Findings on Questions on Race and Hispanic Origin Tested in the 1996 National Content Survey

INTRODUCTION

This report presents preliminary results of testing alternative versions of the questions on race and Hispanic origin in the 1996 National Content Survey (NCS)¹, conducted by the Bureau of the Census as part of its Census 2000 research and testing program. The principal test of questions on race and ethnicity in this program is, however, the 1996 Race and Ethnic Targeted Test (RAETT), also known as the 1996 Census Survey. Findings from the RAETT will be available in the Spring of 1997. The results from both the NCS and the RAETT will be considered in developing questions that will be included in Census 2000. The Office of Management and Budget (OMB) and its Interagency Committee for the Review of Racial and Ethnic Standards will also consider the NCS findings in their review of the Federal standards for the classification of data on race and ethnicity set forth in Statistical Policy Directive No. 15, Race and Ethnic Standards for Federal Statistics and Administrative Reporting.

The NCS, which was conducted from March through June 1996, is the primary vehicle for testing and evaluating the full subject content for Census 2000. Because the results are based on the responses from the households in the national sample that mailed back their questionnaires, the results do not represent the entire national population.

Furthermore, the NCS sample was not designed to detect possible effects of different treatments on relatively small population groups, such as American Indians and Alaska Natives, detailed Asian and Pacific Islander groups (such as Chinese or Hawaiians), or detailed Hispanic origin groups (such as Puerto Ricans or Cubans). In contrast, the RAETT was designed to provide findings for such small population groups. The RAETT, conducted from June through September 1996, focused exclusively on testing and evaluating possible changes to the questions on race and ethnicity² for Census 2000, and to the classifications set forth in OMB Statistical Policy Directive No. 15.

This report focuses on the effects that the following three treatments tested in the NCS have on how people report race and Hispanic origin. The treatments are:

• Adding a multiracial response category in the race question;

• Placing the Hispanic origin question immediately before the race question; and

• Combining both of these changes.

More detailed findings on these and other issues on race and Hispanic origin covered in the NCS (such as preferred terminology) will be available from the Census Bureau by writing to Population Division, Bureau of the Census, Washington, DC 20233-8800.

SUMMARY OF MAJOR FINDINGS

The major findings from the NCS testing of questions on race and Hispanic origin are:

Multiracial or Biracial Response Category

• About one percent of persons reported as multiracial in the versions of the race question that included a multiracial or biracial response category.

• The presence of a multiracial response category in the race question did not have statistically significant effects at the 90-percent confidence level on the percentages of persons who reported as White, as Black, as American Indian³, or as Asian and Pacific Islander. This finding held regardless of the sequence of the race and Hispanic origin questions.

• An apparent decline in the proportion of persons who reported as Asian and Pacific Islander when a multiracial category was included was not statistically significant at the 90-percent confidence level. However, because the relatively small sample size in the NCS might not detect a sizable proportionate decline in the Asian and Pacific Islander population, and because a substantial proportion of the write-ins to the multiracial category included Asian and Pacific Islander responses, one cannot rule out the possibility that adding a multiracial category affects this population.

• Including a multiracial category in the race- followed-by-Hispanic origin-sequence reduced the percentage of persons reporting in the "Other race" category of the race question.

Sequencing of Race and Hispanic Origin Questions

• Placing the Hispanic origin question before the race question significantly reduced nonresponse to the Hispanic origin question.

• Placing the Hispanic origin question before a race question that did not include a multiracial option--

• Reduced the percentage of persons reporting in the "Other race" category of the race question;

• Increased reporting by Hispanics in the White category of the race question.

More information on these and other NCS findings are presented in the sections below on Detailed Findings.

BACKGROUND ON THE OFFICE OF MANAGEMENT AND BUDGET'S STATISTICAL POLICY DIRECTIVE NO. 15

In response to legislative, program, and administrative needs, the OMB issued in 1977 the "Race and Ethnic Standards for Federal Statistics and Administrative Reporting." These standards are now contained in Statistical Policy Directive No. 15. 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 almost 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.⁴

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 related concerns, the OMB solicited public comment on Statistical Policy 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, 1995.⁵

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. A Research Working Group of the Interagency Committee identified several issues requiring research and testing to determine the possible effects on the quality and usefulness of the resulting data. They include:

• A multiracial classification;

• A combined race and Hispanic origin classification;

• A combined race, Hispanic origin, and ancestry classification; and

• New or renamed classifications.

Since the review began, the Census Bureau and several other Federal agencies have conducted research on these issues.⁶ 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). Due to the different sample and methodology in the CPS Supplement, findings from it are not directly comparable with those from the NCS. (For further information, see section on Relationship to Other Research.)

CENSUS BUREAU'S TESTING PROGRAM OF QUESTIONS ON RACE AND ETHNICITY

Overview. The Census Bureau is conducting 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 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 in Chart 1.

