U.S. flag

An official website of the United States government

Skip Header


Results of the 1996 Race and Ethnic Targeted Test

Written by:
Working Paper Number POP-WP018

Disclaimer

This paper reports the general results of research undertaken by Census Bureau Staff. The views expressed are attributable to the authors and do not necessarily reflect those of the Census Bureau.

Acknowledgments

The analysis of data from the 1996 Race and Ethnic Targeted Test (RAETT) and the writing of this report reflect the combined efforts of several co-authors. These co-authors, most of whom are in the Population Division (POP) or the Decennial Statistical Studies Division (DSSD) at the Census Bureau, are listed alphabetically: Claudette Bennett (POP), Manuel de la Puente (POP), Deborah Griffin (DSSD), Brian Harris-Kojetin (on special assignment from the Bureau of Labor Statistics), Roderick Harrison (POP), Joan Hill (DSSD), John Hilton (DSSD), Theresa Leslie (DSSD), and Edna Paisano (POP).

Campbell Gibson (POP) and Henry Woltman (DSSD) provided overall direction for preparation of the report.

The Decennial Management Division coordinated and directed survey operations, including processing of the questionnaires.

Staff in the Population Division made many contributions to the preparation of the report. In particular, Kymberly DeBarros, Rosalyn Green, Emily Lennon, and Barbara Martin provided computer programming support; and, Mary Blankenship, Rhonda Carney, Deborah Carroll, Linda Chase, Mary Kennedy, Debra Niner, Tonya Prince, John Reed, Paula Vines, and Leigh Zarbough provided professional, statistical, and clerical assistance. Numerous reviewers also contributed: Nancy Gordon, Robert Kominski, John Long, Elizabeth Martin, Nampeo McKenney, Gregory Robinson, Daniel Weinberg, Karen Wheeless, and other Census Bureau staff.

 

Highlights of the Race and Ethnic Targeted Test

Since 1977 the federal statistical agencies have used the four racial categories (American Indian or Alaskan Native, Asian or Pacific Islander, Black, and White) and two ethnic categories (Hispanic origin, and Not of Hispanic origin) specified by the Office of Management and Budget (OMB) in Statistical Policy Directive No. 15. These are the minimum categories for reporting race and ethnicity.

In 1994 OMB initiated a comprehensive review of the racial categories prescribed by Directive No. 15, as part of which the Census Bureau conducted three sample surveys: 1) The May 1995 Supplement to the Current Population Survey (CPS) sponsored by the Bureau of Labor Statistics, a final report of which was published in June 1996; 2) the 1996 National Content Survey (NCS), the results of which were published in December 1996; and 3) the 1996 Race and Ethnic Targeted Test (RAETT), the results of which are presented for the first time in the accompanying report.

While the CPS and NCS surveys were designed to be nationally representative, a targeted sample design was used in the RAETT. This design provides larger samples of six targeted populations (Black, American Indian, Alaska Native, Asian and Pacific Islander, Hispanic, and White ethnic) than does a nationally representative sample, by drawing the samples from areas with high concentrations of the specified population groups. Unlike the results of the two earlier samples, the results for the targeted sample in the RAETT can only be generalized to areas of relatively high concentrations of the targeted populations used to select each sample. These areas of high concentration represent 3 percent or less of American Indian, Asian and Pacific Islander, and White ethnic households nationally, 8 percent of Alaska Native households, 10 percent of Black households, and 15 percent of Hispanic households.

SUMMARY OF THE RAETT'S RESEARCH OBJECTIVES AND FINDINGS

Objective: Determine the effects of a multiracial category and of two new instructions in the race question: to "Mark one or more" or to "Mark all that apply."

Findings:
  • Some respondents provided unrequested multiple responses, despite instructions in those panels to mark one box, especially in the Alaska Native and in the Asian and Pacific Islander targeted samples.
  • The options for reporting more than one race did not affect the percentages reporting solely as White, as Black, or as American Indian in their respective targeted samples. However they did affect the percentages of Alaska Native and of Asian and Pacific Islander in their respective targeted samples.
  • The multiracial category affected reporting in the single race category of Asian and Pacific Islander and of American Indian and Alaska Native.
  • The "mark all that apply" option also affected the reporting of Asian and Pacific Islander as a single race in that targeted sample, but the "mark one or more" option did not.
  • When the race and Hispanic origin questions are combined, a high percentage of responses included both Hispanic origin and one of the four major race categories currently allowed under Directive No. 15.
  • Using an illustrative procedure for classifying responses of more than one race into the current OMB race categories eliminated most statistically significant effects of the multiple race reporting options on reporting as Asian and Pacific Islander in that targeted sample, but only reduced the effects in the Alaska Native targeted sample on reporting as American Indian and Alaska Native.

Objective: Determine the effects of collecting information about race, Hispanic origin, and ancestry in a combined, two-part question.

Findings:
  • In every targeted sample, nonresponse to each of two combined questions was significantly lower than nonresponse to the corresponding separate Hispanic origin and race questions.
  • The two combined race and Hispanic origin questions elicited high levels of multiple responses in the Hispanic targeted sample. Over 90 percent of the multiple responses involved Hispanic origin and a race group.
  • When all responses of Hispanic (either Hispanic alone or Hispanic in combination with any other response) are added together, which is the appropriate comparision, there is no statistically significant difference in the percent reporting Hispanic between a combined question and separate questions on Hispanic origin and race.
  • Write-ins to the ancestry component of the two-part question did not provide reporting of the detailed Hispanic origin groups comparable to that provided in separate questions. For the detailed groups of the Asian and Pacific Islander population, comparable information was obtained by one version of the question ("mark one or more" instruction), but not by the other version ("mark all that apply" instruction).

Objective: Determine the effects of placing the Hispanic origin question immediately before the race question.

Findings:
  • This reduced, but did not eliminate, nonresponse to the Hispanic origin question.
  • This also reduced, but did not eliminate, reporting in the "Other race" category on the race question.

Objective: Test alternative terminologies, classifications, and formats in the race question.

Findings:
  • Spelling out "American Indian or Alaska Native" instead of using "Indian (Amer) or Alaska Native" has no effect on reporting as American Indian and Alaska Native in the American Indian targeted sample.
  • Substituting "Native Hawaiian" for "Hawaiian" and listing this category immediately after the "American Indian and Alaska Native" category increased reporting as Hawaiian.
  • Alphabetizing the Asian and Pacific Islander groups after "Native Hawaiian" had no effect on the total percentage reporting as Asian and Pacific Islander in that targeted sample.
  • Given the very small numbers involved, no statistically significant difference was found between reporting as "Guamanian or Chamorro" or as "Guamanian."

Contents

1. Introduction and Research Results
  1.1 Introduction
  1.2 Overview of Survey Design and Methodology
    1.2.1 Survey Design
    1.2.2 Survey Methodology
  1.3 Research Results from the Targeted Samples: Options for Reporting More than One Race
    1.3.1 Responses to the Questions on Race
    1.3.2 Illustrative Approaches to Classifying Data on Race
    1.3.3 Effects of the Options for Reporting More than One Race in the Targeted Samples
  1.4 Research Results from the Targeted Samples: Effects of Placing the Hispanic Origin Question Immediately Before the Race Question
  1.5 Research Results from the Targeted Samples: Effects of Combining the Questions on Race and Hispanic Origin
  1.6 Research Results from the Targeted Samples: Terminology, Classification, and Formatting
  1.7 Relationship to Other Research
 
2. Background
  2.1 The Office of Management and Budget's Statistical Policy Directive No. 15
  2.2 Issues Addressed in the Race and Ethnic Targeted Test
    2.2.1 Reporting More than One Race
    2.2.2 Sequencing the Questions on Race and Hispanic Origin
    2.2.3 Combining the Race and Hispanic Origin Questions With Ancestry Asked in a Second Part
    2.2.4 Terminology and Formatting Issues
 
3. Survey Design and Methodology
  3.1 Design of the Survey
    3.1.1 Experimental Design
    3.1.2 Sample Selection
  3.2 Survey Methodology
  3.3 Editing of Race Data
  3.4 Data Analysis Methods
    3.4.1 Statistics Used
    3.4.2 Panel Comparison Methods

Text Tables

Appendixes

Appendix Text Tables

Appendix Figures

1. INTRODUCTION AND RESEARCH RESULTS

1.1 Introduction

In response to legislative, programmatic, and administrative requirements in the federal government, the Office of Management and Budget (OMB) in 1977 issued the "Race and Ethnic Standards for Federal Statistics and Administrative Reporting," that are set forth in Statistical Policy Directive No. 15. These standards have been used for two decades in decennial censuses, in national surveys of the population, and in data collections to meet statutory requirements associated with monitoring and enforcing civil rights in areas such as housing, mortgage lending, educational opportunities, employment, and voting. The four basic racial categories specified in Directive No. 15 are: American Indian or Alaskan Native; Asian or Pacific Islander; Black; and White. The two specified ethnic categories are: Hispanic origin; and Not of Hispanic origin.1 Persons of Hispanic origin may be of any race.

During the past decade, the standards in Directive No. 15 have come under increasing criticism. Some individuals who report data about themselves, and various users of the data, believe that the categories do not adequately reflect the increasing racial and ethnic diversity of the population of the United States.

As a result of these concerns, OMB initiated a comprehensive review of Directive No. 15 and solicited comments on three topics:2

  • "the adequacy of the current categories;"
  • "principles that should govern any proposed revisions to the standards;" and
  • "specific suggestions for changes that have been offered by various individuals and organizations."

Four public hearings were held and written comments were received from about 800 individuals and organizations. The public comments addressed six issues.

  • "Should the Federal government collect data on race and ethnicity? Should there be standards at all?"
  • "Should Directive No. 15 be revised? Should there be different collection standards for different purposes?"
  • "Should 'race/ethnicity' be asked as a single identification or should 'race' identification be separate from Hispanic origin or other ethnicities?"
  • "Should self-identification or the perception of an observer guide the methods for collection of racial and ethnic data?"
  • "Should population size and geographic distribution of groups be criteria in the final decision on Directive No. 15 categories?"
  • "What should the specific data collection and presentation categories be?"

In 1994, OMB established the Interagency Committee for the Review of Racial and Ethnic Standards. The members of the Interagency Committee include more than 30 agencies that represent the many and diverse federal requirements for data on race and ethnicity. The wide range of evidence to be considered by the Interagency Committee will include the results of three sample surveys and other research.

The first of these surveys is the May 1995 Supplement to the Current Population Survey (CPS) sponsored by the Bureau of Labor Statistics. Preliminary results were announced in a news release in October 1995, and a final report was published in June 1996.3 The second of these surveys is the 1996 National Content Survey (NCS) conducted by the Bureau of the Census. Results from the NCS were published in December 1996.4

The current report presents results from the third of these three surveys, the 1996 Race and Ethnic Targeted Test (RAETT).5 (Another report with more detailed findings from the RAETT is planned for publication later in 1997.) In developing specific questions on race and ethnicity to be included in Census 2000, the Census Bureau will comply with the results of the OMB review of Directive No. 15. The Census Bureau will also consider results from the NCS and RAETT that address issues other than those included in the OMB review.

In contrast to the CPS and NCS, which had sample designs that were close to being nationally representative, a targeted sample design was used in the RAETT. This design provides larger samples of six targeted populations (Black, American Indian, Alaska Native, Asian and Pacific Islander, Hispanic, and White ethnic) than does a nationally representative sample by drawing the samples from areas with high concentrations of the specified population groups.6 Consequently, the targeted samples are themselves not representative of the specified population groups. While targeted samples permit a more meaningful assessment of the effects of different questions on race and ethnicity for relatively small population groups, the results of the RAETT can be generalized only to the portions of the specified population groups residing in areas with relatively high concentrations of the targeted groups, which represent only a small proportion of each specified population group. The targeted sample design of the RAETT is described in Section 1.2 and in more detail in Chapter 3.

There are four purposes of the RAETT. They are discussed in the following paragraphs, but are listed briefly here:

  • to determine the effects of allowing respondents to report more than one race (either by marking a multiracial category and choosing to write in specific races, or by marking more than one of the race categories currently allowed under Directive No. 15);
  • to determine the effects of placing the Hispanic origin question immediately before the race question;
  • to determine the effects of collecting information about race, Hispanic origin, and ancestry in combined, two-part question; and
  • to test alternative terminologies, classifications, and formats in the race question.

The first purpose of the RAETT is to determine the effects of allowing respondents to report more than one race in a self-identification context and to assess the utility of these data for legislative and programmatic needs. The RAETT permitted respondents to report race using a separate multiracial category with write-in lines, or by using a "mark one or more" or a "mark all that apply" approach.

Consideration of these options reflects, in part, the increasing racial and ethnic diversity of the population. Large-scale immigration during the past two decades, particularly from Latin America and Asia, and an increase in interracial marriages have led to a higher proportion of the population being of mixed race or ethnicity. Census data show that there were about 1.5 million interracial couples in 1990. In all but 8 percent of these interracial couples, one spouse (or unmarried partner) was White. In 14 percent of interracial couples, the other 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 also show that the number of children in interracial families increased from less than 500,000 in 1970 to about two million in 1990.7

On the other hand, there is concern that an option to report more than one race would disrupt the historical continuity of the data. Historical continuity of racial and ethnic data is important to many data users, due in part to the increased use of these data in federal legislation, in program development and implementation, and in analyzing changing social and economic conditions of the population. In brief, "These data are required for redistricting for congressional and state elections, for enforcement of federal, state, and local civil rights statutes, for allocation of funds and administration of programs at every level of government and for many related purposes."8 In addition, members of some racial and ethnic communities have argued that an option to report more than one race would provoke internal dissension within their communities and would reduce the official counts of racial and ethnic populations about whom data have been collected using the standards of Directive No. 15.

