of the same race. The study found that the women were more likely to enter two or more specific races than to use a term like "multiracial." (Cantor et al., 1997)

If the Federal standard were to provide for the use of a "multiracial" category, it would be necessary not only to agree on the definition but also to communicate the instructions clearly to respondents as well as interviewers. More emphasis would need to be placed on drafting instructions. The experiences of the States in trying to define the term and the data from the CPS Supplement and the NCS suggest that some confusion exists about the meaning of "multiracial." Absent a generally accepted understanding of the term, confusion could be expected if a "multiracial" category were to be listed among the response options. Most Americans are probably of mixed ancestry, depending on how ancestry is defined, and could confuse ancestry or ethnicity with race. (Also see the discussion in Chapter 4 regarding the concepts of race and ancestry, in regard to the Hispanic population.)

3.6.2.2 Using a Stand-Alone "Multiracial or Biracial" Category or Including a Follow-up Question

The research results indicate that between 1.0 and 1.5 percent of respondents select a multiracial category when offered the opportunity to do so. Providing an option to report by means of a multiracial category with no follow-up question would be responsive to persons who do not want to choose between their different racial heritages. However, since respondents would not be asked to specify their races, it would not be possible to tabulate the responses in the current categories. Concerns about historical continuity of data would not be addressed. While refraining from such a tabulation would be in keeping with self-identification, the responses would provide information of limited utility, particularly for use in health research.

By contrast, a follow-up question would enable the data to be tabulated in the current categories for purposes of historical continuity and trend analysis. Further, with the additional detail, the effects on data for certain groups could be minimized. With a follow-up question, research results suggest that a large percentage of "multiracial" responses could be classified into the categories that have been used since 1977.

A related option would be to use a multiracial category with a write-in. Doing so would take up less space but require more coding than a follow-up question. Conversely, using a follow-up question that specified race categories would take up more space but require less coding.

Another option involves the use of the "Other race" category, as in the decennial census, with a multiracial example. However, the use of this category is offensive to some respondents, and multiracial individuals still would be unable to self-identity in the manner they have requested. With an "Other race" category, a greater amount of coding would be required for the variety of responses.

3.6.3 Using a "Mark One or More" or a
"Mark All That Apply" Instruction in the Race Question

Approximately 0.5 percent of respondents to self-administered surveys, including the 1990 census, already select more than one race, even when asked to select only one. Allowing individuals to report more than one race could increase the accuracy of these data, eliminate some inconsistencies in reporting of race, and improve response rates.

For many Federal agencies, the consequences of implementing the reporting of more than one race could be expected to vary depending on the extent to which responses could be tabulated consistently in accordance with existing racial categories that have been used to meet current legislative mandates. (National Research Council, 1996) If information from multiple responses can be tabulated to the current classifications, the potential for disruption of historical series important to data users would likely be reduced. In particular, such disruption could be minimized if information from persons who have marked multiple boxes could be used to tabulate responses in the race categories currently specified in Directive No. 15. Implementing "mark one or more" or "mark all that apply" approaches would be less burdensome than having to code data from write-ins. The CPS Supplement found that many people provided write-ins that represented ethnicity rather than races, a factor that would unnecessarily increase processing costs. Either of the multiple response approaches could be expected to reduce this type of misunderstanding about the information being asked. Moreover, lengthy definitions of terms would not be needed, whereas if a "multiracial" category were used, instructions would be needed and the wording of the instructions would be extremely important.

3.6.4 Issues Related to Primary and Secondary Data Collections

In many cases, the Federal Government collects data through primary data collections, as in censuses and longitudinal surveys. In primary data collections, agencies rely on essentially two methods for collecting information: by self-identification or by observer identification, which is based on the observer's perception of the most appropriate category in which to report an individual.

With self-identification, individuals would be able to report multiracial backgrounds. In the case of observer identification, however, the observer would have little basis for a realistic assessment of a person's racial background. In this case, a multiple race response option that called for identification of the particular races (including instructions to "mark all that apply") could pose significant data quality problems. This is true today to some extent. For example, American Indians who do not live on or near a reservation are often classified as White or Hispanic.

In other instances, the Federal Government uses secondary data collection, as when it obtains data from institutions and administrative records. Examples include aggregate data collected from colleges and universities on the race and ethnicity of students or degree recipients, or on persons conducting research supported by Federal grants. Reporting could become more burdensome for institutions if individuals who initially provide data to the university were using a multiple response approach. The primary collectors also would need guidance on how to aggregate the raw data into categories specified in the Federal standard.

3.7 Some Implications of Allowing the Reporting of
More Than One Race

3.7.1 Possible Effects on Reporting by
Particular Population Groups

Data available from the CPS Supplement, the NCS, and the RAETT uniformly indicate that adding a multiracial option--whether by means of a multiracial category or providing for multiple-response options--had little effect on the numbers of people who reported as White or as Black. On the other hand, adding a multiracial category had a substantial effect on the reporting in specific racial categories, such as the American Indian (in the CPS Supplement) and the Alaska Native and the Asian and Pacific Islander populations (in the NCS and the RAETT). As noted in section 3.4.1, the Bureau of the Census was able to tabulate substantial percentages of the multiracial responses in the RAETT in the present Directive No. 15 categories using a procedure called the "historical series" approach. (However, there still may be some differences remaining.) Whether this ability to tabulate the data could apply in other contexts needs further investigation.

To the extent that providing a multiracial or a multiple-race response option can change reporting, the affected population could experience some consequences. In the case of the American Indian population, for example, the Bureau of Indian Affairs and the Indian Health Service provide assistance to persons who can prove descent from a member of a federally recognized tribe. Tribal governments have expressed concern that the addition of a multiracial category could affect their ability to identify their members. In the case of health statistics, adding a multiracial category could mean that fewer American Indians/Alaska Natives would be counted for both numerators (number of births to American Indian/Alaska Native mothers) and denominators (total number of American Indian/Alaska Natives). (Carter-Pokras, LaViest, 1996; Hahn, 1992)

3.7.2 Tabulation of Multiple Responses

Whether or not OMB modifies Directive No. 15, some respondents will report more than one race. It is important to ensure that the data are treated uniformly. Accordingly, attention needs to be given to establishing rules for tabulating multiple responses to the race question both for purposes of historical comparability and to ensure consistency across Federal agencies.

An algorithm could be used to tabulate responses in the racial categories that are used currently. For example, one option would be to tabulate responses from a multiracial category in proportion to the distributions for the current single-race categories: with a population of 80 Whites, 10 Asians, and 10 multiracial individuals, the resulting numbers would be 89 Whites and 11 Asians. This algorithm would not change the relative sizes of the single-race categories. However, the tabulation would be arbitrary and could misrepresent the multiracial respondents (if for instance the 10 respondents in the multiracial category were the children of Asian/White unions). (National Research Council, 1996) Moreover, even if this method of tabulation would suffice for some purposes, there are others in which it would be necessary to deal with individual records.

Data from the decennial censuses suggest that the way in which children born into interracial families are identified on the race item does not follow the race and ethnicity distribution of the population. Thus, no simple algorithm could assign a single race based on the races of the parents that adequately matches the race now reported for the children. For instance, while only 12 percent of the United States population is Black, 66 percent of the children of Black and White unions have identified as Black in each census since 1970.

As discussed in section 3.4.1, the Census Bureau developed procedures to address the reduced reporting of only a single race in the RAETT that occurred in some targeted samples when a multiracial category or a multiple-race response option was offered. An algorithm is used in tabulating all multiple race responses. The historical series approach tabulates these responses to the Black, the American Indian and Alaska Native, or the Asian and Pacific Islander category (and to the Hispanic category in two of the RAETT panels). When both the White box and either the "Some other race" or the "multiracial" box were marked, the responses were classified as White. (The extent to which other agencies might be able to implement a similar classification procedure would have to be determined.)

This historical series approach tabulated a large percentage of the multiracial responses in the Directive No. 15 categories. The only targeted sample in which this tabulation did not appear to produce results comparable to the single-race reporting in Panel A was the Alaska Native targeted sample. The historical series noticeably increased the percentages of American Indian and Alaska Native respondents on Panel B (which included a multiracial category) and Panel H (which included a "mark all that apply" instruction); however, the percentages remained lower than on Panel A, which did not offer a multiracial option.

In the cases of the decennial censuses, the Bureau of the Census has not tabulated responses of "Other Race" in the categories specified by Directive No. 15. As noted above, however, the Bureau has developed an algorithm to create a Modified Age-Race-Sex (MARS) file that tabulates responses in the standard race categories to provide data comparable to other statistical systems.

In some cases, the Federal Government already is dealing with the this tabulation issue. In Georgia, Indiana, and Michigan--where the multiracial legislation has general applicability--the requirement to use a multiracial category affects the collection of data on registration certificates for births and deaths and on health survey forms, and it thus affects the reporting of both State and national statistics by race and ethnicity. The National Center for Health Statistics has created a "multiracial" code for vital records from States that have passed such legislation. Multiracial persons are coded by NCHS as "Other" and, before analysis, all such entries are reallocated through an imputation method to the standard race categories, consistent with Directive No. 15. (Carter-Pokras, LaViest, 1996)

A study conducted by the U.S. Department of Education as part of the review of Directive No. 15 found that when categories such as "other" or "multiracial" are used, schools typically aggregate these data into the broad Federal category that is deemed most appropriate by the school staff before reporting the information to the Federal government. (NCES 96-092)

3.7.3 Monetary Costs and Resource Burdens

Efforts were made to obtain estimates of monetary and other resource costs associated with adding a multiracial response option, whether by adding a multiracial category or by allowing for multiple responses to the race question. Several agencies, members of the Council of Professional Associations on Federal Statistics (COPAFS), and State and local data users belonging to the Association of Public Data Users (APDU) provided views.

Some data collections generally would be more costly and difficult if a multiracial category were added (particularly if the changes included a combined format for Hispanic ethnicity). There could be significant costs associated with the disaggregation of the multiracial category into meaningful population groups for enforcement purposes and comparability with a large volume of historical data. Instructions that allowed counting individuals according to more than one race/ethnic group could make it extremely difficult to perform trend analysis. Agencies noted that some of these costs would be ongoing rather than one-time costs.

Costs associated with adding an option to report multiple races could be expected to vary depending on the reporting technique used. If a multiracial category involved a write-in option, for instance, and the responses were assigned to the major groups, the costs for editing and coding entries could be higher than those for fixed categories. Classification algorithms would have to be written, tested, and harmonized across agencies. Further, coding write-in responses could prove more feasible for major statistical agencies with large data processing resources, such as the Bureau of the Census, than for agencies where the collection of racial and ethnic data is only a small portion of their administrative mandate.

In an informal consultation with BLS staff, COPAFS members suggested that in some cases a change in Directive No. 15 would probably mean only minor effects on data systems, Computer Assisted Telephone Interviewing software, and sample management systems. Participants in the discussion noted that a variety of computer-based analytic tools would have to be reprogrammed. In cases where general requirements for data collection apply, changes in industry-wide forms (paper and electronic), electronic data transfer conventions, and computer programs would be needed. Estimates of time range from two to three weeks to reprogram and one to two months to re-estimate models.

COPAFS members also were asked about data systems or software that the organizations would have to revise to accommodate a change. The responses ranged from "only minor changes would be needed" to "significant changes would be required." Members also noted that changing only the nomenclature from that used in Directive 15 would have little effect on cost. However, adding an "Other race" or a multiracial category would be both disruptive and costly. Members said the changes would affect Computer-Assisted Telephone Interviewing software, forms, electronic reporting systems, and resulting databases. The cost would be associated with disaggregating the multiracial category into meaningful groups for enforcement purposes and comparability with a large volume of historical data. Survey processing costs would increase due to the additional editing, coding, and keying of the expanded matrices, and due to the need to redesign the processing systems to account for the additional data. (Tucker, COPAFS, 1996) One participant said the modifications would be handled as part of the massive transition from the 1990 Census to the 2000 census, describing the overall process that occurs once each decade as an arduous one that could be made more complicated by changes to the racial and ethnic categories.

In a meeting with data users from State and local organizations, participants appeared not too concerned about adapting to change. Unless no changes are made to the decennial census, participants noted, they have to rewrite their data analysis programs every ten years--in any event--to conform to the new formats. Participants believed that costs would not be affected to any great extent. (Tucker, APDU, 1996). Most participants ultimately favored an option that would allow for multiple responses to the race question. While recognizing that it would require more work for analysts and data providers, they believed it to be the fairest alternative given our Nation's diverse population. They thought it could be a viable solution, but also expressed interest in having the Federal Government develop rules for tabulating multiple race responses. (Tucker, APDU group, 1996)

Several agencies offered dollar estimates for what it would cost to implement a change in Federal standards that provided for the reporting of more than one race. These ranged from the tens of thousands into the millions of dollars, depending on the approach that might be selected and whether and the extent to which updating of records might be required.

CHAPTER 4. A Combined Race and Hispanic Origin Question

4.1 Background

This chapter addresses the issue of whether there should be a combined race/Hispanic origin question or whether there should be a separate race question and a separate Hispanic origin question. Included in this chapter is a summary of findings from research recently conducted the Bureau of Labor Statistics and by the Bureau of the Census on the effects of using a combined format instead of separate questions. The chapter also presents findings from other relevant research that address the issues associated with a combined format versus separate questions. These issues include concerns about data quality that arise when a separate race question and an Hispanic origin question are used, and approaches that have been tested to address these data quality concerns.

Directive No. 15 calls for collection of information on persons of Spanish origin or culture. This information can be collected using two different formats--either a combined race and Hispanic origin question or two separate questions, one for race and one for Hispanic origin. Both approaches are popular among Federal agencies. The Directive also allows Federal agencies to collect data on race and Hispanic origin using separate questions and then to present the data in the combined format.

Even within the same agency, both formats sometimes are used. For example, almost six out of every ten (56 out of 97) data systems listed in the Directory of Minority Health and Human Services Data Resources which collect information on Hispanic origin do so using the separate format (Department of Health and Human Services, 1995). Slightly more than half (8 out of 15) of the principal data collections at the Department of Justice use the combined format. At the Bureau of Labor Statistics in the Department of Labor, some of the surveys use a combined format while others use two separate questions. The Office for Civil Rights in the Department of Education, the Office of Civil Rights in the Department of Health and Human Services, the Equal Employment Opportunity Commission (EEOC), and the Office of Personnel Management (OPM) use the combined format. In its National Health Interview Survey, the National Center for Health Statistics (NCHS) uses two questions for race (check one or more groups, followed by selection of the group which best represents the person's race), and one question for Hispanic origin. The combined format tends to be preferred for data collections using observer identification.

Briefly, according to the Directive, if data on race and ethnicity are collected using two separate questions, the racial categories are:

--American Indian or Alaskan Native
--Asian or Pacific Islander
--Black
--White

And, for ethnicity:

--Hispanic origin
--Not of Hispanic origin

If the combined format is used, the categories are:

-- American Indian or Alaskan Native
--Asian or Pacific Islander
--Black, not of Hispanic origin
--Hispanic
--White, not of Hispanic origin

The separate questions are designed to provide Hispanic origin information for all persons. The combined format does not allow for collection of Hispanic origin data if a person reports in the American Indian or Alaskan Native category, or in the Asian or Pacific Islander category. When a combined question is used, data on the race of Hispanics is not collected (see OMB Federal Register Notice, June 1994).

4.2 Concepts of Race and Ethnicity

The decennial census categories used to classify data on "race" and "ethnicity" have changed depending on what were considered the population groups of interest. In the 20th century, data on race and ethnicity have sometimes been coded together and at other times have been coded separately. Census researchers Bates, de la Puente, DeMaio, and Martin (1994) have characterized as "official ambivalence" the Federal uncertainty "about whether Spanish-speaking groups should be considered a separate race, or not." For example, the census classified Mexicans as a "race" in 1930, "White" during 1940-1970, and "of any race" they chose in 1980 and 1990. In 1940, persons of Spanish mother tongue were reported. In 1950 and 1960, persons of Spanish surname were recorded. By 1960, all Mexicans, Puerto Ricans and other persons of "Latin descent" were counted as "White" unless they were "definitely Negro, Indian, or some other race (as determined by observation)." In 1970, a separate question on Hispanic origin was added to the census long form (sent to one-sixth of households). In 1980 and 1990, a separate question on Hispanic origin was asked of all households.

Directive No. 15 defines "race" and "ethnicity" as separate concepts. Harry Scarr, then Acting Director of the U.S. Bureau of the Census, stated in his testimony to the Congressional Subcommittee on Census, Statistics and Postal Personnel in April 1993, that although the Bureau treated race and ethnicity as two separate concepts, the "Bureau recognizes that the concepts are not mutually exclusive..." (Scarr, 1994:7). Dr. Scarr's observation has been well documented in the research literature.

Opinion researchers report that respondents in general--not only Hispanics--find questions about "race" and "ethnicity" to be among the most difficult to answer. Tom Smith of the National Opinion Research Center concludes, "Of all basic background variables, ethnicity is probably the most difficult to measure" (Smith, 1983). Although respondents may give different answers to questions about each concept, researchers have observed that respondents do not understand conceptual differences among terms such as "race," "ethnicity," and others such as "ancestry" or "national origin." For example, NCHS reports that interviewers for one of their surveys found that "...the phrase 'origin or descent' was poorly understood by many respondents." (Drury, 1980). Researchers at the Bureau of the Census remark that notions of "race," "ethnicity," and "ancestry" are not clearly distinguished from one another by census respondents and some persons perceive the race, Hispanic origin, and ancestry questions as asking for the same information.

The terms "race" and "ethnicity" are frequently used interchangeably in the United States. For most daily and practical applications, Hispanics are considered a race. Definitions of race and ethnicity in major dictionaries often have considerable overlap. Crews and Bindon (1991) suggest that race is a sociological construct that is poorly correlated with any measurable biological or cultural phenomenon other than the amount of melanin in an individual's skin. Ethnicity, they suggest, is a sociocultural construct that is often, if not always, coextensive with discernible features of a group of individuals. Crews and Bindon cite several human biologists who have advocated vigorously for use of the term "ethnic group" instead of "race" to question hypotheses about the genetic and cultural constituency of groups.

This fluid demarcation between the concepts of "race" and "ethnicity" and the notion that these concepts are a sociocultural construct observed among the general population is also applicable to the Hispanic population. In fact, researchers such as Clara Rodriguez (1992) have noted that this view of race and ethnicity is consistent with the views of many Hispanics. Numerous other researchers have concluded that the racially diverse Hispanic population regards their "Hispanic" identity as a "racial" one.

This view of race and ethnicity among Hispanics has its origins in Latin American culture. For example, Rodriguez (1994) observes that in Latin America, there are a greater number of racial terms for "intermediate" categories. In contrast, the emphasis in the United States has been on constructing "pure" races (e.g., Black and White, and not biracial or multiracial terms). Conceptions of race in Latin America result in the use of more categories since they are based more on ethnicity, national origin, and culture than appearance. Recent studies have found that Hispanics tend to see race as a continuum and use cultural frames of reference when discussing race (e.g., see Bracken and de Bango, 1992; Romero, 1992; Rodriguez and Hagan, 1991).

Unlike the United States where racial formation has evolved from the acceptance and legitimization of the "one-drop" rule, if a person looked "White" in Latin America, then this is what they were, regardless of what their ancestors may have looked like or how much blood of a particular non-White group they may have. Race in the Caribbean and Latin America is often viewed as an individual marker, while in the United States it determines one's reference group (Wright, 1994). Latin American countries tend to have a more social view of race as compared with the genealogically based view in the United States. This more social view of race tends to include other physical and social characteristics besides color (e.g., education, social class, and context), and may lead to overlapping categories and different racial taxonomies (Rodriguez and Cordero-Guzman, 1992; Harris et al., 1993).

4.3 Self-Identification

Studies indicate differences between the racial and ethnic classification assessed by self-identification and (1) proxy identification by other household members, family, or friends, (2) identification by research or survey interviewers, and (3) identification by the personnel of institutions such as funeral homes. Several studies concentrate on the identification of Hispanic origin, while others focus more broadly on the identification of racial and/or ethnic groups, including Hispanics. Substantial differences have been found between how Hispanics identify themselves and how they are identified by interviewers (Rodriguez and Cordero-Guzman, 1992; Falcon, 1994; Tumin and Feldman, 1961; Rodriguez, 1974; Ginorio, 1979; Ginorio and Berry, 1972; Martinez, 1988).

Hahn, Truman, and Barker (1996) examined the consistency of self-perceived identification at first interview and proxy-reported ancestry at a follow-up interview (an average of 10 years later) in the U.S. population. Ten percent of household proxies did not know the backgrounds of sample persons. Proxy reports of ancestry were consistent with self-classification for 55 percent of sample persons. Consistent classification between proxy and sample person was highest for sample persons classifying themselves as Mexican (98 percent); for other Hispanic groups, consistency was 70 percent . Overall, consistency between self- and proxy-identification was high for several European populations, for Asians, and for Hispanics, but low for American Indians.