To develop questions on race and ethnicity to 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 Advisory Committees, and the Interagency Committee. The research included cognitive interviews, focus groups, and classroom experiments.⁷

Background on Race and Hispanic Origin Issues Addressed by the National Content Survey. This section provides background information on two issues addressed by the NCS--adding a multiracial category in the race question and the sequencing of the race and Hispanic origin questions.

Chart 1. Issues Included as Test Objectives in 1996 Census Tests

Multiracial Classification. Currently, Statistical Policy Directive No. 15 does not provide a separate classification for persons who identify with more than one race. A relatively small, but growing, number of people who identify with more than one race have expressed concerns about having either to identify with one race or to report in an "Other race" response category of the race question, such as that included in the decennial censuses.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 group of people who may not choose to identify with a single race.⁹

Opponents of a multiracial classification have expressed concern that it could substantially change the counts in the current racial and ethnic categories. In their view, 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 numbers reporting as multiracial may be small initially, they may increase over time precisely as a result of the Federal Government's establishing the classification.¹¹

Sequencing the Questions on Race and Hispanic Origin. The research on reversing the sequence of the questions on race and Hispanic origin addresses two persistent concerns identified in decennial census evaluations.¹² First, some people see these questions as asking for the same information, and thus do not answer one of the questions. In the 1990 census, the Hispanic origin question was placed several questions after the race question in an attempt to indicate that Hispanic origin represented a different subject than race. Even so, at 10 percent, the nonresponse rate for Hispanic origin 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.¹³

Second, 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 (that is, White, Black, etc.), or they find the race question confusing.¹⁴ In 1990, about 40 percent of Hispanics reported in the "Other race" category.

In a series of sample tests for the 2000 Census, the Census Bureau placed the Hispanic origin question immediately ahead of the race question and included an instruction to answer both questions. This procedure reduced, but did not eliminate, nonresponse to the Hispanic origin question. It also reduced reporting by Hispanics in the "Other race" category, without adversely affecting the nonresponse rate for the race question.¹⁵

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, that include substantial proportions of persons who also identify themselves as being of Hispanic origin. Also, some critics have questioned placing a question designed to identify only one of many ethnic populations in the Nation before a question that provides racial classifications for all persons.

Reverse sequencing 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 the reverse sequencing affects the consistency of responses to the race and Hispanic origin questions. The NCS also provided a test of a treatment where the race question immediately preceded the Hispanic origin question and there is an instruction to answer both questions.

NATIONAL CONTENT SURVEY DESIGN AND METHODOLOGY

The NCS included 6 simple (short) form and 7 sample (long) form panels. Chart 2 summarizes the major features of each of the four simple form panels that were designed to evaluate the effects of a multiracial category and alternative sequencing of the questions on race and Hispanic origin. The chart notes additional modifications to the questions on race and Hispanic origin on each panel that could also affect results.

The specific questions on race and Hispanic origin included in each of the four simple form panels are described below. Panel 1 contains a modified version of the 1990 questions on race and on Hispanic origin. The alternative questions to be tested occur in Panels 2, 3, and 4.

Design of the Mailout Survey and the Reinterview. The NCS sample of 94,500 households was drawn from a universe of housing units in 1990 decennial census mailback areas and represents about 95 percent of the country. It is less representative of American Indians and Alaska Natives, since about 25 percent of this population lived in non-mailback areas in 1990.¹⁶ The NCS sample was allocated among 13 experimental panels.

Description of Questions and Panels

Chart 2. Primary National Content Survey Race and Hispanic Origin Panels

In Panel 1, as in 1990, the question on race is placed first followed by two intervening questions -- age and marital status -- and then the question on Hispanic origin. The race question does not include a multiracial category.

PANEL 1--1990 CENSUS MODIFIED NO MULTIRACIAL CATEGORY

7.	Is this person of Spanish/Hispanic origin? Mark [X] the "No" box if not Spanish/Hispanic.
	[ ]	No, not Spanish/Hispanic
	[ ]	Yes, Mexican, Mexican-Am., Chicano
	[ ]	Yes, Puerto Rican
	[ ]	Yes, Cuban
	[ ]	Yes, other Spanish/Hispanic - Print one group, for example: Argentinean, Colombian, Dominican, Nicaraguan, Salvadoran, Spaniard, and so on. _______________________________________________________

Panel 2 was designed to test the effect that a multiracial category in a race first sequence has on reporting of race and Hispanic origin. The race question includes a multiracial category.

PANEL 2--MULTIRACIAL CATEGORY; RACE-FIRST SEQUENCE (ALSO TERMINOLOGY)

Note: Please answer both Questions 6 and 7.

Panel 3 was designed to examine the effect that reversing the sequence (placing the Hispanic origin question before the race question) has on reporting of race and Hispanic origin. The race question does not contain a multiracial category.

PANEL 3--NO MULTIRACIAL CATEGORY; HISPANIC ORIGIN-FIRST SEQUENCE

Note: Please answer both Questions 5 and 6.