The second purpose of the RAETT is to determine the effects of placing the Hispanic origin question immediately before the race question. In the 1980 and 1990 censuses, in which the race question preceded the Hispanic origin question, a substantial proportion of Hispanics did not identify with any of the racial categories and reported in the "Other race" category.9 Results from the NCS, and other studies show that placing the Hispanic origin question first reduces nonresponse to the Hispanic origin question and reduces the proportion of Hispanics reporting as "Other race." However, some people have expressed concern that a question about a single ethnic group would be asked before a question intended to provide racial identification for all persons. (Hispanics currently represent about 11 percent of the total population, and the Census Bureau projects that they will represent between 22 percent and 26 percent of the total population in 2050.10

The third purpose of the RAETT is to determine the effects of collecting information about race, Hispanic origin, and ancestry in a combined, two-part question.11 Research shows that some Hispanics expect to see "Hispanic" as a response option to the race question, while other Hispanics wish to identify themselves with one of the race categories as well as with Hispanic origin.12 In the two-part question tested in the RAETT, Part A included response categories for reporting race and Hispanic origin,13 and Part B provided write-in lines to report ancestry. One concern about this approach is whether responses to the write-in lines would be as complete for the detailed Asian and Pacific Islander and Hispanic origin groups as when they are listed explicitly in separate questions on race and Hispanic origin.

The fourth purpose of the RAETT is to test alternative terminologies, classifications, and formats in the race question. Examples of terminology issues include Indian (Amer.) versus American Indian. Hawaiian versus Native Hawaiian is a classification issue. An example of formatting issues is whether or not to alphabetize the listing of Asian and Pacific Islander groups in the race question. Chapter 2 provides a more detailed discussion of the purposes of the RAETT.

The RAETT included eight different panels (with eight different questionnaires) that are labeled with letters A through H. The major differences among these eight panels are shown in Table 1-1 and reflect the first three purposes of the RAETT. Panel G differs from Panel B only in the terminologies, classifications, and formats used in the race question, and is included in the analysis only to test these issues. Differences among the panels, including differences in question wording, are shown in more detail in Table 3-2 in Chapter 3. Each of the eight panels was included in each of the six targeted samples (described in Section 1.2) in the survey, except that only Panels A, B, and D were included for the Alaska Native targeted sample. Facsimiles of the questions on race and ethnicity as they appear in each panel are shown in Appendix A.

Table 1-1. Overview of the Eight Survey Panels (Questionnaires)

Panel Multiracial category Race instruction Sequence of race and Hispanic origin questions
A . . . . . . . . . . . . No Mark one Hispanic origin first
B . . . . . . . . . . . . Yes Mark one Hispanic origin first
C . . . . . . . . . . . . No Mark one or more Hispanic origin first
D . . . . . . . . . . . . Yes Mark one Race first
E . . . . . . . . . . . . Yes Mark one Combined
F . . . . . . . . . . . . No Mark one or more Combined
G . . . . . . . . . . . . Yes Mark one Hispanic origin first
H . . . . . . . . . . . . No Mark all that apply Hispanic origin first

1.2 Overview of Survey Design and Methodology

A brief description of the RAETT survey design and methodology is given here to provide a context for the findings, which are presented in the following sections of this chapter.

1.2.1 Survey Design

The RAETT was a mail-out/mail-back survey of households. Because not all households in the survey completed and returned their questionnaires (the overall mail response rate was about 53 percent), the results discussed in this report represent and can be generalized only to responding households.

The RAETT sample of 112,100 households was drawn from census tracts, American Indian reservations, and Alaska Native villages that the 1990 census showed to have high proportions (relative to the nation as a whole) of households in one of six specified racial or ethnic groups: Black, American Indian, Alaska Native, Asian and Pacific Islander, Hispanic origin, or White ethnic.14 (Separate targeted samples were included for American Indians and Alaska Natives because of different issues concerning terminology, as discussed in Section 2.2.4.) For each of these specified population groups, the census tracts that satisfied the "high proportion" criterion became a sampling frame from which a sample of households was selected. The samples are referred to as "targeted" samples because of this design. For example, the Black targeted sample included census tracts that had a relatively high proportion of Black households (in addition to some households of the other specified population groups).

The design of the RAETT used targeted samples because more nationally representative samples, like those used in the CPS and the NCS, do not provide samples of sufficient size to permit the analysis of possible effects of changes in questions on race and ethnicity for smaller population groups. Examples of smaller population groups for which targeted samples are particularly important in obtaining larger samples include American Indian, Alaska Native, and detailed groups within the Asian and Pacific Islander and the Hispanic populations.

Because of the targeted sample design in the RAETT, the results presented here would not be generalizable to the national population, even if all households had completed and returned their questionnaires. More specifically, even with a 100 percent response rate, the results of the RAETT for each targeted sample would generalize only to the population in census tracts in each targeted sample frame. The sample frames represent no more than 15 percent of each specified population group in the nation as a whole. More specifically, the sample frames represent 3 percent or less of American Indian, Asian and Pacific Islander, and White ethnic households, 8 percent of Alaska Native households, 10 percent of Black households, and 15 percent of Hispanic households.

While the results from the RAETT may in some cases be similar among different targeted samples, the sample design of the RAETT does not permit results for different targeted samples to be combined. Because the data can only be analyzed within each targeted sample, the results are reported separately for each one. Within each targeted sample, the results focus on the racial or ethnic group targeted: the Black population in the Black targeted sample, the American Indian and Alaska Native population in the American Indian targeted sample and in the Alaska Native targeted sample, the Asian and Pacific Islander population in the Asian and Pacific Islander targeted sample, the Hispanic population in the Hispanic targeted sample, and the White population in the White ethnic targeted sample. For a more detailed discussion of the sample design of the RAETT, see Chapter 3 and Appendix B.

1.2.2 Survey Methodology

The RAETT was designed to test effects of both a multiracial category and new instructions to mark more than one race category. The four specific race categories currently allowed under Directive No. 15 and "Other race" were included in the race questions tested. Hispanic origin was included among these racial categories in Panels E and F. In Panels B, D, E, and G, an explicit multiracial category was added. New instructions were added to the race question in Panels C and F to "Mark one or more races" and in Panel H to "Mark all that apply."

Early review of the data showed that respondents had marked more than one race category in the panels where they were instructed to only "Mark ONE box." This phenomenon occurred in Panel A, which did not include any of the experimental options, and in Panels B, D, E, and G, which provided a multiracial category. It became apparent that reporting more than one race could not be attributed solely to the experimental treatments, hereafter referred to as unrequested multiple responses.

In this report, rates of reporting more than one race, and the effects of those responses on race distributions, include instances in which the respondent marked more than one race category regardless of the instructions or chose the multiracial category. Differences in the rates of reporting more than one race across panels were used to measure the effects of the experimental treatments.

In order to determine whether observed differences between panels could be due to sampling error, confidence intervals at the 90-percent significance level are used. If observed differences are greater than might be expected due to sampling error, the differences are said to be statistically significant at the 90-percent confidence level. This statement means that there is less than a 10 percent chance that the observed difference would have been due to sampling error alone. All differences or effects reported in the text are statistically significant at the 90-percent confidence level. Statements that a treatment had no effect indicate that such differences were not statistically significant at the 90-percent confidence level. For more information on sampling error in the RAETT, see Section 3.4.

1.3 Research Results from the Targeted Samples: Options for Reporting More Than One Race

As noted earlier, the first purpose of the RAETT is to determine the effects of allowing respondents to report more than one race in a self-administered context, as distinguished from the respondent providing the information to an interviewer as was the case in the CPS supplement. The presentation of findings requires distinguishing several terms for clarity in this report. The term "reporting more than one race" is used as a general term to refer to three different ways that respondents in the RAETT reported more than one race. First, they could mark a response box labeled multiracial (Panels B, D, E, and G), and choose to write in specific races. Second, they could mark two or more response boxes in the race question in response to an instruction to "mark one or more" or "mark all that apply" (Panels C, F, and H). Third, some respondents provided multiple responses even when instructed to "mark one" (Panels A, B, D, E, and G).

The following bullets summarize the most important results of the different options for reporting more than one race. (In Panels E and F, Hispanic origin is treated as a race group.) The racial and ethnic distributions presented are meaningful only in relationship to one another within the experimental design of the RAETT and only within each targeted sample (Black, American Indian, Alaska Native, Asian and Pacific Islander, Hispanic, and White ethnic). These distributions will not correspond to the distributions reported in other surveys or censuses, and hence such comparisons should not be made. Recall also that the findings from the RAETT's targeted samples cannot be generalized to the entire targeted population, only to the portion of it living in areas with relatively high concentrations of the targeted group.

  • Some respondents provided unrequested multiple responses, despite instructions in those panels to mark one box. These responses occurred in panels with and without a multiracial category. In the Alaska Native targeted sample, these responses nearly approached the percentage of those who selected the multiracial category. Unrequested multiple responses were also substantial in the Asian and Pacific Islander targeted sample.
  • The options for reporting more than one race did not affect the percentages reporting solely as White, as Black, or as American Indian in their respective targeted samples. These options elicited the highest reporting of more than one race and had their most important effects in the Alaska Native and the Asian and Pacific Islander targeted samples.
  • In the Asian and Pacific Islander and the Alaska Native targeted samples, the multiracial category affected reporting in the single race category of Asian and Pacific Islander, and of American Indian and Alaska Native, respectively.
  • The "mark all that apply" option also affected the reporting of Asian and Pacific Islander as a single race in that targeted sample, but the "mark one or more" option did not.
  • When the race and Hispanic origin questions are combined, a high percentage of responses included both Hispanic origin and one of the four major race categories currently allowed under Directive No. 15.
  • Using an illustrative procedure for classifying responses of more than one race into the current OMB race categories eliminated most statistical effects of the multiple race reporting options on reporting as Asian and Pacific Islander in that targeted sample, but only reduced the effects in the Alaska Native targeted sample on reporting as American Indian and Alaska Native.

The next three subsections provide more detailed results on the reporting of race in each of the targeted samples. Specifically:

  • Section 1.3.1 describes the combinations of races that were reported in write-ins to a multiracial category or by marking more than one box.
  • Section 1.3.2 summarizes the racial distributions that would result under three of several possible classification procedures that could be used to tabulate responses of more than one race. These procedures provide illustrative examples of racial and ethnic classifications that could be used as a bridge between the old and new classification schemes if an option for reporting more than one race were to be adopted in a revised Directive.
  • Section 1.3.3 presents the statistical effects that the options for reporting more than one race had across each of the targeted samples. Specific panels within each targeted sample are compared to determine the effect of each option on the reporting of race and Hispanic origin.

1.3.1 Responses to the Questions on Race

The reporting of race in each of the targeted samples is shown in the top half of Tables 1-2 through 1-7.15 (These tables are shown at the end of Chapter 1.) In each targeted sample, the proportion of responses for the targeted group alone is shown in bold because this group is the focus in that sample and in our analyses in subsequent sections. The tables also show the proportion of responses in each of the other major race categories, as well as the proportion who reported in the targeted race category in combination with other race categories. These racial and ethnic distributions are representative neither of the general population, nor of the targeted populations. They reflect the responses that were given to the race questions in the different panels of the RAETT, and are provided for descriptive purposes only. Statistical comparisons of the effects of each option for reporting more than one race are given in Section 1.3.3.

As noted earlier, respondents could report more than one race either by marking more than one race response box, or, on some panels, by marking a separate multiracial response box and choosing to write in specific races. Some respondents marked more than one box even when instructed to mark only one, and even when they were given a separate multiracial response box. As noted below, the proportion of "unrequested" multiple responses was sometimes quite high and approached the proportion of responses in the multiracial category. The entries in the tables labeled "reporting more than one race" include all these ways of responding.

Across the targeted samples, the proportion of responses of more than one race ranged from less than 3 percent (in the White ethnic and the Black targeted samples) to about 13 percent (in the Asian and Pacific Islander and the Alaska Native targeted samples). Multiple responses were even higher (17 percent or more) in the Hispanic targeted sample on the panels with the combined race and Hispanic origin questions. These responses usually involved marking both the Hispanic box and one race box.

The Black targeted sample

Less than 3 percent of the race responses in the Black targeted sample included more than one race (Table 1-2).

The American Indian targeted sample

Between 2 percent and 5 percent of the race responses in the American Indian targeted sample included more than one race on the panels with separate race and Hispanic origin questions (Table 1-3). About 7 percent of the responses indicated more than one race in the combined race and Hispanic origin question on Panel F. About one-fourth of these responses included the marked boxes for American Indian and Alaska Native and for Hispanic.

The Alaska Native targeted sample

In the Alaska Native targeted sample, reporting more than one race represented about 9 percent of the responses to the race question in Panel D, and about 13 percent in Panel B (Table 1-4). Providing unrequested multiple responses was relatively high, representing about 5 percent and 6 percent of the responses in Panels A and B, respectively. These three panels were the only ones administered in the Alaska Native sample, because this was one of the more difficult populations to target, and dividing the sample across more panels would have compromised the ability of the RAETT to detect the effects of any of the experimental treatments.

Most of the respondents in the Alaska Native sample who reported more than one race on Panel B (12 percent out of a total of 13 percent) provided a write-in that included an American Indian and Alaska Native entry; these were most frequently combined with an "Other race" or multiracial category entry (about 6 percentage points), with a White entry (about 5 percentage points), or with an Asian or Pacific Islander entry (about 1 percentage point). The pattern was similar on Panel D: 8.0 percentage points of the total of 8.7 percent who reported more than one race provided an American Indian or Alaska Native write-in, most frequently with a White entry (about 4 percentage points), an Asian or Pacific Islander entry (about 2 percentage points), or an "Other race" or "multiracial" entry (about 1 percentage point).

The Asian and Pacific Islander targeted sample

The proportion of responses in the Asian and Pacific Islander targeted sample that included more than one race was between about 4 percent in Panel A and about 12 percent in Panel D (Table 1-5). The most frequent responses of more than one race across all panels included an Asian and Pacific Islander entry combined with a White, an "Other race" or a multiracial category entry.

The Hispanic targeted sample

On panels with separate race and Hispanic origin questions, reporting more than one race on the race question in the Hispanic targeted sample ranged from about 2 percent to about 5 percent (Table 1-6). Reporting of more than one race was substantially higher (18 percent to 19 percent) on the panels with the combined race and Hispanic origin questions (Panels E and F), primarily because of responses that included both a race and a Hispanic origin. The overwhelming majority of the multiple responses in Panels E and F involved reporting as Hispanic and as White.

The multiple responses on Panel E were almost entirely unrequested multiple responses; only about 1 percent of the responses involved the multiracial category. These unrequested multiple responses suggest that even if the option is not offered, a relatively high percentage of Hispanics will use multiple marks to report both a race and a Hispanic origin when they do not have separate questions that enable them to do so.