In another study comparing self- and interviewer-identification (Drury, Moy, and Poe 1980), researchers compared respondents' self-identified ancestry, including Hispanic categories as well as races, with classification at the same time by an observer (as White, Black, or other). Among self-identified Hispanic groups, between 86 percent and 100 percent were identified by interviewers as White, the remainder as Black or other. A more recent study of the U.S. population (Hahn, Truman, and Barker 1996) compared respondents' self-identified ancestry with race as determined by the interviewer. Among respondents who self-identified as Mexican, 95 percent were classified as White, 5 percent as other; among respondents who self-identified as members of other Hispanic populations, 84 percent were classified as White, 15 percent as Negro. Overall, studies consistently indicate that interviewers are effective in identifying Whites and Blacks, moderately effective in identifying the members of Hispanic groups, and poor in identifying Asians and American Indians.

Other studies have focused on identification by personnel of institutions such as funeral homes. Hahn, Mulinare, Teutsch (1992) compared the race and ethnicity on the birth and death certificates of all U.S. infants born from 1983 through 1985 who died within a year. Among infants designated as Hispanic at birth, 20 percent of Mexicans, 48 percent of Puerto Ricans, and 67 percent of Cubans were likely to have another designation at death; for all Hispanic infants who had different designations on birth and death certificates, more than half were classified as non-Hispanic (White or Black) on death certificates. Observer identification may result in underestimation of mortality for some racial and ethnic groups. For example, when data on Hispanic origin from the birth certificate was used instead of the death certificate, estimates of Hispanic infant mortality were 8.9 percent higher than those based on the death certificate (Hahn 1992).

Similar discrepancies have been reported for U.S. adults. Poe et. al., (1993) found that Hispanics were misclassified as non-Hispanic on 19 percent of death certificates. Other studies have also found significant misclassification of Hispanics (Sorlie 1993; Lindan 1990; Massey 1980).

4.4 Some Alternative Formats for Questions

Several alternative formats for questions to collect data on Hispanic origin have been suggested in public comments. Directive No. 15 currently allows two formats for questions on race and ethnicity: a combined format option (referred to as Alternative 1 for the discussion in this section), and two separate questions (Alternatives 2 and 3). Hispanic can be chosen independently of race only when it is a separate question. Alternative 1: Combined format (allowed under Directive No. 15)

--American Indian or Alaskan Native
--Asian or Pacific Islander
--Black, Not of Hispanic Origin
--Hispanic
--White, Not of Hispanic Origin

Alternative 2: Two separate questions with race question first (allowed under Directive No. 15)

--American Indian or Alaskan Native
--Asian or Pacific Islander
--Black
--White
--Hispanic origin
--Not of Hispanic origin

Alternative 3: Two separate questions with Hispanic origin question first (Allowed under Directive No. 15)

--Hispanic origin
--Not of Hispanic origin
--American Indian or Alaskan Native
--Asian or Pacific Islander
--Black
--White

The following two formats are commonly used outside the Federal Government: Alternative 4:

--American Indian or Alaska Native
--Asian or Pacific Islander
--Black
--Hispanic
--White

Alternative 5:

--Non-Hispanic American Indian or Alaska Native
--Non-Hispanic Asian or Pacific Islander
--Non-Hispanic Black
--Hispanic
--Non-Hispanic White

Variations of these have also been suggested in public comments. For example, some suggested that a "multiracial" category could be followed by a list of categories to select, or a line could be provided to specify the categories. Another alternative which was tested in the Race and Ethnic Targeted Test combined the concepts of race, ethnicity, and ancestry in a two-part single question.

4.5 Research on Data Quality

This section summarizes research that has examined the quality of data on race and Hispanic origin obtained through a separate question for race and a separate question for Hispanic origin. The major data quality measures examined by this research include the reporting of "other race" by Hispanics (section 4.5.1), item nonresponse for race (section 4.5.2), item nonresponse for Hispanic origin (section 4.5.3), and inconsistent reporting in both the race and Hispanic-origin items (section 4.5.4). The chapter then turns to measures that have been proposed and tested for addressing the data quality concerns just cited (section 4.5.6).

4.5.1 Reporting in the "Other Race" Category by Hispanics

Evaluations of the results from the 1980 Census, the 1980 Current Population Survey, the 1990 Census, the 1990 Panel Study of Income Dynamics, and the 1991 Current Population Survey have shown that approximately 40 percent of Hispanics select the "Other Race" category (Denton and Massey, 1989; Tienda and Ortiz 1986; Rodriguez 1992). Research also shows that the use of the "Other Race" category varies by Hispanic subgroup and geography (Rodriguez, 1989; Tucker et al., 1996). Almost all (98 percent) of respondents who classified themselves as "Other Race" in the 1990 Census were Hispanic (U.S. General Accounting Office, 1993:26). This has raised concern among researchers that Hispanics do not identify with the racial categories usually offered. Reporting in the "Other Race" category by Hispanics occurs because, as noted earlier, some Hispanics do not identify with the major race groups. For this reason these members of the Hispanic population report in the "Other Race" category and many register their Hispanic origin in the "Other Race" write-in line when available. (For example, see Kissam et al., 1993). In the 1996 National Content Survey, between 25 percent and 43 percent of Hispanics reported in the "Other Race" category depending on whether the Hispanic origin question was placed before or after the race question (Harrison et al., 1996).

4.5.2 Item Nonresponse in the Race Question

Relatively high item nonresponse to the race question among Hispanics is another reporting issue associated with the use of a separate question to collect information on Hispanic origin and race. The item nonresponse to the race question varies depending on the mode of data collection. In self-administered surveys such as the 1996 National Content Survey (NCS), the item nonresponse rate for the race question is much higher than in interviewer-administered surveys. For example, in the NCS, the item nonresponse rate for the race question ranged from 1.1 percent to 2.2 percent for non-Hispanics, and from 31 percent to 36.5 percent for Hispanics. (Harrison et al., 1996). In interviewer-administered surveys, item nonresponse to the race question is much lower. For example, item nonresponse for the race question in the 1994 National Health Interview Survey was 0.4 percent, and on the Current Population Survey, less than one tenth of one percent of Hispanics were missing information on race.

4.5.3 Item Nonresponse in the Hispanic Origin Question

The General Accounting Office concluded that "the results from the 1990 census showed that the Hispanic origin item continues to pose one of the more significant data quality challenges for the Bureau in terms of allocation rate" (GAO, 1993:24). The Hispanic origin question had the highest nonresponse rate of any question of the 1980 and 1990 censuses, suggesting that some people regarded the question as not applicable, redundant, or unclear. Information was missing from 10 percent of the 1990 census short forms (McKenney, 1992). For the more detailed sample questionnaires, the allocation rate for nonresponse was 3.5 percent . Non-Hispanic respondents contributed substantially to the high nonresponse rate for the Hispanic origin item. The 1990 Content Reinterview Survey found that 94 percent of non-respondents to the Hispanic origin item were non-Hispanic.

In the Census Bureau's 1996 National Content Survey, item nonresponse to the Hispanic origin question ranged from 5.2 percent to 8.6 percent depending on whether the Hispanic origin question was placed before or after the race question (Harrison et al., 1996).

Item nonresponse to the Hispanic origin item is considerably lower in interviewer-administered surveys than in self-administered surveys. For example, the item nonresponse rate from the Current Population Survey for the Hispanic origin variable was 0.6 percent for the first 6 months of 1995. In the 1994 National Health Interview Survey, Hispanic origin was missing for 1.2 percent of sample persons. On the other hand, some data systems that collect information based on observer-identification have considerably higher nonresponse for the Hispanic origin data items. Examples include 15 percent for the National Hospital Ambulatory Medical Care Survey, 30 percent for the National Home and Hospice Care Survey, and 75 percent for the National Hospital Discharge Survey, all conducted by the National Center for Health Statistics. (DHHS, 1995).

4.5.4 Reporting Inconsistency

The General Accounting Office concluded that "the Content Reinterview Survey for the 1990 Census showed generally good response consistency for both the race and Hispanic origin questions" (GAO, 1993, p. 22). However, of those who said they were "Other Hispanic," only 64 percent answered similarly in the reinterview study. In the race question, only 36 percent of those who said on the Census form that they were of "Other Race" reported similarly when reinterviewed. Those reporting as American Indians also were more likely to change their response. Reporting race generally was less consistent for multiple-race persons, Hispanics, foreign-born persons, and persons who did not read or speak English well (OMB Federal Register Notice, 1995: 44675).

The 1996 National Content Survey compared responses from mailback survey forms to the responses provided in the telephone reinterview (Harrison et al., 1996). Approximately 3 percent of Hispanics reported inconsistently on the mailback survey forms and telephone reinterview when two separate questions on race and ethnicity were used. Using an Hispanic origin question first with no multiracial category, 2.9 percent of Hispanics reported inconsistently. Inconsistency was not reduced for Hispanics when the order of the questions on race and Hispanic origin was changed (2.9 percent). Among Hispanics, inconsistency was highest (3.8 percent) when Hispanic origin was asked first and the race question included a multiracial category. Use of a multiracial category in the 1996 National Content Survey did not have a statistically significant effect on the consistency with which persons reported Hispanic origin (Harrison et al., 1996).

Information on reporting consistency is also available from other surveys. For example, Hahn, Truman and Barker (1996) found that 58 percent of respondents to the first National Health and Nutrition Examination Survey and subsequent Epidemiologic Follow-up Study were consistent in self-classification over the follow-up period. In another study Johnson et al. (1995:15) found that 40 percent of mixed-race and Hispanic respondents changed the way they reported their racial and ethnic background depending on the context, social situation, options on application forms or "perceived advantages in applying for scholarships, loans, school admissions, housing and employment." Changes in self-awareness and identification were also responsible for changes in reported identity. Hispanics with two Hispanic parents were much less likely (12.5 percent) to have ever identified themselves differently.

4.6 Measures to Correct Misreporting in the Race Question and the Hispanic Origin Question

The reporting issues just described -- reporting in the "Other race" category, item nonresponse to the race question, item nonresponse to the Hispanic origin question, and inconsistency of reporting -- result from having a separate race and a separate Hispanic origin question. Two important measures have been used and tested to address these reporting concerns while keeping two separate questions: placement of the Hispanic origin question before the race question, and providing respondents with written instructions to respond to both the race question and the Hispanic origin question.

Bates, de la Puente, Martin and DeMaio (1994) analyzed and summarized multiple replications of five major Census Bureau studies on decennial census race and Hispanic origin questions to determine the effects of question order and instructions on reporting in the race question and the Hispanic origin question. Based on this analysis and on qualitative information obtained through focus groups and in-depth personal interviews, the authors conclude that the evidence consistently shows that placement of the Hispanic origin question before the race question provides a more restrictive frame of reference for race reporting and thus respondents (mostly Hispanics) are less likely to report in the "Other Race" category and more likely to select one of the major race groups listed in the race question. Further, restricting the frame of reference for race reporting also results in reductions in item nonresponse to the race question. Although these measures substantially reduced reporting in the "Other Race" category, reduced item nonresponse for the race question among Hispanics, and reduced item nonresponse to the Hispanic origin questions by non-Hispanics, these measures did not entirely eliminate the reporting problems.

For example, in the National Content Survey, "Other Race" reporting by Hispanics went from 40 percent when the race question was placed before the Hispanic origin question down to 20 percent when the Hispanic origin question was placed before the race question. The comparable percentages in the Appeals and Long Form Experiment were 53 percent when the race question was placed before the Hispanic origin question and 26 percent when the Hispanic origin question was placed before the race question. The declines in "Other Race" reporting by Hispanics in the other three Census Bureau studies were more modest. (Bates et al., 1994).

Bates, de la Puente, Martin, and DeMaio (1994) report that the inclusion of instructions to aid reporting had positive effects. For example, the Alternative Questionnaire Experiment (AQE) used a two-question format to gather data on race and Hispanic origin, and included an instruction in some panels that read "Fill in the NO circle if not Spanish/Hispanic" next to the question text on Hispanic origin. Results from the AQE demonstrate that adding this instruction alone reduced nonresponse to the Hispanic origin question from 19 percent to 8 percent. Combining the instruction with asking the ethnicity question prior to race resulted in a nonresponse rate of 5 percent. These findings suggest that instructions can help reduce, but not eliminate, nonresponse to the Hispanic origin question.

Bates, de la Puente, Martin and DeMaio (1994) also conducted multivariate analyses to improve understanding of the effects of question order and instructions on race reporting by Hispanics. Four variables hypothesized to affect race reporting by Hispanics were included in the analyses: place of birth (native or foreign-born), recency of arrival in the United States, educational level, and English proficiency. The results from the multivariate analyses are mixed. The authors concluded that the effect of question ordering on the reporting of race among Hispanics does not seem to be influenced by time in the United States, education, or knowledge of English. The authors added that data from at least two of the five Census Bureau studies considered indicated that Hispanic response to the race question may be conditioned by recency of arrival in the United States (Bates et al., 1994).

Unlike the Census Bureau tests examined in the Bates, de la Puente, Martin and DeMaio (1994) study, the 1996 National Content Survey also examined the effects of sequencing on the reporting of race and Hispanic origin using race questions that provided a "multiracial" category as one of the response options. Findings from this test are in line with the results reported by Bates et al. (1994).

In the 1996 National Content Survey panels where the race question did not include a multiracial category as a response option, "Other Race" reporting by Hispanics significantly declined from about 43 percent when the Hispanic origin question was placed after the race question to approximately 25 percent when the Hispanic origin question was placed before the race question. "Other Race" reporting also declined among Hispanics when the Hispanic origin question was placed before the race question that included a multiracial category as a response option, but the decline was not statistically significant. In panels where the race question included a multiracial response option, reporting of "Other Race" by Hispanics declined from about 33 percent when the Hispanic origin question was placed after the race question to about 25 percent when the Hispanic origin question was placed before the race question (Harrison et al., 1996). It is important to note that these declines in "other race" reporting were reduced, but not eliminated, by reversing the order of the Hispanic origin and race questions.

Placing the Hispanic origin question before the race question in the 1996 National Content Survey reduced item nonresponse rates for the race question among Hispanics, but these reductions were not statistically significant and item nonresponse rates for the race question remained relatively high (Harrison et al., 1996).

The sequencing of the Hispanic origin question and the race question was also one of the major research objectives of the Race and Ethnic Targeted Test (RAETT). The findings from the RAETT on this issue echo those of studies just discussed. In the Hispanic targeted sample, asking the Hispanic origin question before the race question reduced item nonresponse to the Hispanic origin question from about 10 percent to about 7 percent. Placing the Hispanic origin question before the race question had no effect on the item nonresponse rate for the race question in the Hispanic targeted sample.

In the RAETT, reductions in the reporting as "Other Race" and "Multiracial" and an increase in the reporting as "White" in the Hispanic targeted sample were detected when the Hispanic origin question was asked before the race question. More specifically, in the Hispanic targeted sample in Panel D (race question first), about 56 percent of respondents reported as White, about 25 percent reported as "Other Race", and about 3 percent reported as "Multiracial." In contrast, when the Hispanic origin question was placed before the race question (Panel B), approximately 67 percent reported as White, 16 percent reported as "Other Race", and 2 percent reported as "Multiracial."

4.7 The Effects of Combining the Race Question and
the Hispanic Origin Question into a Single Question

A combined question on race and Hispanic origin was tested in the 1995 CPS Supplement and in the RAETT.

4.7.1 Results From the May 1995 CPS Supplement on Race and Ethnic Origin

Having a separate versus combined race and ethnicity question appears to have a significant effect on the percentage of persons who identify as Hispanic. In the May 1995 Current Population Survey (CPS) Supplement, significantly more people identified as Hispanic when they were asked a separate question on Hispanic origin than when Hispanic origin was combined with the race question (See Table 4.1). (Because an interviewer collects the data, either in person or by telephone, multiple responses are much less likely to occur.) In particular, 10.6 percent of the respondents who received a separate question (panels 1 and 2 combined from Table 4.1) identified as Hispanic compared with 8.1 percent of the respondents who were given the combined race and ethnic origin question (panels 3 and 4 combined from Table 4.1), (Tucker et al., 1996).

Note: Detail may not add to totals due to rounding. From Table 3, Tucker, et al. (1996).

Additionally, it is important to note that some specific Hispanic subgroups may respond differently than others to separate race and ethnicity questions versus a combined race and ethnicity question (See Table 4.2). In particular, the proportions of respondents who report Mexican, Cuban, and "Other Hispanic" national origins differed significantly depending on the type of race and ethnicity question. Specifically, the respondents who identify as Hispanic in a combined race and ethnicity question (as in panels 3 and 4 combined from Table 4.2) are composed of a greater percentage of people with Mexican national origin (66 percent) than the respondents who identify as Hispanic in a separate ethnicity question (about 60 percent in panels 1 and 2 combined from Table 4.2). In contrast, the respondents who identify as Hispanic in a separate question are composed of a greater percentage of people with Cuban and "Other Hispanic" national origins (about 4 percent Cuban and 13 percent "Other Hispanic" in panels 1 and 2 combined from Table 4.2) than the respondents who identified as Hispanic from the combined race and ethnicity question (about 2 percent Cuban and 9 percent "Other Hispanic" in panels 3 and 4 combined from Table 4.2). In other words, Hispanics of different national origins differ in how likely they are to identify themselves as Hispanic depending upon whether they are asked a separate Hispanic question or a combined race and Hispanic origin question (Tucker et al., 1996).

Note: Detail may not add to totals due to rounding. From Table 11, Tucker, et al. (1996).

In the May 1995 CPS supplement, analyses of the effect of a separate versus combined race and ethnicity question showed that there were no significant differences in the percentage of people identifying as Black, Asian or Pacific Islander, or American Indian (See Table 4.3). However, the number of American Indians in the sample was too small for drawing reliable conclusions for that population. The percentage of people identifying as White was influenced by whether there was a separate Hispanic question or not, with 75.22 percent (panels 3 and 4 combined from Table 4.3) of the respondents identifying as White when Hispanic was included in the list of races compared with 79.81 percent who identified as White when Hispanic origin was a separate question (panels 1 and 2 combined from Table 4.3). Thus, including Hispanic as a category in the race question will likely lower the proportion of people currently identifying as White only and the proportion of persons classified as "Other." These findings were also reflected in the analysis of the differences in respondent reporting between the CPS race question and the May 1995 CPS Supplement race questions (see Tucker et al., 1996).

Note: Adapted From Table 5, Tucker et al., (1996).

By using respondents' Hispanic national origin from the CPS and examining their racial identification in the May 1995 CPS Supplement, further insights are gained into how subgroups of Hispanics identify depending upon whether they are asked separate race and ethnicity questions or a combined race and ethnicity question (See Table 4.4). As can be seen in Table 4.4, a sizeable percentage of respondents with Hispanic national origins do not identify as Hispanic in a combined race and ethnicity question (panels 3 and 4). Specifically, 11 percent of respondents with a Mexican national origin identified as White when having to choose between White and Hispanic in the combined race and ethnicity question. Similarly, 23 percent of respondents with other Hispanic national origins identified as White when there was a combined race and ethnicity question and a majority of respondents of Cuban origin identified as White even though the Hispanic category was offered in the combined question (Tucker et al., 1996). This pattern of racial identification for Mexican-origin and Cuban-origin respondents is consistent with the findings of the 1990 Panel Study of Income Dynamics conducted by the Institute for Survey Research at the University of Michigan. For Hispanics reporting a single race when given a list of racial categories that included "Latino," 88 percent of Cubans reported as White and 9 percent as Latino, compared with Mexicans, 56 percent of whom reported as White and 35 percent of whom reported as Latino (Duncan et al., 1992). Bates, et al. (1996) found that Cubans, compared with other Hispanic groups, were most likely to report their race as White when the race question followed a question on Hispanic origin.

4.7.2 Results From the Race and Ethnic Targeted Test

Two versions of a combined race, Hispanic origin, and ancestry question were tested in the RAETT. Both versions provided check boxes for "White," for "Black, African Am., or Negro," for "Indian (Amer.) or Alaska Native" (with a write-in line for tribal affiliation), for "Asian or Pacific Islander," for "Hispanic" and for "Some other race." One version (Panel E) also included the category "Multiracial or biracial." A second version (Panel F) did not contain a multiracial category but rather instructed respondents to "Mark one or more boxes to indicate what this person considers himself/herself to be." Both versions, E and F, were followed by a question which asked respondents to write in their "ancestry or ethnic group" in the space provided.

Panels E and F were compared with the corresponding panels that contained a separate race question and a separate Hispanic origin question. These were Panel B (containing a multiracial category like Panel E) and Panel C (containing a multiple response option like Panel F). The major findings from these panel comparisons are presented below.

4.7.2.1. Reporting of Hispanic Origin

A combined race and Hispanic origin question must, of necessity, produce fewer Hispanic only responses or fewer responses in at least one of the major race groups, than a separate race question and a separate Hispanic origin question. If all individuals who select the Hispanic category alone or in combination with another race group are tabulated as Hispanic (termed "all-inclusive Hispanic"), such a tabulation could provide similar information to that which would be obtained if separate questions on race and Hispanic origin were used.