Panel 4 combines features of both Panels 2 and 3. In Panel 4 the race question contains a multiracial category (as in Panel 2) and the Hispanic origin question is placed before the race question (as in Panel 3). Panel 4 is intended to examine the combined effect of adding the multiracial category and reversing the sequence of the questions.

PANEL 4--MULTIRACIAL CATEGORY; HISPANIC ORIGIN-FIRST SEQUENCE (ALSO TERMINOLOGY)

Note: Please answer both Questions 5 and 6.

The analysis of race and Hispanic origin in this report relies primarily on 4 of these 13 panels. Each of the four questionnaire versions was mailed to almost 6,000 households, 72 percent of which responded. The results presented below therefore are based on approximately 18,000 households. Because five of the sample (long) form panels included questions on race and Hispanic origin that were similar or identical to the four simple form panels, some additional analyses using these panels were conducted and are presented in the Detailed Findings Sections of this report. These additional analyses should be used with caution since previous census questionnaire research has sometimes found that sample (long form) data does not detect differences that are statistically significant in comparable simple (short form) data.

Since the sample excluded households outside of 1990 census mailback areas, and some households did not complete and return a questionnaire, results from the NCS cannot be generalized to the entire national population. Distributions of characteristics in the NCS, including race and Hispanic origin, therefore, are not expected to be identical to the distributions from the 1990 census. For more detailed information on the NCS sample design, see Appendix C.

The NCS included computer-assisted telephone reinterviews of households. The reinterviews were conducted in May and June 1996, and used the race and Hispanic origin questions contained in the 1990 census, modified for telephone interviewing, with one exception: one-half of the households who completed forms that provided a multiracial response option in the original survey were also given this multiracial option in the reinterview. Whenever possible, the respondent in the telephone reinterview was the household member who completed and mailed back the questionnaire. Write-in responses to the race question on the mail-return questionnaires and comparable follow-up questions from the telephone reinterview were coded and later collapsed into the major race categories presented in Appendix B. The results in this report are based on edited data files, where write-ins to the race question that were inconsistent with the category box checked were used to assign the response to the category appropriate for the write-in.¹⁷

This report seeks to answer three questions about the experimental treatments:

• Did the treatment affect the proportions who reported in the major categories of the race question on the mailback form? the Hispanic origin question?

• Did the treatment affect the proportions who answered the race question on the mailback form? the Hispanic origin question?

• Did the treatment affect the proportions who did not report the same race on both the reinterview and the mailback form? the same Hispanic origin?

The NCS was designed to detect differences across panels in the proportions reporting in categories of the race and Hispanic origin questions. The report addresses the first question by comparing, across panels, the proportions (excluding nonresponses) who reported in each category of the race question and of the Hispanic origin question. The race categories that were compared were obtained by collapsing the 16 detailed race categories on the forms without a multiracial category into 5 categories and collapsing the 17 detailed race categories on the forms with a multiracial category into 6 categories. These major categories are presented in the detailed tables in Appendix B.

The second question is addressed by comparing, across panels, nonresponse rates for the race question and the Hispanic origin question. The nonresponse rate is simply the proportion of persons who did not answer that question on their mailback form.

The third question can be addressed by comparing, across panels, the rate of disagreement for specific response categories (e.g., multiracial) as well as the overall rate of disagreement for a question (e.g., race). These statistics compare the responses from the mailback forms to the responses given in the reinterview. The overall rate of disagreement for a question is the proportion of persons whose responses to that question on the reinterview and the mailback form were inconsistent. The rate of disagreement for a specific category is the proportion of persons who reported inconsistently--that is, reported the category on either the mailout form or in the reinterview, but not on both.

For this report, separate analyses were conducted of the reporting of both race and Hispanic origin by foreign-born and native-born persons, and also of the reporting of race by Hispanics and non-Hispanics.¹⁸ Prior research consistently has shown that foreign-born persons and persons of Hispanic origin have distinct patterns of reporting race and ethnicity. For example, Hispanics have high levels of nonresponse on the race question, and non-Hispanics on the Hispanic origin question. Foreign-born persons report race less consistently than do the native-born.¹⁹

Throughout this report, statements that a treatment had an effect indicate that the differences in proportions, item nonresponse rates, or rates of disagreement cited 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 (see Appendix C).

DETAILED FINDINGS: MULTIRACIAL CATEGORY IN THE RACE QUESTION

Would adding a multiracial category to a race question which precedes the Hispanic origin question affect the reporting of race or Hispanic origin? To answer this, we compared the percentage of persons reporting in each of the race and Hispanic origin categories in Panel 2, which included a multiracial category, with the comparable percentage in Panel 1, which did not.²⁰ The effects of adding a multiracial category to a race question in the alternative sequence, where the Hispanic origin question immediately preceded the race question, were identified by comparing results from Panel 4 with those from Panel 3. These comparisons are summarized in Chart 3.