The White ethnic targeted sample

About 2 percent or less of the responses in the White ethnic targeted sample included more than one race (Table 1-7).

1.3.2 Illustrative Approaches to Classifying Data on Race

Presenting data from multiple responses is inherently more complex than doing so when all responses represent single categories. This section describes three different approaches to tabulating the multiple responses given in the RAETT. We stress that these approaches are presented for illustrative purposes only, and that their presentation does not imply that these or any other approaches should be adopted for classifying the reported data. Instead, these methods simply provide examples of classification procedures that could be developed as bridges between the current and any new classification scheme. The examples might help in indicating the feasibility of constructing such bridges, as well as challenges that might have to be addressed. If some option for reporting more than one race were to be adopted in a revised Directive, then some approach for classifying these responses into the current OMB categories might be useful for some data users.

The distributions generated by these three classification schemes are shown on the bottom of Tables 1-2 through 1-7.

Single race approach

The first of these three approaches is called the "single race" approach. This approach would classify responses that reported only one of the four races specified in Directive No. 15 or "Other race" into the designated category. Reports of more than one race (however they reported) would be classified into a separate "multiple race" category. This approach would classify each response into one and only one race category. For example, a response of White and Asian would be classified as "multiple race", while one of Asian only would be classified as Asian and Pacific Islander.

The racial distribution under the "single race" approach would approximate the distribution which might result from the addition of a "multiple race" category to those allowed under Directive No. 15. Since the "multiple race" category could draw from any of the four major race categories, the percentages for the major race categories under the "single race" option could be smaller than the percent that would report under current Directive No. 15, without a "multiple race" category.

Historical series approach

The second approach is called the "historical series" classification. A historical series might be useful to some data users, including federal agencies that use data on race and ethnicity to monitor and enforce civil rights legislation, because it emphasizes classification into the race categories that have been used to monitor changes under extant legislation.

It is important to stress that there is no reason to believe that the historical series approach would classify all respondents who reported more than one race into the single race category that they would select if required to report only one race. For example, a respondent who reported White and Chinese would be classified in the historical series approach in the Asian and Pacific Islander category; however, it is possible that this respondent would report as White or as "Other race" if required to report in only one race category. The classifications used for the historical series approach would include:

  • White -- White only, White and "Other race," White and multiracial category (no additional races specified);
  • Black -- Black only, Black and White, Black and "Other race," Black and multiracial category (no additional races specified);
  • Asian and Pacific Islander (API) -- API only, API and White, API and "Other race," API and multiracial category (no additional races specified);
  • American Indian and Alaska Native (AIAN) -- AIAN only, AIAN and White, AIAN and "Other race," AIAN and multiracial category (no additional races specified);
  • Other -- Other race; and
  • Multiple race -- multiracial category (no additional races specified); combinations of two race categories other than White and "Some other race;" and three or more race categories.

For Panels E and F, in which the race and Hispanic origin questions were combined into one question, Hispanic origin is included in the list as though it is a race group, and is treated accordingly. For these panels the Hispanic classification is defined as:

  • Hispanic -- Hispanic only, Hispanic and White, Hispanic and "Other race," Hispanic and multiracial category (no additional races specified).

The "historical series" approach classifies each response into one and only one classification in a set of mutually exclusive and exhaustive categories that add to 100 percent. For example, the historical series approach would classify a response of Black and White in the category for Blacks, and classify a response of Black and American Indian in the category for multiple race reporting.

All- inclusive approach

The third approach is called the "all-inclusive" approach. This approach classifies each response into all of the race or Hispanic origin categories reported. As a result, a response reporting a number of racial categories would be counted in each of them, and the sum of the percentages for each of the racial categories currently allowed under Directive No. 15 and "Other race" could exceed 100 percent. For example, a response of Black and American Indian would be tabulated as Black and also be tabulated as American Indian. For any given targeted sample and panel, this method yields the largest percentage of the targeted racial or ethnic group. For the Hispanic category in Panels E and F, with a combined race and Hispanic origin question, the all-inclusive category would be most comparable to the Hispanic results from the other panels, which used two separate questions.

Results of the illustrative approaches

The results of using these different approaches in each targeted sample are shown in the bottom of Tables 1-2 through 1-7. Recall again that the distributions based on these approaches are descriptive of these targeted samples only and are not representative of the targeted populations.

Historical series distributions are compared statistically in Section 1.3.3 to help analyze the effects of the different options for reporting more than one race.

In the Black targeted sample, the percentage of responses reporting Black as a single race ranged from 71 percent in Panel D to about 74 percent in Panel G (Table 1-2); it was 72 percent on Panel A. The percentages tabulated as Black using the historical series approach ranged from 72 percent to 74 percent, while the percentage tabulated as Black in the all-inclusive approach ranged from 72 percent to 75 percent.

In the American Indian targeted sample, the proportion of responses that reported American Indian or Alaska Native as a single race was 36 percent on Panel A, and ranged from about 33 percent in Panel B to 37 percent in Panel C (Table 1-3). The percentages of American Indian and Alaska Native responses tabulated using the historical approach and the all-inclusive approach both ranged from about 36 percent to about 41 percent in this targeted sample.

In the Alaska Native targeted sample, the proportion of responses of American Indian or Alaska Native as a single race ranged from 65 percent in Panel B to 79 percent in Panel A (Table 1-4). Tabulating responses in the Alaska Native targeted sample according to the "historical series" described above would assign to this category 84 percent of the responses in Panel A, 76 percent of the responses in Panel B, and 75 percent in Panel D. The all-inclusive approach yielded very similar results.

In the Asian and Pacific Islander targeted sample, the single race reporting of Asian and Pacific Islander ranged from about 58 percent in Panel E to about 65 percent in Panel A (Table 1-5). The historical series approach classified as Asian and Pacific Islander about 62 percent of the responses in Panel E, about 68 percent the responses in Panel A, and about 73 percent of those in Panel C. The corresponding percentages from the all-inclusive approach were about 63 percent (Panel E), 68 percent (Panel A), and 74 percent (Panel C).

In the Hispanic targeted sample, the major focus is on reporting of ethnicity. A combined race and Hispanic origin question must, of necessity, produce lower reporting in the Hispanic category or in one of the major race groups than separate race and Hispanic origin questions. On separate questions, many of the responses include both a Hispanic and a race entry, so the total is more than 100 percent. Consequently, the most similar tabulations are the all-inclusive approach (which classified about 75 percent as Hispanic on Panels E and F) and the percent reporting as Hispanic on the separate question on the other panels (between 74 percent and 78 percent). The historical series classified about 74 percent of the responses on both Panels E and F as Hispanic.

In the White ethnic targeted sample, the percentage reporting White as a single race ranged between about 94 percent (Panel E) and 97 percent (Panel G). The percentages tabulated as White using the historical series approach and the all-inclusive approach were similar, ranging between 96 and 99 percent (Table 1-7).

1.3.3 Effects of the Options for Reporting More Than One Race in the Targeted Samples

The effects of the multiracial category were determined primarily by statistically comparing responses to Panel B with those to Panel A. Although Panels D and G included a multiracial category, they also included other changes, so the A and B comparison shows the effect of the multiracial category without confounding effects. Panel C was compared to Panel A to determine the effects of the "mark one or more" instruction. Panel H was compared to Panel A to determine the effects of the "mark all that apply" instruction.

As noted earlier, these targeted samples were designed to test more adequately the effects of reporting more than one race. Often, nationally representative samples cannot provide a sufficiently large sample of relatively small population groups to determine the possible effects of these options.

In what follows, statistically significant differences between panels in the percentages reporting in a particular category indicate that data for that targeted population may be affected by the option for reporting more than one race. However, one cannot determine from these results the true size of the effect in any of these populations because the targeted samples are neither representative of the populations targeted, nor of the population in general. The results of the statistical comparisons of panels in the RAETT must be interpreted only within each particular targeted sample and with these qualifications in mind.

General findings

The results from the RAETT revealed that neither the multiracial category (Panel B) nor the multiple response options (Panels C and H) had any statistically significant effects in the White ethnic, Black, or American Indian targeted samples on the percentages who reported a single race of White, Black, or American Indian and Alaska Native, respectively. In contrast, the results indicated that the multiracial reporting options affected single-race reporting in the Asian and Pacific Islander and the Alaska Native targeted samples. There was evidence that including a multiracial category on the race question statistically affected the reporting in detailed Hispanic origin groups, but not the total reporting of Hispanic.

These results are described in more detail below. Findings are reported for targeted samples where the options for reporting more than one race had statistically significant effects. The detailed tables in Appendix C provide more information, including standard errors of estimates and confidence intervals for differences in estimates between pairs of panels for all targeted samples.

Effect of a multiracial category on race reporting

Adding a multiracial category to the race question (Panel B) reduced the reporting of Asian or Pacific Islander as a single race in the Asian or Pacific Islander targeted sample from 65 percent (in Panel A) to 60 percent (Table 1-8). Much of this decline is attributable to a drop in reporting as Hawaiian, from about 9 percent in Panel A to about 6 percent on Panel B. At the same time, reporting of more than one race increased from 4 percent in Panel A to 11 percent in Panel B.

In the Alaska Native targeted sample, the reporting of American Indian or Alaska Native as a single race was also lower in Panel B (65 percent) than in Panel A (79 percent) (Table 1-4). In the American Indian targeted sample, however, the multiracial category had no statistically significant effect on the reporting of American Indian or Alaska Native (Table 1-3).

Effect of the multiple response options on race reporting

One of the multiple response options affected the distribution of races reported in the Asian and Pacific Islander targeted sample, but neither option had an effect on reporting of the targeted populations in any of the other targeted samples in which these options were tested. In the Asian and Pacific Islander targeted sample, the "mark all that apply" (Panel H) instruction reduced reporting of Asian and Pacific Islander as a single race from 65 percent in Panel A to 58 percent (Table 1-8). In contrast, the "mark one or more" instruction (Panel C) did not affect the total percentage of responses with a single race of Asian and Pacific Islander (about 65 percent). Reporting as Hawaiian dropped on both panels offering multiple response options. However, this decline was counterbalanced by the higher percentage in the "Other Asian and Pacific Islander" category in Panel C, which included the instruction to "mark one or more," and thus did not affect the total percentage reporting as Asian or Pacific Islander. The corresponding increase in the "Other Asian and Pacific Islander" category on Panel H (about 8 percent), which included the instruction to "mark all that apply," only partially counterbalanced its decline in reporting as Hawaiian.16

In summary, of the three options for reporting more than one race, only the instruction to "Mark one or more..." had no effect on the total reporting as Asian and Pacific Islander in that targeted sample. Therefore, this option may be the one least likely to affect the historical continuity of data on race and ethnicity that some federal agencies use to monitor and enforce civil rights. However, as noted earlier, neither multiple response option was tested in the Alaska Native targeted sample, where unrequested multiple responses and selection of the multiracial category had effects on reporting as American Indian and Alaska Native.

Effects of the options for reporting more than one race on reporting of Hispanic origin

In the Hispanic targeted sample, nonresponse to the Hispanic origin question was higher in Panel B (about 8 percent), which included a multiracial category on the race question, than in Panel A, which did not (about 6 percent) (Table 1-9). The multiple response options did not affect nonresponse to the Hispanic origin question.

None of the options for reporting more than one race affected the total percentage of responses of Hispanic to the separate Hispanic origin question. However, reporting as Mexican was higher in Panel B than in Panel A, and reporting as Puerto Rican was lower in both Panel B and Panel H, which offered the option to "mark all that apply," than in Panel A (Table 1-9).17

Comparison of alternative racial classification procedures

Statistical comparisons of the historical series approach were made for panels in targeted samples that showed statistically significant differences in single race reporting. These comparisons were conducted to determine whether the differences remained when some responses reporting more than one race were classified in the current OMB race categories and "Other race" using the historical series approach. That approach serves here only as an illustration of the many alternative racial classification procedures that could be developed for analytical purposes.

For the Asian and Pacific Islander targeted sample, the historical series approach eliminated the statistically significant differences in the proportions of responses classified as Asian and Pacific Islander when Panel B and Panel H were compared to Panel A (Table 1-5). However, the proportion of responses classified as Asian or Pacific Islander using the historical approach was significantly higher in Panel C than in Panel A.

Similar analyses were conducted in the Alaska Native targeted sample. Although the historical series classification reduced, it did not eliminate the statistically significant differences in the proportions of responses classified as American Indian and Alaska Native on Panels B and D compared with Panel A (Table 1-4).

1.4 Research Results from the Targeted Samples: Effects of Placing the Hispanic Origin Question Immediately Before the Race Question

The second purpose of the RAETT is to determine the effects of placing the Hispanic origin question immediately before the race question. These effects were determined by comparing Panel D (race question first) and Panel B (Hispanic origin question first). Both of these panels also contained a multiracial category.

The analyses in this section focus on the Hispanic targeted sample, and the findings are reported for this sample only. However, the detailed tables in Appendix C show the comparisons of these two panels for all targeted samples.

Consistent with prior research, the analyses of the Hispanic targeted sample in the RAETT showed that placing the Hispanic origin question before the race question reduced, but did not eliminate nonresponse to the Hispanic origin question and reporting in the "Other race" category on the race question.18 Further details are given below.

Effect of the sequence of questions on nonresponse to the race and Hispanic origin questions

Asking the Hispanic origin question before the race question reduced the nonresponse rate to the Hispanic origin question in the Hispanic targeted sample from about 10 percent in Panel D (where the race question preceded the Hispanic origin question) to about 7 percent on Panel B (Table 1-9). The Hispanic first sequence had no effect on the nonresponse rate to the race question in the Hispanic origin targeted sample.

Effect of the sequence of questions on race reporting

In the Hispanic targeted sample, asking the Hispanic origin question before the race question reduced reporting of more than one race and as "Other race," and increased reporting as "White." In Panel D (race question first), about 56 percent of responses were White, about 25 percent were "Other race," and about 5 percent reported more than one race (Table 1-10). In contrast, in Panel B (Hispanic origin question first), about 67 percent reported as White, about 16 percent reported as "Other race," and about 4 percent reported as more than one race.

