Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 205
Measuring Racial Discrimination 10 Measurement of Race by the U.S. Government Since the first U.S. census in 1790, statistics on race have been aprominent part of the nation’s censuses and surveys. The Constitution requires the federal government to conduct a census of the country’s population every 10 years for use in the allocation of seats in the House of Representatives. Although the uses of the data, the definitions of race, and the methods of data collection have changed, there continues to be intense interest in census data on race and, more recently, on ethnicity (see Anderson, 1988, 2000, for a history of the census). Today, these data are an integral part of the nation’s economic and social policies. Race and ethnicity statistics are used in important and politically sensitive areas, such as the enforcement of civil rights and antidiscrimination laws, and determination of voting districts. For example, state legislatures rely on census race and ethnicity data for geographic areas as small as individual blocks to ensure representation of black and other nonwhite voters within the new boundaries of voting districts that are revised every 10 years. Statistics on race and ethnicity are also used by federal regulators as statistical evidence in employment discrimination lawsuits, as a means of determining whether banks discriminate against minorities when they award home mortgages, and in class action court cases alleging racial discrimination. To meet these and other information requirements, the U.S. statistical system has changed considerably over the more than 200 years of the country’s existence. Nevertheless, problems in the collection of accurate data on race and ethnicity remain. As the population has changed, so have the country’s views about defining race. On the one hand, recent news re-
OCR for page 206
Measuring Racial Discrimination ports have continued to focus on immigration and the country’s heritage as a “melting pot” of many races and cultures. On the other hand, prejudice toward disadvantaged racial groups continues to exist, and many members of such groups live in lower economic and social circumstances than the rest of the population. Because the federal government has responsibility for providing information on all groups within the country’s population, the statistical system continues to struggle with questions about the number of races for which data are to be collected, how to define and enumerate them accurately, what labels to apply to them, and how to classify persons of multiracial background. In addition, experience has shown the considerable difficulty involved in enumerating the Hispanic population, which appears to bridge both ethnicity and race concepts (see Chapter 2). This chapter first provides a brief history of the federal government’s collection of data on race and ethnicity. It then reviews the standards for government collection of data on race and ethnicity issued by the U.S. Office of Management and Budget (OMB) in 1977 and the revision of those standards in 1997. Next we summarize race and ethnicity data collected in the 2000 census, paying special attention to data for those who selected more than one race. We then discuss some of the issues involved in interpreting and using the new multiple-race data and briefly review research under way in the federal statistical system to resolve those issues. Finally, we make recommendations for continued collection of data on race and ethnicity with categories that are responsive to changing concepts of race among groups in the U.S. population. We further stress the need for sustained research by federal agencies to develop best practices for the measurement of race, to gain knowledge of how different groups report race and of changes in such reporting over time, and to inform users of the meaning of different measures of race and ethnicity. HISTORY Article 1, Section 2, of the U.S. Constitution, written in 1787, requires a census every 10 years to determine the number of people living in each of the states. The requirement for data on race grew out of the struggle in the Constitutional Convention over the distribution of power between the North and the South. Because most of the country’s slave population lived in southern colonies, the Founding Fathers searched for a way to balance sectional power. The language adopted at the convention—and included in the Constitution—was that “Representatives and direct Taxes shall be apportioned among the several States…by adding to the whole Number of free Persons, …excluding Indians not taxed, three fifths of all other Persons.” The first census in 1790 collected the data required by the Constitution—on free white men (over and under age 16), free white women, and
OCR for page 207
Measuring Racial Discrimination other free persons (who were black)—by direct enumeration or enumerator observation of race. Indians (not taxed) were excluded from the census counts. As time progressed, the labels changed; skin color was frequently introduced to identify racial differences; and in some censuses, more detail about the amount of nonwhite blood was listed—for example, whether Indians were full-blooded and whether blacks were mulatto, quadroon, or octoroon. After the Civil War, when slavery was outlawed by the Thirteenth Amendment, Section 2 of the Fourteenth Amendment provided for a count of the “whole number of persons in each State, excluding Indians not taxed.” Nevertheless, the data system continued to count the number of whites, blacks, and Indians. When large numbers of immigrants from Asia and from Eastern and Southern Europe began coming to the United States, the demand for more information on race, ancestry, ethnicity, and languages rose, and new categories were added. A category for Chinese was added in 1870, and, as more immigrants from East Asia came to this country, other categories (e.g., Japanese, Hindu, and Korean) were added as well. In recognition of the fact that increasing numbers of Hispanic immigrants had established themselves and their families in this country, the 1970 census added a separate question on Hispanic origin.1 As can be seen from Table 2-1 (see Chapter 2), the number of racial categories continued to grow, and by 1990 there were 15 separate categories plus a separate question with four categories for Hispanic origin (Mexican, Puerto Rican, Cuban, other Hispanic). In addition, methods of collecting census data changed, with mail increasingly substituting for direct enumeration. By 1960 the Census Bureau began some data collection by mail, and data collection by telephone was used both to conduct entire surveys and to supplement mail collection. These changes obviously made enumerator observation of race impossible. As a result, the manner in which race was determined in government censuses and surveys changed. Today, the household member responding to the questionnaire or survey is asked to identify his or her own race/ethnicity and in some cases that of other members of the household as well.2 In the 1960s and 1970s, civil rights laws—such as the Civil Rights Act of 1964 banning discrimination in employment and public accommodations, the Voting Rights Act of 1965, and the Fair Housing Act of 1968—were passed to prohibit the exclusion of disadvantaged racial groups from social and economic privileges. The emergence of the new legislation re- 1 Table 2-1 shows that the 1930 census included Mexicans as a race, but the category was dropped in later censuses. (See Chapter 2 for a fuller explanation.) 2 The issue of self-identification versus interviewer observation is discussed more fully in Chapter 2.