Findings on the Multiracial Category Observed in Both Sequences. Slightly more than one percent of those reporting a race identified as multiracial on the panels that included this category in the race question (Table A). Hispanics were more likely than the total population to identify as multiracial (6.7 percent in Panel 2 and 10 percent in Panel 4; Tables 11 and 12).

One concern about any new category is whether respondents understand the category and report consistently. These issues can be addressed for the multiracial category by looking at the write-ins to the category and the consistency of reporting as multiracial on the reinterviews and the mailback forms.

Chart 3. Multiracial Category Test Objective.

(Examine the effects of adding a multiracial category to the race question on responses to the questions on race and Hispanic origin.)

Virtually all persons (98 percent) in Panels 2 and 4 who checked the multiracial category provided a write-in. About half (55 percent) of these write-ins involved two or more different race groups (such as, White and Black),²¹ and about one-third involved a race and an Hispanic origin group (for example, White and Hispanic). The remainder of the write-ins (12 percent), when coded, represented only one of the race or ethnic classifications specified in Statistical Policy Directive No. 15.

The vast majority (over 80 percent) of the multiple write-ins to the multiracial category included White. This is consistent with research on interracial and inter-ethnic marriages and households, which usually involve one White spouse (92 percent) or parent (86 percent).²² About 30 percent of the write-ins involved an Asian or Pacific Islander response, about 25 percent involved a Black response, and about 7 percent included an American Indian response.²³

Given that only one percent of persons reported as multiracial, the effects on larger race groups are small. For example, even if all those who reported as multiracial and who included White among their write-ins had instead reported a single race of White, the increase in the White proportion of the population would have been only one percent. The comparable increase for Blacks would have been three percent. In contrast, if the Asian and Pacific Islander write-ins to the multiracial category had been reported solely as Asian and Pacific islander, the proportion of the population in that category would have increased by more than 10 percent.

Those who reported as multiracial on their mail return forms were very likely to do so when reinterviewed: the rates of disagreement were less than 1 percent on Panels 2 and 4 with the multiracial option--lower than the comparable rates for the White and the "Other race" categories (Tables 2 and 4).

The addition of a multiracial category had no statistically significant effect on the percentage of persons who reported as White, as Black, as American Indian, or as Asian or Pacific Islander in either the race-first or the Hispanic origin-first sequence (Table A). However, relatively small sample size in the NCS might not detect effects that were substantively important for small populations.

There were, however, apparent declines in the proportions reporting as Asian and Pacific Islander (from 4.0 percent in Panel 1 to 2.7 percent in Panel 2, and from 3.4 percent in Panel 3 to 2.8 percent in Panel 4) (Table A). These apparent declines seemed substantial, but did not reach statistical significance at the 90-percent confidence level, due perhaps to the relatively small sample size in the NCS. The apparent declines, coupled with the substantial proportion of write-ins to the multiracial category that involved an Asian or Pacific Islander group, prevent drawing a firm conclusion that adding a multiracial category would not reduce reporting as Asian and Pacific Islander.

The RAETT may provide more definitive results on the possible effects of a multiracial category on the Asian and Pacific Islander and on the American Indian or Alaska Native populations.

Findings on the Multiracial Category Unique to the Sequence with Race First. Adding the multiracial category to the race question in the race-first sequence had two effects on the reporting of race. First, doing so reduced reporting as "Other race", from 3.3 percent in Panel 1 to 1.7 percent in Panel 2 (Table A). Second, the treatment increased nonresponse to the race question modestly among non-Hispanics. Persons who reported as not Hispanic on the reinterview had higher nonresponse to Panel 2, which included a multiracial category, than to Panel 1, which did not (2.2 percent and 1.1 percent, respectively) (Table 15). Although statistically significant, this effect was not substantively important because these nonresponse rates are very low. The multiracial category, therefore, did not affect the total nonresponse rate to the race question.

Generally, Hispanics have had the highest rates of nonresponse to the race question. The multiracial category did not reduce their very high nonresponse levels (over 25 percent on both panels) (Table 15).

Adding the multiracial category also had no effect on the consistency of race reporting. The overall rate of disagreement was low (under 4 percent) in both Panels 1 and 2 (Table 9). (The rate of disagreement measures the inconsistency of reporting.) Adding the multiracial category to the race question did not affect reporting in the Hispanic origin categories (Table A). It also did not affect nonresponse rates or the consistency of reporting in the Hispanic origin question. The overall rate of disagreement was low (under 3 percent) for the Hispanic origin question in both Panels 1 and 2 (Table 10).

Findings on the Multiracial Category Unique to the Sequence with Hispanic Origin First. Adding the multiracial category to the race question in this sequence reduced the percentage of Hispanics who identified as Black on the race question. Among persons who identified as Hispanic on the reinterview, 0.6 percent reported as Black on Panel 4, where a multiracial category was available, compared with 2.7 percent on Panel 3, where it was not (Table 12). This effect is based upon a very small number of cases, however, and should be treated with caution.