1.5 Research Results from the Targeted Samples: Effects of Combining the Questions on Race and Hispanic Origin

The third purpose of the RAETT is to determine the effects of collecting information about race, Hispanic origin, and ancestry in a combined, two-part question. The RAETT tested two versions of a combined question. Both provided response boxes for the current OMB race groups, for Hispanic origin, and for "Some other race." Both also included a write-in line for American Indian or Alaska Native tribe. One version (Panel E) included a multiracial category, while the other (Panel F) included an instruction to "mark one or more." Both were followed by Part B of the question, which asked respondents to report their "ancestry or ethnic group" on write-in lines provided for that purpose. One objective of the ancestry write-in was to determine how detailed Asian and Pacific Islander and Hispanic origin groups would be reported.

The effects of combining the race and Hispanic origin questions were determined by comparing Panels E and F with the corresponding panels that included separate race and Hispanic origin questions -- Panel E with B for the multiracial category, and Panel F with C for the "mark one or more" option. The detailed tables in Appendix C show the results of the statistical tests for these panel comparisons for all targeted samples. As noted earlier, the statistical comparisons of panels in the RAETT are meaningful only within each targeted sample. Significant differences between panels indicate that reporting in particular categories may be affected by combining the race and Hispanic origin questions. However, one cannot determine from these results the true size of the effect in any of these populations because the targeted samples are neither representative of their populations, nor of the population in general. The section focuses on results for the Hispanic targeted sample.

The most important results include:

  • In every targeted sample, nonresponse to each combined question (Panels E and F) was significantly lower than nonresponse to the corresponding separate Hispanic origin and race questions (Panels B and C).
  • The combined race and Hispanic origin questions elicited high levels of multiple responses in the Hispanic targeted sample on both Panels E and F (17 percent and 19 percent). An additional 1 percent of responses in Panel E were for the multiracial category. Over 90 percent of the multiple responses involved Hispanic origin and a race group.
  • When all responses of Hispanic (either Hispanic alone or Hispanic in combination with any other response) are added together, which is the appropriate comparison, there is no statistically significant difference in the percent reporting Hispanic between a combined question and separate questions on Hispanic origin and race.
  • The ancestry write-ins to the two-part, combined race, Hispanic origin, and ancestry question on Panel E did not provide percentages of either the detailed Asian or Pacific Islander groups or of the detailed Hispanic groups in the respective targeted samples for these populations comparable to those provided by the separate race and Hispanic origin questions that list these groups separately (Panel B). In contrast, the write-ins to the ancestry component of the combined question on Panel F provided a distribution of detailed Asian or Pacific Islander groups in the Asian or Pacific Islander targeted sample statistically similar to that on the corresponding separate race question (Panel C) where these groups are listed separately, but did not do so for the detailed Hispanic groups.
Effect of a combined question on race and Hispanic origin on the reporting of Hispanic Origin

A combined race and Hispanic origin question must, of necessity, produce lower reporting in the Hispanic category or in one or more of the major race groups than separate race and Hispanic origin questions. On separate questions, many of the responses include both race and Hispanic origin entries, so that the total is more than 100 percent. Because the all-inclusive approach uses all information reported to classify responses as Hispanic or as not Hispanic, it could provide similar data to that which would be obtained if separate questions on race and Hispanic origin were used. Therefore, the most direct comparison of Hispanic reporting in the separate question is the all-inclusive approach in the combined question.

There were no statistical differences between the all-inclusive approach for the combined panels with the appropriate panels containing separate questions. 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. If one were to classify as Hispanic only single responses of Hispanic (responses that did not include any multiple entries), then a much lower proportion (about 57 percent) of responses would be Hispanic in Panels E and F.

Effect of a combined question on race and Hispanic origin on the reporting of race

Reporting as White alone (94 percent) was lower on the combined question (Panel E) than on the corresponding separate race question (Panel B, about 97 percent) in the White ethnic targeted sample (Table 1-7). However, there were no significant differences between Panels C and F in reporting as White alone in this targeted sample, with both about 96 percent. In addition, both combined questions reduced reporting as Asian and Pacific Islander in the Asian and Pacific Islander targeted sample from 60 percent in Panel B to 58 percent in Panel E, and from about 65 percent in Panel C to 60 percent in Panel F (Table 1-5).

In summary, a combined race and Hispanic origin question will produce lower reporting in the Hispanic category alone or in one of the major race groups alone than separate race and Hispanic origin questions where responses as Hispanic and in the race categories can total more than 100 percent. The pattern of declines in reporting as White and as Asian or Pacific Islander on the combined questions suggests that the pool of persons who report as Hispanic and in these two race groups on separate questions split on a combined question, with some reporting as Hispanic, some in the race group, and some marking more than one, even when this option is not offered. In contrast, the absence of significant declines in reporting as Black and as American Indian in the respective targeted samples for those populations suggests that people in those targeted samples who report as Hispanic on a separate Hispanic origin question are relatively small in number or are more likely to report their race than their Hispanic origin on a combined question.

Nonresponse to the two-part combined question on race, Hispanic origin, and ancestry

In every targeted sample, the nonresponse rates to the combined question were considerably lower than the nonresponse rates to the separate questions on race and Hispanic origin (Table 1-11). In the Hispanic targeted sample, the nonresponse rate to the race question was about 13 percent in both Panels B and C compared with about 1 percent in Panels E and F. Nonresponse to the Hispanic origin question was over 6 percent on Panels B and C in the Hispanic targeted sample.

Reporting of detailed Asian and Pacific Islander and detailed Hispanic groups in the ancestry write-ins

An important question for evaluating the two-part, combined race and Hispanic origin questions (Panels E and F) is whether "ancestry or ethnic group" write-ins in Part B provided data on the detailed Asian and Pacific Islander groups and the detailed Hispanic origin groups comparable to the data from the separate race and Hispanic origin questions.19

About 4 percent of the Asian and Pacific Islander responses in that targeted sample did not include write-ins to the ancestry component of the combined question on Panel E, which included a multiracial category. The percentages obtained from the ancestry write-ins for most specific Asian and Pacific Islander groups were lower on the combined race and Hispanic origin question with the multiracial category (Panel E) than on the corresponding separate race question (Panel B).20 There were significant differences in the proportion of responses for all of the detailed Asian and Pacific Islander groups using the ancestry write-in in Panel E compared with the detailed listing on a separate question with the exception of those for Hawaiian, Vietnamese, and Japanese in Panel B (Table 1-8).

About 3 percent of Asian and Pacific Islander responses in that targeted sample, did not provide write-ins to the ancestry component of the combined question on Panel F, which included the instruction to "mark one or more." The ancestry write-ins on this panel provided percentages for most specific Asian and Pacific Islander groups similar to those on the panel with the corresponding separate race question (Panel C), with the exceptions of Guamanians and of Other Asian and Pacific Islander (Table 1-8).

About 11 percent (Panel E) and about 13 percent (Panel F) of the Hispanic responses in the Hispanic targeted sample did not include write-ins to the ancestry component of the combined question. Because of these relatively high rates of nonresponse to the ancestry component of the combined question, the percentages of each specific Hispanic origin group obtained from these write-ins were consistently lower than those obtained from the response boxes on the separate Hispanic origin questions on Panels B and C (Table 1-9).21

1.6 Research Results from the Targeted Samples: Terminology, Classification, and Formatting

The fourth purpose of the RAETT is to test alternative terminologies, classifications, and formats in the race question. The RAETT included a questionnaire (Panel G) designed to test these issues. These options and the results pertaining to them are presented below. To determine the effect of these options, comparisons were made between Panels B and G, both of which included a separate multiracial category.

Spelling out "American Indian"

On the experimental panels except Panel G, the category was termed "Indian (Amer.) or Alaska Native." Panel G spelled out "American Indian or Alaska Native," but the changes in Panel G had no effect on reporting as American Indian and Alaska Native in the American Indian targeted sample (Table 1-12).

Substituting "Native Hawaiian" for "Hawaiian" and listing this category immediately after the "American Indian and Alaska Native" category

Substituting "Native Hawaiian" for "Hawaiian" and listing this category immediately after the "American Indian and Alaska Native" category increased reporting as Hawaiian from about 6 percent (Panel B) to about 7 percent in Panel G (Table 1-13 ). This increase in Panel G might reflect the use of "Native Hawaiian" instead of "Hawaiian" for the category, its placement directly after the category for "American Indian and Alaska Native" instead of in the midst of the list of Asian and Pacific Islander groups, or both changes. The percentages on Panels B and G, which both included a multiracial category, were lower, however, than the percentage reporting as Hawaiian on Panel A (9 percent), which did not include a multiracial category.

Alphabetizing the Asian and Pacific Islander groups after "Native Hawaiian"

Alphabetizing the Asian and Pacific Islander groups after "Native Hawaiian" had no effect on the total percentage reporting as Asian and Pacific Islander in that targeted sample. Reporting as Samoan was lower on Panel G (0.5 percent) than on Panel B (1.4 percent) (Table 1-13). However, the specific change in Panel G that might have produced this result is not apparent.

Using "Guamanian or Chamorro" instead of "Guamanian"

Given the very small numbers involved, no statistically significant difference was found between reporting as "Guamanian or Chamorro" on Panel G and as "Guamanian" on Panel B (Table 1-13).

1.7 Relationship to Other Research

The RAETT was one of three major surveys conducted to test alternative questions on race and ethnicity as part of OMB's review of Statistical Policy Directive No. 15. The other two were: the National Content Survey (NCS), a mail-out survey conducted in March 1996 by the Bureau of the Census to test and evaluate the full subject content for Census 2000; and the May 1995 Race and Ethnicity Supplement to the Current Population Survey (CPS), sponsored by the Bureau of Labor Statistics.

The sample for the NCS was 94,500 households drawn from census mailback areas representing about 95 percent of the country. The tests of alternative questions on race and ethnicity were primarily based on 4 of the 13 panels of the NCS. The results of these tests were based on the 18,000 households that responded on these four panels (about 72 percent of the approximately 24,000 households to which the four questionnaire versions were mailed). The CPS Supplement interviewed about 60,000 households in a Computer Assisted Telephone Interview (CATI) (about 80 percent of the sample) or Computer Assisted Personal Interview (CAPI) (20 percent of the sample).

In contrast to the CPS and NCS, which had sample designs that within some limitations were close to being nationally representative, the sample design of the RAETT targeted specific populations in order to obtain larger samples of small populations (e.g., American Indians and Alaska Natives) than nationally representative samples provide.22 The targeted samples that the RAETT provided are more likely to be sufficiently large to detect effects that different questions on race and ethnicity might have on the targeted populations. However, the results for the targeted sample can only be generalized to the areas of relatively high concentrations of the targeted populations used to select each sample. Due to the high nonresponse rate in the RAETT, even such generalizations should be made with caution.

The RAETT tested two different approaches to allowing respondents to report more than one race. The first approach, a multiracial category, was tested in the NCS and the CPS Supplement. The second approach, tested only in the RAETT, provided instructions to "mark one or more" or to "mark all that apply."

Use of the options to report more than one race was highest and had the most important effects in the Asian and Pacific Islander and in the Alaska Native targeted samples. The RAETT also found that notable percentages of respondents marked more than one box on panels (such as Panels A, B and E) that instructed them to "Mark ONE box." These unrequested multiple responses nearly approached the percentage of respondents who selected the multiracial category on Panel B in the Alaska Native targeted sample and were also substantial in the Asian and Pacific Islander targeted sample.

Given these high rates of unrequested multiple responses, the Census Bureau manually identified and tallied unrequested multiple responses on two panels of the NCS. About 0.5 percent of the responses on these two panels were unrequested multiple responses, a level consistent with informal estimates of unrequested multiple responses to the 1990 race question. This suggests that the higher rates observed in several targeted samples of the RAETT reflect the greater likelihood of unrequested multiple responses in the targeted populations, rather than an increase since 1990 in providing unrequested multiple responses.

The RAETT results were consistent in two important ways with findings from previous research on adding a multiracial category to the race question. First, the RAETT found that neither the multiracial category nor the options to mark more than one box had any statistically significant effects on the percentages of persons who reported as White or as Black in their respective targeted samples. The CPS Supplement and the NCS found that a multiracial category had no effects on reporting of race in these two categories.

Second, the RAETT found that the panels with the multiracial reporting options had significant effects on reporting by two groups for whom either the CPS or the NCS also provided some evidence of possible effects: American Indians and Alaska Natives, and Asians and Pacific Islanders. The CPS found lower reporting as American Indian, Eskimo, or Aleut on panels that included a multiracial category on the race question. The sample of this population in the NCS was too small to detect differences that the multiracial category or other treatments might have had on reporting as American Indian, Eskimo, and Aleut.

The RAETT included separate targeted samples for American Indians and, on selected panels, for Alaska Natives, and different results were found for the two samples. The RAETT found that none of the multiracial reporting options significantly affected reporting in the targeted sample for American Indians. Differences in the sample designs might contribute to the differences between the RAETT and the CPS findings for this population: rates of intermarriage and reporting as multiracial might be higher among American Indians living in urban areas (more likely to be in the CPS sample) than among those living on or near reservations (targeted in the RAETT).

The targeted sample of Alaska Natives received two of the questionnaires that included a multiracial category, and the questionnaire for Panel A. Reporting more than one race in the Alaska Native sample was among the highest observed in all the targeted samples, and contributed to statistically significant declines in reporting as American Indian and Alaska Native. The use of a combined American Indian and Alaska Native category on Panels B and D, in contrast to the separate categories for American Indians, for Eskimos, and for Aleuts on Panel A, may also have contributed to the decline.

The RAETT found that both adding a multiracial category to the race question and providing instructions to "mark all that apply" reduced reporting of Asian and Pacific Islander as a single race in the targeted sample for that population. However, providing instructions to "mark one or more" had no such effect. The NCS report noted that the declines observed in reporting as Asian and Pacific Islander on panels that included a multiracial category on the race question were not statistically significant, but that analysis of write-in responses suggested this effect might be significant in a study providing a larger sample of Asians and Pacific Islanders.

Taken together, the results from the CPS Supplement, the NCS, and the RAETT suggest that providing multiracial reporting options would not affect the percentages reporting as White or as Black, but may well affect reporting in populations with higher intermarriage rates, most notably American Indians and Alaska Natives, and Asians and Pacific Islanders.