OCR for page 208
Measuring Racial Discrimination quired the collection of race and ethnicity data to monitor compliance. Thus, the purpose of racial categorization shifted politically from denying opportunities to disadvantaged racial groups to ensuring compliance with civil rights laws and promoting antidiscrimination policies (Anderson and Fienberg, 1999b; Nobles, 2000). Many nonwhite advocacy groups lobbied for the federal government to continue collecting data on race and ethnicity to ensure civil rights protection for their groups. For example, several Asian American groups insisted that their specific categories be added to the 1980 and 1990 census questions on race. Government statistical agencies also undertook research on the response effects on race/ethnicity reporting of wording, questionnaire design, data collection techniques, and other aspects of survey design in an effort to obtain better data, expand coverage of minorities, and improve scientific survey methods (Tucker and Harrison, 1995; Tucker and Kojetin, 1996). As it became clear that some groups within the population, especially disadvantaged racial groups, continued to be counted less accurately than others, increasing attention was focused on the problems caused by the differential undercount of these populations and how to overcome those problems.3 STANDARDS FOR THE COLLECTION OF RACE AND ETHNICITY DATA The 1977 OMB Standards Because of the need for consistent data based on uniform definitions for use in connection with civil rights legislation and monitoring of equal treatment, as well as for other public policy uses of race/ethnicity data, in 1977 OMB developed and issued to federal statistical agencies a set of standards for the collection of such data (Nobles, 2000). Statistical Directive Number 15 established a classification system that included four major categories for race—white, black, Asian or Pacific Islander, and American Indian or Alaskan Native—and two for ethnicity—Hispanic and non-Hispanic.4 It 3 For several decades, controversy about the census focused primarily on population coverage and the differential undercount of nonwhite groups. Despite the special steps that have been taken to improve their response rates, more nonwhite than white people have been missed in the census. Because the number of people counted can affect apportionment for the House of Representatives as well as allocations of funds to states and local governments, the undercount issue has been surrounded by political controversy. (See Anderson and Fienberg, 1999b, for a discussion of the issues involved, and National Research Council, 2004, for an evaluation of the problem in the 2000 census.) 4 The census and other surveys also include various racial subcategories, such as Japanese, Chinese, and Vietnamese.