Since the Hispanic Black population represents a relatively small percentage of the entire Black population, adding a multiracial category in the Hispanic origin-first sequence did not reduce the percentage of all Blacks in Panel 4 compared with Panel 3. Additional analysis of the long form panels found no statistically significant decline in the percentage of Hispanics who reported as Black on either the Hispanic-first or the race-first panels.

Table A. Multiracial Category -- Percent Distribution of Mail Returns by Race and Hispanic Origin: 1996 National Content Survey

In the Hispanic origin-first sequence, adding the multiracial category reduced inconsistent reporting among Whites in the race question. The disagreement rates fell from 3.8 percent in Panel 3 to 1.8 percent in Panel 4 (Table 4). This treatment also reduced inconsistent reporting for the "Other race" category in the race question. The disagreement rates dropped from 3.9 percent on Panel 3 to 1.7 percent on Panel 4 (Table 4).

Finally, this treatment (Panel 4) had a lower rate of disagreement for the race question than Panel 3 (Table 9). Inconsistency was low (under 5 percent) on both panels, however.

Studies consistently have shown that persons, especially Hispanics, who switch from the "Other race" to the White category, or vice-versa, when reinterviewed account for much of the inconsistent reporting in both categories.²⁴ Adding the multiracial category and reversing the sequence of the race and Hispanic origin questions each reduced "Other race" reporting. The combined treatment seems to have also reduced inconsistent reporting in both the "Other race" and the White categories.

The panel that included the multiracial category in the Hispanic origin-first sequence had a lower percentage who reported as Hispanic in the Hispanic origin question -- 6.9 percent in Panel 4, compared to 9.0 percent in Panel 3. Conversely, 93.2 percent reported as non-Hispanic in Panel 4, and 91.0 percent in Panel 3 (Table A). The drop was particularly sharp among Mexicans, falling from 5.6 percent in Panel 3 to 3.2 percent in Panel 4 (Table A). This difference was detected in the native-born population, where the percentage reporting as Mexican decreased from 5.0 percent in Panel 3 to 2.7 percent in Panel 4 (Table 18). No decrease was observed for the foreign-born population (Table 19).

In additional analyses of responses to comparable sample form panels, neither the multiracial category nor differences in the wording of the Hispanic origin question on Panels 3 and 4 caused a statistically significant decline in the proportion of Mexicans or of Hispanics on those forms. (The long form panels permitted separate analysis of each treatment.) In addition, analysis of reinterview data ruled out the possibility that significantly different proportions of Mexicans were sampled in Panels 3 and 4. The drop in reporting as Hispanic, and particularly as Mexican, on Panel 4 is thus unexplained. Further research using multivariate analyses to examine the interaction between the multiracial category and sequencing, may help to explain the observed effects.

Finally, in the Hispanic origin-first sequence, the multiracial category had no statistically significant effect on the consistency with which persons reported Hispanic origin (Table 10). The overall rate of disagreement was low (under 3 percent) in both Panels 3 and 4 (Table 10).

Summary of Findings on the Multiracial Option in the Alternative Sequence. Adding the multiracial category to the race question in the race-first (1990 census) sequence reduced reporting as "Other race" and marginally reduced the response rate for the race question among non-Hispanics. It did not affect the very high nonresponse rates of Hispanics to the race question. The treatment had no statistically significant effects upon reporting in the Hispanic origin question.

Adding a multiracial category in the Hispanic origin-first sequence reduced inconsistent reporting in the "Other race" and White categories of the race question. This treatment may also reduce the proportion of Hispanics who report as Black on the race question. This finding is based on a very small number of cases, however, and must be interpreted with caution.

This treatment also had lower reporting as Hispanic and as Mexican. However, further analyses of comparable long forms suggests that neither the addition of the multiracial category nor the different terminology and question wording for the Hispanic origin question on Panel 4 can account for this drop. Further research, including multivariate analyses, may help to explain this decline.

DETAILED FINDINGS: SEQUENCING OF THE RACE AND HISPANIC ORIGIN QUESTIONS

To determine if placing the Hispanic origin question before the race question affects reporting in each question when the race question does not include a multiracial category, we compared the percentage of persons reporting in each of the race or Hispanic origin categories on Panel 3 with the corresponding percentages in Panel 1. To identify effects when the Hispanic origin question is asked first and the race question includes a multiracial category, we compared the percentage of persons reporting in each race and Hispanic origin categories in Panel 4 with those in Panel 2. These comparisons are summarized in Chart 4.