The RAETT found no statistically significant differences between the "all-inclusive Hispanic" tabulation for the combined question on panels E and F and the appropriate panels containing a separate Hispanic origin question and a separate race question. Specifically, panels B and E, which both contained a multiracial category, and panels C and F, which both contained the instruction to "mark one or more," all had responses ranging from 74 percent to 76 percent. However, if one were to tabulate as Hispanic those who selected only the Hispanic category, then a much lower percent (about 57 percent) of responses would be Hispanic in panels E and F.

Table 4.5 shows that the percentages reporting the specific Hispanic origins Mexican, Puerto Rican, Cuban and Other were quite different on panels E and F than on panels A, B, and C. This is most likely an artifact of the way the data were collected and tabulated. In panels A, B, and C, respondents were asked to check boxes with the labels shown in Table 4.5. In panels E and F, respondents were asked in a separate question to write in their ancestry or ethnic group. These write-in groups were tabulated (for those who marked only the Hispanic category) and are shown in table 4.5. Those who consider themselves both Hispanic and something else are not included in counts shown for the specific Hispanic origins for panels E and F; they are included only in "Hispanic (only or in combination)." In addition, if Hispanic only respondents wrote in two different Hispanic origins they are counted in "other Hispanic" in Panels E and F. In panels A, B, and C, the instructions appeared to ask Hispanic respondents to select one Hispanic origin category, although some may have marked multiple categories. A tabulation using the "historic series" approach or the "all-inclusive" approach would shed additional light on this issue.

Note: Adapted From Table 1-9 in Bureau of the Census (1997)

4.7.2.2 Reporting of Multiple Races

The combined race, Hispanic origin, and ancestry question (Panels E and F) elicited relatively high percentages of multiple responses in the Hispanic targeted sample. Table 4.6 shows that in Panel E, where a multiracial category was provided and respondents were instructed to mark one box, 18 percent of respondents in the Hispanic targeted sample selected more than one category. In Panel F, where there was no multiracial category and respondents were instructed to "Mark one or more boxes," 19 percent of respondents in the Hispanic targeted sample selected more than one category.

The relatively high rates of multiple responses in the Hispanic targeted sample on Panel E suggests that substantial percentages of Hispanics wish to report a race as well as their Hispanic origin, and will check more than one category even when they encounter a question that instructs them to choose one or the other. Additional support for this conclusion can be found in the fact that more than 92 percent of multiple responses in Panels E and F in the Hispanic targeted sample marked the Hispanic box or provided Hispanic write-in entries.

(NA) Not available. Note: Adapted From Tables 1-1R, 1-2R, 1-3R, 1-4R, 1-5R, 2-1R, 2-2R, 2-3R, 2-4R, 2-5R, 8-1R, 8-2R, 8-3R, 8-4R, 8-5R in Bureau of the Census (1997)

4.7.2.3 Summary of Findings

Inherently, a combined race and Hispanic origin question will result in lower reporting in the Hispanic origin category alone, or in one of the major race groups alone than separate race and Hispanic origin questions where race and Hispanic origin are independent. The RAETT found patterns of declines in reporting as Hispanic alone, as White alone, and as Asian and Pacific Islander alone in the combined questions. This suggests that there are respondents who will report as Hispanic and as White or as Asian and Pacific Islander when they encounter separate questions on race and Hispanic origin. However, when faced with a combined question, some of these respondents will report as Hispanic, some will report as White or as Asian and Pacific Islander, and some will mark more than one of these categories, even when the option of doing so is not offered. In contrast, the absence of significant changes in reporting as Black or as American Indian in the respective RAETT targeted samples for those populations suggests that the numbers of respondents in the Black and American Indian targeted samples who report as Hispanic when separate Hispanic origin and race questions are offered are relatively small or that they are more likely to report their race rather than their Hispanic origin in a combined question.

When Hispanic is offered as an option in the combined question, a number of Hispanic respondents will select both Hispanic and a race, even when instructed not to do so.

4.8 Public Sentiment

The Hispanic origin ethnicity category was included in Directive No. 15 to meet the requirements of Public Law 94-311, which called for improving data on persons of Spanish culture or origin. During discussions of the content of the 1990 Census, the Interagency Working Group on Race and Ethnicity concluded that a combined race and Hispanic origin question would not meet program needs and could result in an undercount of the Hispanic origin population (Bureau of the Census, 1988).

During 1994, several national Hispanic organizations supported the incorporation of the term "Hispanic" into a combined "Race/Ethnicity" question (Kamasaki, 1994; Olguin, 1994; Blackburn-Moreno, 1994). Both the National Council of La Raza (NCLR) and ASPIRA Assoc. Inc. argued that additional research should be conducted before any change is made. The Mexican American Legal Defense and Education Fund (MALDEF) saw the lack of a uniform definition of Hispanic throughout the Federal Government and differential undercounts of Hispanics as more important problems (Carbo, 1994). A few public comment letters sent in response to OMB's August 28, 1995, Federal Register Notice showed some support for Hispanic as a racial category, but none of these letters of support were from an Hispanic surnamed individual or Hispanic organization.

In a book chapter published in January 1997, the NCLR president, Raul Yzaguirre stated that he does not support the inclusion of Hispanic origin as a racial category, but does support further testing of both the Hispanic origin and race questions. He also added that: "Before large-scale changes are made, however, it is critical that the Census Bureau and the Office of Management and Budget determine which version of the questions accommodates the largest number of respondents and provides the most accurate data." (Yzaguirre 1997: 89).

The majority of Hispanics in the May 1995 CPS Supplement preferred the combined question. It has been argued that they did not know the impact of combining the questions on the population count of Hispanics (Torres, 1996:4). This concern appears to be based on the comparison of the percent reporting Hispanic using the separate question format with the percent reporting Hispanic only using the combined question with a multiple response option. As described in Section 4.7.2.1, approximately the same percent report as Hispanic when data are tabulated in the all inclusive Hispanic category (the total of those who mark Hispanic either alone or in combination with other categories) in the combined format as report "Hispanic" in a separate question format.

A concern expressed by some is that the use of the combined format may affect aggregate statistics about the Hispanic population since Cubans tend to have higher socioeconomic and health status than other Hispanics. Two examples were therefore calculated. When the results from the May 1995 CPS Supplement are applied to 1994 data on unemployment by Hispanic subgroup, it is estimated that the 1994 unemployment rate for Hispanics would have changed relatively little--from 10.9 percent to 11.2 percent if the combined format (and Hispanic alone category) had been used. The percent of Hispanics with a regular source of primary health care in 1991 did not change in these calculations (61.8 percent using separate questions and 61.4 percent using the combined format).

4.9 Additional Cost Concerns

If OMB were to change the choice Federal agencies currently have to collect Hispanic origin data using either the combined format or two separate questions, there would be a sizeable number of large data systems for which data collection forms, computer programs, interviewers' and coders' manuals, and other related materials would have to be changed. For example, both the separate and combined formats are used within the Department of Health and Human Services, (DHHS, 1995). Fifty-eight percent (56 out of 97) of the DHHS data systems listed in the Directory which do collect Hispanic origin data use the separate format.

The Indian Health Service (IHS) in the Department of Health and Human Services prefers that "Hispanic" be retained as a separate ethnic category. Many American Indians and Alaska Natives are of Hispanic origin and have Spanish surnames, especially in the West and Southwest. They state that if "Hispanic" were to be considered as a racial category (even if there were a "mark all that apply" approach built in), it is probable that the identity of many American Indians and Alaska Natives would be masked by responses to the Hispanic category. If "Hispanic" is retained as an ethnic category, however, Indians will still be able to identify with both backgrounds. Based on findings from the 1990 Census and the May 1995 Current Population Survey supplement, IHS expects that although the reductions in reporting as American Indian, strictly from an alternative that would include Hispanic as a racial category, would be less than from the adoption of a stand-alone multiracial category (or a multiracial category with a follow-up question); the reduction would, nonetheless, be serious.

The Health Care Financing Administration (HCFA) uses the combined format to collect information on race and Hispanic origin for Medicare beneficiaries. If the decision were made to use only two separate questions to collect data on race and ethnicity, HCFA would have to perform a 100% survey of Medicare beneficiaries. To revise HCFA's race/ethnicity categories for future beneficiaries, HCFA would have to negotiate payment to the Social Security Administration to collect this information on Social Security beneficiaries at enrollment. The cost of changing HCFA's data systems to accept new codes if a combined format were to be used would be minimal.

Similarly, the Equal Employment Opportunity Commission (EEOC) currently uses only the combined question format to collect data on race and ethnicity. The instruction booklets for completing all EEOC employment reports have a section on race/ethnic identification which provides guidance on conducting visual surveys and maintaining postemployment records as to the race/ethnic identity of employees. Thus, the costs associated with a requirement to use only the two question format would extend beyond simple computer programming, and the expenses would be greater than the minimal costs that some states have recently encountered when implementing state legislative requirements for a multiracial category.

CHAPTER 5. Other Possible Changes

5.1 Background

This chapter considers suggestions for changes in how data on certain population groups should be classified and for other improvements or clarifications. The issues discussed cover four areas: establishment of new categories for specific population groups, terminology, format, and instructions. The chapter's sections correspond to specific racial and ethnic categories, and all of the issues related to that category or subcategory are discussed together.

It should be noted that while Directive No. 15 uses the term "Alaskan Native," the term used in Federal law and generally preferred is "Alaska Native." For this reason the term appears as "Alaska Native" throughout those sections dealing with this group except where the reference is specifically to the category in Directive No. 15.

5.2 Specific Suggestions

In addition to the proposals discussed in Chapters 3 and 4, the following fifteen suggestions for changes were examined during the current review of Directive No. 15:

Changes related to American Indians and Alaska Natives

• Should the term "American Indian" or "Native American" be used?
• Should the term "Alaska Native" or "Eskimo and Aleut" be used?
• Should a distinction be made between federally recognized and nonfederally recognized tribes?
• What is the best way to elicit tribal affiliation?
  
• Should the definition be changed to include Indians indigenous to Central America and South America?

Changes related to Asians and Pacific Islanders

• Should the "Asian or Pacific Islander" category be split into two categories? If yes, how should this be done?
• Should specific groups be listed under the "Asian or Pacific Islander"category?
• Should the term "Guamanian" or "Chamorro" be used?

Changes related to Hawaiians

• Should the term "Native Hawaiian" or "Hawaiian" be used?
• Should Hawaiians continue to be included in the "Asian or Pacific Islander" category; be reclassified and included in an "American Indian or Alaska Native" category; or be established as a separate, new category?

Other terminology issues

• Should the term "Black" or "African American" be used?
• Should the term "Hispanic" or "Latino" be used?
• Should more than one term be used in either case?

Other new category issues

• Should an Arab or Middle Eastern category be created? If yes, how should it be defined?
• Should a Cape Verdean category be created?

5.3 Evaluation of the Possible Effects of Suggested Changes

5.3.1 Changes related to American Indians and Alaska Natives

The following suggested changes to Directive No. 15 as they relate to American Indians and Alaska Natives are discussed in this section:

• Should the term "American Indian" or "Native American" be used?
• Should the term "Alaska Native" or "Eskimo and Aleut" be used?
• Should a distinction be made between federally recognized and nonfederally recognized tribes?
• What is the best way to elicit tribal affiliation?
• Should the definition be changed to include Indians indigenous to Central America and South America?

Currently, the "American Indian or Alaskan Native" category is used to classify data on "a person having origins in any of the original peoples of North America, and who maintains cultural identification through tribal affiliation or community recognition."

5.3.1.1 Should the term "American Indian" or "Native American" be used?

"American Indian" is the term used in Directive No. 15 to identify the descendants of the indigenous population of North America. The term has generally been used over the past several decades to identify this population group and is recognized by members of this group. In general usage, the term "American Indian" includes individuals who are members of tribes that may or may not be recognized by the Federal Government. Federally recognized tribal governments include only members of their tribe and may use their own Indian name for their tribal name. Further, while Federally recognized tribal governments have their own criteria to determine tribal membership, such membership is not required by Directive No. 15. As a result, the number of individuals enumerated in this category exceeds the number of individuals who hold official membership in recognized tribal governments. Most Federal programs do not require membership in federally recognized tribes for program eligibility. For example, to be eligible for Indian Health Service (IHS) programs, a person need only prove descent from a member of a federally recognized tribe; blood quantum and membership are not relevant. It has also been the practice to classify Canadian Indians in this category.

The term "Native American" has been in use since the 1960s. There are other indigenous groups besides American Indians and Alaska Natives (e.g., Hawaiians) in the United States and areas under U. S. Government jurisdiction. Technically, "Native American" is a term that does not apply exclusively to American Indians and Alaska Natives. Its use may also lead to some confusion in that individuals who are not descended from indigenous populations but who were born in the United States may consider themselves to be "Native Americans" and may select this category erroneously. The May 1995 CPS Supplement on Race and Ethnicity found that more than half of those identifying as American Indian or one of the Alaska Native groups preferred "American Indian" or "Alaska Native" but a third chose "Native American." (Tucker et al., 1996) Public comments from tribal governments to OMB indicated a clear preference for the term "American Indian."

In the RAETT's American Indian targeted sample, American Indians continued to write in a tribal affiliation across all panels, A through H, that used the combined category "Indian (Amer.) or Alaska Native" with the instruction, "Print name of enrolled or principal tribe." On Panels B through H, some respondents used write-in entries such as "Amer. Indian," "American Indian," "American Ind.," and "Indian Amer." to indicate that they are American Indian rather than Alaska Native, but did not provide a specific tribal entry. The percentages ranged from 6.5 percent on Panel H to less than 1 percent on Panel A. There also were write-ins, such as "Amer-Indian-Navajo," in which respondents indicated first that they are American Indian, before writing in the tribal affiliation.

In the RAETT, which drew its American Indian targeted sample from areas in close proximity to reservations, reinterviews were conducted with respondents in households with at least one person who identified as American Indian. This group indicated they preferred the term Native American (52 percent) to American Indian (25 percent). The remaining respondents indicated they had no preference (16 percent), preferred both terms (6 percent), or preferred another term (2 percent).

Measurement. Measurement issues--discussed for each of the options presented in this chapter--relate to self-identification, quality and consistency of data, and implementation.

The use of self-identification allows more people to identify as American Indian than are members of tribes. This includes people who are or who have ancestral ties to American Indians but do not meet tribal enrollment requirements. The term "Native American" attracts persons who were born in the United States as well as persons with American Indian and/or Alaska Native ancestry.

To improve reporting of American Indian tribes in the decennial census, the instruction "Print name of enrolled or principal tribe" was tested and then included in the 1990 census race question. The instruction helped to reduce the rate of nonreporting of tribe from about 20 percent in 1980 to 13 percent in 1990. This improvement occurred in reservation areas but not in off-reservation areas. (1990 CPH-L-99, "American Indian Population by Tribe, for the United States, Regions, Divisions, and States: 1990" and unpublished tables)

The use of self-identification rather than observation by an enumerator provides more complete data on American Indians but with limitations. The consistency of reporting as American Indian is low among persons with both American Indian and White ancestry. In decennial census data collection and tabulation there has been no distinction between federally recognized tribes and nonfederally recognized tribes. The federally recognized tribal governments, as well as the Department of the Interior's Bureau of Indian Affairs, would like the American Indian and Alaska Native definition limited to enrolled tribal members of federally recognized tribes. The Indian Health Service favors a distinction between federally recognized tribes and nonfederally recognized tribes. IHS is only responsible for federally recognized tribes; however, a separate count for nonfederally recognized tribes indicates the potential IHS service population if the tribes were to receive Federal recognition.

Some have suggested using a follow-up question to ask if a person is enrolled in the tribe reported in the race question. An enrollment question has not been included in the decennial census because there are no statutory requirements for tribal enrollment data and because of space constraints on the census questionnaire. Also, tribal governments that responded to the Bureau of the Census Survey of Census Needs of Non-Federal Data Users did not indicate that they needed tribal enrollment data.

The 1980 Census Supplementary Questionnaire for American Indians (Reservations and the Historic Areas of Oklahoma) asked a follow-up question on whether the person was enrolled in the tribe reported. There were a total of 336,280 American Indians on all reservations and 113,280 American Indians in the historic areas of Oklahoma (excluding urbanized areas) reported. For those on reservations, 87 percent were enrolled and 7 percent did not answer the question. For the historic areas of Oklahoma (excluding urbanized areas), 51 percent were enrolled and 11 percent did not answer. To determine whether a tribal enrollment question should be asked in the future, more extensive research will be needed on how to improve the reporting of such enrollment, particularly given the relatively high nonresponse rates in the past.

Data production. Data production issues--discussed for each of the options presented in this chapter--relate to coding, editing, and adjustment needs.

A change in the name of the American Indian category would not change the way American Indians are tabulated and would raise no data production issues. However, the introduction of the term "Native American" could be misinterpreted as meaning "anyone born in the United States," with the result that some respondents would be misclassified. While the instruction asking for "enrolled or principal tribe" might indicate the focus of the category, it might also lead to a large number of write-in answers that would need to be coded.

Analytic. Analytic issues--discussed for each of the options presented in this chapter--relate to comparability over time and aggregation.

On the face of it, a change in the name of a group should not lead to a change in results if the definition of that group is not changed. To the extent that native-born individuals mistakenly check this category and are not identified in the coding or editing procedures, however, it is possible that using the term "Native American" would result in data that are not compatible with historical series.

Cost. While there are no direct costs associated with a change in name, there are important, if unmeasurable, indirect costs related to misclassification and the cascading effect on data analysis.

Legislative or program needs. Any approach collecting accurate data for this category would meet legislative and programs needs for most Federal agencies. The exception is the Bureau of Indian Affairs, which needs data only for federally recognized tribes and their members. Most Federal agencies use special tabulations of American Indians and Alaska Natives as one group, but data are also tabulated by tribe for some users.

5.3.1.2 Should the term "Alaska Native" or "Eskimo and Aleut" be used?

While Directive No. 15 uses "Alaskan Native," the preferred term is "Alaska Native." This is reflected in Public Law 92-203, the Alaska Native Claims Settlement Act (ANSCA) of 1971, and subsequent legislation. The Indian Health Service, the Bureau of Indian Affairs, and the Bureau of the Census prefer and use "Alaska Native."

In the RAETT Alaska Native targeted sample, most Alaska Natives--83 percent on Panel B and 88 percent on Panel D--reported a specific tribe or corporation when the panel used the combined category "Indian (Amer.) or Alaska Native" with the instruction, "Print name of enrolled or principal tribe." The "tribe not reported" rates on these panels were 14 percent and 12 percent, respectively. On Panels B and D, 21 percent and 15 percent of respondents, respectively, wrote in "Alaska Native" by itself. These respondents indicated they were Alaska Native rather than American Indian, but did not provide a specific tribal or corporation affiliation. In addition, on Panels B and D, some respondents reported "Eskimo" (10 percent and 15 percent, respectively) and "Aleut" (2 percent and 1 percent, respectively) without reporting a specific tribal or corporation affiliation.

In the RAETT reinterview for the Alaska Native targeted sample, respondents in households with at least one person who identified as Eskimo or Aleut indicated, by answering "yes" or "no" to each, that their tribal entry was an ethnic group (63 percent), a tribe (55 percent), a land corporation (55 percent), a nation (30 percent), or something else (22 percent). Respondents who said their tribal entry was something else provided examples such as "born in Alaska, indigenous people, Eskimo group, or Eskimos, self government, and people. In reinterview households with at least one person who identified as Eskimo, 88 percent of the respondents indicated that Eskimo was an acceptable term to them. Respondents who said Eskimo was acceptable but who preferred another term to Eskimo provided examples such as Inupiat, Yupik, Alaska Native, and American Indian. In reinterview households with at least one person who identified as Aleut, all respondents indicated that Aleut was an acceptable term to them.

In Alaska, the terms Alaskan Indian, Eskimo, and Aleut were in general use before 1971. Beginning with the passage of ANCSA in 1971, the term Alaska Native came into use and has been used since. Alaska Native includes Alaskan Indians (Athabascans, Tlingits, and Haidas), Eskimos (Inupiat, Yupiks, etc.), and Aleuts (who primarily live on Kodiak Island and in the Aleutian chain) covered by ANCSA. Under ANSCA, Alaska Native does not include children who were born after 1972, but such persons do identify with the term despite the legal distinction. ANCSA established regional and village corporations that have membership requirements. It is also important to distinguish among the tribes that comprise the Alaska Native population. Alaska Native tribal governments and the State of Alaska have stated that they would find census data more useful if tribes were distinguished for Alaska Natives as they are for American Indians. These tribes are just as distinct politically, culturally, and linguistically as are the American Indian tribes in the lower 48 states.

Focus groups and cognitive interviews with Alaska Natives found that Alaska Natives are reporting in the combined category, "American Indian or Alaska Native," and are reporting a tribe. Also, statements indicated that the use of the term "Eskimo" may be offensive to some people. If the combined category is used, the term "Eskimo" as a descriptor would not be used.