OCR for page 209
Measuring Racial Discrimination also encouraged self-identification as the preferred method of collecting data on race. Respondents were instructed to choose only one race. The standards were required to be used in censuses and surveys conducted by the federal government, as well as for federal administrative records and research (U.S. Office of Management and Budget, 1977). Although the same definitions were also used in private surveys (especially those financed by the federal government, such as the General Social Survey), private surveys sometimes collected less detailed data on race, combined categories into broader groups, or formulated the race questions somewhat differently. By the 1990 census, questions had been raised about the continued relevance of the 1977 standards. Many population changes had occurred since 1977, and the population of disadvantaged racial groups had grown considerably. In fact, the rate of population increase for blacks, American Indians, Eskimos, and Aleuts, as well as for Asians and Pacific Islanders, between 1980 and 1990 had been higher than the rate for the white population. In addition, questions began to be raised about how to enumerate race for children born of interracial unions. Statistical agencies had initiated research on the effects of differences in question wording and placement. They believed research was required on how to define race and ethnicity, which labels to attach to the various categories, and what to do about the rising number of multiracial individuals. The issues addressed in that research were discussed widely with many population groups (e.g., Arabs, Cape Verdeans, Muslim West Asians, and Creoles) who wanted separate categories for population groups not yet included in the census categories and increased detail about countries of origin and languages used. These groups actively campaigned to add their categories to the census. Congressional hearings were held in 1993 (by the House Subcommittee on Census, Statistics, and Postal Personnel), and OMB decided to undertake a complete review of the 1977 standards. Other kinds of issues were also raised. Many groups, concerned about children of interracial marriages, argued that they should not be forced to select the race of only one of their parents and asked for a new multiracial category. Other groups, worried about the use of racial categories as the basis for antidiscriminatory action, feared that use of a multiracial category would dilute data needed for the nation’s civil rights programs (Anderson and Fienberg, 2000; U.S. Office of Management and Budget, 1997). Research by Federal Statistical Agencies on Race and Ethnicity OMB established an interagency committee to assist it in its review of the 1977 standards. That committee established an interagency research working group, chaired jointly by the Census Bureau and the Bureau of Labor Statistics (BLS), to develop an agenda for the specific questions to be
OCR for page 210
Measuring Racial Discrimination addressed and the methods to be used for determining how changes might affect the measurement of race and the quality of the data obtained. The working group reviewed the criticisms and suggestions made thus far and developed a research agenda focused in particular on how to enumerate people who identify themselves as multiracial, whether to add new racial categories, whether to change the terminology used for racial categories, and whether to combine race and Hispanic origin in one question or have separate questions (Tucker and Harrison, 1995). In addition, the National Research Council’s Committee on National Statistics conducted a workshop on these issues (National Research Council, 1996). In May 1995, a special supplement to the Current Population Survey (CPS) was undertaken, focused primarily on three issues: (1) the ability for respondents to select a multiracial category; (2) whether Hispanic should be added to the list of races or whether, as in the past, a separate question on ethnicity should be used; and (3) use of such alternative race/ethnicity labels as black, African American, or Negro, and Hispanic, Latino, or Spanish. In the 1996 National Content Survey and the Race and Ethnic Targeted Test, the Census Bureau explored multiracial response options, the combining of Hispanic origin with race, and race wording issues. Other statistical agencies were also involved. For example, the National Center for Education Statistics (NCES) explored how race was recorded in schools, and the National Center for Health Statistics (NCHS) reviewed the determination of race in the administrative records with which it dealt. The following results were among the most important findings from the CPS supplement study: The number of respondents identifying themselves as Hispanic was higher when a separate question on Hispanic origin was followed by another question on race than when Hispanic origin was combined with the race question (see Chapter 2 for discussion). In the two test panels (each comprising 15,000 households) in which a question on whether respondents were or were not Hispanic or Latino was followed by a separate question on race, the inclusion of a multiracial category in the race question had little effect on the percentage identifying themselves as Hispanic—10.79 percent reported Hispanic origin when there was no multiracial category in the race question, as compared with 10.41 percent who reported Hispanic origin when the race question included a multiracial category. In the two test panels of 15,000 respondents each in which Hispanic, Latino, or “of Spanish origin” was included as a racial category instead of as a separate ethnicity question, smaller percentages reported Hispanic origin—7.5 percent identified themselves as Hispanic when there was no mul-
OCR for page 211
Measuring Racial Discrimination tiracial category, as compared with 8.6 percent when a multiracial category was included (Tucker and Kojetin, 1996). Not surprisingly, the number selecting “other” for race was smaller when the Hispanic question was combined with race. The count of the white population was smaller when Hispanic was listed as a category in the race question, apparently because a number of Cubans who identified themselves as white when separate race and ethnicity questions were asked could not do so when Hispanic was listed only as a part of the race question (Tucker and Kojetin, 1996). Despite these results, it is interesting to note that when asked for their preference, a substantial majority of Hispanics preferred to have the Hispanic question included with race. This was true both for those panels with a separate ethnicity question and for those panels with Hispanic as a racial category, although the former group had a somewhat lower percentage (Tucker and Kojetin, 1996). As mentioned in Chapter 2, different question formats can affect responses to questions on race and ethnic origin. Moreover, there are different perspectives on race and ethnicity even within the Hispanic population, making it difficult to interpret data on race and Hispanic origin from surveys (de la Garza et al., 1992; Denton and Massey, 1989; Harris, 2002). The BLS test also provided other information. Only a small group (less than 2 percent) identified themselves as multiracial. The fact that respondents could select a multiracial category had little effect on other racial categories, with the American Indian/Alaskan Native group as a possible exception. Although no firm conclusions could be drawn from this test about preference for the use of the term African American rather than black or for Native American rather than American Indian, a sizable minority of each group preferred those terms (Tucker and Kojetin, 1996). A year later the Census Bureau undertook two surveys to explore some of these issues. The National Content Survey tested the effects of the addition of a multiracial category, placement of the Hispanic origin question, and combinations of those changes. The results of the National Content Survey were similar to those of the CPS supplement. They showed that Only about 1.0 percent chose the multiracial category, and the choice had no statistically significant effect on the other racial groups, with the possible exception of the Asian and Pacific Islander category. Nonresponse for Hispanics was significantly reduced when the Hispanic origin question came before the race question. Also, this placement increased the number of Hispanics identifying themselves as white in the racial category.