Findings on Sequencing Regardless of Including a Multiracial Category. Placing the Hispanic origin question before the race question reduced nonresponse to the Hispanic origin question. This finding is consistent with prior research,²⁵ and held whether or not the multiracial category was included in the race question (7.6 percent on Panel 2 compared with 5.2 percent on Panel 4; 8.6 percent on Panel 1 compared with 5.5 percent on Panel 3; Table 10). Nonresponse to the Hispanic origin question also dropped for the native-born population (Table 24).

This treatment did not significantly affect the consistency of Hispanic origin reporting, whether or not the race question included a multiracial category. The rates of disagreement were low (less than 3 percent) in all four panels (Table 10).

The Hispanic origin-first sequence did not affect nonresponse to the race question or the consistency with which race was reported. Rates of disagreement were low (less than 5 percent) on all four panels (Table 9). This treatment also did not affect the percentages of persons reporting as White, as Black, as American Indian, or as Asian and Pacific Islander (Table B).

Findings on Sequencing that Depended on the Presence or Absence of a Multiracial Category in the Race Question. On the panels without a multiracial category in the race question, the Hispanic origin-first sequence reduced reporting as "Other race" from 3.3 percent on Panel 1 to 1.7 percent on Panel 3 (Table B). The drop primarily reflected a decrease in the percentage of Hispanics who reported as "Other race," from 42.9 percent on Panel 1 to 24.9 percent on Panel 3 (Table 13). This treatment also increased reporting by Hispanics in the White category of the race question from 52.5 percent in Panel 1 to 72.1 percent in Panel 3 (Table 13). Prior research has established that many Hispanics select the "Other race" category because they wish to report their Hispanic identity when responding to the race question. When the Hispanic origin question immediately precedes the race question, some Hispanics might feel they have already expressed their Hispanic identity in the Hispanic question, and need not do so again by reporting as "Other race" in the race question.

On the panels without the multiracial category in the race question, the Hispanic origin-first sequence reduced nonresponse to the Hispanic origin question for the foreign-born from 9.5 percent in Panel 1 to 3.9 percent in Panel 3 (Table 24).

Chart 4. Alternative Sequencing Test Objective.

(Examine the effects of sequencing on the responses to the Hispanic origin and race questions.)

Summary of Findings on Sequencing. The results indicate that the Hispanic origin-first sequence increased response rates to the Hispanic origin question in the NCS, as it has in previous studies, without increasing nonresponse to the race question. Moreover, it had no statistically significant effects on the consistency of reporting in either the race or the Hispanic origin questions.

Placing the Hispanic origin question before the race question without a multiracial category reduced the tendency of Hispanics to report in the "Other race" category in the race question. On the panels with a multiracial category, the Hispanic origin-first sequence did not significantly reduce "Other race" reporting below that observed in the race-first sequence panel. This, at least in part, is because in the race-first sequence, the multiracial category also had reduced reporting as "Other race," and the alternative sequence combined with the multiracial category did not achieve further reductions.

SUMMARY OF NATIONAL CONTENT SURVEY FINDINGS ON ALTERNATIVE TEST QUESTIONS

This section summarizes the results for each of the alternative questionnaires tested in the NCS.

What are the effects of a sequence that places the race question first when the race question includes a multiracial response option (Panel 2 compared with Panel 1)? Insofar as the NCS can provide evidence, the results suggest that adding a multiracial category to the race question in the race-first (1990 census) sequence does not significantly affect the overall percentages of persons reporting in the White, Black, American Indian, or Asian or Pacific Islander categories.

Although the apparent decline in the proportion who reported as Asian and Pacific Islander in Panel 2 was not statistically significant at the 90-percent confidence level, its size suggests that further analysis would be helpful, some of which will be possible with the RAETT. Moreover, the substantial proportion of write-ins to the multiracial category involving an Asian or Pacific Islander population suggest that the possibility cannot be ruled out that adding a multiracial category to the race item would reduce the proportion of persons who report as Asian or Pacific Islander.

Reporting of "Other race" also dropped on Panel 2, suggesting that some persons who would have reported in that category may instead report as multiracial when given that option.

The multiracial option may increase the small proportion of non-Hispanics who do not answer the race question, but does not affect the very high nonresponse rate for race by Hispanics. The effect for non-Hispanics is too small to affect the total nonresponse level.

The NCS findings suggest that reporting in the Hispanic origin question would not be significantly affected by including a multiracial option in the race question in a race-first sequence.

Table B. Alternative Sequencing -- Percent Distribution of Mail Returns by Race and Hispanic Origin: 1996 National Content Survey

What are the effects of a sequence that places the Hispanic origin question first when the race question does not include a multiracial option (Panel 3 compared with Panel 1)? The NCS and prior research conducted by the Census Bureau suggest that placing the Hispanic origin question before the race question reduces nonresponse to the Hispanic origin question, without increasing nonresponse to the race question. The improved reporting was primarily by non-Hispanics, who have made up the bulk of the nonrespondents in previous studies. Both the native and the foreign born populations also had improved reporting to the Hispanic origin question in this sequence.