A historical series approach to classifying based on information from the multiple responses and the write-ins eliminated most of the statistically significant effects that the options for reporting more than one race had on the percentages classified as Asian and Pacific Islander in that targeted sample. However, this approach reduced but did not eliminate the effects observed in the Alaska Native targeted sample on reporting as American Indians and Alaska Natives. Of the three options for reporting more than one race, only the instruction to "mark one or more" had no effect on the total reporting of Asian and Pacific Islander in that targeted sample; consequently, it might be the option least likely to disrupt the historical series of data on race and ethnicity that some federal agencies rely upon to monitor and enforce civil rights. This option was not tested in the Alaska Native targeted sample, however.

The RAETT also tested two very different strategies for addressing problems that some respondents seem to have in answering separate race and Hispanic origin questions: the first was sequencing the Hispanic origin question immediately before the race question, and adding an instruction on the importance of answering both questions; the second was combining the race and Hispanic origin questions into a single question. The NCS tested sequencing, and the CPS also tested a combined race and Hispanic origin question.

The RAETT results on placing the Hispanic origin question first were consistent with prior research, including the NCS. The Hispanic origin-first sequence reduced nonresponse to the Hispanic origin question, without affecting rates of response to the race question.23 It also reduced, but did not eliminate, reporting as "Other race" by Hispanics.

The combined race and Hispanic origin questions had far lower nonresponse rates in every targeted sample than did the corresponding separate race and Hispanic origin questions (Panels B and C), including those on which the Hispanic origin question was placed first. However, in the Hispanic targeted sample, reporting of more than one race was high (over 18 percent) on the combined panels, and this substantially reduced reporting of Hispanic as a single category (from about 75 percent to about 57 percent). The May 1995 CPS Supplement found a comparable decline (about 20 percent) in reporting as Hispanic on the panels that included a combined race and Hispanic origin question.

However, on separate questions, responses include both race and Hispanic origin entries, and total more than 100 percent. Reporting in the Hispanic category alone or in one or more of the major race groups alone will necessarily be lower on a combined question where responses can total only 100 percent. The all-inclusive approach uses all information reported to classify responses as Hispanic or not as Hispanic and therefore provides data more comparable to that obtained from separate questions on race and Hispanic origin. There were no statistical differences between the all-inclusive approach for the combined panels with the appropriate panels containing separate questions. In the May 1995 CPS Supplement, the percentage of Hispanics increased with the use of responses to follow-up questions to the combined question, but never equaled the percentage reporting as Hispanic on the separate race and ethnicity questions.

The write-ins to the ancestry component of the combined question with the instruction to "mark one or more boxes" in the Asian and Pacific Islander targeted sample provided percentages of the detailed Asian and Pacific Islander groups comparable to those provided by the separate race question that lists these groups separately. However, it did not do so for the detailed Hispanic origin groups, in part because nonresponse to the ancestry component was relatively high (over 10 percent) among those who marked the Hispanic category in the Hispanic targeted sample. When the combined race and Hispanic origin question included a multiracial category, the write-ins to the ancestry question provided lower percentages of several detailed Asian and Pacific Islander groups, as well as of the detailed Hispanic origin groups, in their respective targeted samples than did the separate race and Hispanic origin questions.

A combined race and Hispanic origin question that allows multiple responses thus may be able to provide estimates of the total Hispanic origin population comparable to a separate Hispanic origin question if the responses are classified using an approach like the historical series tabulation, but probably not comparable information on the detailed Hispanic origin groups. A combined race and Hispanic origin question with a multiracial category seems less able to provide estimates of the total Hispanic origin population comparable to a separate Hispanic origin question, and also provides lower estimates of several specific Hispanic origin groups and of several specific Asian and Pacific Islander groups than the corresponding separate questions that list these groups.

---------------

The terms Hispanic origin and Hispanic are used interchangeably in this report.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, p. 44676.

"A Test of Methods for Collecting Racial and Ethnic Information," USDL 95-428, U. S. Department of Labor, Bureau of Labor Statistics, October 26, 1995. Clyde Tucker, Ruth McKay, Brian Kojetin, Roderick Harrison, Manuel de la Puente, Linda Stinson, and Ed Robison, "Testing Methods of Collecting Racial and Ethnic Information: Results of the Current Population Survey Supplement on Race and Ethnicity," Statistical Note Series, No. 40, U.S. Department of Labor, Bureau of Labor Statistics, June 1996.

Roderick Harrison, Manuel de la Puente, and Claudette Bennett, "Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey," Population Division Working Paper Series, U.S. Department of Commerce, Bureau of the Census, December 1996. The NCS is known also as the U.S. Census 2000 Test.

The RAETT is known also as the 1996 Census Survey.

The term "White ethnic" includes persons whose ancestry was European (such as British, German, Italian, or Polish, but excluding Spanish), Canadian, or American.

Claudette Bennett, Nampeo McKenney, and Roderick Harrison, "Racial Classification Issues Concerning Children in Mixed-Race Households," paper presented at the Annual Meeting of the Population Association of America, San Francisco, CA, April 1995.

Barry Edmonston and Charles Schultze (eds.), Modernizing the U.S. Census, National Academy of Sciences, Washington, DC, 1995 p. 140.

Roderick Harrison, Manuel de la Puente, and Claudette Bennett, "Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey," Population Division Working Paper Series, U.S. Department of Commerce, Bureau of the Census, December 1996; Nancy E. Bates, Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente, "Questionnaire Effects on Measurements of Race and Spanish Origin," Journal of Official Statistics, Vol. 11, pages 433-459, 1996.

10 Jennifer Day, "Population Projections of the United States by Age, Sex, Race, and Hispanic Origin: 1995 to 2050," Current Population Reports, P25-1130, U.S. Department of Commerce, Bureau of the Census, Washington, DC, 1996, p. 13.

11 The combined, two-part question discussed in this paragraph is often referred to, for simplicity, as a combined question in the remainder of this report when referring to the collection of data on race and Hispanic origin. When the collection of data on ancestry is the focus, it is referred to as a two-part question.

12 Eleanor Gerber and Manuel de la Puente, "The Development and Cognitive Testing of Race and Ethnic Origin Questions for the Year 2000 Decennial Census," Paper presented at the Census Bureau's Annual Research Conference, Arlington, VA, March 1996.

13 The instruction in Panel E (with a multiracial category) is to "mark one box," and the instruction in Panel F (without a multiracial category) is to "mark one or more boxes."Part A includes a write-in line for "enrolled or principal tribe" for the "Indian (Amer.) or Alaska Native" category. The eight panels (questionnaires) in the RAETT are described in Chapter 3.

14 Census tracts are statistical subdivisions of counties and usually have populations between 2,500 and 8,000. The race or ethnicity of a household is classified as the race or ethnicity of the householder, which may not be the same as the race or ethnicity of other household members. See Chapter 3.

15 In Panels E and F, Hispanic origin is treated as a race.

16 The increases in the "Other Asian and Pacific Islander" category on both Panels C and H reflect an edit that assigned persons who marked two or more "Other Asian and Pacific Islander" categories (but no categories other than Asian and Pacific Islander) to "Other Asian and Pacific Islander." The same edit was applied to the multiple responses to the race question in Panels A and B, but it produced no significant increase in reporting as "Other Asian and Pacific Islander."

17 Because these effects were not observed in Panel G, which also had a multiracial category, they should be interpreted with caution.

18 Nancy E. Bates, Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente, "Questionnaire Effects on Measurements of Race and Spanish Origin," Journal of Official Statistics, Vol. 11, pages 433-459, 1996.

19 In the later report, we will examine the "ancestry or ethnic groups" reported in the write-ins to the combined question for ethnic groups.

20 The distributions of detailed Asian and Pacific Islander groups in the Asian and Pacific Islander targeted sample on the combined panels (Table 1.8) were constructed by assigning any person who marked the Asian and Pacific Islander box (and only that box) and who wrote in one of the specific Asian or Pacific Islander groups listed on the separate race question to that specific group. Persons who wrote-in more than one Asian and Pacific Islander group were assigned to the "Other API" detailed category, which is consistent with how those who marked two or more Asian and Pacific Islander boxes (but no boxes for groups other than Asians and Pacific Islanders) were classified on the panels with the separate race question.

21 A comparable procedure was used to obtain distributions of detailed Hispanic origin groups from the ancestry write-ins on Panels E and F: anyone who marked the Hispanic box (and only that box) and who wrote in one of the specific Hispanic origin groups listed on the separate Hispanic origin question was assigned to that specific group. Persons who wrote in more than one Hispanic group were assigned to the "Other Hispanic" category.

22 Due to their designs and to nonresponse to the surveys, neither the NCS nor the CPS supplement samples were fully representative of the national population, and their distributions for race and Hispanic origin were not expected to match distributions from the 1990 census or from current surveys. For more detailed discussion of the NCS sample, see pages 7-13 of Roderick Harrison, Manuel de la Puente, and Claudette Bennett, "Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey," Population Division Working paper Series, U.S. Department of Commerce, Bureau of the Census, December 1996.

For a more detailed discussion of the sample for the CPS Supplement, see Clyde Tucker, Ruth McKay, Brian Kojetin, Roderick Harrison, Manuel de la Puente, Linda Stinson, and Ed Robison, "Testing methods of Collecting Racial and Ethnic information: Results of the Current population Survey Supplement on Race and Ethnicity," Statistical Note Series, No. 40, U.S. Department of Labor, Bureau of Labor Statistics,June 1996.

23 Nancy E. Bates, Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente, "Questionnaire Effects on Measurements of Race and Spanish Origin," Journal of Official Statistics, Vol. 11, pages 433-459, 1996.

2. BACKGROUND

2.1 The Office of Management and Budget's Statistical Policy Directive No. 15

In response to legislative, program, and administrative needs, the Office of Management and Budget (OMB) issued in 1977 Statistical Policy Directive No. 15, "Race and Ethnic Standards for Federal Statistics and Administrative Reporting." The racial classifications set forth in the Directive are American Indian or Alaskan Native, Asian or Pacific Islander, Black, and White. The ethnic classifications specified are "Hispanic origin" and "Not of Hispanic origin." The standards have been used throughout the Federal Government for two decades: in two decennial censuses, in various surveys of the population, in data collections to meet statutory requirements associated with monitoring and enforcing civil rights, and in other administrative reporting for Federal programs.1

During the past several years, the standards have come under growing criticism from those who believe that the minimum set of categories no longer reflects the increasing racial and ethnic diversity of the Nation's population. In response to this and other concerns, the OMB solicited public comment on Directive No. 15 in a Federal Register notice published on June 9, 1994, and held four public hearings in July 1994. The OMB summarized the comments it received in a second Federal Register notice issued on August 28, 1995.2

The OMB established an Interagency Committee for the Review of Racial and Ethnic Standards. Its members, drawn from more than 30 agencies, represent the many and diverse federal needs for data on race and ethnicity, including needs arising from statutory requirements. The OMB Research Working Group of the Interagency Committee identified several issues requiring research and testing to determine the possible effects of suggested changes on the quality and usefulness of the resulting data. They include:

  • a reported more than one race classification;
  • a combined race and Hispanic origin classification;
  • a combined race, Hispanic origin, and ancestry classification; and
  • revised or renamed classifications.

Since the review began, the Census Bureau and several other federal agencies have conducted research on these issues.3 One of these research projects was a Supplement on Race and Ethnicity to the May 1995 Current Population Survey (CPS), sponsored by the Bureau of Labor Statistics (BLS).4 Due to the different sample design and methodology in the CPS Supplement, findings from it are not directly comparable with those from the 1996 National Content Survey (NCS) or 1996 Race and Ethnic Targeted Test (RAETT).

2.2 Issues Addressed in the Race and Ethnic Targeted Test

The Census Bureau conducted extensive research, including the NCS and the RAETT, to examine issues arising from the OMB review of Statistical Policy Directive No. 15, as well as on issues resulting from evaluations of 1990 census data on race and ethnicity. The issues on race and ethnicity included in the NCS and RAETT are listed on the following page in Table 2-1.

To develop questions on race and ethnicity that address these issues, the Census Bureau implemented an extensive program of consultation and research. The consultations involved ongoing meetings and other communications with numerous data users, an expert panel of researchers, the Census Bureau's Advisory Committees, and the OMB Interagency Committee. The research included cognitive interviews, focus groups, and classroom experiments.5

Background information on issues covered in this report, which are limited to issues included in the OMB review of Directive No. 15, is discussed in the remainder of Section 2.

2.2.1 Reporting More Than One Race

Currently, Directive No. 15 does not provide a separate response category for persons who identify with more than one race. Instead, it states that these persons "should use the single category which most closely reflects the individual's recognition in his or her community."6

Table 2-1. Issues Included as Test Objectives in 1996 Census Tests

Issue 1996 National Content Survey (NCS) 1996 Race and Ethnic Targeted Test (RAETT)
*Terminology included in RAETT, but not as a test objective.
QUESTION DESIGN
Reporting more than one race:    
"Multiracial or biracial" category X X
More than one box approach-
   Mark one or more boxes
  X
   Mark all that apply   X
Combined race, Hispanic origin, and ancestry with    
       "Multiracial or biracial" category   X
       Check one or more boxes   X
Alternative sequencing of race and Hispanic origin questions    
       With "Multiracial or biracial" category X X
       Without "Multiracial or biracial" category X  
CATEGORIES
Combine "Indian (Amer.) and Alaska Native" with space for tribal affiliation   X
Native Hawaiian   X
TERMINOLOGY, FORMATTING, AND RELATED ISSUES
Alphabetize Asian and Pacific Islander groups listed on the race question   X
Spell out "American" in category "American Indian or Alaska Native"   X
Terminology:    
     Black, African Am. or Negro* X  
     Spanish/Hispanic/Latino* X  
     Guamanian or Chamorro   X
     "Some other race" X  

A relatively small but growing number of people have expressed concerns about having either to identify with one race or to report in the "Other race" response category in the race question, such as that included in the decennial censuses.7

Proponents for separately identifying race by reporting more than one race argue that the current Directive No. 15 requires persons of mixed racial background to deny their full heritage and perhaps even to choose between their parents. In addition, they find reporting in the "Other race" category demeaning.8 Census data indicate that the number of children in interracial families grew from less than one-half million in 1970 to about two million in 1990, indicating a growing number of people who may identify with more than one race.9 Although the "multiracial" issue is often thought of as involving primarily Blacks and Whites, children in such unions have consistently represented about one-fifth of all children in interracial families. In 1990, about 34 percent of all children in interracial families with at least one White partner were American Indian, and 45 percent were Asian. Conversely, in 1990 more than one-half of the children with at least one American Indian parent lived in families where the other parent was White, as did about 20 percent of the children with at least one Asian parent.