Measurement. As in the case of American Indians, the use of self-identification allows more people to identify as Alaska Native than are members of tribes or corporations. However, Directive No. 15 (which uses the term Alaskan Native) makes no reference to ANCSA, with the result that individuals not included in the legal definition only identify themselves as Alaska Native.

Data production. If Alaska Natives are asked to designate an enrolled or principal tribe, there will be data tabulation and production implications for the decennial census. For example, a list of the tribes will have to be developed; a determination will have to be made about which tribes to list in tabulations; and editing and coding routines will have to be refined to correct for multiple spellings or misspellings of tribal names.

Analytic. If Alaska Natives are asked to report their tribal affiliation, it would still be possible to aggregate them into the groups (American Indian, Eskimo, and Aleut) used previously in the decennial census.

Cost. The data production needs discussed above will increase the cost of the decennial census to collect and report results by specific tribe.

Legislative or program needs. Using the term Alaska Native and asking for the enrolled or principal tribe would meet legislative and program needs for most Federal agencies. It would not meet the needs of the Bureau of Indian Affairs to differentiate, at a minimum, between tribes that are or are not recognized by the Federal Government. It also would not allow for an absolute accounting of who is a member of a recognized tribe.

5.3.1.3 Should a distinction be made between federally recognized and nonfederally recognized tribes?

In public comments to OMB, the federally recognized tribal governments would like the American Indian and Alaska Native definition limited to enrolled tribal members. In decennial census data collection and tabulation there has been no distinction between federally recognized tribes and nonfederally recognized tribes. Because self-identification is used in the decennial census, it is not possible to distinguish between those individuals who have formally registered with a specific tribe and those who only claim an ancestral tie. To meet requirements of tribes, according to the Bureau of Indian Affairs and the Indian Health Service, it is preferable that data be collected for both members and nonmembers alike, but that a distinction be made between the two groups.

Measurement. Currently, aside from the decennial census, most data collection follows Directive No. 15 and uses the "American Indian or Alaskan Native" category or a combined American Indian, Eskimo, and Aleut category without asking for any tribal affiliation. The 1980 and 1990 decennial censuses used three separate categories--American Indian, Eskimo, and Aleut. For persons who identified as American Indian, tribal affiliation was asked. The continued use of the category "American Indian or Alaska Native" does not impose an implementation problem for Federal agencies.

Data production. Aside from data collections that ask for enrolled or principal tribe, there are no data production issues. However, when tribal affiliation is asked, many coding and editing issues come into play. These issues are not new and are well known to the agencies for which tribal affiliation is an important factor.

Analytic. To the extent that data production related to coding and editing tribal affiliation identifies and reclassifies respondents who erroneously checked this racial category, no longer asking this information will inflate the number of American Indians.

Cost. There are some costs associated with coding and editing tribal affiliation.

Legislative or program needs. Using the category "American Indian or Alaska Native" and asking for the enrolled or principal tribe would meet legislative and program needs for most Federal agencies, except for the Bureau of Indian Affairs, which needs data on tribal members of federally recognized tribes.

5.3.1.4 What is the best way to elicit tribal affiliation?

American Indians have been asked in most decennial censuses to report their tribal affiliation. In the 1990 census, the instruction, "Print name of enrolled or principal tribe," improved reporting of tribal affiliation.

Given the relatively large number of Alaska Natives who also specify tribal affiliation and the extent of negative reaction to the term "Eskimo," careful consideration needs to be given to its continued use in either the name of the category or as an example. The use of the combined category "American Indian or Alaska Native" and the instruction, "Print name of enrolled or principal tribe," would address both points.

See section 5.3.1.2 above for a discussion of the measurement, data production, analytic, cost, and legislative or program needs issues related to this topic.

5.3.1.5 Should the definition of the "American Indian or Alaskan Native" category be changed to include Indians indigenous to Central America and South America?

Currently, the definition for the "American Indian or Alaskan Native" category does not include Indians indigenous to Central America and South America. In the 1990 census, members of Central American tribes (1,688) and South American tribes (3,133) comprised less than 0.3 percent of the total American Indian population (1,878,285). Given these small numbers, no major difficulties occur with the current classification and collection method if the category were to be expanded. Even if the census numbers include these tribes, the count would have to be much larger, at least 50,000 or more, to appear in any Federal data collection other than the decennial census. (1990 CPH-L-99, "American Indian Population by Tribe, for the United States, Regions, Divisions, and States: 1990")

It should be noted that in the development work that formed the basis for the current categories, some members of the FICE Ad Hoc Committee thought that the definition should refer to "original peoples of the Western Hemisphere" so as to include South American Indians. Ultimately, the Ad Hoc Committee decided that including South American Indians might present data problems for Federal agencies concerned with federally recognized tribes or Indians eligible for U. S. Government benefits.

Given that the Central and South American Indian population in the United States is so small, no significant issues arise with respect to measurement, data production, analytic, cost, or legislative or program needs.

5.3.2 Changes related to Asian and Pacific Islanders

The following suggested changes to Directive No. 15 concerning Asian and Pacific Islanders are discussed in this section:

• Should the "Asian or Pacific Islander" category be split into two categories? If yes, how should this be done?
• Should specific subgroups be listed under the current category?
• Should the term "Guamanian" or "Chamorro" be used?

5.3.2.1 Should the "Asian or Pacific Islander" category be split into two categories? If yes, how should this be done?

The issue is whether to retain the current Asian or Pacific Islander category, or to split the category into two separate categories, one for Asians and one for Pacific Islanders. The argument in favor of such a split is that the current category places together peoples who have few social or cultural similarities. It is argued that having separate categories for Asians and Pacific Islanders would result in more homogeneous groups, which would increase the comprehensibility and logic of the entire classification scheme. In addition, the two resulting groups are dissimilar on a number of measures. For example:

• Education--Although approximately the same numbers of Asians and Pacific Islanders graduate from high school, far fewer Pacific Islanders (about 11 percent of persons 25 years of age and older) than Asians (about 40 percent) obtain bachelors degrees.

• Income and employment--According to 1990 census data, 5.2 percent of Asians over age 16 were unemployed, compared with 7.3 percent of Pacific Islanders. Median household income was $41,583 for Asians and $33,955 for Pacific Islanders.

• Poverty--The poverty rate was 13.7 percent for Asians and 16.6 percent for Pacific Islanders. (Fernandez, 1996)

Aggregating Asians and Pacific Islanders separately is not problematic in decennial census data as currently collected, since separate data are available for each population group. Other data collections do not provide the opportunity to collect data separately for Asians and Pacific Islanders. In these instances, since Pacific Islanders are a small group numerically, their inclusion does not strongly affect the statistics for Asians. For example, the poverty rate for the entire Asian and Pacific Islander category is 13.8 percent, as compared with 13.7 percent for Asians alone. Because Pacific Islanders were only 365,000 of the Asian and Pacific Islander total of 7,274,000 reported in the 1990 census (Fernandez, 1996), however, the situation of Pacific Islanders is frequently masked. For this reason it is possible to argue that users could make better use of data if there were separate Asian and Pacific Islander categories. Given their relatively small numbers, however, there is the question of whether Pacific Islanders are a large enough population group to warrant a separate category.

A complicating factor is the request to separate Hawaiians from other Pacific Islanders, and to include them in the American Indian category (see section 5.3.3.2). If Hawaiians are not counted with other Pacific Islanders, the remaining "Non-Hawaiian Pacific Islander" group becomes very small. About 60 percent (211,000) of the Pacific Islanders are Hawaiians (Fernandez, 1996). The remaining 154,000 Pacific Islanders may be too small a group to justify a separate category. A residual "Asian and Non-Hawaiian Pacific Islander" category might confuse Hawaiian respondents, since the word Hawaiian would occur in two places in the question, and could prove difficult for other respondents to comprehend. For these reasons it is possible to argue that the Pacific Islander category, assuming it meets some minimum threshold, should only be considered as a stand-alone category if Hawaiians continue to be included in that category.

With such small numbers, it might become difficult to obtain adequate sample data for Pacific Islanders at the State or other local level if the category were to stand alone. Unless it uses a methodology that calls for oversampling for Pacific Islanders, any national survey using a random sample of the general population would expect to find three Pacific Islanders per 2,000 cases. A study would have to have a sample in excess of 20,000 respondents to obtain thirty respondents without using a stratified sample. It is unlikely that Federal agencies could afford to plan a study calling for such a national sample in order to have reliable data for a separate Pacific Islander category.

In addition, only a few agencies, such as the Department of Education in its assessment of reading proficiency, currently collect data separately on Asians and Pacific Islanders. In a number of cases, the numbers of Pacific Islander students were too small to permit statistically significant estimates. For example, although the percentage of Pacific Islander students at or above a "proficient" reading level in fourth grade in 1994 could be determined nationally, sample sizes were too small to permit reliable estimates for the Northeast, Southeast, Central, and West regions of the United States. Estimates were published only for three of the fifty States, and the estimate for California was flagged for interpretation with caution (Campbell, et al., 1996).

Currently, Directive No. 15 defines a member of the Asian or Pacific Islander category as a person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands (including, for example, China, India, Japan, Korea, the Philippine Islands, and Samoa). This definition does not clearly distinguish Asian from Pacific Islander areas. For example, by some definitions, Japan (an Asian country) could be considered a Pacific Island, and many of the peoples of the Philippines (also considered part of Asia) share linguistic and cultural features in common with Polynesians, Micronesians, and Melanesians. Further, the definition does not provide guidance about the classification of some groups. For example, Australian aborigines and the Papuan cultures of the South Pacific might be considered to be Pacific Islanders, although they have few social or linguistic affinities with the Polynesian, Micronesian, and Melanesian peoples otherwise included in the group.

Data production. Since the decennial census already codes and edits the Asian and Pacific Islander groups separately, data production in this case should not be affected by separating the Asian or Pacific Islander category. In data collection procedures that require a write-in for national origin, additional coding and editing would be required. Regardless of the size of the data collections at the national level, splitting this category will cause production difficulties for States with large populations of the two groups.

Analytic. Whenever a new category is established there are comparability discontinuities. In this case the discontinuities should be minor. It would be possible to recreate the antecedent category simply by adding the two categories together. Of greater difficulty would be trying to recreate data for earlier surveys using the two categories. Where population counts are large enough (as in the case of the decennial census), it should be a simple matter of disaggregation. In smaller studies, however, even those that oversampled for Asian and Pacific Islanders, splitting may be impossible.

Splitting the Asian or Pacific Islander category would have an additional effect in those areas where Asian and Pacific Islander populations have intermarried (such as Hawaii). Individuals with both Asian and Pacific Islander ancestry, who currently are able to respond in a single category, would have to choose between the two categories. They might respond as "other race" or as "multiracial," if such a category were available. Thus, comparisons over time would be more difficult, inasmuch as certain individuals might no longer report either as Asian or as Pacific Islander.

Cost. There would be substantial costs to requiring all Federal agencies to collect data on Asians and Pacific Islanders separately, particularly for the larger samples that would be required to produce statistically significant data for the small residual Pacific Islander category. Additional decennial census costs would be marginal for data collection and processing, since Asian and Pacific Islander groups are handled separately now. Additional costs would be incurred in the preparation and dissemination of new data products containing the split categories.

Legislative or program needs. Data on Asian and Pacific Islander populations are needed for apportionment in those States with large Asian or Pacific Islander populations. Splitting the Asian or Pacific Islander category into two categories might have an impact on apportionment for State legislative seats in States that have large populations of both groups.

5.3.2.2 Should specific groups be listed under the Asian or Pacific Islander category?

The issue of whether to list specific groups in this category is important only for the decennial census, as most agencies do not collect data on separate Asian and Pacific Islander groups on a regular basis. A brief history puts this issue into perspective.

The 1980 Census contained a listing of Asian and Pacific Islander groups. The Census Bureau conducted several tests to see if Asian or Pacific Islander reporting would suffer if the specific groups were not listed and if a write-in line was provided instead. These tests indicated that data quality was the same or better in questions that did not list the groups separately. The 1986 National Content Test used the original 1980 version of the question, a modified version with a shorter list of subgroups, and a "short" version with a write in box for specifying nationality after responding to the Asian or Pacific Islander category. The original 1980 version had an item nonresponse rate of 5.3 percent, the modified short list version an item nonresponse rate of 2.7 percent, and the short version an item nonresponse rate of 1.6 percent. The Bureau of the Census found the item nonresponse for the 1980 version was unacceptably high: "...traditionally, the race nonresponse rate has been small under two percent." An additional test in Chicago also found that the short-question version produced better results than the original 1980 version. (Minutes and Report of Committee Recommendations, Census Advisory Committee, April 21 and 22, 1988.) For 1990, the Census Bureau recommended using the Asian or Pacific Islander category in the short form, in combination with a write-in box where all Asian and Pacific Islander groups could supply detailed data. However, citizen groups objected to this plan, and they were able to bring Congressional pressure to bear to restore the original list of Asian and Pacific Islander groups.

The arguments in favor of and against listing specific groups remain essentially the same as they were in 1988. An issue paper dated November 10, 1988, described the case for listing the Asian and Pacific Islander groups in terms of relations between the Census Bureau and the Asian and Pacific Islander community, which might have a negative impact on Asian or Pacific Islander participation in the census. The arguments in favor of listing the groups included: (1) strong opposition and outrage in the Asian and Pacific Islander community could actually lead to poorer reporting of race; (2) intense emotional feelings have the potential of affecting the overall enumeration (therefore, coverage in the census); and (3) opposition was creating divisiveness among racial and ethnic groups.

The groups that advocated the listing of the Asian and Pacific Islander groups were also concerned that the proposed 1990 version, which would have required all Asian and Pacific Islander persons to write in a group, could not produce detailed statistics on each group in a timely manner.

The current arguments against listing the subgroups are again the same as those made in 1988. A Census Bureau paper dated August 19, 1988, discussed the anticipated problems with listing the Asian and Pacific Islander groups. It noted that the listing approach would affect the accuracy of the racial data for Asians and Pacific Islanders as well as for Whites, Blacks, American Indians, Eskimos and Aleuts in the following ways (based on 1980 census and 1990 census test experience):

• Nonresponse rate for the race item would be higher.

• Misreporting by Asians or Pacific Islanders (for example, groups not listed such as Cambodians or Laotians reporting in the Vietnamese category; Asians and Pacific Islanders misreporting in the category of "Other race" due to a lack of understanding of the category "Other API").

• More misclassifications by Black and White persons (for example, ethnic groups such as Italian, West Indian, and Greek writing in an entry in the "Other race" box instead of using the appropriate category).

• More misreporting in the "Other race" category due to confusion about the intent of the question and lack of understanding of categories.

These drawbacks are still likely to occur in formats that list the Asian and Pacific Islander groups, as reflected in the National Content Survey and other recent Census Bureau tests.

It is important to note that a number of these drawbacks pertain to the reactions of other groups to a question that lists countries of origin only for Asians and Pacific Islanders. In 1988, the Bureau of the Census reported to the Minority Advisory Committee:

"The national origin groups listed in the race question caused confusion among respondents, and some racial groups protested that they were not specifically identified in the question. For example, some European and Black ethnic groups misinterpreted the race question; they also marked off the "Other" race category and wrote in their ethnic identification. That was not the question's intent, and the misreporting required a very expensive corrective operation both in the field and in the data processing offices." (Minutes and Report of the Minority Advisory Committee Recommendations, April 21-22, 1988)

The effectiveness of the question for other groups should be of concern in a decision about the listing of Asian or Pacific Islander groups in the decennial census.

An additional consideration, as before the 1990 census, is space. Although the format of the census instrument has changed from a grid to a booklet, space remains at a premium. This makes it difficult to add additional categories (such as persons from the countries of the former Soviet Union that should report in the Asian or Pacific Islander category) to the question to represent a changing Asian and Pacific Islander population.

Measurement. It is clear from the discussion above that the listing of Asian and Pacific Islander groups negatively affects general data quality with an item nonresponse rate more than four times higher than when group data are collected in a write-in format. The listing also has an effect on other racial categories, when respondents look for a relevant specific listing and then use the "Other race" category to supply ethnic or ancestral data.

The RAETT tested two variations in listing the groups that make up this category: listing them in alphabetical order and not listing them in alphabetical order. The results of this methodological difference are reported in Table 11-4R, "Terminology Issue: Comparison of Panel B (Without Alphabetization of Asian and Pacific Islander) and Panel G (With Alphabetization of Asian and Pacific Islander) for the Asian and Pacific Islander Targeted Sample, By Race: 1996 RAETT." Of the ten groups listed (Chinese, Filipino, Hawaiian, Korean, Vietnamese, Japanese, Asian Indian, Samoan, Guamanian, and Other Asian and Pacific Islander), five reported higher numbers with alphabetization and five reported higher numbers without. However, only two groups recorded a statistically significant difference at the 90-percent confidence level, one under each option. This seems to indicate that the manner in which the list is shown has no consistent effect on the category as a whole.

Data production. Part of the resistance to the short version of the census race question prior to 1990 (without the Asian and Pacific Islander subgroups) came from doubts that the Census Bureau would be able to code write-in responses in a timely manner. According to a Government Accounting Office report on the controversy, "[d]elays in the publication of detailed Asian and Pacific Islander data after the 1980 census resulted in concerns about how the data from the 1990 census would be processed." The Census Bureau's plans to put new technology in place came too late to ease this concern (GAO, 1993). With the automated coding operation that is now in place, this argument in favor of listing Asian and Pacific Islander groups can no longer be made.

Editing may also be necessary if the list of Asian and Pacific Islander groups remains in the decennial census race question. Tests conducted during the 1980s found that recently migrated groups that were not listed did not use the "other" write in as intended, but rather filled the circle next to a closely related group, crossed out the group's name, and wrote in their own country of origin. For example, Laotians and Cambodians (not listed separately) filled the circle by the category "Vietnamese" and then crossed out "Vietnamese." The Bureau of the Census estimates that 6 percent of those reporting as Vietnamese did so in error. The exact figures are not known because most of the editing was done directly on the questionnaires, in the regions or in the processing centers, and records were not kept of these changes.

Analytic. Splitting the Asian or Pacific Islander category would not create a comparability problem if the definitions of the two groups remain the same. However, if Hawaiians are removed, the resulting groups would not be comparable over time.

5.3.2.3 Should the term "Guamanian" or "Chamorro" be used?

In November 1995, the Bureau of the Census released a report on a focus group involving twelve Chamorro speakers held in the Washington, DC area. In the conclusion to the report, the author states that "the term Chamorro should probably be substituted for Guamanian on the questionnaire.... All focus group participants indicated that they preferred Chamorro to Guamanian, although with varying degrees of intensity." It should be noted, however, that the sample underrepresented Chamorros born in the United States and non-Chamorro speakers. (Levin, 1995)

In the RAETT reinterview for the Asian and Pacific Islander targeted sample, respondents in households with at least one person who identified as Guamanian indicated they preferred Guamanian (58 percent), Chamorro (20 percent), had no preference (18 percent), or preferred both (4 percent). Respondents also indicated that Guamanian (72 percent) and Chamorro (79 percent) were acceptable terms to them.

There are no measurement, data production, analytic, cost, or legislative or program needs issues related to the current method of data collection.

5.3.3 Changes related to Hawaiians

Changes to Directive No. 15 as they relate to Hawaiians discussed in this section include:

• Should the term "Native Hawaiian" or "Hawaiian" be used?
• Should Hawaiians continue to be included in the "Asian or Pacific Islander" category; be reclassified and included in the "American Indian or Alaska Native" category; or be established as a separate, new category?

5.3.3.1 Should the term "Native Hawaiian" or "Hawaiian" be used?

Two questions are raised by this issue. The first is how best to identify individuals who trace their ancestry to the people who lived in what is now the State of Hawaii prior to the arrival in 1778 of Captain James Cook. The second is how to help respondents differentiate between these individuals and others who are born in Hawaii but who are not descended from the indigenous people.

In the vital statistics system for the State of Hawaii, births are counted as Hawaiian if either parent is Hawaiian or part Hawaiian. The State is also developing a register of individuals who can trace their ancestry back to someone living in Hawaii before Captain Cook's 1778 visit to the Hawaiian Islands. Directive No. 15 itself does not provide guidance on this level of detail. Publications from the 1990 census use the term "Hawaiian." The RAETT results shed some light on this issue as four panels include a "Hawaiian" category and two include a "Native Hawaiian" category.

The RAETT tested the term "Native Hawaiian" in Panels D and G. The results of this test are reported in Table 7-4R, "Sequencing Issue in: Comparison of Panel D (Race Question First) and Panel B (Hispanic Origin Question First) for the Asian and Pacific Islander Targeted Sample, by Race: 1996 RAETT" and Table 11-4R, "Terminology Issue: Comparison of Panel B (Without Alphabetization of Asian and Pacific Islander) and Panel G (With Alphabetization of Asian and Pacific Islander) for the Asian and Pacific Islander Targeted Sample, by Race: 1996 RAETT." While no table specifically looks at the results using "Hawaiian" versus "Native Hawaiian," it is possible to get an idea whether the terminology used affects the results. In Table 7-4R no statistical difference in the reporting of Hawaiians is shown, while in Table 11-4R a statistical difference in the reporting of Hawaiians is shown.