Placing the Hispanic origin question first did not reduce the historically high proportion of Hispanics who do not answer the race question. The Hispanic origin-first sequence reduced reporting as "Other race", as some Hispanics apparently chose one of the specific racial categories. The sequence did not affect reporting in any other categories of the race question or of the Hispanic origin question.

What are the effects of combining a race question with a multiracial response option and a sequence that places the Hispanic origin question first (Panel 4 compared with Panel 3 and with Panel 2)? The major effects of combining both changes (the multiracial response option and the Hispanic origin-first sequence) on one panel are similar to those observed when a multiracial category is added in the race-first sequence (Panel 2) or when sequencing is changed without adding a multiracial category (Panel 3). Specifically, reporting as "Other race" in Panel 4 is statistically similar to the lower levels in Panel 2 and in Panel 3. There is some evidence that combining both changes may have some additional effects beyond those observed when each change is made separately. (For example, combining both changes reduced inconsistent reporting in the "Other Race" and White category might also reduce the proportion of Hispanics who report as Black.) Additional research, including multivariate analyses, would be needed, however, to identify more fully the joint effects of adding multiracial category to the race question in the Hispanic origin-first sequence.

RELATIONSHIP TO OTHER RESEARCH

The findings reported here from the NCS constitute one of many sources of information to be considered by the Office of Management and Budget and other agencies in their review of Statistical Policy Directive No. 15. Changes, if any, to the Directive could result in modifications of the race and Hispanic origin questions for the 2000 census.

It is important to emphasize that one would not expect results from different studies to agree fully, especially when they differ in sample designs and modes of data collection, and in the format, wording, and context of race and ethnicity questions. In particular, some results from the NCS differ from the Race and Ethnicity Supplement to the May 1995 Current Population Survey (CPS). That Supplement, sponsored by the Bureau of Labor Statistics, interviewed 60,000 households. About one-fourth of the households (those included for the CPS for the first time) were interviewed in person; the remainder were interviewed by telephone.

On two of the four panels on the Supplement, the race question included a multiracial category. About 1.6 percent of respondents on these panels selected it. Both the CPS and NCS results thus suggest that, currently, less than 2 percent of respondents nationally might select a multiracial category when it is offered. Given the NCS findings discussed earlier, one cannot rule out the possibility that adding a multiracial category might reduce the proportion of persons reporting as Asian or Pacific Islander. The proportion of persons reporting in the American Indian or Alaska Native category dropped on the CPS panels with the multiracial category. However, the report²⁶ notes that it is difficult to draw reliable inferences about effects on this population given the very high rates of disagreement²⁷ for the American Indian or Alaska Native category in the Supplement (over 25 percent on all panels).

The NCS was not designed to provide sufficiently large samples of American Indians to detect such effects. One would also expect some differences in findings from a study based on responses to a self-administered questionnaire compared with those from a telephone or an in-person interview. Such differences in mode may help explain, for example, why much higher proportions of write-ins to the multiracial category on the NCS than on the CPS clearly represented two or more major race categories or a race and an Hispanic origin category.

Differences in the results of these and other studies are thus not so much inconsistent as they are complementary. For example, the CPS results may provide a good indication of effects that might occur in national telephone and in-person surveys, while those in the NCS and RAETT might do the same for self- administered questionnaires. The OMB has emphasized that the classifications of data on race and ethnicity in Directive No. 15 must be applicable across both (interview and mail return) modes, as well as in administrative records systems.

Differences in the effects found in each mode, as well as differences in the uses of data generated in each context, need to be assessed in considering the results of the research and testing for the OMB review of Directive No. 15. Data from the RAETT will be particularly important for detecting possible effects that the NCS could not in the context of self-administered questionnaires. This is because the RAETT targeted specific populations - - including American Indians and Alaska Natives, as well as specific Asian, Pacific Islander, and Hispanic origin groups.²⁸

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

¹ The National Content Survey is also known as the U.S. Census 2000 Test.

² Questions on ethnicity include Hispanic origin and ancestry.

³ The NCS was not designed to detect differences among questionnaire versions for this population. The RAETT sample was designed to do this. The mail return data examined for this report included, on average, only about 50 American Indians per panel, and no persons who reported as Alaska Native.

⁴ The decennial census collects greater detail on race and ethnicity than the Directive No. 15 categories; however, as required by the Directive, the detail can be combined into the standard categories.

⁵ Federal Register Notice, Standards for the Classification of Federal Data on Race and Ethnicity, Office of Management and Budget, Federal Register Vol. 59, No. 123 (59FR 29831-35), Thursday, June 9, 1994, pages 29831-29835.

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

⁶ Federal Register Notice, Standards for the Classification of Federal Data on Race and Ethnicity, Office of Management and Budget, Federal Register Vol. 60, No. 166 (60FR 44674-93), Monday, August 28, 1995, pages 44674-44693.

⁷ Gerber, Eleanor, and Manuel de la Puente. 1996. "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.