OCR for page 212
Measuring Racial Discrimination A second Census Bureau test was conducted in the 1996 Race and Ethnic Targeted Test. This test was focused on measuring the effects of changes in the race and ethnic standards, especially on smaller population groups, such as American Indians and Alaskan Natives, Asians, and Hispanic subgroups (e.g., Puerto Ricans and Cubans). It also tested a “mark more than one” format for the race question, finding that the number of respondents reporting Hispanic origin did not decline in the combined race/ethnicity question format compared with the two-question format when respondents could check more than one race. Moreover, response rates were higher for the combined format than for the separate race and ethnicity questions (Hirschman et al., 2000). In addition, NCES undertook an investigation of how race and ethnic classifications are used in the public schools. It found that 55 percent of all public schools record the race and ethnicity of students only when they first enroll in school, and about one-quarter collect these data each year. Some 45 percent of schools ask parents to select one of the OMB race/ethnic categories for their children, 17 percent ask them to select from a list used by the school district, and in some cases parents may write in their own category. Interestingly, more than one-fifth (22 percent) of the public schools use teacher or administrator observation to determine the race/ ethnicity of students. The proportion determined by observation is much higher in the Northeast (44 percent) (National Center for Education Statistics, 1996). The 1997 OMB Revised Standards Building on these results, and following public comment and hearings, Statistical Directive Number 15 was revised in 1997 to define the categories to be used in the 2000 census and for other government surveys (U.S. Office of Management and Budget, 1997). The new standards—Standards for Maintaining, Collecting and Presenting Federal Data on Race and Ethnicity—included three major changes. First, five racial categories were to be used in measuring race: black or African American, white, Asian, American Indian and Alaskan Native, and Native Hawaiian or Other Pacific Islander. Second, there was a requirement that respondents be permitted to select more than one race. Third, the question on ethnicity was to be simplified by asking respondents whether they were Hispanic or Latino, and the ethnicity question was to be asked before the race question. Although the standards were to be used by all federal agencies in the future, agencies were permitted to add categories when more detailed data were needed. In issuing the new standards, OMB emphasized that “the categories represent a socialpolitical construct designed for collecting data on the race and ethnicity of
OCR for page 213
Measuring Racial Discrimination broad population groups in this country, and are not anthropologically or scientifically based” (U.S. Office of Management and Budget, 1997:16). The new standards were used in the 2000 census, with the Census Bureau adding the category “other” to the five racial categories established by the standards. GOVERNMENT DATA ON RACE AND ETHNICITY The 2000 Census In the 2000 census, nearly 275 million people or almost 98 percent of the total population identified themselves as one race only, whereas 2.4 percent or 6.8 million people selected two or more races. Of those who chose one race only, 75 percent identified themselves as white and 12.3 percent as black or African American (see Table 10-1). The Asian population, which had grown by 48 percent between 1990 and 2000, was the next largest group identifying with one race (3.6 percent), followed by American Indians and Alaskan Natives (0.9 percent) and Native Hawaiian and other Pacific Islanders (0.1 percent). These results demonstrate the remarkable increase in the country’s racial diversity, both because the white population has not increased as rap- TABLE 10-1 Race and Hispanic Origin Population in the United States, 2000 Race and Hispanic or Latino Origin Number Percent of Total Population Race Total population 281,421,906 100.0 One race 274,595,678 97.6 White 211,460,626 75.1 Black or African American 34,658,190 12.3 American Indian and Alaskan Native 2,475,956 0.9 Asian 10,242,998 3.6 Native Hawaiian and Other Pacific Islander 398,835 0.1 Some other race 15,359,073 5.5 Two or more races 6,826,228 2.4 Hispanic or Latino Origin Total population 281,421,906 100.0 Hispanic or Latino 35,305,818 12.5 Not Hispanic or Latino 246,116,088 87.5 SOURCE: Data from U.S. Census Bureau (2001b).