Opponents for reporting more than one race and some federal agencies have expressed concern that the addition of such a procedure could substantially change the counts in the current racial and ethnic categories. They fear that such changes could disrupt the historical continuity of data important to monitoring and enforcing civil rights and equal employment opportunity legislation.10 Some researchers also note that although the number of persons who currently report more than one race is small, including the option on future forms may result in a rapid increase in reporting due to a shift in racial identification similar to that experienced by the American Indian population between the 1970 and 1980 and the 1980 and 1990 decennial censuses.11 To allow some possibility for comparison to the current classifications, some agencies have emphasized the importance of having respondents who report more than one race to specify their races or indicate a primary racial identification if permitted to report more than one. From write-in entries, agencies could then develop procedures to either edit or allocate write-in entries into current OMB categories.

Two different approaches were used in the RAETT for race reporting by respondents who identify with more than one race. The first approach includes a separate "Multiracial or biracial" category with write-in lines in the race question. The second approach uses modified instructions for the race question to read "Mark one or more races" on two panels in the RAETT. On one of the panels using the second approach, the instructions to respondents were to "Mark all that apply." The second approach has the advantage of preserving detailed data about racial identification which might not be captured by a single "Multiracial or biracial" response category with write-in lines; however, race data which allow more than one response can make data capture on the form more complex. The "Multiracial or biracial" category was included in both the NCS and the RAETT. The multiple mark option appeared only in the RAETT.

To determine where best to place the separate "Multiracial or biracial" category in the race question, the Census Bureau tested two options: (1) placing the "Multiracial or biracial" category between the "Other Asian or Pacific Islander" and "Some other race" response categories; and (2) placing the "Multiracial or biracial" category after the "Some other race" response category. The results of cognitive research revealed that when the "Multiracial or biracial" category and its write-in lines were placed between the "Other Asian or Pacific Islander" and "Some other race" response categories, which also had a write-in line, some respondents found it difficult to find and distinguish the "Multiracial or biracial" category and its write-in lines. Other respondents used the write-in lines for the "Multiracial or biracial" category to provide their entry for the "Some other race" category because they were confused by the proximity of these write-in lines. Subsequent cognitive testing indicated that the placement was clearer to the respondents if the "Multiracial or biracial" category and its write-in lines were placed last, after the "Some other race" response category and its write-in line.

Several factors had to be considered in determining the number and length of write-in lines required for the "Multiracial or biracial" category. These factors included the availability of space on the census form and public perception. Cognitive research showed that when one write-in line was provided, some respondents thought they were to provide only one race group; others thought that the single write-in line was an attempt to record only one of their racial backgrounds. Too many write-in lines in the race question confused some respondents.

Cognitive interviews also showed that some respondents thought that the term "Multiracial" meant more than two races. Based on cognitive research and consultation with researchers and members of the Race and Ethnic Advisory Committees (REACs), the category was modified to read "Multiracial or biracial," to convey to respondents that the category could be used by persons who identify with two or more race groups.12

The second approach to reporting one or more race, as noted earlier, was to allow respondents to select more than one race. Some have argued that this approach has the advantage of preserving detailed data about the racial identifications which might not be captured by a single multiracial category with write-in lines. On the other hand, data collected using a question which allows for more than one response can present difficulties during data capture and analyses.

Cognitive research was used to develop instructions to the race question that would best convey to respondents that they could mark one or more than one race in the race question. Initial instructions to the race question were "Mark [X] ONE box for the race that the person considers himself/herself to be. Persons who identify with more than one race may mark more than one box and write the race they most identify with in Box A below." These instructions proved unfeasible because respondents did not understand them and had problems finding the box to which the instructions applied. In addition, some respondents expecting a separate multiracial category were disappointed that this response category was not provided. Persons who identified with more than one race also objected to the instructions to specify a single race. Two different sets of instructions for the race question with a multiple mark option were developed, each of which were tested in the RAETT. One version asks respondents to "Mark [X] one or more races to indicate what this person considers himself/herself to be."13 The second instructs respondents to "Mark all that apply. Mark the race(s) that this person considers himself/herself to be."

2.2.2 Sequencing the Questions on Race and Hispanic Origin

The research on how to best sequence the race and the Hispanic origin questions addressed two persistent concerns identified in decennial census evaluations.14 First, some respondents see these questions as asking for the same information and thus do not answer one of the questions. In the 1990 census, the race and the Hispanic origin questions did not immediately follow each other. Rather the race question was placed first, followed by questions on age and marital status then by the Hispanic origin question, in an attempt to indicate that Hispanic origin represented a different subject than race. Even so, the 10 percent nonresponse rate to the Hispanic origin question was high. A study of 1990 census content reinterview data showed that most of the people who did not answer the Hispanic origin question were non-Hispanics.15

Research from the 1990 census and cognitive studies has shown that some Hispanics view themselves racially as Hispanic and do not identify with one of the specific racial categories (White, Black, etc.), or they find the race question confusing.16 In 1990, about 40 percent of Hispanics reported in the "Other race" category.

Some data users have raised concerns about whether placing the Hispanic origin question first would affect reporting for groups such as American Indians and Asians and Pacific Islanders, which include substantial proportions of persons who identify themselves as also being of Hispanic origin. Some critics have raised concerns about placing a question designed to identify only one of many ethnic populations in the Nation before one that provides racial identification for all persons.

The placement of the race and the Hispanic origin questions was tested in the NCS to replicate the earlier studies in a survey with both a large sample and a reinterview to permit evaluations of whether placing the race question immediately before the Hispanic origin question, or placing the Hispanic origin question before the race question, affects the consistency of responses to the race and to the Hispanic origin questions. The NCS also provided a test of a treatment where the race question immediately preceded the Hispanic origin question with instructions to answer both questions. Results from the NCS indicated that placing the Hispanic origin question before the race question significantly reduced nonresponse to the Hispanic origin question, and this occurred whether there was or was not a separate "Multiracial or biracial" response category. This finding was consistent with previous census research.17 In addition, the Hispanic origin first sequence did not affect nonresponse to the race question or the consistency with which race was reported.

2.2.3 Combining the Race and Hispanic Origin Questions With Ancestry Asked in a Second Part

As part of the review of Directive No.15, the Bureau of the Census included in its research and testing program for the 2000 census, a two-part, combined race, Hispanic origin, and ancestry question. Although Directive No. 15 provides the option of collecting race and Hispanic origin data using a single question, the Census Bureau has used two separate questions to collect race and Hispanic origin data to provide more complete information about the Hispanic and non-Hispanic populations. However, over the years suggestions have been made to combine the race question and the Hispanic origin question. There is considerable division on this issue.18

Census Bureau studies and other research show that some respondents, particularly Hispanics, view Hispanic origin as a race rather than an ethnic group.19 A substantial proportion of Hispanics reported as "Other" in the 1980 census and as "Other race" in the 1990 census and provided a Hispanic origin write-in to the race question. This pattern, as well as cognitive research and debriefings of Hispanic respondents, indicate that many Hispanics expect to see "Hispanic" as an option on the race question, and do not identify with any of the other categories. Other Hispanics do identify with one of the race groups, as well as with Hispanic origin, and wish to remain able to do so.

At the December 1994 meeting of the Census Bureau's Race and Ethnic Advisory Committees (REAC), the Committee on American Indians and Alaska Natives recommended a combined race and Hispanic origin question as a way of eliminating the overlap of data for persons who now report as American Indian in the race question and Hispanic in the Hispanic origin question. During the 1993 Congressional Hearings on Race and Ethnicity, the National Council of La Raza stated that combining the Hispanic origin category into the race question merited research. At these hearings, other researchers raised the possibility of a combined race, Hispanic, and ancestry item, but also said the issue should be researched. In subsequent deliberations with the OMB, the Panel of Experts, and the Census Bureau's REAC's, combining ancestry into the question was also discussed as one possible way to collect data on other ethnic groups, such as Arab Americans, Cape Verdeans, and German Americans, who have requested separate classifications in Directive No. 15.

Some Federal agencies, in accordance with the minimum requirements of Directive No. 15, already collect and report data using a combined race and Hispanic origin question. This format does not provide information on the race of Hispanic persons and may "therefore reduce the utility of the four racial categories by excluding from them persons who otherwise would be included."20 Counts of Hispanics and/or of persons in the current race categories may therefore be adversely affected in a combined format. In the May 1995 CPS Supplement on Race and Ethnicity, the BLS found that the proportion of persons reporting as Hispanic were about 20 percent lower on the panels with a combined race and Hispanic origin question than on the panels where these were two separate questions. About 50 percent of Cuban respondents and about 25 percent of "Other Spanish/Hispanic" respondents reported as White on the combined question.

A combined race and Hispanic origin question may affect the counts of race groups and of persons of Hispanic origin. Some Hispanics were concerned that they no longer would be able to select a race category if they select a Hispanic origin category. Persons who may identify mainly with a racial group but have Hispanic ancestry also object because they would have to choose only one classification. A concern of data users is that they would no longer have the overlap of race groups and of persons of Hispanic origin which provided the fullest count for each racial group and the Hispanic origin population.

Research done since 1987 suggested that placing a Hispanic origin category in the race question and adding a write-in line for ancestry may reduce the problem of nonresponse to the Hispanic origin and ancestry questions.21 The ancestry response would allow for tabulating and customizing data for "emerging ethnic" groups and for documenting the race of persons who report more than one race. Another benefit is to eliminate the overlap between race groups and Hispanic groups by having respondents choose the group with which they most closely identify as a way to minimize Hispanics selecting the "Other race" category which is not normally allowed by Directive No. 15.22

The Census Bureau used cognitive interviews to develop a combined race, Hispanic origin, and ancestry question. In the initial draft of the question, four relevant concepts were included, along with a complex instruction. The question read:

What is this person's race, ancestry, ethnic group or national origin? Mark [X] one box for the race or ethnic group with which this person most closely identifies himself/herself. Print a more specific ancestry, ethnic group or national origin in the space provided below.

The initial combined question asked respondents to state their "race," "ancestry," "ethnic group," or "national origin." Additionally, the question contained long, complex instructions. Each race or ethnic group had its own write-in space with examples. This question format, in addition to being difficult to understand by respondents, also posed a problem of space on the form. The instructions for this question proved to be overwhelming to almost all respondents, chiefly because of the mixture of concepts, the complexity of the instructions, and the cluttered format of the question.23 In subsequent versions, we found that persons understood the terms race and ancestry. Most respondents recognized the term "ethnic group" but did not distinguish it from the concept of "race." In the context of the combined question, "ethnic group" was viewed as almost completely identical to the concept of "race." "National origin" was originally included because prior research indicated that Asian and Hispanic immigrants primarily identified with their countries of origin.24 The introduction of this term created difficulties. For some respondents, the term "national origin" meant the country of birth while for others it referred to citizenship, but not place of birth. As a result, "ancestry" and "national origin" are two quite different concepts. For example, White respondents often see their "ancestry" as a European country, but their "national origin" as "America." The term "national origin" was later dropped from further testing.

The combined question on race, Hispanic origin, and ancestry in the RAETT is in two parts. Part A instructs respondents to "Mark [X] ONE box for the race or origin that this person considers himself/herself to be." After conducting cognitive research, it was determined that the phrase "race or origin" should be included in the instructions to help respondents understand the intent of the question. Early cognitive interviews contained the word "ethnic" in the instructions. However, some respondents did not understand the use of the term "ethnic," so it was deleted from the final instructions. The reference to "race" or "origin" was deemed most appropriate since the instructions for the combined question tell respondents to select the race or origin they most closely identify with from the four racial groups and Hispanic origin groups specified in Directive No. 15, along with an additional category labeled "Multiracial or biracial."25

Part B of the combined question instructs respondents to provide information on their "ancestry" or "ethnic group" and features a list of ancestry examples and two write-in lines. The major objective of Part B of the combined question was to determine if the detailed Asian and Pacific Islander and Hispanic origin groups shown in separate race and Hispanic origin questions would report in a combined format using write-in lines. Another objective was to determine if respondents to the "Multiracial or biracial" and "Some other race" response categories would provide meaningful write-ins about their racial identification. Cognitive interview participants were able to provide this information. American Indian and Black participants found the question redundant, but nonetheless were able to respond to it.

2.2.4 Terminology and Formatting Issues

One aspect of the review of Directive No. 15 is to examine terminology that racial and ethnic groups prefer when identifying themselves. Racial and ethnic identification is a social process that is changing for some proportion of each of the racial and ethnic groups identified in Directive No. 15. Evidence of these changes is illustrated in the comments and recommendations received by the OMB during its review of Directive No. 15. Suggestions to use terms in common usage for the Black, American Indian and Alaska Native, Guamanian, and Hispanic populations, along with suggestions to reclassify the Hawaiian populations, are addressed in the RAETT and are discussed below.

Black, African American, or Negro. The 1990 census used the terms "Black or Negro" in the race item. This terminology was an issue in only a few areas of the country during the 1990 Census. The Census Bureau is now using the terms "African American" and "Black"interchangeably in its news releases and other communications with the public.

The use of "African American" has increased since 1990. Some data users argue that "African American" is the more appropriate terminology and should replace "Black or Negro" in the race item for the 2000 census. Others suggest that "African American" be used along with "Black," and that the term "Negro" should be dropped.