In neither comparison is the issue of using the Hawaiian or the Native Hawaiian terminology the only issue under consideration. Therefore, it is hard to interpret these results conclusively. On the one hand, the term "Hawaiian" does not appear to cause any confusion in the minds of respondents. But on the other hand, the term "Native Hawaiian" may not cause confusion either, and it might more clearly define the population the term is aimed at enumerating.

In the RAETT reinterview for the Asian and Pacific Islander targeted sample, respondents in households with at least one person who identified as Hawaiian indicated that they preferred Hawaiian (48 percent), Native Hawaiian (35 percent), had no preference (10 percent), or preferred another term (0.5 percent). Respondents also indicated that Native Hawaiian (84 percent) and Hawaiian (95 percent) were acceptable terms to them.

There are no measurement, data production, analytic, cost, or legislative or program needs issues related to this decision regardless of which option is selected.

5.3.3.2 Should Hawaiians continue to be included in the "Asian or Pacific Islander" category; be reclassified and included in the "American Indian or Alaskan Native" category; or be established as a separate, new category?

In the public comments, some Native Hawaiians expressed a preference for the option of being included with American Indians and Alaska Natives in a category for indigenous peoples of the United States, possibly called "Native Americans." They said that including them in the large "Asian and Pacific Islander" category resulted in data that do not accurately reflect their social and economic conditions. For example, Pacific Islanders have relatively high poverty rates. They also have health issues and educational needs different from Asians. American Indian Tribal organizations opposed this option. Other comments against this option ranged from the term "Native" can "mean any persons born in a particular area" to the "data would be less useful than currently for policy development, trend analyses, and needs assessment;" and "not useful for health research."

Inclusion of Hawaiians in a category with American Indians and Alaska Natives would have a major impact on the picture of the social and economic conditions of American Indians and Alaska Natives; while Hawaiians make up 2.9 percent of the Asian and Pacific Islander category, they would represent 9.7 percent of a reconstituted "American Indian or Alaskan Native" category. (For detail on the State of residence of Hawaiians, see Table 5.1.)

A separate Hawaiian category also was proposed. In addition, it was suggested that "Hawaiian" be changed to "Hawaiian, part-Hawaiian," because most native Hawaiians are part Hawaiian and many, in the past, have categorized themselves as "White." Those for this option say that it provides specific information for policy development, trends analyses, needs assessments, program evaluation, and civil rights enforcement. However, because Hawaiians are a small geographically concentrated population, this option may create a problem for surveys in states outside the Pacific Region. In most states there are not enough Hawaiians to form a sampling pool large enough to obtain findings that are significant in any way.

The 1990 census reported 211,014 Hawaiians, or slightly less than 0.01 percent of the total population of the United States. Hawaiians are a highly concentrated population: almost two-thirds (138,742) reside in the State of Hawaii. The second highest concentration is in California, which has more than one-sixth (34,447) of all Hawaiians. The third highest concentration is in the State of Washington, which has about 2.5 percent (5,423) of all Hawaiians.

Another option, not suggested, but always available, is for local areas with large Hawaiian or part Hawaiian populations to have a separate classification. If Hawaiian is not included in the minimum list of OMB categories, it could still be used by states, local governments, or federal agencies with a specific need for this category.

What category should include Hawaiians may be a question of the alternative bases for classification and intent. If the categories used are intended to classify the races as a function of geography, the individuals of Hawaiian ancestry should remain as a sub-category of the Asian or Pacific Islander category.

If, on the other hand, the goal is to classify the indigenous people of what is now the United States of America, then individuals of Hawaiian ancestry should be moved. However, this also raises a question about the other groups that are indigenous to various territories that are part of the United States - e.g., Guam, Micronesia, and the Virgin Islands. While a distinction could be made based on the fact that Hawaii is a State, this is nonetheless an issue that will likely need to be addressed in a future, if not in this, revision of the Federal standards.

More important, however, is the issue of whether classifying individuals of Hawaiian ancestry into the same category as the American Indians confuses matters regarding legal status. American Indians have a special legal status with the Federal Government as a result of treaties and legislation. It is important, if individuals of Hawaiian ancestry are categorized as "Native Americans," that linkage to this special legal status be addressed and not left to interpretation or litigation.

¹Percent Hawaiian Ancestry of a proposed American Indian, Eskimo, and Aleut racial category equals the number of Hawaiians divided by the sum of Hawaiian Ancestry plus the current American Indian, Eskimo, and Aleut racial category.

²Percent Hawaiian Ancestry of the Asian and Pacific Islander racial category equals Hawaiian Ancestry divided by the Asian and Pacific Islander racial category.

The RAETT sheds some light on the number of individuals selecting the Hawaiian category under various reporting options. Table 1-4R [Multiracial Issue: Comparison of Panel A (No Multiracial Category) and Panel B (With a Multiracial Category) for the Asian and Pacific Islander Targeted Sample, by Race: 1996 RAETT,] and Table 6-4R [Multiracial Issue: Comparison of Panel C ("Mark One or More" Instruction) and Panel H ("Mark All That Apply" Instruction) for the Asian and Pacific Islander Targeted Sample, by Race: 1996 RAETT,] show that the addition of an option to report multiple races results in a lower reporting of Hawaiian only. Many Hawaiians select a multiple race option. Without a multiple reporting option, 9.20 percent of the Asian and Pacific Islander targeted sample report as Hawaiian (Panel A Table 1-4R). When a "Multiracial" category is offered (Panel B), the proportion selecting "Hawaiian" (only) drops to 5.48 percent. Table 6-4R shows that the proportion reporting Hawaiian (only) is 4.66 percent when the instruction is to "mark one or more" races (Panel C) and is 3.87 percent when the instruction is to "mark all that apply" (Panel H). The two panels in which multiple responses were allowed also showed an increase in the proportion reporting as "Other Asian and Pacific Islander," 9.93 percent in Panel C and 7.57 percent in Panel H. This increase is due in part to recoding done by the Bureau of the Census to prepare tabulations for the RAETT. If "Hawaiian" and any other Asian or Pacific Islander category were marked, the respondent was classified as "Other Asian and Pacific Islander." A more complete analysis of the multiple race reporting on RAETT among Hawaiians could provide additional insights.

Measurement. The measurement of individuals of Hawaiian ancestry in the decennial census or in those studies that identify this group would not be affected by reclassification of Hawaiians since there is no change in how Hawaiian ancestry is determined. However, such reclassification of those with Hawaiian ancestry would have substantial impact on the data consistency for both the resulting "Asian or Pacific Islander" category and the expanded "American Indian" category in the more typical cases where detail for individuals of Hawaiian ancestry is not collected/reported separately. It is likely that there would be no consistency across the classification change. It would be impossible to say with certainty whether differences in characteristics over time in either resulting category were a consequence of real change or of the new categorization of those with Hawaiian ancestry. Informing the data user about the discontinuity could be accomplished by footnotes. Data users interested in a time series would require additional information or special tabulations in the absence of specific subcategory data, which may not always be possible to produce

Data production. Data production would not be affected by moving individuals of Hawaiian ancestry; the group would not be defined differently, but moved to a different tabulation category. Of more importance would be the need for a redesign of the published tables at the subcategory level, as well as the need for explanatory footnotes. Analytic. While there should be no effect on who is reporting as being of Hawaiian ancestry, a change would have a major impact on the comparability over time of the aggregated, larger racial categories. While this population is small in number, Hawaiians make up just under 3 percent of the current "Asian or Pacific Islander" category but would make up almost 10 percent of a newly broadened category that would include American Indians, Alaska Natives, and those of Hawaiian ancestry. Casual data users looking up information in an almanac or a statistical publication might be misled by the change. Researchers using race as a major analytic variable in longitudinal time series might have to adjust their time series.

Cost. The costs associated with reclassifying Hawaiians are hard to calculate. They include, but are not limited to, discarding current forms; the preparation of new forms and instructions; an educational campaign to inform people filling out forms as well as data users of the change; the need to check submissions over the short run to make sure the change has been properly made; and the fact that data for the next few years may be inaccurate as a result of misclassifications.

Legislative or program needs. Current legislative and program needs related to individuals of Hawaiian ancestry will be unaffected by this change. However, legislative and program needs related to American Indians would be affected by the need for an additional analytic step to account for the change. For example, Census figures from 1990 show a median family income of $21,750 for American Indians and Alaskan Natives with 31 percent of the individuals in this population below the poverty line. Median family income in 1990 for Asian and Pacific Islanders was $41,251, and 14 percent were below the poverty line (1990 Census of Population, Social and Economic Characteristics: United States, 1990CP-2-1). These figures for Hawaiians (a very small proportion of the Asian or Pacific Islander category) were much closer to those for the Asians than to those for American Indians -- $37,269 and 14 percent. Asians, however, are considerably more likely to have completed college (37.7 percent) than either Hawaiians (11.9 percent) or American Indians (9.3 percent).

In addition, moving individuals of Hawaiian ancestry to the American Indian category could affect apportionment at the State legislative district level in local areas or States where the reclassification affects the resulting Asian and Pacific Islander or American Indian counts.

5.3.4 Other terminology issues

Other issues Related to Directive No. 15 concerning terminology covered in this section are:

• Should the term "Black" or "African American" be used?
• Should the term "Hispanic" or "Latino" be used?
• Should more than one term be used in either case?

5.3.4.1 Should the term "Black" or "African American" be used?

The terms used to identify population groups do not necessarily invalidate the categorization scheme, but they may have marginal effects on nonresponse rates and misreporting. They also could cause resentment among some respondents. Smith (1992) notes that the terms can be important because they are used by the particular group's members to indicate the achievement of standing in the greater community. In the case of Blacks, disagreements over terms can result among persons of different ancestries. Among Blacks of African-American heritage, a growing proportion express a preference for "African-American" over the term "Black" (Lavrakas, Schejbal, and Smith, 1994). On the other hand, Blacks with roots in the Caribbean or Africa do not identify with the term "African-American" (Denton and Massey, 1989; Billingsly, 1993).

Options that were investigated with respect to the Black category included using only Black, as currently, or using African-American instead.

Measurement. Testimony given at hearings held by OMB on proposed changes to Directive No. 15 stressed the importance of having categories that are generally understood and with which people could identify. This is a fundamental requirement if the principle of self-identification is to be honored. Moreover, supplying the Federal Government's definitions for the various population groups will be particularly important for recent immigrants.

The terms used for classification have to be both familiar and acceptable to the respondent. For instance, focus group participants from the Association of Public Data Users (APDU) believed that Jamaicans would resist identifying as African-American, but that they would identify as Black. If only African-American were offered, Jamaicans might turn to the "Other" category. This underscores the need for supplying a comprehensive definition of the category to interviewers and respondents.

The May 1995 CPS Supplement asked Black respondents to choose the term they preferred. Keeping in mind that their choices may have been influenced by the terminology in the race and ethnicity questions they already had received, "Black" was the term most preferred. However, while 44 percent chose "Black," almost as many in total selected either "African-American" (28 percent) or "Afro-American" (12 percent), while 9 percent gave no preference (Tucker et al., 1996). Additional analysis of the CPS Supplement data revealed that preference was dependent on respondents' demographic characteristics. Young and well-educated Blacks were most likely to prefer "African-American" or "Afro-American." The results of the National Content Survey generally coincide with the results from the CPS Supplement. "Black" was preferred by 45 percent of those identifying as Black, while 33 percent preferred "African-American."

As noted, problems could arise if terms are not defined or if certain national groups feel excluded by the terms. This may be a particular problem for example, for Caribbean Blacks.

The context in which data collection occurs must be considered when changing terminology. Again, mode of data collection will affect the way choices can be presented. Where observer identification is necessary, clear coding rules will need to accompany any changes in terminology. More precise population group definitions in instructions and data collection instruments will help State and local governments as well as private-sector organizations.

Data production. To the extent that some Blacks do not identify with the terminology provided, they may not respond or may check the "Other race" category when it is offered. In this case, specific answers would have to be coded. Better instructions and definitions may reduce this problem.

Analytic. Because there is diversity in the Black community, the terminology used to measure this population needs to be encompassing. Denton and Massey (1989) found that it is important to capture the complete ethnic identities of Blacks when studying living patterns. For example, they documented that Caribbean Blacks were less segregated from Hispanics than they were from other Blacks.

A number of Federal agencies have expressed concern that changes will make it difficult, if not impossible, to recreate or to aggregate data to the categories they currently are using. These agencies do not object to greater detail but do worry that aggregation to the current categories might not be possible. Their concern is that some Blacks (or Hispanics) no longer would identify with the same category if terminology were changed. Both the Department of Defense and the Federal Bureau of Investigation suggested that part of the Black population, especially recent immigrants, could be misclassified if "African-American" were to replace "Black." Furthermore, some of the public comment suggests that the term "American" should not be used in this category, given that it is not used in other categories such as Asian.

Cost. The costs involved in changing terminology would small relative to some of the other possible changes. These costs would come from the development of new instructions, new definitions, and new forms designs. Some costs may be incurred for additional statistical adjustment and estimation procedures beyond those usually employed after each decennial census if distributions change as a result of new terminology. Changes in terminology should not increase costs much for those outside the Federal Government since these changes would be incorporated in the transition made to accommodate the new data from Census 2000.

Social costs may result whether changes are made or not made. Depending upon the decision, different interest groups may be unhappy.

Legislative or program needs. Many Federal agencies will expect to be able to make comparisons to past data series regardless of any changes. To the extent that changes in terminology prevent such comparisons, this will be a problem that must be resolved. However, the problems in this particular case are expected to be minimal relative to other possible changes. A survey of public school systems conducted by NCES (1996) found that a majority (55 percent) did not believe changing to "African-American" would be a problem, while 10 percent said it would be a significant problem. About 30 percent believed it would create some problems.

5.3.4.2 Should the term "Hispanic" or "Latino" be used?

The issues with respect to terminology for the Hispanic category are somewhat different. Many Hispanics prefer to identify with their country of origin. As Hahn (1994) points out, "Hispanic" is a term created by the Federal Government and is not traditionally used by peoples with origins in Central and South America. In fact, the term appears to be a compromise among the various groups. Some researchers suggest using "Latino" instead (Hayes-Bautista and Chapa, 1987) while others are comfortable with "Hispanic" (Trevino, 1987). In either case some groups might mistakenly be included or excluded. For example, Italians might identify as Latino, but Filipinos would not. In addition to the broad category identifier, knowledge of the particular Hispanic subgroup is often desirable (Farley, 1993). The National Council of La Raza, for example, supports the collection of the respondent's subgroup.

In the case of Hispanic origin, possibilities include (1) using only Hispanic; (2) collecting Hispanic subgroup designation or country of origin; or (3) using other terms instead of Hispanic, such as "Latino," "Chicano," and "Of Spanish Origin." In addition, instructions could be given for the respondent to mark "No" if not Hispanic. If an Hispanic subgroup is asked for, an "Other" category might be provided along with a space to specify the group.

Measurement. In the CPS Supplement, the term "Hispanic" was chosen by 58 percent of the respondents, and "Latino" and "Of Spanish Origin" were each selected by 12 percent. Another 10 percent indicated they had no preference, while 8 percent chose some other term. More than 60 percent of Mexicans and Puerto Ricans chose "Hispanic," compared with a little over 40 percent among the other subgroups. Hispanics over age 50 were less likely than younger ones to prefer "Hispanic." They were more likely than the others to choose "Of Spanish Origin" or "Some other term." Again, the results from the National Content Survey paralleled the CPS Supplement findings. The term "Hispanic" was preferred by 47 percent of the respondents, "Spanish" by 21 percent, and "Latino" by 13 percent.

Differences in specific terms or subgroup identifiers might not be recognized by neutral observers, but they can be very important to the individual respondent. Even if observers could classify Hispanics correctly, identifying the particular subgroup (e.g., Puerto Rican, Cuban, Mexican, or other Hispanic) or distinguishing when someone is both Black and Hispanic (e.g., the Caribbean Blacks spoken of by Billingsly, 1993). Hahn, Truman, and Barker (1996) also found that even some proxies had trouble with this task. Clearly, the quality of data will suffer when proxies or observers cannot correctly determine race and ethnicity, but respondents themselves are not always consistent in their responses to these questions. McKenney, et al. (1991) found this in examining reinterview data from the 1990 census. Overall, inconsistency was found to be low, but it was greatest for Hispanics who had been in this country for a long period of time or those who were born here, who only spoke English, and who said they were "Other Spanish" when asked to indicate their subgroup. The Hispanics of higher socio-economic status also show some inconsistency (Hazuda et al., 1986). Those who are not Hispanic do not consistently mark "No" unless provided with an instruction to do so (Bates, 1991).

Kissam, Herrera, and Nakemoto (1993) concluded that "Hispanic" or "Latino" would be better than "Spanish," but that asking for national origin would be even better, particularly for recent immigrants. The use of several terms or complicated instructions can be difficult both for recent immigrants and the illiterate. The effects of specific terms or the question format differ by mode of survey. Personal visits can overcome these problems best, but many surveys are no longer done this way. Mail surveys do lay out the alternatives clearly for respondents, but this mode assumes literacy. Telephone surveys may be most affected by wording and format.

Data production. As with Blacks, to the extent that some Hispanics do not identify with the names of the categories provided, they may not respond or may check the "Other" category when it is offered (either in the Hispanic origin question or the race question). When more detailed information on Hispanics is collected, the write-in answers in the "Other" category must be coded. Editing of open-ended responses may be required. Imputation will be needed for those who do not identify with the terms provided and who leave the question blank. This may be a particular problem for Hispanics failing to give a subgroup. This editing is on top of that resulting from Hispanics failing to respond to the race question and non-Hispanics not answering the ethnicity question.

To the extent that the failure to answer the race and ethnicity questions because of disagreement with the terms is not random, both the Blacks and the Hispanics that do answer the questions will not be representative. This would be an additional source of error affecting statistical distributions including the counts of subgroups. Weighting adjustments would be needed, but could be carried out only if the necessary information is available.

Analytic. One methodological point that those studying the Hispanic community agree on is that more detailed information about respondents' origins is needed. This is certainly true for substantive analysts, although some Federal agencies may not need this level of detail to carry out their specific mandates. Researchers stress that a simple "yes-no" question is not sufficient for analyzing differences in the diverse Hispanic community. Gimenez (1989) concluded that a global identification is not useful because Hispanics are so heterogeneous. The members of APDU who were interviewed indicated that they often must distinguish between different Hispanic subgroups in their work in local communities. Wong and McKay (1992) argued that comparisons across Hispanic subgroups actually are more important than comparisons of Hispanics with Blacks, Whites, and Asians. Kleinman (1990), in looking at health outcomes, came to the same conclusion.

The 1990 census did request Hispanic subgroup. Whether or not Hispanic subgroup is ascertained, the Hispanic community is so diverse that the terminology used needs to be encompassing. To the extent that some Hispanics cannot identify with the terms used, a part of this diverse population might be missed. Furthermore, with the increasing Hispanic immigration, subgroups might need to be tracked and terminology might need to change more rapidly than in the past in order to provide the same level of knowledge.

Cost. Most of the same issues discussed for Blacks apply in this case, with two additional ones. More space on forms would have to be allocated if information on Hispanic subgroups is desired. The amount of open-ended coding in the race question probably would be affected more by changes in terminology for Hispanics than for Blacks.

Legislative or program needs. Federal agencies will have the same concerns about changes in categories for Hispanics as they do about changes for Blacks.

5.3.4.3 Should more than one term be used for Black or for Hispanic?

One possible solution to the problems arising from the choice of terms the Black and Hispanic categories is the use of more than one term in the names of the categories. If several terms were used, respondents who identified with any one of the terms could select the category. Options considered as part of this review included (1) some combination of "Black," "African-American," and "Negro" and (2) some combination of "Hispanic," "Latino," "Chicano," and "Of Spanish Origin."

Measurement. If several terms are used (or, possibly, with just a change in terms), the current definitions might need revision. For example, a recommendation was offered at the Workshop on the Federal Standards for Racial and Ethnic Classification, held by the National Academy of Sciences, to use the term "African-American" in addition to the term "Black" (1996). The evidence from the CPS Supplement suggests that using both Black and African-American would satisfy most of the respondents in that category. The same would be true for using several terms in the Hispanic origin question. In both cases, the populations identifying with each category could be more diverse. At that point, the identification of subgroups might become more critical for analytic purposes.

The Hispanic origin question in Panel 3 of the NCS read, "Is this person of Spanish/Hispanic origin?" Additionally, in Panel 3 the Hispanic origin question came immediately before the race question and the race question did not offer a multiracial category as a reporting option. The Hispanic origin question in Panel 4 of the NCS read, "Is this person Spanish/Hispanic/Latino?" Further, as in Panel 3, the Hispanic origin question in Panel 4 came immediately before the race question but, unlike Panel 3, the race question in Panel 4 offered a multiracial category as a reporting option.

The NCS found that Panel 4 (where the race question included the multiracial category) had a lower percentage of respondents who reported as Hispanic in the Hispanic origin question compared with Panel 3--6.9 percent in Panel 4 compared with 9.0 percent in Panel 3. This decline was particularly pronounced among Mexicans, declining from 5.6 percent in Panel 3 to 3.2 percent in Panel 4.

Additional analyses of responses to comparable panels were conducted to determine whether the decline in Hispanic origin identified by these data is due to the fact that a multiracial category was included in the race question or to the change in the wording of the Hispanic origin question ("Spanish/Hispanic origin" in Panel 3, and Spanish/Hispanic/Latino" in Panel 4). These analyses revealed that neither the multiracial category in the race question nor differences in the wording of the Hispanic origin question was associated with a statistically significant decline in the proportion of Mexicans or of Hispanics in those panels 3 and 4. Moreover, additional analyses using NCS reinterview data ruled out the possibility that significantly different proportions of Mexicans were sampled in Panels 3 and 4.

Given these analyses, it is not clear whether the decline in the percentage who reported as Hispanic in Panel 4 relative to Panel 3, particularly among the Mexican subgroup, is due to the presence of the multiracial category in the race question, the wording of the Hispanic origin question, the placement of the Hispanic origin question before the race question, or the confluence of these factors. Thus, the drop in reporting as Hispanic, and particularly as Mexican, on Panel 4 remains unexplained.

Data production. If several terms were used for the Hispanic origin and Black categories, it is possible that the coverage of these populations would be improved. A significant number of Hispanics, however, might still choose an "Other race" category or not answer the race question, as demonstrated by the NCS and the CPS Supplement.

Analytic. The use of several terms may increase the diversity of those comprising the Black and Hispanic populations. Thus, their characteristics may be different than would be the case if only one term were used. In fact, while a more complete picture of these groups may result, that picture could be confusing. Subgroup differences might be more important.

Cost. Again, costs will be small compared to some of the other changes being considered, and these costs are for the same items already mentioned. However, costs for open-ended coding are likely to be reduced if multiple terms are used, because the residual or "Other" category will be chosen less often.

Legislative or program needs. The use of several terms for Blacks and Hispanics still could produce a lack of comparability with earlier data. Slightly larger population counts may result for the groups from the use of multiple terms. The effects could be more pronounced in some local areas than in others, depending on the diversity of the population.

5.3.5 Other new category issues

Public comment included suggestions to add other population groups to the minimum set of categories currently used for all data collection and reporting by the Federal Government. Some of the issues raised (summarized in OMB's August 1995 Federal Register Notice) were: adding categories for White ethnic groups; adding a category for persons of Arab or Middle Eastern descent; adding a category for Creoles; and adding a category for Cape Verdeans. The discussion below focuses on issues surrounding the addition of categories for Arab or Middle Easterner and for Cape Verdean. There were a number of public comments which requested that categories for European-Americans and for German-Americans be included in the revised Directive. This issue was not addressed in the research program. However, such data are available from the ancestry question on the decennial census.

5.3.5.1 Should an Arab or Middle Eastern category be created and, if so, how should it be defined?

The argument for creating a separate category for persons of Arab or Middle Eastern descent is similar to that made for persons of Hispanic descent: they are a diverse population group having some language and cultural characteristics in common. Like Hispanics, persons of Arab or Middle Eastern descent can be of any race. Many are White but there also are many of Black and other racial descent. The number of persons (1.6 million, or 0.7 percent of the U.S. population in 1990) who report in one of the ancestries that the Census Bureau has shown under the heading of "North Africa and Southwest Asia" (a very broad, geographically based categorization) exceeds that of many of the groups shown on the decennial census form. (An alternative to adding an ethnic group would be a short-form question on ethnicity/ancestry--replacing or in addition to the Hispanic origin question--with space for a write-in of specific, less common ancestries.) It has been suggested that in order to track problems related to discrimination against Arabs or Middle Easterners, some way of identifying them separately is necessary. Then, if a pattern of problems can be discerned, a case could be made to alter legislation in which specific protected groups are identified. It is also contended that recent Arab and Middle Eastern immigrants have the same problems as those from Asia, Central or South America, or Africa.

Some believe that having a separate category for persons of Arab or Middle Eastern descent would more easily qualify them for program benefits aimed at the socially and economically disadvantaged. On the other hand, an article in American Demographics states that, while it is true that Arab Americans suffer from stereotyping and negative press, it is equally true that they are younger, more educated, and more affluent than the average American. ("The Arab-American Market," American Demographics, January 1994)

Currently there is no recognized common identity for this population group-- neither a generally accepted name nor a common description. One characteristic that many Arab or Middle Easterners have in common is the Moslem religion; but many others are of other religious backgrounds as, for example, Lebanese Christians. Because of the separation of church and state in the United States, data are not collected on religious affiliations. Conversely, many Moslems do not have race or geographic origin in common--they come from Asia, Sub-Saharan Africa, etc. If the category were called or included "Middle Easterner" in its title, would it include persons from a non-Arab state such as Israel?

While a name and a definition could be imposed for this suggested new category, in a decennial census respondents need to understand clearly the concepts and the definitions of the classifications without necessarily having to read a definition. The public comment showed there is no agreement about the Middle Eastern countries to be included; this is further confused by the fact that there are Arab countries in North Africa and that the Middle East includes Israel, a non-Arab country.

The research to develop a definition and a commonly understood name (and the information campaign that would be required to inform the public of the new category) would be difficult to undertake in time for the 2000 census. While such research has not always been carried out prior to including a category in the decennial census, such a decision without research would be hard to rationalize given the intensive research on other issues surrounding race and ethnicity.

The requisite research could allow consideration of incorporating a new classification that would identify persons of Arab and Middle Eastern descent in a future classification system. The 1990 census indicates that this is a growing population group with a high proportion of foreign-born and recent immigrants. According to a Census Bureau report (1990 CP-3-2), 40 percent of persons of Arab ancestry are foreign-born and half of these foreign-born came to the United States between 1980 and 1990.

Measurement. No research has been conducted on the quality and consistency of reporting of persons of Arab or Middle Eastern descent on the race item on previous decennial censuses. Directive No. 15 instructs persons of Middle Eastern or North African descent to report their race as "White." However, it is not known how well this instruction is followed--or even if persons know that such a definition exists. Over the years there has been confusion about how persons of these ancestries should respond--"Asian," "White," or "Other race." Requests for consideration of adding an Arab or Middle Eastern category have not been consistent in the suggested name and the criteria for the definition of what geographic area should be encompassed.

Even in 1990 census reports, the definition of Arab was not consistent. Two reports on ancestry, Ancestry of the Population in the United States (1990 CP-3-2) and Detailed Ancestry Groups for States (1990 CP-S-1-2), used different definitions of "Arab," which resulted in different counts of persons. A comparison is presented in Table 5.2.

Note: See paragraph preceding Table 5.2 for definitions of the codes.

The data on ancestries that are marked "X" on Table 5.2 were shown separately in the respective reports. Ancestries marked "#," including the specific reporting of "Arab" as an ancestry, were grouped and shown as a balance category, "Other Arab," in Ancestry of the Population in the United States. In contrast, in Detailed Ancestry Groups for States, "Arab" was shown as a separate category, not grouped with other ancestries. In this latter report, the ancestries that are marked with an asterisk on Table 5.2 were combined into a balance category called "Other North African and Southwest Asian, n.e.c. (not elsewhere classified)."

Table 5.3 presents data from Detailed Ancestry Groups for States. It shows the number of persons reporting in any of the categories listed, as well as the number who reported specifically as "Arab" or "Middle Eastern." The report carries a footnote stating that these two categories are "a general type response, which may encompass several ancestry groups" (no further explanation is provided).

Given the lack of a generally understood concept, should the term Arab or Middle Eastern be used and be defined as persons whose "mother tongue" or culture was Arabic? Or should the category be based upon a strict geographic definition (and if so, which countries should be included)? Public comment included the following suggested names: Middle Eastern; Middle Easterner; Arab American; Middle Eastern or Arabic heritage; Arab American and other Middle Eastern; and West Asian. In any case, implementation would require a consensus building effort to arrive at appropriate terminology and a definition. In addition, the implementation of such a category on a 100-percent basis would require more instruction than is typically given on a 100-percent item in the decennial census. The closest approximation would be a listing such as that given on the 1990 census long form ancestry item.

¹ Includes persons who reported the following ancestries: Algerian, Alhucemas, Arab, Armenian, Assyrian, Bahraini, Berber, Egyptian, Iranian, Iraqi, Israeli, Jordanian, Kalday, Kurdish, Kuwaiti, Lebanese, Libyan, Middle Eastern, Moroccan, North African, Omani, Palestinian, Rio de Oro, Saudi Arabian, South Yemen, Syrian, Tunisian, Turkish, United Arab Emirates, and Yemeni.

Source: U.S. Bureau of the Census, 1990 Census of Population, Supplementary Reports, Detailed Ancestry Groups for States, 1990 CP-S-1-2.

Data production. If a separate category specifically for Arab or Middle Eastern were presented on the decennial census form, no further coding would be necessary. However, it would be advisable to compare the reported race to any other information collected in the decennial census (e.g., country of birth and ancestry, if these data are collected), to be able to understand the reported information better.

Analytic. The addition of a racial category in which persons of Arab or Middle Eastern descent might respond could reduce the total number of Whites counted in the next census. If this category were generally understood and only persons who previously responded "White" reported into this new category, one could compare the numbers of Whites between censuses (or other Federal data collections) by adding the Arab and Middle Eastern numbers to the numbers of persons reporting White to approximate the numbers of Whites in previous collections. However, the number of persons considering themselves to be Arab or Middle Eastern who actually reported in the White category is unknown; in the 1980 and 1990 censuses, many may have reported into the "Asian or Pacific Islander" category rather than the "White" category. If this was the case, then adding the numbers of persons reporting into a new "Arab or Middle Eastern" category to those reporting "White" could result in a higher number of "Whites" overall.

If an ethnic category were added, rather than a racial category, there would be no reduction in the numbers of any racial category. Before such an addition could be made, however, there would have to be agreement on how the new category would be defined. As the public comments have indicated, this is not an easy task.

Cost. The cost of collecting information about persons of Arab or Middle Eastern descent from the decennial census is not known. Components of the cost are the cost of adding a specific category to the form itself and then the cost of analyzing the resultant data to determine its quality and usefulness. The cost of tabulations of data would incrementally increase with the addition of a new category. As Table 5.2 indicates, the 1990 census reports did tabulate Arab or Middle Easterner, but under two different definitions.

Legislative or program needs. At this time, there are no extant Federal legislative needs or specific program rule requirements for data on Arabs or Middle Easterners. Persons who have requested that this information be collected in the 2000 census and other Federal data collections make the argument that the information is needed in order to make a case for changes in civil rights and related legislation. An example of this contention appeared in a public comment, which erroneously held that under current civil rights legislation "a Korean shopkeeper is protected but a neighboring Arab or Middle-Eastern shopkeeper is not" (letter received by OMB during public comment period). Others would argue that current civil rights laws provide for a means of seeking redress for discrimination.

5.3.5.2 Should a Cape Verdean category be created?

Cape Verde is a country consisting of a number of islands off the west coast of Africa at about 15 degrees latitude. For many years the islands were a Portuguese colony. The population of the islands is generally a mix of Black and White. As an island nation, its population depended on the ocean for economic survival. As skilled seaman, many islanders immigrated to New England to take part in the whaling industry. According to a Census Bureau report, Ancestry of the Population of the United States (1990 CP-3-2), 71 percent of all persons of Cape Verdean ancestry are native-born, and 18 percent are foreign-born and are not citizens. (Thus, the proportion of Cape Verdeans who are native-born is lower and the proportion of foreign-born noncitizens is higher than for the total U.S. population: for the total U.S. population 92 percent were native-born and 5 percent were foreign-born and were not citizens.)

As of the 1990 census, 51,000 persons reported Cape Verdean ancestry or ethnicity (0.02 percent of the total U.S. population). They are a population that is concentrated in four Northeastern states; 86 percent of persons who reported Cape Verdean ancestry lived in Massachusetts (58 percent), Rhode Island (20 percent), Connecticut (6 percent), and New York (2 percent). Another 5 percent of the Cape Verdean ancestry population resided in California. While they are a very small percentage of the U.S. population as a whole, they made up 1.0 percent of the Rhode Island population, 0.5 percent of the Massachusetts population, and 0.1 percent of the Connecticut population.

Measurement. Discussion with respect to this population group is limited because the only previous measures come from the ancestry/ethnicity questions in the census long forms of 1980 and 1990. This discussion assumes that if there were a separate ethnic category, about the same numbers of people would report as Cape Verdean as in the 1990 ancestry question.

Because a distinct ethnic category for such a small and geographically concentrated population group may not be possible, even on the decennial census, the Cape Verdean population might also find acceptable a multiracial or "Other race, specify" category that required specification of the respondent's component races. This question, combined with the use of the ethnicity/ancestry question that was tested as one of the options in the RAETT, may be a feasible and acceptable form of reporting. The addition of a multiracial category on other Federal forms would allow persons to report as multiracial (Cape Verdean) on these as well. If achieving a count of Cape Verdeans on a Federal form at the national level through the race question is desired, then an educational program would be required in order to inform persons that they can report this way. However, there has been no research concentrated on this population group; hence, it is not known how they would report given race classifications such as "multiracial" or "Other race, specify."

Perhaps the most satisfactory solution for counting Cape Verdeans is a local one. The four states with the highest numbers of Cape Verdeans in their populations (Massachusetts, Connecticut, New York, and Rhode Island) could find some means to count them for local and state purposes--for example in school administrative records systems, in employment and unemployment data, and in vital records systems. If guidance is given on how to aggregate this population into the Federal categories, there should be little impact for the State's record systems.

Data production. Cape Verdeans often write in "Cape Verdean" after marking the "Other race" category.

Analytic. In the absence of specific research, it is unclear how other race categories would be affected if a separate Cape Verdean category were established.

Cost. The cost of collecting information about Cape Verdeans by adding a new category in the decennial census is not known. If such information were collected on a 100-percent basis, the cost would be significantly higher than was experienced in coding responses to the ancestry item on the long form sample of one-sixth of all households.

Legislative or program needs. Currently, there are neither Federal legislative needs nor programmatic needs for these data on the national level. State-level program needs for information on Cape Verdeans are likely to exist in those states where there are significant concentrations of this population.

CHAPTER 6. Recommendations and Major Findings

6.1 Summary of Recommendations and Major Findings

Research conducted as part of the review of Directive No. 15 has produced a considerable amount of information about the issues covered in this report. The sources of this information have included public comments gathered from hearings and responses to two Office of Management and Budget (OMB) Notices published in the Federal Register, opinions of experts in the area of race and ethnicity, small-scale ethnographic and cognitive laboratory studies, and several national tests sponsored by Federal agencies. This section presents the recommendations of the Interagency Committee for the Review of the Racial and Ethnic Standards to OMB for how Directive No. 15 should be changed. It also summarizes the major research findings for the issues addressed by the recommendations. These findings are based on estimates from sample surveys.

The recommendations concern options for reporting by respondents, formats of questions, and several aspects of specific categories, including possible additions, revised terminology, and changes in definitions. Instructions for interviewers, the wording of questions, and specifications for tabulations are not addressed in the recommendations. The need for separate guidelines covering these topics is discussed at the end of the chapter. As in the current Directive No. 15, the recommendations are designed to provide minimum standards for Federal data on race and ethnicity. The recommendations continue to permit the collection of more detailed information on population groups to meet the needs of specific data users, provided the additional detail can be aggregated to comply with the minimum standards.

6.1.1 Recommendations concerning reporting more than one race:

• When self-identification is used, a method for reporting more than one race should be adopted.
• The method for respondents to report more than one race should take the form of multiple responses to a single question and not a "multiracial" category.
• When a list of races is provided to respondents, the list should not contain a "multiracial" category.
• Two acceptable forms for the instruction accompanying the multiple response question are "Mark one or more ..." and "Select one or more...."
• If the criteria for data quality and confidentiality are met, provision should be made to report, at a minimum, the number of individuals identifying with more than one race. Data producers are encouraged to provide greater detail about the distribution of multiple responses.
• The new standards will be used in the decennial census, and other data producers should conform as soon as possible, but not later than January 1, 2003.

The multiracial population is growing, and the task of measuring this phenomenon will have to be confronted sooner or later. Adopting a method for reporting more than one race now means that the demographic changes in society can be measured more precisely with a smaller discontinuity in historical data series than would occur in the future. Moreover, while technical concerns should not govern the decision, new procedures will be needed in any event, given that at least 0.5 percent of respondents to the 2000 Census are likely to select more than one race even if told to select only one. Allowing respondents in Federal data collections to select more than one race will be consistent with the trend toward this option at the state level, and may encourage the states to conform to a Federal standard.

Methods for reporting more than one race have been tested in both self-administered and interviewer-administered settings with similar results. This change will involve costs, but they are likely to be manageable and probably would be incurred eventually. The counts for Whites and Blacks, at least in the short term, will not likely be affected by allowing the reporting of more than one race; for populations whose counts could be affected, the information can be recovered to some degree with tabulation procedures. Standardized tabulation rules need to be developed by the Federal agencies working in cooperation with one another. When results from data collection activities are reported or tabulated, the number selecting more than one race should be given, assuming that minimum standards for data quality and confidentiality are met. Data producers are encouraged to provide greater detail about the distribution of multiple responses.

Allowing multiple responses is preferable to establishing a multiracial category, given the lack of legislative need for a specific count of the multiracial population and some of the drawbacks associated with the use of that category. There is no general consensus for a definition of "multiracial," as reflected in the public comment and in current state legislation requiring a multiracial category. A multiracial category is more likely to be misunderstood by respondents, resulting in greater misreporting. If a multiracial category were to be used (with write-in lines or a follow-up question), it would require either more space or more coding. An "Other" category with a multiracial example may be less likely to produce accurate data, may be offensive, and will require coding. Although self-identification should be greatly encouraged, its use is not always feasible or appropriate. When observer identification is used, determining a multiracial background by observation may be difficult, if not impossible.

Since data producers will be given until 2003 to conform to the new standards, additional research could be conducted in the context of the different data collection initiatives. This research might estimate the effects in the different settings and evaluate methods for data tabulation to meet users' needs. This date was chosen because information from Census 2000 will be available then for use in conjunction with other Federal data collections. It is expected, however, that data producers will begin using the new standards as soon as possible.

6.1.1.1 Findings concerning a method for reporting more than one race

Findings favoring adoption of a method for reporting more than one race:

• Between 1 and 1.5 percent of the public select a multiracial category when offered an opportunity to do so.
• The opportunity to identify with more than one race promotes self-identification, may increase self-esteem, and may reduce nonresponse to the race question,.
• The multiracial population has grown over the past 25 to 30 years.
• Some multiracial individuals strongly advocate the change.
• Some states have already begun allowing individuals to identify with more than one race using a multiracial category.
• Approximately 0.5 percent of respondents to self-administered surveys, including the 1990 census, selected more than one race even when asked to select only one race.
• Allowing individuals to report more than one race may provide a more complete report of a changing society.
• Allowing individuals to report more than one race could increase the accuracy of racial reports, and some inconsistencies in racial reporting may be eliminated.
• The counts for Whites and Blacks, at least in the near term, are unlikely to be affected.
• The counts for affected races can, to some degree, be recovered using various tabulation procedures.
• Test results in self-administered surveys and interviewer-administered surveys have produced similar estimates of individuals who are likely to report more than one race.
• The process for reapportionment and redistricting is not likely to be affected.

Findings not favoring adoption of a method for reporting more than one race:

• There is a potential for lowering counts for some groups, such as American Indians and Alaskan Natives and Asians and Pacific Islanders.
• Advocacy groups for some populations have strongly opposed the change.
• Time series and other analyses will have to account for the change.
• Alternative tabulations will be needed to carry out some program requirements, and this may be in conflict with the principle of self-identification.
• The effects of survey mode (self-administered or administered by interviewer, over the telephone or in person) may be accentuated, and data quality may suffer if instructions for reporting more than one race are not as successfully communicated to the respondent in some modes as in others.
• Enforcement of the Voting Rights Act might be affected by the reporting of more than one race.
• Only a subset of multiracial individuals may choose to identify with multiple races, so estimates for this population might be questioned.
• Data processing systems may have to be modified to incorporate tabulation procedures for reporting more than one race.
• Data collection instruments, instructions, and procedures will have to be modified, and more emphasis will need to be placed on the creation of instructions for respondents.
• Observer, and possibly proxy, identification could be operationally difficult to implement.
• There are no Federal legislative requirements for information about the multiracial population.