OCR for page 214
Measuring Racial Discrimination idly as the nonwhite population and because the sizes and mix of many disadvantaged racial groups have changed. For example, 100 years ago, 87 percent of the population was white, 12 percent was black, and only about 1 percent was from some other group. By 2000 the white population had declined to 75 percent; the black population, at 12.3 percent, had risen only slightly as a proportion of those who identified themselves as belonging to a single race; and the Asian population had become a significant racial category (Anderson, 2000). But whites and blacks do not fully explain the changing race/ethnicity makeup of the American population. The 2000 census counted nearly 15 million people, or 5.5 percent of those identifying with a single race, in the “some other race” category. Further breakdowns by the Census Bureau show that a very large number of those identifying with “some other race” (14.9 million people) were Hispanics, who responded to the race question by selecting the “other” category (U.S. Census Bureau, 2001b). This result suggests that many Hispanics do not identify with the census racial categories and underscores the need for more research on how to measure racial identification more accurately, especially for the Hispanic population. In the 2000 census, the number of Hispanics of any race was close to the number of blacks (including those identified as black only and black together with some other race; see Table 10-1). In the decade between 1990 and 2000, the number of Hispanics in the population increased at a much faster rate than was the case for blacks: The rate of increase was 57.9 percent for Hispanics, 3.5 times the 15.6 percent rate for blacks. The multiracial population counted in the 2000 census was small—only 6.8 million people or 2.4 percent of the total population. Those identified as belonging to more than one race in response to the race question were young (4 percent of the population under age 18); only about 4 million (1.9 percent of the population over age 18) were adults (U.S. Census Bureau, 2001b). The group included many Hispanics; in fact, 2.2 million Hispanics selected more than one race—nearly one-third of the 6.8 million who reported two or more races. This finding suggests once again that Hispanics have differing conceptions of race and ethnicity and are not certain how to respond to the racial categories on the census questionnaire (see Chapter 2). Categories checked by the multiracial population varied, with such combinations as white and American Indian, white and Asian, white and black, and white and other. As part of its work on the new American Community Survey, planned to replace the decennial census long form in 2010, the Census Bureau fielded a large (700,000 household) survey in 2000 to provide data for comparison and analysis with the 2000 census long form. Estimates for race in the new survey, called the C2SS, differed in some ways from those in the census,
OCR for page 215
Measuring Racial Discrimination TABLE 10-2 Household Data on Race and Ethnicity in Census 2000 and C2SS Race/Hispanic Origin Census 2000 C2SS C2SS–Census Total 273,643,273 273,643,269 –4 White alone 206,127,572 211,867,275 5,739,703 Black or African American alone 32,939,206 32,256,169 –683,037 American Indian and Alaskan Native alone 2,400,916 2,117,034 –283,882 Asian alone 10,037,229 10,453,603 416,374 Native Hawaiian and other Pacific Islander alone 388,153 436,612 48,459 Some other race alone 15,053,131 10,700,143 –4,352,988 Some other race Hispanic 14,600,195 10,107,129 –4,493,066 Two or more races 6,697,066 5,812,433 –884,633 Two or more races/Hispanic 2,181,583 1,643,812 –537,771 NOTES: Numbers for both the census and the C2SS are for persons in households. The last column is the difference between C2SS and the 2000 Census (i.e., C2SS minus Census). The CS22 did not cover group quarters. SOURCE: Unpublished data from the Racial Statistics Branch, U.S. Census Bureau. even when corrected for differences in coverage (see Table 10-2).5 The C2SS found slightly more whites and Native Hawaiian and other Pacific Islanders than did the census and somewhat smaller estimates of multiple races as well as for the “other race” category. Those identifying themselves as being of two or more races amounted to 2.1 and 2.4 percent of the household population in the C2SS and the census, respectively. Those reporting “some other race” were a smaller group—amounting to 3.9 percent in the C2SS versus 5.5 percent in the census. Once again, Hispanics appear to account for most of the difference, as the C2SS showed more white Hispanics and fewer “some other race” Hispanics than did the census. Unfortunately, the questionnaires were designed by different groups, and the layout of the race questions was somewhat different in the two surveys, with the C2SS race questions being printed horizontally across the questionnaire and the cen- 5 This survey was designed to cover a large sample of households. It excluded group quarters (e.g., people living in nursing homes and in prisons and students living away at school). Because the census is intended to cover all people in the country, it includes those living in group quarters. Table 10-2 therefore includes separate data for persons in households so that the results of the C2SS can be compared with those of the 2000 census. It should be noted that the C2SS, as a sample survey, is subject to sampling error.