⁸ The Census Bureau has an exemption from the Office of Management and Budget to use the "Other race" categories.

⁹ Bennett, Claudette, Nampeo McKenney, and Roderick Harrison. 1995. "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 Federal Register Notice, Standards for the Classification of Federal Data on Race and Ethnicity, Office of Management and Budget, Federal Register Vol. 59, No. 123 (59FR 29831-35), Thursday, June 9, 1994, pages 29831-29835.

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

¹² McKenney, Nampeo R., and Arthur Cresce. 1990. "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.

¹³ McKenney, Nampeo, Claudette Bennett, Roderick Harrison, and Jorge del Pinal. 1993. "Evaluating Racial and Ethnic Reporting in the 1990 Census." Paper presented at the 1993 Joint Statistical Meeting of the American Statistical Association, San Francisco, CA, August 1993.

¹⁴ de la Puente, M. and R. McKay. 1995. "Research Improves Questions." Civil Rights Journal, Vol. 1, No. 1.

de la Puente, M. and R. McKay. 1995. "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 on the American Statistical Association: Section on Survey Research Methods, Vol. 1.

¹⁵ Bates, Nancy E., Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente. 1996. "Questionnaire Effects on Measurements of Race and Spanish Origin." Journal of Official Statistics (11) 1433-459.

¹⁶ U.S. Department of Commerce, Bureau of the Census, "Note to Correspondents," CB91-232, issued in 1991.

¹⁷ Write-ins to the race question were not used to edit the data when the multiracial category was marked or where the respondent provided two or more write-ins. We also analyzed unedited versions of the data, but present only results from the edited files in this report. The results from the unedited file were generally consistent with those reported here. Most differences can be clearly attributed to the editing.

¹⁸ These population groups are based on responses provided to the Hispanic origin and nativity questions collected in the reinterview. Persons not providing a response to these reinterview questions are excluded from these analyses.

¹⁹ McKenney, Nampeo, Claudette Bennett, Roderick Harrison, and Jorge del Pinal. 1993. "Evaluating Racial and Ethnic Reporting in the 1990 Census." Paper presented at the 1993 Joint Statistical Meeting of the American Statistical Association, San Francisco, CA, August 1993.

²⁰ The difference between the race question immediately preceding the Hispanic origin question and the inclusion of intervening questions could also contribute to any differences observed between the race distribution on Panels 2 and 1. The use of the term "Latino" in addition to the term "Spanish" and "Hispanic" in the Hispanic origin question may contribute to the differences in Hispanic origin reporting among the panels. We, therefore, cannot attribute the effects measured solely or entirely to the addition of the multiracial category.

²¹ This includes write-ins of ethnic groups that were coded to major categories. For example, "English and Black" was coded as White and Black, and in this section should be understood to represent two different race groups.

²² Bennett, Claudette, Nampeo McKenney, and Roderick Harrison. 1995. "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.

²³ These percentages sum to over 100 percent because they are not mutually exclusive: up to four entries to the multiracial write-in line were coded for each respondent who selected this reponse option.

²⁴ McKenney, Nampeo, Claudette Bennett, Roderick Harrison, and Jorge del Pinal. 1993. "Evaluating Racial and Ethnic Reporting in the 1990 Census." Paper presented at the 1993 Joint Statistical Meeting of the American Statistical Association, San Francisco, CA, August 1993.

²⁵ Bates, Nancy E., Elizabeth A. Martin, Theresa J. DeMaio, and Manuel de la Puente. 1996. "Questionnaire Effects on Measurements of Race and Spanish Origin." Journal of Official Statistics (11) 1433-459.

²⁶ Tucker, Clyde, Ruth McKay, Brian Kojetin, Roderick Harrison, Manuel de la Puente, Linda Stinson, and Ed Robison. 1966. Testing Methods of Collecting Racial and Ethnic Information: Results of the Current Population Survey Supplement on Race and Ethnicity, BLS Statistical Notes No. 40.

²⁷ Disagreement between the race reported in the CPS Supplement and in the original CPS interview.

²⁸ The RAETT, conducted in June 1996, had a sample of 114,000 housing units drawn from urban and rural areas of the country with different concentrations of racial and ethnic populations. It, therefore, contained census tracts and blocks with high concentration of households including American Indians, Alaska Natives, Asians, Pacific Islanders, Blacks, Hispanics, White ethnic groups, and persons who identify with more than one race or origin.

Source: U.S. Census Bureau, Population Division,
Special Populations Staff
Questions? / 1-866-758-1060
Last Revised: July 09, 2008 at 02:50:56 PM

People | Estimates | Projections | Glossary | Help

Skip this navigation

Census Bureau Links:   Home · Search · Subjects A-Z · FAQs · Data Tools · Catalog · Census 2000 · Quality · Privacy Policy · Contact Us

Page Last Modified: July 09, 2008