Results from the NCS reinterviews found that "Black" is still the preferred term (45 percent). A slightly higher proportion (33 percent) of Black NCS respondents than respondents in the BLS Supplement preferred African American. A similar proportion (about 3 percent) of Black respondents preferred the term "Negro" in the NCS, the BLS Supplement, and earlier census studies. The Black respondents to the May 1995 CPS Supplement preferred the term Black (44 percent) over African American (28 percent) or Afro-American (12 percent).26 A 1991 survey conducted by the Joint Center for Political and Economic Studies showed that 72 percent of Black Americans preferred the term "Black," 15 percent preferred "African American," 3 percent preferred "Afro American," and 2 percent preferred "Negro." African American was least favored in the South, by older residents, and by persons who had not gone beyond high school. A 1993 study conducted by the National Opinion Research Center showed similar results.27

Cognitive research conducted by Wingerd for the Census Bureau indicated that foreign born Blacks from the Caribbean and Africa did not believe that "African American" referred to them. "Black" was more acceptable to them.28 Lavrakas, Schejbal, and Smith's study of ethno-racial terminology noted that most Blacks used "Black" to refer to their race in 1993, but the proportion had declined from earlier surveys.

As a result of earlier findings, cognitive research and consultations, two of the NCS panels and all of the RAETT panels have the category "Black, African Am., or Negro" in the race question.

Latino versus Spanish or Hispanic origin. The term "Spanish/Hispanic" has endured a lot of criticism over the years, particularly among those who view this as an imposed terminology, rather than one arising within the group it designates. Some argue that "Latino" is more appropriate and should be substituted for one or both of the "Spanish/Hispanic" terms. They contend that "Latino" may be more likely to be chosen by members of the group to which it is meant to apply and that the term is increasingly popular, particularly in California and large Midwestern and northeastern cities, such as Chicago. Advocates for the use of the term Latino suggest that it is a Spanish language term, and therefore, should be more recognizable to Hispanics. On the other hand, others argue that "Latino" is not universally acceptable. Most research show that Hispanics prefer to identify themselves using their national origin. The Census Advisory Committee on the Hispanic Origin Population recommended that the Bureau of the Census "add Latino to the Spanish/Hispanic designation" in a separate Hispanic origin question.

Results from telephone reinterviews for the NCS found that Hispanic respondents preferred the term "Hispanic," (47 percent), followed by "Spanish" (21 percent). About 19 percent of Hispanics had no preference, and another 13 percent preferred the term "Latino." However, many Hispanic respondents found the term "Latino" (68 percent) and "Spanish" (46 percent) unacceptable. Their reasons included not knowing or not having heard the term previously, and a preference for a specified Hispanic subgroup, such as Cuban, Mexican and so forth. The Hispanic respondents to the May 1995 CPS Supplement strongly preferred the term "Hispanic" (58 percent) over "Latino" or "of Spanish origin"(12 percent each).

In cognitive research, the Census Bureau tested the addition of "Latino" to the Spanish/Hispanic origin question. Respondents did not seem to have a problem with this change. In some formats, the length of the category's name makes the wording of the category awkward because it has to end using origin. In two of the NCS panels the phrase "Spanish/Hispanic/Latino" is used in the separate Hispanic origin question. In the RAETT, all of the panels have the category "Hispanic, Latino, or Spanish origin.

Native Hawaiian. Creating a separate category for Native Hawaiians, and including it in the same category as American Indian was included among the issues, that the OMB solicited public comments on in its review of Directive No.15. A broad spectrum of Native Hawaiian groups and individuals have requested and supported the change for at least two reasons. First, they argue that Native Hawaiians, like American Indians, were the original or indigenous inhabitants of territories that the United States acquired; unlike other Asian or Pacific Islander groups, they did not immigrate to the United States. Second, they argue that separate data are needed to understand and address the distinct social, economic, and health conditions found in the Native Hawaiian population, and that these data are often lost when aggregated into the Asian or Pacific Islander category. In addition, some groups and individuals have made similar arguments concerning the indigenous people of Guam and American Samoa, and have suggested that they also should be included in any classification created for indigenous populations. Some federal legislation and programs for American Indians and/or Native Americans currently include American Indians and Alaska Natives, as well as Native Hawaiians.

Recommendations pertaining to the classification of Hawaiian have been received from two of the four Census Bureau Race and Ethnic Advisory Committees. Members of the Asian and Pacific Islander Committee recommended that a category "Native Hawaiian" be placed immediately following the Eskimo and Aleut categories in the race question, thereby deleting it from the Asian and Pacific Islander category. They also suggested that the current category "Hawaiian" be removed from the Asian and Pacific Islander category in the race question, relabeled "Native Hawaiian," and listed after the Aleut category. The members of the American Indian and Alaska Native Committee also suggested that the Bureau of the Census test the listing of Native Hawaiian as a separate race category.

The Census Bureau examined national data from the 1990 decennial census on selected demographic, social, and economic characteristics for the Hawaiian population and for the Pacific Islander population excluding Hawaiians. It also examined these characteristics for a category combining Hawaiians with American Indians and Alaska Natives. Results would probably vary for states with large concentrations of each of the groups.

The national analysis showed that Hawaiians are 3 percent of the total Asian and Pacific Islander population, but about 59 percent of the total Pacific Islander population. Removing Hawaiians from the Pacific Islander population would affect the demographic profile of this population, such that the remaining group would be a somewhat younger population, with a larger proportion of families being married couples, and a larger proportion of its members less educated with lower median households and family incomes and a larger proportion of persons and families in poverty. Table 2-2 shows selected social, demographic and economic characteristics for the combined Pacific Islander category, the category once Hawaiians are removed, and Hawaiian only characteristics.

The 1990 census data showed that Hawaiians would comprise about 9 percent of the total population of a combined American Indian, Alaska Native, and Hawaiian category. Since American Indians and Alaska Natives comprised the majority of this category, general social and economic characteristics, such as age and educational attainment, would not be affected. However, because Hawaiians have higher median family and household incomes than American Indians and Alaska Natives including Native Hawaiian in the American Indian or Alaska Native category would have a higher median household income and a higher family median income.

Table 2-2. Selected Characteristics of the Asian and Pacific Islander Population

(Data for age and type of family based on 100-percent tabulations. Remaining data based on sample)
Characteristic Total Pacific Islander Pacific Islander, Hawaiian removed Hawaiian
Source: Edward Fernandez, "Comparisons of Selected Social and Economic Characteristics Between Asians, Hawaiians, Pacific Islanders, and American Indians (Including Alaska Natives)," Population Division Working Paper No. 15, 1996.
Percent in selected age groups
Under 5 years 10.6 11.7 9.8
18 years and over 63.8 61.7 65.4
65 years and over 4.0 2.9 4.8
Percent of families
Married couple 73.0 75.7 71.2
Female householder 19.4 16.5 21.3
Male householder 7.7 7.8 7.5
Percent, Educational attainment of persons 25 years and over
High school or more 76.1 70.9 79.5
Bachelor's degree or more 10.8 9.1 11.9
Income and poverty in 1989 dollars
Median household $31,980 $27,140 $34,830
Median family $10,340 $8,780 $11,450
Percent of persons in poverty 16.1 20.4 13.9

The term "Native Hawaiian" was tested in the RAETT, which was designed to provide information on relatively small population groups. The term "Native Hawaiian" appears in Panels D and G in the RAETT and it was included immediately after the category for American Indian or Alaska Native. The RAETT telephone reinterview included preference questions regarding this term.

Guamanian or Chamorro. The Census Bureau conducted cognitive interviews with persons who identify with the territory of Guam to determine the most appropriate term to be used in the RAETT. Both Chamorro and Guamanian were determined to be acceptable terms. Recently, "Chamorro" has become more preferable to some, much like "African American" has in the Black population. The cognitive research indicated that some preferred "Guamanian" and others "Chamorro." Younger and more educated respondents preferred "Chamorro" and older respondents preferred "Guamanian."

The term "Guamanian or Chamorro" was tested in Panels D and G in the RAETT. It was not tested on the NCS.

American Indian or Alaska Native Category. The Census Advisory Committee on the American Indian and Alaska Native Populations recommended using a combined category "Indian (Amer.) or Alaska Native" with the instruction to "Print the name of the enrolled or principal tribe." The Committee concluded that using such a format would enable the Bureau of the Census to capture the different Eskimo and Aleut tribes (e.g., Inupiats, Yupiks, etc.) as well as American Indian tribes. Alaska Native and State officials requested separate counts of the Eskimo tribes to complete the State's redistricting plan in 1990, but they could not be obtained from the 1990 race question.

In 1995, the Census Bureau conducted a small test of a combined "Indian (Amer.) or Alaska Native" category to determine if reliable data on American Indian and Alaska Native tribes could be captured. Results indicated that respondents do provide write-ins of American Indian or Alaska Native tribes. However, these data also suggest that respondents will use the write-in line to provide data other than for tribe, such as their racial or Hispanic identity.

Cognitive research was conducted with American Indians and Alaska Natives in selected areas, such as Albuquerque, New Mexico, and Fairbanks, Alaska. Results indicated that the term "enrolled" was generally understood because of the requirement for Alaska Natives enrollment in Regional Corporations, but some noted that this does not necessarily reflect ethnicity. Concerns were expressed that less than full-blooded American Indians might feel they would have to report as non-Indian, and also that some persons identify with more than one tribe.

The RAETT tested a combined "Indian (Amer.) or Alaska Native" category in seven of the eight panels. The objective was to determine if American Indian tribal enrollment and Alaska Native tribes could be captured on the same write-in line.

American Indian. Past research has shown that the use of "American" in any category for the race question or Hispanic origin question may attract misreporting by persons who want to emphasize that they or their children are Americans. However, members of the Census Bureau Race and Ethnic Advisory Committees, as well as other data users, have requested that the term "American Indian" be considered as an alternative to the 1990 abbreviated version, "Indian (Amer.)." For this reason the term "American Indian" was included in one panel of the RAETT.

Alphabetization of the Asian and Pacific Islander Subgroups. The Census Bureau's Asian and Pacific Islander Advisory Committee recommended that the current list of Asian and Pacific Islander subgroups be alphabetized to assist respondents in finding their specific subgroup. This recommendation was implemented in Panel G of the RAETT.

Table 2-3 lists issues tested and indicates where results are discussed in the report.

Table 2-3. Summary of Issues Covered in the RAETT

ISSUE Panels tested
Multiracial:  
       Separate category B
       Mark one or more C
       Mark all that apply H
Combined race and Hispanic origin question with ancestry asked in a second part E,F
       With multiracial category E
       With mark one or more F
Alternative sequencing of race and Hispanic origin question D
Terminology:  
       Native Hawaiian G
       Guamanian or Chamarro D
Formatting:  
       Combined Indian (Amer.) and Alaska Native B through H
       Spelling out of "American" D
       Alphabetization of Asian and Pacific Islander category D

---------------

The decennial census collects greater detail on race and ethnicity than the Directive No. 15 categories; but, as required by the Directive, the detail can be combined into the standard categories.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 59, No. 123 (59FR 29831-35), Office of Management and Budget, Thursday, June 9, 1994, pages 29831-29835.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

Clyde Tucker, Ruth McKay, Brian Kojetin, Roderick Harrison, Manuel de la Puente, Linda Stinson, and Ed Robison, "Testing Methods of Collecting Racial and Ethnic Information: Results of the Current Population Survey Supplement on Race and Ethnicity," Statistical Note Series, Number 40, Department of Labor, Bureau of Labor Statistics, June 1996.

Eleanor Gerber and Manuel de la Puente, "The Development and Cognitive Testing of Race and Ethnic Origin Questions for the Year 2000 Decennial Census," Paper presented at the Census Bureau's Annual Research Conference, Arlington, VA, March 1996.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

The Census Bureau has an exemption from the Office of Management and Budget to use the "Other race" category.

"Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

Claudette Bennett, Nampeo McKenney, and Roderick Harrison, "Racial Classification Issues Concerning Children in Mixed-Race Households," Paper presented at the Annual Meeting of the Population Association of America, San Francisco, CA, April 1995.

10 "Standards for the Classification of Federal Data on Race and Ethnicity", Federal Register, Vol. 59, No. 123 (59FR 29831-35), Office of Management and Budget, Thursday, June 9, 1994, pages 29831-29835.

11 "Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

12 Eleanor Gerber and Manuel de la Puente, "The Development and Cognitive Testing of Race and Ethnic Origin Questions for the Year 2000 Decennial Census," Paper presented at the Bureau of the Census Annual Research Conference, Arlington, VA, March 17-19, 1996.

13 There were minor wording changes to the instructions when the combined race, Hispanic origin, and ancestry question was tested. The instructions to the race question were changed to say, "Mark [X] one or more boxes to indicate what this person considers himself/herself to be."

14 Nampeo McKenney and Arthur Cresce, "Identification of Ethnicity in the United States: The Census Bureau Experience," Paper presented at the Annual Meeting of the Population Association of America, Toronto, Canada, May 1990.

15 Nampeo McKenney, Claudette Bennett, Roderick Harrison, and Jorge del Pinal, "Evaluating Racial and Ethnic Reporting in the 1990 Census," Paper presented at the Joint Statistical Meeting of the American Statistical Association, San Francisco, CA, August 1993.

16 Manuel de la Puente and Ruth McKay, "Research Improves Questions," Civil Rights Journal, Vol. 1, No. 1, 1995.

Manuel de la Puente and Ruth McKay, "Developing and Testing Race and Ethnic Origin Questions from the Current Population Survey Supplement on Race and Ethnic Origin," Proceedings of the 1994 Annual Meeting of the American Statistical Association: Section on Survey Research Methods, Vol. I, 1995.

17 Nancy E. Bates, Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente, "Questionnaire Effects on Measurements of Race and Spanish Origin," Journal of Official Statistics, Vol. 11, pages 433-459, 1996.

18 "Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

19 Clara E. Rodriguez, "Challenges and Emerging Issues: Race and Ethnic Identity Among Latinos," Proceedings from Workshop on Race and Ethnicity Classification: An Assessment of the Federal Standards for Race and Ethnicity Classification. Committee on National Statistics, National Academy of Science, February 17-18,1994; Ruth McKay and Manuel de la Puente, "Research Improves Questions," Civil Rights Journal, Vol. 1, No. 1, 1995; Manuel de la Puente and Ruth McKay, "Developing and Testing Race and Ethnic Origin Questions from the Current Population Survey Supplement on Race and Ethnic Origin," Proceedings of the 1994 Annual Meeting of the American Statistical Association: Section on Survey Research Methods, Vol. 1, 1995.