OCR for page 216
Measuring Racial Discrimination sus questions printed vertically. Census Bureau staff believe this difference in layout may account for much of the difference in the results between the two surveys. Race in Other U.S. Government Surveys The wording of race and ethnicity questions and their placement on questionnaires have often differed among government surveys. Self-enumeration is used to the extent possible, but that does not mean each person covered in a survey always responds for himself or herself. In many of the major government surveys, the questions are asked of only one person, the reference person, who responds to all questions for all members of the household. The agencies currently are working to implement the revised race and ethnicity standard so that all future surveys will include a question about Hispanic/Latino origin before the race question is asked, and all respondents will be permitted to select more than one race. The agencies anticipate that all major surveys will have complied with the new standard by 2003, if not before. One federal government survey, the Health Interview Survey (HIS), collected by the Census Bureau for NCHS, has a 20-year history of asking respondents to select one or more racial categories. In addition, those who select multiple racial categories in the HIS are asked to indicate which of those races “would best represent your race.” These data will be especially useful to the statistical system for understanding issues of data presentation and development of multiracial historical series. Ongoing Research The Census Bureau and BLS continue to plan and carry out research designed to study issues associated with the collection of data on race and ethnicity. In particular, they plan to conduct research on racial identification of children in surveys and on potential effects of the mode of data collection on responses on race. In May 2002 BLS tested the new race question in a supplement to the CPS. The question wording varied by age—information for household members aged 12 and older was obtained about the race the person considered himself or herself to be, while information for younger household members was obtained about the race the respondent considered the child to be. The new race question, with the wording variation by age, was used in the CPS beginning in January 2003. In addition, the Census Bureau and BLS will conduct field tests in an attempt to determine the effect of different modes of data collection on race responses. These tests will include collection by computer-assisted telephone
OCR for page 217
Measuring Racial Discrimination interviewing, computer-assisted personal interviewing, and personal enumeration by use of paper and pencil. The two agencies have agreed on the following wording for the race question on the tests: “Please choose one or more races that [you/name] consider[s] [yourself/himself/herself] to be: White, Black or African American, American Indian or Alaskan Native, Asian, or Native Hawaiian or Other Pacific Islander.” Cooperative efforts are also under way among the statistical agencies to agree on the wording of race and ethnicity questions to achieve uniformity across most of the government surveys. ISSUES IN THE REPORTING OF DATA ON MULTIPLE RACES Because of the new OMB standard on enumerating the multiracial population, government statistical agencies must address a number of issues. There may be important reasons for some surveys to use differing approaches, but many of the same issues must nonetheless be addressed. For example, how and how often should multiple races be included in ongoing releases and other publications? How can confidentiality be maintained when samples of those selecting particular multiracial categories are small? How can the agencies ensure that all of the multiracial categories that are published are statistically reliable? How should the new multiracial data be linked to the old single-race data for purposes of historical analysis? How should the data be mapped across various sources, and how will data users identify mismatched data? Finally, for agencies that rely on administrative records data for some or all of their data, how can consistency between the survey and the administrative data be maintained? These are all important questions, and many statisticians and analysts within the federal statistical agencies have been examining alternative approaches to addressing them. Many are still under study, and much will depend on the purpose of the analysis to be undertaken (Tucker et al., 2000).6 In preparation for a review of statistical agency action on the new racial guidelines, representatives of seven government agencies held discussions with one another: the Census Bureau, BLS, NCHS, NCES, the Bureau of Justice Statistics (BJS), the National Science Foundation, and the statistical policy group at OMB. These discussions revealed that considerable progress has already been made toward implementing the new guidelines in the government’s surveys, although several problem areas remain. A brief dis- 6 Tucker et al. focus on alternative methods of linking the new race data to data collected under the old guidelines. They describe a number of alternative approaches and conclude that “it is likely that which method is best at matching a reference distribution for outcome measures will depend on the outcome being examined” (2000:21).