6.1.1.2 Findings concerning different formats for reporting more than one race

Multiracial category

• Definitions and terminology for the category would have to be generally understood and accepted by the public.
• Persons may identify with two or more races, but may not choose to respond as "multiracial."
• Using a multiracial category with a write-in would take up little space but require more coding.
• Using a multiracial category with a follow-up question specifying races would take up more space but require less coding.
• A multiracial category with a write-in works well for self-administered data collections but would not be appropriate for interviewer-administered surveys, which would need a follow-up question.
• Multiracial is sometimes misinterpreted by respondents as also meaning multiethnic.
• The presence of a multiracial category may affect reporting by Hispanics on the Hispanic origin question.

Select one or more races

• Only one question is needed.
• With fewer write-ins, less coding is required.
• It is not necessary to select terminology and develop a definition if a "multiracial" category is not being added.
• Instructions would be needed, and their wording would be extremely important.
• Some respondents already select more than one race even when asked to mark only one.
• Tabulating a multiple response option may be more straightforward and consistent across Federal agencies than tabulating write-in responses would be.

An "Other" category with examples that include multiracial

• Public comment indicated that an "Other" category is offensive to some respondents.
• A greater amount of coding of responses would be required. Multiracial individuals will not be able to express adequately their own identity. A smaller proportion of respondents may report "other" compared with the other options for reporting more than one race

6.1.2 Recommendations concerning a combined race and
Hispanic ethnicity question:

• When self-identification is used, the two question format should be used, with the race question allowing the reporting of more than one race.
• When self-identification is not feasible or appropriate, a combined question can be used and should include a separate Hispanic category co-equal with the other categories.
• When the combined question is used, an attempt should be made, when appropriate, to record ethnicity and race or multiple races, but the option to indicate only one category is acceptable.

The two question format allows Hispanics both to identify as Hispanic and to provide information about their race. It provides a complete distribution simply and continuity with past data is more likely to be maintained. Data on Hispanic subgroups can be obtained more easily with this format. The two question format should be used in all cases involving self-identification. When self-identification is not possible (e.g., the respondent is incapacitated), a combined format could be used. The recording of both Hispanic ethnicity and a race should be encouraged. The recording of only one identification, however, should be left as an option.

6.1.2.1 Findings concerning whether race and Hispanic origin should be combined into a single question

Findings favoring a single question:

• Respondents may not confront what they may consider to be redundant questions.
• The concepts of "race" and "ethnicity" are difficult to separate.
• Reporting by Hispanics in the "Other" race category may be reduced.
• Some Hispanics and data users have expressed support for a combined question.
• The number of respondents using write-ins for the race question may be reduced.
• Inconsistencies in Hispanic reporting may be reduced.
• Self-identification for Hispanics may be enhanced.

Findings not favoring a single question:

• Some Hispanics want to identify their race in addition to Hispanic origin.
• Some Hispanics, including the Census Hispanic Advisory Committee and most Hispanic organizations, oppose a single, combined question.
• "Hispanic" is not considered a race by some respondents and users.
• The reporting of Hispanic subgroups will be awkward with a single question.
• A single, combined question may have a differential effect on reporting by Hispanic subgroups.
• A single, combined question will increase the need for additional tabulations as a result of multiple responses.
• Time series and other analyses will have to account for the change.
• The historical continuity of economic or demographic statistics for Hispanics may be affected.
• Additional tabulations may be needed for administrative reporting, and this might infringe on self-identification.

6.1.2.2 Findings concerning different formats if race and Hispanic origin are combined in a single question

A combination of race, ethnicity, and ancestry

• More responses will need to be coded and edited.
• Some Hispanic respondents may not provide subgroup detail, reducing the counts of specific subgroups and increasing the "other Hispanic" group.
• Ancestry would be collected for the entire population on every data collection and not just the Census long form, but the distribution may change from that with a separate ancestry question.
• The question may be too difficult for some respondents.

A question with an Hispanic category allowing multiple responses

• Only a single question is needed.
• Hispanic origin would be a category co-equal with race.
• Some Hispanics prefer to indicate both their Hispanic origin and race.

A question with an Hispanic category allowing only one response

• The count of Hispanics may be reduced, since some Hispanics may select a category other than Hispanic.
• Hispanic origin would be co-equal with race.
• Observer and proxy identification could be more difficult.
• For those reporting Hispanic, no race is obtained.

6.1.3 Recommendations concerning the retention of both
reporting formats:

• The two question format should be used in all cases involving self-identification.
• The current combined question format should be replaced with a combined format which includes a co-equal Hispanic category for use, if necessary, in observer identification.

The two question format for collecting data on Hispanic origin and race is considered superior to the single question format, and it should be used in all cases involving self-identification. The single question format should only be used where self-identification is not possible. In these cases, a single question in the form of the combined question discussed above can be used, but, again, data collectors should be strongly encouraged to record both ethnicity and race to provide more complete information about the individual. Attempts to obtain proxy responses (from family or friends) as opposed to using observer identification also should be encouraged in order to promote data accuracy.

Findings favoring retention:

• Both formats are being used by Federal agencies; a number of large administrative data bases use the combined format.
• Some data collection instruments and procedures as well as processing systems currently being used will have to change if only one format is retained.
• Time series and other analyses would have to account for the change.

Findings not favoring retention:

• The two formats do not produce comparable data.
• The combined format allowed in Directive 15 does not produce a complete distribution of Hispanic origin by race.

6.1.4 Recommendation concerning the ordering of
the Hispanic origin and race questions:

When the two question format is used, the Hispanic origin question should precede the race question.

All research findings point to placing the Hispanic origin question before the race question. Hispanics appear less confused by the race question and do not select the "Other" race category as often when this sequencing is used. This reduces the amount of data editing and coding needed. Furthermore, non-Hispanics are more likely to give a response to the Hispanic origin question.

Findings favoring the race question appearing first:

• Current time series or other analyses would have to take account of a change in question sequencing.
• Even if the Hispanic origin question were to appear first, some Hispanic respondents will not answer the race question or will select "Other" race in the decennial census.

Findings favoring the Hispanic origin question appearing first:

• The meaning of the race question will be clearer, especially to Hispanics.
• Non-Hispanics will be more likely to give a response to the Hispanic origin question.
• Data editing and coding should be reduced.

6.1.5 Recommendation concerning adding Cape Verdean
as an ethnic category:

• A Cape Verdean ethnic category should not be added to the minimum data collection standards.

Given the small size and geographic concentration of this population, the analytical power gained by a separate identification at the national level would be minimal compared to the costs, especially for sample surveys. Even without a separate category, however, the ability to report more than one race may allow Cape Verdeans to express their identity. An ancestry question would allow Cape Verdeans to identify themselves for the purposes of estimating population size. States with a significant Cape Verdean population can collect data for state and local purposes.

Findings favoring the addition of a Cape Verdean ethnic category:

• It would respond to complaints that discrimination against Cape Verdeans is difficult to assess without a separate category for data on this population.
• Cape Verdean is easily defined.
• Some Cape Verdeans favor the addition of the category.
• Data may be useful for administering some state and local programs.
• The number of write-ins in an "Other" category may be reduced.
• The principle of self-identification would be supported. The picture of society would be more complete.

Findings not favoring a Cape Verdean ethnic category:

• This population is concentrated in certain states that could collect data at the local level.
• There is no specific Federal requirement for information about Cape Verdeans.
• Little research has been done on the effects of adding Cape Verdean to the list of ethnic categories.
• Time series and other analyses would have to account for the change.
• Cape Verdeans could be accommodated if the reporting of more than one race were allowed, although additional tabulations would be needed.
• The ancestry question on the decennial census provides an opportunity for individuals to identify their Cape Verdean ancestry.

6.1.6 Recommendation concerning the addition of
an Arab or Middle Eastern ethnic category:

• An Arab or Middle Eastern ethnic category should not be added to the minimum data standards.

The definition of Arab or Middle Eastern ethnicity is problematic. At least three approaches--linguistic, geographic, and religious--have been proposed. More space would be needed on questionnaires, and Arab or Middle Eastern ethnicity can be obtained from an ancestry question. States with a significant Arab or Middle Eastern population can collect data for state and local purposes. Given the small size and geographic concentration of this population, the analytical power gained by a separate identification at the national level would be minimal compared to the costs, especially for sample surveys.

Findings favoring the addition of an Arab or Middle Eastern ethnic category:

• It would respond to complaints that discrimination against Arabs or persons from the Middle East is difficult to assess without a separate ethnic category.
• Some Arabs or Middle Easterners favor a separate ethnic identification.
• It may address the difficulty some Arabs or Middle Easterners have in responding to the race question.
• Data may be useful for administering some state and local programs.
• The number of write-ins for an "Other" category may be reduced.
• The principle of self-identification would be supported. The picture of society would be more complete.
• Arabs and Middle Easterners are racially mixed and, hence, similar conceptually to the Hispanic community.

Findings not favoring the addition of an Arab or Middle Eastern ethnic category:

• An Arab or Middle Eastern ethnicity is difficult to define. States having concentrations of Arabs or Middle Easterners could collect data at the local level.
• An Arab or Middle Eastern ethnicity question would require more space.
• There are no Federal requirements for information about Arabs or those from the Middle East.
• Little research has been done on the effects of adding an Arab or Middle Eastern ethnic category.
• Time series or other analyses would have to account for the change.
• Arab or Middle Eastern ethnicity can be obtained with an ancestry question on the decennial census.

6.1.7 Recommendation concerning the addition of
any other categories to the minimum set:

• No other racial or ethnic categories should be added to the minimum set of categories.

Additional racial and ethnic categories would require more space with little analytical value added. States can collect data at the state and local level for groups concentrated in their areas. The current Directive permits the collection of this greater detail. Some of these groups would be accommodated by allowing the reporting at the Federal level of more than one race. Given the small size and geographic concentration of these populations, the analytical power gained by a separate identification at the national level would be minimal compared to the costs, especially for sample surveys.

Findings favoring the addition of other categories:

• Such an addition would respond to complaints that discrimination cannot be assessed without separate categories.
• Some states and local areas have diverse populations and need additional detail for administrative purposes.
• The picture of society would be more complete.
• Some groups favor the creation of their own categories.
• The number of write-ins in an "Other" category may be reduced.
• The principle of self-identification would be supported.

Findings not favoring the addition of other categories:

• There are no specific Federal requirements for information on other population groups.
• States having concentrations of certain population groups could collect data at the local level to meet their requirements.
• Little research has been done on the effects of additional categories.
• A long list would require more space on all data collection instruments, not just the decennial census forms.
• Time series and other analyses would have to account for the change.
• Some of these categories would be accommodated by allowing the reporting of more than one race.
• The current Directive permits the collection of more detailed data on population groups, provided the detail can be aggregated into the minimum set of categories.

6.1.8 Recommendation concerning changing the term
"American Indian" to "Native American":

The term American Indian should not be changed to Native American.

The term "Native American" may confuse those born in the United States, and the count of American Indians may become less accurate. "Native American" is a term which could include more than American Indians. American Indians are divided on which term they prefer, but most tribal organizations prefer "American Indian."

Findings favoring the change:

• Some find the term to be a more accurate description of this indigenous population.
• Some American Indians expressed a preference for the term "Native American."

Findings not favoring the change:

• American Indian tribal governments prefer to retain the term "American Indian."
• The term "Native American" often is interpreted by respondents to mean "born in this country."
• The accuracy of the counts of American Indians may be affected by a change in terminology.
• Time series and other analyses would have to account for the change in terminology.
• "Native American" is confusing, since it refers to groups other than American Indians.

6.1.9 Recommendation concerning changing the term "Hawaiian" to "Native Hawaiian":

• The term "Hawaiian" should be changed to "Native Hawaiian."

Although the term "Native Hawaiian" may be misinterpreted by respondents to mean "born in Hawaii," there is little evidence to suggest this would be as likely as in the case of "Native American." Furthermore, the preponderance of the public comments on this issue favored using "Native Hawaiian."

Findings favoring the change:

• Hawaiians are an indigenous people to what is now the United States.
• Public comment indicated a preference for the use of the term "Native Hawaiian."
• The review found no compelling evidence that counts of this group would be affected.

Findings not favoring the change:

• "Native Hawaiian" may be misinterpreted by respondents to mean "born in Hawaii."
• The accuracy of counts of Hawaiians may be affected.
• Time series and other analyses could have to take account of the change.
• Some research findings indicated that more Hawaiians appear to prefer "Hawaiian" to "Native Hawaiian," but both were acceptable terms.

6.1.10 Recommendation concerning the classification of Hawaiians:

• Hawaiians should continue to be classified in the Asian or Pacific Islander category.

Although Hawaiians are an indigenous people, they are geographically linked to other Pacific Islanders. Furthermore, other groups, such as the American Samoans and the Guamanians, requested a similar change, with the result that the meaning of the Pacific Islander classification would likely be affected. Hawaiians are divided on which classification should be used. The historical continuity of data on the economic characteristics of Pacific Islanders would be affected.

Findings favoring classification with other indigenous populations.

• Hawaiians are an indigenous people.
• Like Alaska, and unlike American Samoa or Guam, Hawaii is a state.
• Hawaiians account for approximately ten percent of the indigenous population of the United States.
• Some Hawaiians favor classification in the same category as the American Indians and Alaska Natives.

Findings favoring continued classification as Asian/Pacific Islander:

• Geographically, Hawaiians should be classified with other Pacific Islanders.
• Time series and other analyses would not have to account for the change in classification.
• The administration of Federal programs for the indigenous population might be affected by the change.
• Other groups, such as the Samoans and the Guamanians, also have requested reclassification out of the Asian/Pacific Islander category. These changes, along with a change for Hawaiians, would effectively eliminate the Pacific Islander category.
• The historical continuity of economic and demographic statistics for Pacific Islanders as well as American Indians could be affected by a change in classification.
• American Indian tribal governments are opposed to the change, because it might affect the quality of the data for American Indians.
• There appears to be no clear preference on the part of Hawaiians--some Hawaiians favor classification in the American Indian category, and still others favor a separate Native Hawaiian category.
• Except for the proportion of college graduates, Hawaiians resemble Asians more than American Indians in terms of economic status.

6.1.11 Recommendations concerning the use of Alaska Native instead of Eskimo and Aleut:

• "Alaska Native" should replace the term "Alaskan Native."
• Alaska Native should be used instead of Eskimo and Aleut.
• The Alaska Native response option should be accompanied by a request for tribal affiliation when possible.

"Alaska Native" is the term preferred by this population (as compared to "Alaskan Native"). Alaska Native, accompanied by a request for tribal affiliation, provides more accurate and complete data.

Findings favoring the use of Alaska Native:

• The term "Eskimo" is offensive to some respondents.
• Alaska Native, accompanied by a request for tribal affiliation, provides more accurate data for administrative purposes.
• "Alaska Native" is the term preferred by this population.

Findings not favoring the use of Alaska Native:

• The terms "Eskimo" and "Aleut" are acceptable to most Alaska Natives.

6.1.12 Recommendations concerning the classification of South and Central American Indians:

• South and Central American Indians should be classified as American Indian.
• The definition of the "American Indian or Alaska Native" category should be modified to include the original peoples from South and Central America.

The classification of South and Central American Indians as American Indian is consistent with how the Canadian Indians are classified, but the definition of the category would need to be changed accordingly. While the effects on the count of American Indians will be minimal, South and Central American Indians may find it easier to answer the race question.

Findings favoring a more inclusive American Indian classification:

• Classification in the American Indian category would be consistent with how the Canadian Indians in the United States have been classified using the current categories.
• The consistency of the classification of American Indians will be increased.
• It would be easier for South and Central American Indians to answer the race question.
• The effects of this change on the population count and other data on American Indians will be minimal.
• Some South and Central American Indians may prefer being classified as American Indian.

Findings not favoring a more inclusive American Indian classification:

• Little research has been done on the potential effects of changes.
• Some South and Central American Indians may prefer being classified as White.
• The reclassification may have a small effect the administration of Federal programs for American Indians.

6.1.13 Recommendations concerning the term or terms to be used for the name of the Black category:

• The name of the Black category should be changed to "Black or African American."
• The category definition should remain unchanged.
• Additional terms, such as Haitian or Negro, can be used if desired.

Substantial numbers of this population identify with one of the two terms, Black and African-American. If the two terms are connected by an "or," Caribbean Blacks can identify with the category. Other terms, such as "Negro" and "Haitian," can be used, but they should not be required. Since a relatively small number of Blacks identify with "Negro" and "Haitian," the term "Black or African American" is likely to be sufficient.

Findings favoring using "Black":

• Time series and other analyses will be unaffected.
• A plurality of Blacks prefer this term.
• This term does not cause much confusion for respondents, such as Caribbean Blacks.
• For most Blacks, it is not an offensive term.
• Some respondents find "African-American" a confusing term because the term could exclude Caribbean Blacks or include anyone from Africa, including Whites.
• Some public comment indicated an objection to the use of "American" in "African-American," because it connotes nationality and is not used in the names of the other categories, except for the American Indian category.

Findings favoring using "African American" or "Afro-American":

• A large proportion of Blacks favor one of these terms.
• For most Blacks, these are not offensive terms.
• The terms are commonly used and there seems to be a general consensus about the population group in the United States for which the term is intended.

Findings favoring another term:

• "Negro" may be favored by older Blacks.
• "Colored" may be favored by some Blacks in the South.

Findings favoring use of more than one term:

• Using more than one term is more inclusive and could achieve more complete coverage of the Black population.
• Nonresponse to the race question among Blacks may be reduced.
• Write-ins are less likely.

6.1.14 Recommendations concerning the term or terms to be used for Hispanic:

• The term used should be "Hispanic."
• The definition of the category should remain unchanged.
• Additional terms, such as Latino or Spanish Origin, can be used if desired.

A majority of Hispanics prefer the "Hispanic" term. "Hispanic" is a term with which most of this population is now familiar. Other terms, such as "Latino" or "Spanish Origin," can be used to achieve more complete coverage of the Hispanic population. There is some evidence, however, that using the term "Latino" may result in the inclusion of some unintended population groups, so it should not be a part of the minimum standard.

Findings favoring using Hispanic:

• A majority of Hispanics favor this term.
• Time series and other analyses are likely to be unaffected.
• Most Hispanics are familiar with this term.
• The inclusion of other terms, such as "Latino," might have the effect of including unintended population groups.

Findings favoring using the term "Latino":

• Some Hispanics favor this term.
• Some Hispanics are more familiar with this term than with "Hispanic" or other terms.

Findings favoring using the term "Spanish Origin":

• Some respondents of Spanish or European descent prefer this term.
• Some Hispanics may be more familiar with this term than with other terms.

Findings favoring another term:

• The term "Chicano" may be favored by Hispanics in the Southwest region of the United States.

Findings favoring use of more than one term:

• Nonresponse of Hispanics to the Hispanic ethnicity question may be reduced.

6.2 Comparison of the Current Standards with the Recommended Standards

This section summarizes the differences between Directive No. 15 and the recommended changes. The current standards are presented in Section 6.2.1. Section 6.2.2 shows how the current standards would be changed if the recommendations were to be adopted by the Office of Management and Budget. In the latter case, the Interagency Committee's recommended changes are presented in bold type so that they can be more readily compared to the current standards.

6.2.1 The Current Standards in Directive No. 15

The basic racial and ethnic categories for Federal statistics and program administrative reporting are defined as follows:

• a. American Indian or Alaskan Native. A person having origins in any of the original peoples of North America, and who maintains cultural identification through tribal affiliation or community recognition.

  b. Asian or Pacific Islander. A person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands. This area includes, for example, China, India, Japan, Korea, the Philippine Islands, and Samoa.

  c. Black. A person having origins in any of the black racial groups of Africa.

  d. Hispanic. A person of Mexican, Puerto Rican, Cuban, Central or South American or other Spanish culture or origin, regardless of race.

  e. White. A person having origins in any of the original peoples of Europe, North Africa, or the Middle East.

To provide flexibility, it is preferable to collect data on race and ethnicity separately. If separate race and ethnic categories are used, the minimum designations are:

• Race:
  --American Indian or Alaskan Native
  --Asian or Pacific Islander
  --Black
  --White
  Ethnicity:
  --Hispanic origin
  --Not of Hispanic origin

When race and ethnicity are collected separately, the number of White and Black persons who are Hispanic must be identifiable, and capable of being reported in that category.

If a combined format is used to collect racial and ethnic data, the minimum acceptable categories are:

• American Indian or Alaskan Native
  Asian or Pacific Islander
  Black, not of Hispanic origin
  Hispanic
  White, not of Hispanic origin

The category which most closely reflects the individual's recognition in his community should be used for purposes of reporting on persons who are of mixed racial and/or ethnic origins.

In no case should the provisions of this Directive be construed to limit the collection of data to the categories described above. However, any reporting required which uses more detail shall be organized in such a way that the additional categories can be aggregated into these basic racial/ethnic categories.