20 "Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44674-44693.

21 "Standards for the Classification of Federal Data on Race and Ethnicity," Federal Register, Vol. 60, No. 166 (60FR 44674-93), Office of Management and Budget, Monday, August 28, 1995, pages 44679.

22 Jorge del Pinal and Susan J. Lapham, "Impact of Ethnic Data Needs in the United States, Challenges of Measuring an Ethnic World: Science, Politics, and Reality," Proceedings of the Joint Canada-United States Conference on the Measurement of Ethnicity, April 1-3, 1992, Ottawa, Canada, pages 447-475.

23 For more details on the aspects of the combined question see Gerber and de la Puente (1996).

24 Ruth McKay and Manuel de la Puente, "Cognitive Research in Designing the CPS Supplement on Race and Ethnicity," Paper presented at the Bureau of the Census 1994 Annual Research Conference, Arlington (Rosslyn), VA, March 19-23, 1995.

25 The final question also has a category, "Some other race," for respondents who do not identify with any of the categories offered. This is consistent with the categories used in the 1990 census race question.

26 Clyde Tucker, Ruth McKay, Brian Kojetin, Roderick Harrison, Manuel de la Puente, Linda Stinson, and Ed Robison, "Testing Methods of Collecting Racial and Ethnic Information: Results of the Current Population Survey Supplement on Race and Ethnicity," Statistical Note Series, No. 40, Department of Labor, Bureau of Labor Statistics, June 1996.

27 Joint Center for Political and Economic Studies 1992. Annual Report, Washington, DC; Paul Lavraskas, Judith A. Schejbal and Tom Smith, "The Use and Perception of Ethno-Racial Labels: "African-American" and/or "Black," Paper presented at the Bureau of the Census, 1994 Annual Research Conference and CASIC Technologies Interchange. Arlington (Rosslyn), VA, March 20-23, 1994.

28 Judith Wingerd, "Urban Haitians: Documented/Undocumented in a Mixed Neighborhood" Ethnographic Evaluation of the 1990 Census, Report No. 7, Final Report for Joint Statistical Agreement 90-10.

3. SURVEY DESIGN AND METHODOLOGY

This section provides general information on the design of the survey, the methodology for the survey, and the editing of race data, and the data analysis methods used. Additional detail on the survey design and methodology is provided in Appendix B.

3.1 Design of the Survey

3.1.1 Experimental Design

The 1996 Race and Ethnic Targeted Test (RAETT) is the major vehicle for testing alternative versions of the race and Hispanic origin questions for Census 2000. The primary objectives of the RAETT were to test the effects of:

  • adding a multiracial response category or permitting multiple responses to the race question;
  • sequencing the Hispanic origin question immediately before the race question; and
  • combining race, Hispanic origin, and ancestry into one question.

To test these objectives, eight questionnaires (panels) were included in this survey: one control panel and seven experimental panels. Each of the experimental panels was designed to assess one or more of the proposed changes, as shown in Table 3-1. For example, a comparison of Panel B to Panel A allows us to evaluate the effect of including a multiracial category, while a comparison of Panel B to Panel D allows us to evaluate the effect of asking the race and Hispanic origin questions in a different sequence. Table 3-2 on the following page provides a detailed description of the question design features for these eight panels. 

Table 3-1. Experimental Design

Order of race and Hispanic origin questions Options for reporting more than one race
None Multiracial category Mark one or more/ Mark all that apply
Hispanic origin asked first Panel A Panel B, Panel G Panel C, Panel H
Race asked first   Panel D  
Race and Hispanic origin asked together in one question   Panel E Panel F

Table 3-2. Race and Hispanic Origin Question Design Features by Panel/1

Separate race, Hispanic origin questions Combined race, Hispanic origin, ancestry question Separate race, Hispanic origin questions
Panel A Panel B Panel C Panel D Panel E Panel F Panel G Panel H
/1 Terminology for the Black and Hispanic origin population is consistent across all panels. All forms have consistent sequencing of sex, age, and relationship as the first three questions.

/2 See Appendix B for the modifications to the 1990 Census race and Hispanic origin questions.
Modified 1990 Census race question2) "Multiracial or biracial" category "Mark one or more races..." instruction "Multiracial or biracial" category "Multiracial or biracial" category "Mark one or more boxes..." instruction "Multiracial or biracial" category "Mark all that apply" instruction
Separate categories:
"Indian (Amer.)" "Eskimo" "Aleut"
Combined category, "Indian (Amer.) or Alaska Native" Combined category, "Indian (Amer.) or Alaska Native" Combined category, "Indian (Amer.) or Alaska Native" Combined category, "Indian (Amer.) or Alaska Native" Combined category, "Indian (Amer.) or Alaska Native" Combined category and spell out "American Indian or Alaska Native" Combined category, "Indian (Amer.) or Alaska Native"
"Hawaiian"; "Guamanian" categories "Hawaiian"; "Guamanian" categories "Hawaiian"; "Guamanian" categories "Native Hawaiian"; "Guamanian or Chamorro" categories Combined category, "Asian or Pacific Islander" Combined category, "Asian or Pacific Islander" "Native Hawaiian"; "Guamanian or Chamorro" categories "Hawaiian"; "Guamanian" categories
No alphabetization No alphabetization No alphabetization No alphabetization No alphabetization No alphabetization Alphabetize Asian and Pacific Islander groups No alphabetization
Modified 1990 Census Hispanic origin question2) Modified 1990 Census Hispanic origin question Modified 1990 Census Hispanic origin question Modified 1990 Census Hispanic origin question Combined question Combined question Modified 1990 Census Hispanic origin question Modified 1990 Census Hispanic origin question
1995 test census sequence: Hispanic origin followed by race Hispanic origin followed by race Hispanic origin followed by race Race followed by Hispanic origin Combined question Combined question Hispanic origin followed by race Hispanic origin followed by race

3.1.2 Sample Selection

It was critical that the RAETT sample allow inferences to be made for the following population groups:

  • (White) European, Canadian, or American ancestry groups;
  • Blacks;
  • Hispanics;
  • American Indians;
  • Asians or Pacific Islanders; and
  • Alaska Natives.

To accomplish this, six independent sampling frames were created based on 1990 race, Hispanic origin, and ancestry data. In order to obtain samples that included a higher proportion of the targeted population groups than would result from a stratified national sample, only areas with high proportions of households for each special targeted population group were included in the sampling frames. For example, the Hispanic targeted sample would contain only areas with high proportions of Hispanic households. Appendix B details the criteria used to maintain a sufficiently large sampling frame while maintaining a high proportion of the targeted population groups. The six sampling frames represent only a fraction of the total housing units in the United States. In some instances the sampling frames are limited to 15 or fewer states. Table 3-3 provides data on the number of states and housing units in each sampling frame. For each targeted sample, the final column in Table 3-3 provides an approximation of the number of occupied housing units (households) in each sampling frame (e.g., Black) as a proportion of occupied housing units in the United States containing such persons. For example, the 10 percent in the last column of Table 3-3 for the Black targeted sample means that the sampling frame for the Black targeted sample only contained approximately 10 percent of the Black households in the United States.

Table 3-3. Characteristics of the Sampling Frames

Targeted sample States Housing units
(in thousands)
Households in sampling frame as a percent of total U.S. households (containing race/ancestry group)
White ethnic 29 156 1
Black 34 1,495 10
Hispanic 15 1,190 15
American Indian 18 35 2
Asian or Pacific Islander 8 119 3
Alaska Native Alaska only, 20 villages 2 8

An independent, systematic sample of housing units was selected from each of these six frames. When a housing unit was selected, the next seven housing units were also taken, thus forming fairly homogenous clusters of eight housing units. The eight housing units in the clusters were then randomly assigned to the eight panels such that each housing unit was assigned to only one panel. The sample was allocated to the eight panels as shown in Table 3-4. Since the RAETT analysis is based only on questionnaires returned by mail, the sample allocation was designed so the expected number of returned questionnaires across panels within each targeted sample would be approximately the same; variation in mail response rates was expected across targeted samples. As shown in Table 3-4, the total sample consisted of 112,100 housing units.

Table 3-4. Mailout Sample Size (Housing Units) by Panel and Targeted Sample

Targeted sample Panels Total
A B C D E F G H
(NA) Not applicable. Panel not included for the Alaska Native targeted sample.
White ethnic 2,710 2,710 1,355 2,710 2,710 2,710 1,240 1,355 17,500
Black 4,126 4,126 2,063 4,126 4,126 4,126 1,794 2,063 26,550
Hispanic 4,126 4,126 2,063 4,126 4,126 4,126 1,794 2,063 26,550
American Indian 2,450 2,450 1,225 2,450 2,450 2,450 1,150 1,225 15,850
Asian or Pacific Islander 3,660 3,660 1,830 3,660 3,660 3,660 1,740 1,830 23,700
Alaska Native 650 650 (NA) 650 (NA) (NA) (NA) (NA) 1,950
Total 17,722 17,722 8,536 17,722 17,072 17,072 7,718 8,536 112,100

3.2 Survey Methodology

In order to maximize the number of questionnaires returned by mail, a mailout strategy developed in testing after the 1990 Census was used in the RAETT. A prenotice letter (advising the household that a questionnaire would arrive shortly) was mailed to all sampled housing units on June 14, 1996. This was followed by the initial questionnaire mailout on June 18. The RAETT census day was June 22, and a reminder card was sent on June 26. Finally, in an attempt to maximize response rates, a replacement questionnaire was mailed on July 16. The replacement questionnaire was only sent to the households that had not returned the initial questionnaire.

Because the Hispanic targeted sample contained many Spanish speaking households, the eight forms were translated into Spanish. Each household in the Hispanic targeted sample was mailed both English and Spanish forms, and the respondents could choose which form to fill out and return. Just over 11,000 forms were completed and returned in the Hispanic targeted sample; of these, almost 38 percent were Spanish forms.

The response rates for the six targeted samples are provided in Table 3-5. The response rate is the ratio of the total number of questionnaires returned to the total mailed questionnaires that could be delivered by the United States Postal Service. Results provided in this report are based on responses from persons in households who filled out and returned a form. These results apply only to those who responded to the mailout questionnaire and cannot be generalized to persons in households who did not complete and return a questionnaire.

Table 3-5. Mail Response Rate by Targeted Sample

Targeted sample Mail response rate
(percent)
Number of returns
White ethnic 71.3 12,471
Black 47.4 12,577
Hispanic 44.1 11,714
American Indian 53.1 8,411
Asian or Pacific Islander 55.2 13,081
Alaska Native 34.0 663
Total 52.6 58,917

3.3 Editing of Race Data

For the mail return form that did not contain an option for reporting more than one race in the race question (Panel A), the 16 race categories were collapsed into the following five categories for analysis: White, Black, American Indian or Alaska Native, Asian or Pacific Islander, and Other race. The mail return forms containing the option for reporting more than one race (Panels B through H) varied in the number of race categories each had; some forms had six categories while others had 15 categories. Regardless, the various race categories were collapsed into six categories for analysis: the same five categories plus multiracial category/multiple race. An unrequested multiple response category was also included in some analyses to reflect those respondents who checked more than one race category when the instructions said to mark only one. The five panels with an unrequested multiple response category were A, B, D, E, and G. Those respondents who reported more than one race were aggregated in several ways using the write-in entries. These different methods are discussed in detail in Appendix B.

For the mail return race data, if only one specific race were provided by a respondent and it did not agree with its associated major race category, it was reclassified into the appropriate major race category. For example, a write-in entry coded as Cape Verdian provided in the "Other Asian or Pacific Islander" group would have been reassigned to the "Other race" category. Cases in which the respondent provided two or more write-ins for a category were not reclassified. Most write-in entries were computer coded using a master file built from the 1990 census; however, those entries that could not be coded by the computer were coded by expert clerical coders. A more detailed discussion of the response coding techniques is in Appendix B.

3.4 Data Analysis Methods

3.4.1 Statistics Used

The effects of different race and Hispanic origin questions on the mail return questionnaires were evaluated by comparing item nonresponse rates and the distribution of responses across panels. The race distributions, excluding item nonresponse, were compared at the category level using the proportion within the category. The item nonresponse rate for a panel gives the proportion of persons who did not answer the question. The distributions of responses are compared to determine if patterns of response differ across panels.

3.4.2 Panel Comparison Methods

To test for differences between two panels, estimates for the percent in a category and the item nonresponse were calculated, as well as the standard errors of those estimates, and the standard error of the difference between the two estimates. The standard error of the estimates measures the amount of variation in the estimates due to sampling. Because data from the RAETT are based on a sample survey and not on a complete census of households in each sampling frame, the results are subject to sampling error. Standard errors are not included in text tables as they are in the detailed tables in Appendix D.

Confidence intervals at the 90 percent level were used to test for significant differences. The standard method of constructing confidence intervals was employed, i.e., the standard error of the difference was multiplied by a value from the Student's t-distribution and this product was added to and subtracted from the estimate of the difference. If zero is included in the confidence interval, then no significant difference exists (i.e., the apparent difference may be due to sampling error). The t-distribution value was 1.645 for most comparisons. Some comparisons required an adjustment to account for the effect of multiple comparisons in order to maintain an overall confidence level of 90 percent. The adjustment was used for two comparisons: Panel A to Panel C; and Panel A to Panel H. The effect of using an adjustment factor is to make the individual comparison tests more conservative (i.e., less likely to detect a significant difference), while maintaining the overall confidence level for the two comparisons. For those comparisons where a multiple comparison adjustment was needed, a value of 1.95 was used.

Throughout this report, statements that a treatment had an effect indicate that the differences in the percent in the category or in the item nonresponse were statistically significant at the 90 percent confidence level. Conversely, statements that a treatment had no effect indicate that such differences were not statistically significant at the 90 percent confidence level. Since the sample design for each targeted sample used equal probability of selection methods, unweighted data were used in all the analyses in this report. Usually, weights are used if one desires to make inferences (e.g., population totals) about the target population; however, in the RAETT only estimates of population proportions were made.

Page Last Revised - March 12, 2022
Is this page helpful?
Thumbs Up Image Yes Thumbs Down Image No
NO THANKS
255 characters maximum 255 characters maximum reached
Thank you for your feedback.
Comments or suggestions?

Top

Back to Header