OCR for page 218
Measuring Racial Discrimination cussion of the most important of these follows, along with some proposed solutions. Publication and Release of Data Recognizing the problems inherent in dealing with the 63 racial categories that can be developed from information in the 2000 census, OMB, in its discussion of the 1997 standards, identified the need for further research by federal statistical agencies on methods for reporting the numbers of people who selected more than one race (U.S. Office of Management and Budget, 1997). In addition to developing publication rules for multiracial data from their own surveys, the federal statistical agencies needed to ensure that data required for enforcement and monitoring of civil rights could be made available. While the statistical agencies were engaged in the research necessary to develop rules for the publication of multiracial data for their own publications, OMB issued Bulletin Number 00-02 as guidance to the government’s executive branch on the aggregation and allocation of racial data for civil rights monitoring and enforcement (U.S. Office of Management and Budget, 2000). Rules for Combining Multiracial Data for Civil Rights Cases OMB Bulletin 00-02 lists the five single-race categories and four additional multiracial categories—American Indian or Alaskan Native and white, Asian and white, black or African American and white, and American Indian or Alaskan Native and black or African American. The guidance establishes aggregation rules for agencies to determine counts of multiracial groups, providing for the “collection of information on any multiple race combinations that comprise more than one percent of the population of interest,” one example being that, “in Hawaii, there may well be combinations of racial groups that meet this threshold such as Native Hawaiian or Other Pacific Islander, and Asian” (U.S. Office of Management and Budget, 2000:1-2). Moreover, allocation rules for civil rights monitoring and enforcement provide that, at the aggregate level, multiracial responses combining a nonwhite race with white are to be allocated to the nonwhite race for analysis purposes. Responses that include two or more nonwhite races are allocated to the nonwhite race on which the alleged discriminatory behavior was based. When action requires assessing disparate impact discriminatory patterns (see Chapter 3), the patterns are to be analyzed “based on alternative allocations to each of the minority groups” (U.S. Office of Management and Budget, 2000:2). These guidelines have been criticized by civil rights advocates and racial theorists as arbitrary rules for collecting and tabulating racial data (see Harris, 2002; Harrison, 2002; see also Chapter 2).
OCR for page 219
Measuring Racial Discrimination Publication of Multiracial Survey Data The research conducted by the government’s statistical agencies has focused on rules for deciding how much multiracial data to publish. In the past, when sample sizes permitted, most agencies published data on whites, blacks, Hispanics, and a category of “other races.” In the future, in accordance with the new OMB standards, efforts will be made to publish data separately for Asians, a category that increased in size in the 2000 census. The expectation is that data for the Hispanic population will be improved considerably once the change in placement of the Hispanic/Latino question has been fully implemented in government surveys because the new procedure should result in better coverage of that population. For example, the CPS research mentioned above showed that use of a separate Hispanic origin question placed before the race question significantly reduced Hispanic nonresponse (Tucker and Kojetin, 1996). However, most of the agencies still have not decided how or how often to publish data for the multiracial categories. Because the multiracial population thus far appears to be quite small, sample sizes for most household surveys will make monthly publication consistent with confidentiality rules difficult if not impossible. The probability is that those who select multiple races in the surveys will most often be combined with those placed in the category for “other races” or included in the total. This will surely be the case for regular monthly publication of data from such surveys as the CPS and the HIS. The agencies will publish data for the multiracial category separately when tests show that the numbers involved are sufficiently large to make the data reasonably accurate. For most of the important, large household surveys, therefore, multirace data will be published at best on a quarterly basis and in some cases only on a semiannual or annual basis. Some agencies, such as NCHS, have indicated that they will attempt to make multiracial data available by pooling the data over several years because confidentiality restrictions will make it impossible to publish them more frequently. BJS believes that samples of criminal events collected in the Crime Victimization Survey are far too small to warrant separate publication of multiracial data, although the bureau intends to collect such data. It is unlikely, therefore, that the socioeconomic aspects of the multiracial categories will achieve much prominence for some time to come.7 7 Each of these surveys has a different sample design and somewhat different use of interviewers and telephone responses. The methods employed in the collection of data also differ, because the government uses a variety of techniques—paper-and-pencil personal interview collection, computer-assisted personal interview collection, and computer-assisted telephone collection. For repetitive, time-series surveys, the household person responding to the survey also can differ at various times, depending on the subject matter and which household members are available for the interview. Each of these processes, as well as others, can affect the quality and consistency of the data obtained.