6.2.2 Recommended Standards

The minimum categories for data on race and ethnicity for Federal statistics and program administrative reporting are defined as follows:

• a. American Indian or Alaska Native. A person having origins in any of the original peoples of North and South America (including Central America), and who maintains cultural identification through tribal affiliation or community recognition.

  b. Asian or Pacific Islander. A person having origins in any of the original peoples of the Far East, Southeast Asia, the Indian subcontinent, or the Pacific Islands. This area includes, for example, China, India, Japan, Korea, the Philippine Islands, Hawaii, and Samoa.

  c. Black or African-American. A person having origins in any of the black racial groups of Africa.

  d. Hispanic. A person of Mexican, Puerto Rican, Cuban, Central or South American or other Spanish culture or origin, regardless of race.

  e. White. A person having origins in any of the original peoples of Europe, North Africa, or the Middle East.

To provide flexibility and assure data quality, it is preferable to collect data on race and ethnicity separately. When race and ethnicity are collected separately, ethnicity should be collected first. Persons of mixed racial origins can, but are not required to, report more than one race. If race and ethnicity are collected separately, the minimum designations are:

• a. Race:
  --American Indian or Alaska Native
  --Asian or Pacific Islander
  --Black or African-American
  --White
  b. Ethnicity:
  --Hispanic origin
  --Not of Hispanic origin

When the data are reported, a minimum of one additional racial category, designated "More than one race," must be included, if the criteria for data quality and confidentiality are met, in order to report the aggregate number of multiple race responses. Data producers are encouraged to provide greater detail about the distribution of multiple responses. Terms such as "Haitian" or "Negro" can be used in addition to "Black" and "African-American." Terms such as "Latino" or "Spanish origin" can be used in addition to "Hispanic."

If a combined format must be used to collect racial and ethnic data, both race and ethnicity or multiple races should be collected when appropriate, although the selection of one category will be acceptable. If a combined format is used, the minimum categories are:

• American Indian or Alaska Native
  Asian or Pacific Islander
  Black or African-American
  Hispanic
  White

When the data are reported, a minimum of two additional categories, designated "Hispanic and one or more races" and "More than one race," must be included if the criteria for data quality and confidentiality are met and both race and ethnicity and multiple races were collected.

In no case should the provisions of this Directive be construed to limit the collection of data to the categories described above. In fact, the collection of subgroup detail is encouraged. However, any reporting required which uses more detail shall be organized in such a way that the additional categories can be aggregated into these minimum categories for data on race and ethnicity.

6.3 Recommendations for Further Research

A great deal of research has been conducted over the past few years to provide information on which to base possible revisions to Directive No. 15. More research still is needed. Most immediately, research should be conducted by the affected agencies both to evaluate the effects of the proposed changes and to consider methods for accommodating them. A phased implementation period of up to five years has been proposed to allow agencies to make changes in data collection instruments and procedures, as well as in processing and tabulation systems. To assist the agencies, OMB should issue guidelines on data tabulation and reporting, instructions for interviewers, and suggested wording for questions by January 1, 1999.

Tabulation methods are particularly important in the case of reporting more than one race, and Federal and state agencies are encouraged to work together, under the auspices of OMB, to develop methods that would produce consistent results for program purposes and for comparisons with historical data. These guidelines would be particularly useful for those charged with civil rights enforcement. In addition, much thought should be given to the appropriate way to tabulate multiple responses for official purposes. Because instructions can have a profound effect on data quality, instructions for respondents and interviewers that will effectively communicate the intention of the race and Hispanic origin questions should be developed. Other aspects of questionnaire design, including question wording, also should be addressed by the guidelines.

Some important issues have not been resolved during this period of review and a number of questions are left unanswered. For example, conceptual bases for defining Arab or Middle Eastern ethnicity should be explored. The differences between the concepts of "race," "ethnicity," and "ancestry" have not been satisfactorily determined. More intensive study of small populations such as Hawaiians, Cape Verdeans, and Creoles should be undertaken. In many cases, this work would have to be done in local areas where these population groups are concentrated. In the future, there will be the opportunity to examine why some people choose to select more than one race while others, with the same characteristics, do not. Also, more research is needed on inconsistencies in reporting race and ethnicity over time. More thought should be given to the current use of geographic origin in the definition of racial categories. Building on considerable progress the Census Bureau has made, the search for a single question that satisfactorily captures both race and ethnicity should be continued.

REFERENCES

Almey, M., Pryor, E.T., and White, P.M. (1992), "National Census Measures of Ethnicity in the Americas," in Proceedings of the International Union for the Scientific Study of Populations' Conference on "The Peopling of the Americas," Vera Cruz, Mexico, V. 3, pp. 5-22.

Bates, N. (1991), "The 1990 alternative questionnaire experiment: Preliminary report of the 100-percent items," Paper written for the 1990 Decennial Census.

Bates, N., Martin, E.A., DeMaio, T.J., and de la Puente, M. (1996) "Questionnaire Effects on Measurements of Race and Spanish Origin," Journal of Official Statistics, Vol. 11, pages 433-459.

Bates, N., de la Puente, M., DeMaio, T.J., and Martin, E.A. (1994), "Research on Race and Ethnicity: Results from Questionnaire Design Tests," Paper presented at the Census Bureau's Annual Research Conference, Rosslyn, VA.

Bennett, C., McKenney, N., and Harrison, R. (1995), "Racial Classification Issues Concerning Children in Mixed-Race Households," paper presented at the Annual Meeting of the Population Association of America, San Francisco, California.

Billingsly, A. (1993), Climbing Jacob's Ladder: The enduring Legacy of African-American Families, New York: Simon & Schuster.

Blackburn-Moreno R. (1994), Letter from ASPIRA Assoc. Inc. to OMB.

Bracken, K., and de Bango, G. (1992), "Hispanics in a racially and ethnically mixed neighborhood in the greater metropolitan New Orleans area," Ethnographic evaluation of the 1990 decennial Census report #16, Prepared under Joint Statistical Agreement 89-45 with Hispanidad '87, Inc, Bureau of the Census, Washington D.C.

Bureau of the Census (1997), "Results of the 1996 Race and Ethnic Targeted Test," Staff in Population Division and Decennial Statistical Studies Division, Population Division Working Paper No. 18.

Bureau of the Census (1996), "Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey," Population Division Working Paper No. 16.

Bureau of the Census (1995), Statistical Abstract of the United States: 1995 (115th edition) Washington, D.C. 1995.

Bureau of the Census, Ancestry of the Population in the United States, 1990, CP-3-2.

Bureau of the Census, Detailed Ancestry Groups for States, 1990 CP-S-1-2.

Bureau of the Census, American Indian Population by Tribe, for the United States, Regions, Divisions, and States: 1990, 1990 CPH-L-99.

Bureau of the Census (1988), "Assessment of Implications of Listing Asian and Pacific Islander Groups in the 1990 Census Race Item," August 19.

Bureau of the Census (1988), "Issue Paper on the 1990 Census Race Question," November 10.

Bureau of the Census (1988), "Minutes and Report of Committee Recommendations, Census Advisory Committee of the American Indian and Alaska Native, Asian and Pacific Islander, Black and Hispanic Populations for the 1990 Census," April 21.

Bureau of the Census (1982), Persons of Spanish origin by State: 1980, U.S. Department of Commerce, Bureau of the Census (PC80-1-7).

Campbell, J.R., Donahue, P., Reese, C., and Phillips, G. (1996), "NAEP 1994 Reading Report Card for the Nation and the States: Findings from the National Assessment of Educational Progress and Trial state Assessment," Office of Educational Research and Improvement, U.S. Department of Education.

Cantor, D., Kerwin, J., and Schecter, S. (1996), "Evaluation of changes to the race question on birth certificates." Proceedings of the Section on Survey Research Methods, American Statistical Association.

Carbo, 1994. Testimony by MALDEF to the Office of Management and Budget, Public Hearings on Standards for the Classification of Federal Data on Race and Testimony.

Carter-Pokras, O., and LaVeist, T. (1996), "Research and Politics: The Classification of Multiracial Persons," paper presented at the American Public Health Association, New York.

Crews, D.E., and Bindon, J.R. (1991), "Ethnicity as a taxonomic tool in biomedical and biosocial research," Ethnicity and Disease, 1:42-49.

Denton, N.A., and Massey, D.S. (1989), "Racial identity among Caribbean Hispanics: The effect of double minority status on residential segregation," American Sociological Review, 54, 790-808.

Department of Health and Human Services (1995), Directory of Minority Health and Human Services Data Resources, September.

Drury, T.F., Moy, C.S., and Poe, G.S. (1980), "Going beyond interviewer observations of race in the National Health Survey," in Classification Issues in Measuring the Health Status of Minorities, Hyattsville, MD: U.S. Public Health Service, NCHS.

Duncan, G.J., Hill, M.S., Lepkowski, J.,. Garza, R.G., Falcon, A., Garcia, C., and Garcia, J. (1992), Panel Study of Income Dynamics, 1968-1990 [Latino Sample, 1990]. [Computer file]. Ann Arbor, MI: Survey Research Center [producer]. Ann Arbor, MI: Consortium for Political and Social Research [distributor].

Elias-Olivares, L and Farr, M (1991), "Sociolinguistic Analysis of Mexican-American Patterns of Non-Response to Census Questionnaires," Ethnographic Exploratory Research, Report #16. Center for Survey Methods Research, Bureau of the Census, Washington DC.

Falcon A. (1994), "Puerto Ricans and the politics of racial identity," in The Psychological development and creative expression of racial and ethnic identity, eds. Harris H, Griffith E, Blue H, New York: Routledge.

Farley, R. (1993), "Questions about race, Spanish origin and ancestry: Findings from the Census of 1990 and Proposals for the census of 2000," Testimony at hearings of the Subcommittee on Census, Statistics and Postal Personnel, House of Representatives, April 14.

Federal Register Notice (1995), "Standards for the Classification of Federal Data on Race and Ethnicity," Office of Management and Budget, Federal Register, 60FR44674-44693, August 28, 1995.

Fernandez, E.W. (1996), "Comparisons of Selected Social and Economic Characteristics Between Asians, Hawaiians, Pacific Islanders, and American Indians (including Alaskan Natives)," Population Division Working Paper No. 15, U.S. Bureau of the Census.

General Accounting Office (1993), Census Reform: Early Outreach and Decisions Needed on Race and Ethnic Questions, GAO Report GGD-93-96.

Gerber, E., and de la Puente, M. (1996), "The Development and Cognitive Testing of Race and Ethnic Origin Questions for the Year 2000 Decennial Census," Proceedings of the Bureau of the Census 1996 Annual Research Conference, Arlington, Virginia.

Gimenez, M.E. (1989), "Latino/Hispanic - who needs a name? The case against a standardized terminology," International Journal of Health Services, 19, 557-571.

Ginorio, A. (1979), "A comparison of Puerto Ricans in New York with native Puerto Ricans and Native Americans on two measures of acculturation: gender roles and racial identification," Doctoral dissertation, Fordham University, New York, NY.

Ginorio, A., and Berry, P.C. (1972), "Measuring Puerto Ricans' perceptions of racial characteristics," (Summary) Proceedings of the 80th Annual Convention of the American Psychological Association, 7:287-288.

Hahn, R.A. (1994), "The coding of race and ethnicity in federal health statistics; problems and prospects," paper presented at a Workshop on Race and Ethnicity Classification: An Assessment of the Federal Standard for Race and Ethnicity Classification, National Research Council, Washington, DC, Feb. 17-18.

Hahn, R. A. (1992), "The State of Federal Health Statistics on Racial and Ethnic Groups," The Journal of American Medical Association, 267, 268-271.

Hahn, R., Mulinare, J., and Teutsch, S. (1992), "Inconsistencies in Coding of Race and Ethnicity Between Birth and Death; A New Look at Infant Mortality in the United States, 1983-1985," Journal of American Medical Association, 267:259-63.

Hahn, R.A., Truman, B.I. and Barker, N.D. (1996), "Identifying ancestry: The reliability of ancestral identification in the United States by self, proxy, interviewer, and funeral director," Epidemiology, 7, 75-80.

Harris, M., Consorte, G., Josildeth, L., Joseph, Byrne, B. (1993), "Who are the Whites? Imposed Census categories and the racial demography of Brazil," Social Forces; 72(2), 451-462.

Hayes-Bautista, D.E., and Chapa, J. (1987), "Latino terminology: Conceptual bias for standard terminology," American Journal of Public Health, 77, 61-68.

Hazuda, H.F., Comeaux, P.H., Stern, M.P., Haffner, S.M., Eifler, C.W., and Rosenthal, M.A. (1986), "A comparison of three indicators for identifying Mexican Americans in epidemiologic research; Methodological findings from the San Antonio Heart Study," American Journal of Epidemiology, 123, 96-112.

Johnson, T., Jobe, J.B., O'Rourke, D., Sudman, S., Warnecke, R.B., Chavez, N., and Golden, P. (1995), "Racial and ethnic self-identification in birth certification," paper presented at the American Sociological Association. Washington DC.

Kamasaki C. (1994), "An Hispanic assessment of the Federal standard for race and ethnicity," paper presented at the Workshop on Race and Ethnicity Classification: An Assessment o the Federal Standard for Race and Ethnicity. Committee on National Statistics, Washington.

Kissam, E., Herrera, E., and Nakamoto, J.M. (1993), "Hispanic responses to Census enumeration forms and procedures," Report prepared for the Bureau of the Census, Suitland, MD.

Kleinman, J.C. (1990), "Infant mortality among racial/ethnic minority groups," Morbidity and Mortality Weekly Report CDC Surveillance Summary, 39 (3), 31-39.

Lavrakas, P.J., Schejbal, J.A., and Smith, T.W. (1994), "The use and perception of ethno-racial labels: African-American' and/or Black,'" Proceedings of the 1994 Annual Research Conference, 505-515.

Levin, M.J. (1995), "Report on Focus Group Research for Race Item for the 2000 Census: Guamanian Versus Chamorro (or Both)," Report prepared for the Bureau of the Census, Suitland, MD.

Lindan, C.P., Hearst, N., Singleton, J.A., Trachtenberg, A.I., Riordan, N.M., Tokagawa, D.A., and Chu, G.S. (1990), "Underreporting of Minority AIDS Deaths in San Francisco Bay Area, 1985-86," Public Health Reports 105(4), 400-404.

Martin, E., DeMaio, T.J., and Campanelli, P.C. (1990), "Statement on cognitive research. director of the Center for Survey Methods Research. U.S. Bureau of the Census," Workshop on Race and Ethnicity Classification: An Assessment of the Federal standard for race and ethnicity classification. Committee on National Statistics. National Research Council. Washington DC,1994.

Martinez, A.R. (1988), "The effects of acculturation and racial identity on self-esteem and psychological well-being among young Puerto Ricans," Ph.D. Dissertation. City University of New York.

Massey, J.T. (1980), "A Comparison of Interviewer Observed Race and Respondent Reported Race in the National Health Interview Survey," paper presented at Annual Meeting of the American Statistical Association. Houston, Texas.

McKenney, N. and Cresce A. (1992). "Measurement of Ethnicity in the United States: Experiences of the U.S. Census Bureau." Published in the Proceedings of the Joint Canada-United States Conference on the Measurement of Ethnicity, pages 173-222.

McKenney, N.R., Bennett, C.E., Harrison, B., and del Pinal, J. (1991), "Evaluating racial and ethnic reporting in the 1990 Census," Washington, DC: U.S. Bureau of the Census. <

McKay, R.B., and de la Puente, M. (1995), "Cognitive Research in Designing the CPS Supplement on Race and Ethnicity," Proceedings of the Bureau of the Census 1995 Annual Research Conference, Rosslyn, Virginia, 435-445.

McKay, R.B., Stinson, L.L., de la Puente, M. and Kojetin, B.A. (1996), "Interpreting the Findings of the Statistical Analysis of the CPS Supplement on Race and Ethnicity," Proceedings of the Bureau of the Census 1996 Annual Research Conference, Rosslyn, Virginia, 326-337.

National Center for Education Statistics, Racial and Ethnic Classifications Used by Public Schools, NCES 96-092, pp. 14-15.

National Council of La Raza (1995). Letter to the Office of Management and Budget.

National Research Council, Committee on National Statistics (1996), Spotlight on Heterogeneity: The Federal Standards for Racial and Ethnic Classification(Summary of a Workshop), National Academy Press, Washington, D.C.

Office of Management and Budget (1995), "Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register; 60:44674-44693, August 28, 1995.

Office of Management and Budget (1977), "Race and Ethnic Standards for Federal Statistics and Administrative Reporting."

Olguin F. (1994) Letter from the Mexican American Legal Defense and Educational Fund to OMB.

Poe, G.S., Powell-Griner, E., McLaughlin, J.K., Robinson, K., and Placek, J. (1993), "A comparability study of the responses to the 1986 National Mortality Follow-back Survey to similar demographic items from the death certificate," US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Center for Health Statistics, Vital and Health Statistics, Data Evaluation and Methods Research, Series 2, No. 118.

Rodriguez, C.E., (1994) "Challenges and Emerging Issues: Race and Ethnic Identity Among Latinos," Committee on National Statistics, National Research Council, Washington, DC.

Rodriguez, C.E. (1992), "Race, culture and Latino otherness' in the 1980 Census," Social Science Quarterly, 71(4), 930-937.

Rodriguez, C. (1989), Puerto Ricans: Born in the USA, Boston: Urwin and Hyman. (Reissued by Westview Press in Boulder, CO in 1991).

Rodriguez, C. (1974), "Puerto Ricans: Between Black and White," Journal of New York Affairs, 1(4), 92-101.

Rodriguez, C.E., and Cordero-Guzman, J.M. (1992), "Place race in context," Ethnic Racial Studies, 15, 523-43.

Rodriguez, N., and Hagan, J. (1991), "Investigating Census coverage and content among undocumented: An ethnographic study of Latino immigrant tenants in Houston," Ethnographic evaluation of the 1990 Decennial Census Report #3, Prepared under Joint Statistical Agreement 89-34 with the University of Houston, Bureau of the Census, Washington D.C.

Romero, M. (1992), "Ethnographic evaluation of behavioral causes of Census undercount of undocumented immigrants and Salvadorans in the Mission District of San Francisco, California," Ethnographic evaluation of the 1990 Decennial Census Report #18. Prepared under Joint Statistical Agreement 89-41 with the San Francisco State University Foundation, Bureau of the Census, Washington D.C.

Scarr, H.A. (1994), Statement of Harry A. Scarr, Acting Director, Bureau of the Census. at the Hearings before the Subcommittee on Census, Statistics, and Postal Personnel of the Committee on Post Office and Civil Service, House of Representatives, 103 Congress, first Session, June 30, 1993, Serial No. 103-7, pp. 3-17.

Smith, T. (1992), "Changing racial labels: From Colored' to Negro' to Black' to African American,'" Public Opinion Quarterly, 56, 496-514.

Smith, T.W. (1983), "Problems in ethnic measurement: Over-, under- and misidentification," Proceedings of American Statistical Association.

Sorlie, P.D., Backlund, E., Johnson N.J., and Rogot, E. (1993), "Mortality by Hispanic Status in the United States," Journal of a American Medical Association 270: 2464-68.

Subcommittee on Census, Statistics and Postal Personnel of the Committee on Post Office and Civil Service House Representatives (1993), Hearings on Review of Federal Measurements of Race and Ethnicity.

Tienda, M., and Ortiz, V. (1986), "Hispanicity and the 1980 Census," Social Science Quarterly, 67, 3-20.

Torres, J. (1996), "Report Says Combining Race and Ethnicity in Census Lowers Latino Response," Hispanic Link Weekly Report, June 3, p. 4.

Trevino, F.M. (1987), "Standardized terminology for Hispanic populations," American Journal of Public Health, 77, 69-72.

Tucker, C. (1996), "Summary of Responses on Assessing Possible Changes in the Collection of Statistics on Race and Ethnicity," Survey of Council of Professional Association on Federal Statistics members.

Tucker, C. (1996), Summary of Focus Group with State and Local Association of Public Data Users Members.

Tucker, C., McKay, R., Kojetin, B., Harrison, R., de la Puente, M., Stinson, L., and Robinson, E. (1996), "Testing methods of collecting racial and ethnic information: Results of the Current Population Survey Supplement on Race and Ethnicity," Bureau of Labor Statistics Statistical Notes, No. 40.

Tumin, M., Feldman, A. (1961), Social class and social change in Puerto Rico, Princeton: Princeton University Press.

Wong, F.Y., and McKay, D.R. (1992), "Hispanic subgroups and intraracial comparisons (letter)," American Journal of Public Health, 82 (12), 1691.

Wright, L. (1994), "One Drop of Blood," New Yorker, 7/25/94, pp. 46-54.

Yzaguirre, R., and Perez, S.M. (1997), "Accurate racial/ethnic data should drive category review," in Double exposure: poverty and race in America, eds. C. Hartman. and M.E. Sharpe, Armonk, NY.

The Budget  |  Legislative Information  |  Management Reform/GPRA
Grants Management
Financial Management  |  Procurement Policy  |  Information & Regulatory Policy
Contact the White House Web Master

Privacy Statement