OCR for page 220
Measuring Racial Discrimination Time-Series Data Research is under way in several of the statistical agencies on methods for developing historical series using the old and new racial categories. The alternative—to announce a break in the series—is unattractive, especially for those who analyze trends. BLS and NCHS have undertaken research to work the new data backwards so that they can be interpreted as a single series, before and after 2000. Several data sources are being used in this research to bridge the new and old racial categories. BLS has arranged for all respondents to the National Longitudinal Survey to be asked the race question a second time in order to obtain data for the same people using the old and new racial categories. BLS has also undertaken research to use information collected in special supplements to the CPS to introduce a CPS historical series in 2003, using population weights for both 1990 and 2000. Many of the agencies will also make use of the time series developed in the HIS to help develop a bridge to the old racial categories. For many years, the HIS has permitted respondents to select one or more races, and for those who do so, ask a follow-up question to determine the race with which the respondent identifies most closely. These data should provide a reasonable foundation for developing historical data for the NCHS health surveys and may also assist other agencies in linking data reported under the 1977 and 1997 standards Administrative Data Although education and employment data may be available as early as 2005, it appears unlikely that data collected from other forms and administrative records will provide information on multiple races in the near future. BJS, for example, works with probation offices, jails, and state correction agencies to collect data from their records, but in most cases these forms include very limited racial data. In the case of the vital records system developed through cooperation between NCHS and the states, the problem is that data on race either are not present at all or are subject to considerable understatement. For example, the race of a child is not recorded on the birth certificates in most states. Information on race on death certificates is usually furnished by physicians or funeral directors, who may have little knowledge of the deceased. The result is that racial information on death certificates may be inaccurate or not reported at all. Population Controls Census population estimates, together with up-to-date data on immigration, emigration, and births and deaths, are used as controls for all gov-
OCR for page 221
Measuring Racial Discrimination ernment sample surveys to weight the sample data to totals that represent the population groups the sample has been selected to represent. It is clear that much of the success in the handling of the post-2000 census data on race and ethnicity will depend on the manner in which the new population controls are developed. If there are coverage or estimation errors in the population weights, the data from the surveys will reflect those errors. Methods for developing population counts are especially complex because the Census Bureau must develop these controls not just for the country as a whole but also for states and, for some surveys, for a number of individual areas, some of them quite small. The quality of these population counts—and the detail in which they are developed—can affect the presentation of data from all of the nation’s household surveys, as well as private research that rely on these data. SUMMARY, CONCLUSION, AND RECOMMENDATIONS Data constitute an important tool in defining discrimination and in assisting in the reduction of inequities in treatment based on race. This is especially true of statistical information on race and ethnic categories. Although the country has been collecting such information for more than 200 years, and scientific advances have greatly improved the data collected by the federal government, race and ethnicity data remain difficult to define, and racial categories are frequently not well understood. The federal government’s collection of data on race has changed over time, in part reflecting changing conceptions of race in the United States. In 1997, OMB revised standards for the collection of data on race and ethnicity in the 2000 census and other government surveys. The changes resulted in more realistic categories and labels and permitted respondents to select more than one race. Although government standards are not always consistent with or comparable to scholarly discussions of the meaning of race, the collection of such data is useful. Conclusion: Data on race and ethnicity are necessary for monitoring and understanding evolving differences and trends in outcomes among groups in the U.S. population. Differences in ancestry, language, and culture, as well as societal attitudes toward race and ethnic differences, influence how people identify with race. The growing number of Hispanics and individuals who identify with more than one race adds to the complexity of measuring racial self-identification. In addition, different respondents interpret questions differently, which can affect the accuracy of their responses (e.g., Hispanics have differing conceptions of race and ethnicity).
OCR for page 222
Measuring Racial Discrimination Because the nation’s statistical system is highly decentralized, the data produced by the system may lack consistency and vary considerably in quality. We lack information about such differences among racial subgroups. In addition, survey practices and data collection methods differ considerably depending on the type of survey and the kind of respondent. Federal government guidelines make self-identification of race by respondents the preferred means of collection, “except in instances where observer identification is more practical, e.g., completing a death certificate” (U.S. Office of Management and Budget, 1997:8). Even so, in many federal government surveys—including the census, from which the country’s most comprehensive data on race and ethnicity are developed—the household member who responds to the questions identifies the race and ethnicity of all members of the household. In many respects, the changes in race/ethnicity categories incorporated into the 2000 census are useful. During the next decade, the federal government needs to further improve race and ethnicity data. Recommendation 10.1. The federal government and, as appropriate, state and local governments should continue to collect data on race and ethnicity. Federal standards for race categories should be responsive to changing concepts of groups in the U.S. population. Any resulting modifications to the standards should be implemented in ways that facilitate comparisons over time to the extent possible. Recommendation 10.2. Data collectors, researchers, and others should be cognizant of the effects of measurement methods on reporting of race and ethnicity, which may affect the comparability of data for analysis: To facilitate understanding of reporting effects and to develop good measurement practices for data on race, federal agencies should seek ways to test the effects of such factors as data collection mode (e.g., telephone, personal interview), location (e.g., home, workplace), respondent (e.g., self, parent, employer, teacher), and question wording and ordering. Agencies should also collect and analyze longitudinal data to measure how reported perceptions of racial identification change over time for different groups (e.g., Hispanics and those of mixed race). Because measurement of race can vary with the method used, reports on race should to the extent practical use multiple measurement methods and assess the variation in results across the methods.
Representative terms from entire chapter: