660 REVISÃO REVIEW

The concept and measurement of race and their relationship to public health: a review focused on Brazil and the United States

O conceito e mensuração de raça em relação à saúde pública no Brasil e nos Estados Unidos

Claudia Travassos 1 David R. Williams 2

Abstract “Race is a social construct, but as for other as- pects of social stratification, with biological 1 Departamento de Race has been widely used in studies on health consequences” 1. Informações em Saúde, and healthcare inequalities, especially in the The notion that health is influenced by the Centro de Informação Científica e Tecnológica, United States. Validity and reliability problems social position of individuals has been known Fundação Oswaldo Cruz, with race measurement are of concern in public for many centuries. Nancy Krieger 2 notes that Rio de Janeiro, Brazil. health. This article reviews the literature on the since Hippocrates the relationship between 2 Department of Sociology and Survey Research Center, concept and measurement of race and compares health and social position has been acknowl- Institute for Social Research, how the findings apply to the United States and edged. It has also been shown that social dis- University of Michigan, Brazil. We discuss in detail the data quality is- parities in mortality exist for almost all causes Ann Arbor, U.S.A. sues related to the measurement of race and the of death in most societies, and these disparities Correspondence problems raised by measuring race in multira- have been increasing in recent decades in sev- Claudia Travassos cial societies like Brazil.We discuss how these is- eral developed countries 3. Av. Brasil 4365, Rio de Janeiro, RJ sues and problems apply to public health and Race has been used extensively in the med- 21045-900, Brasil. make recommendations about the measure- ical and public health literature, especially in [email protected] ment of race in medical records and public the United States, to measure social differences health research. in health outcomes and treatment, and its use has increased in recent decades. In the US, Race; Medical Sociology; Equity; Review Litera- there is a vast literature that relates race to dis- ture; Skin Color parities in health outcomes, which shows that race is an important predictor of health status. “Blacks” in the US are disadvantaged compared to “Whites” on most indicators of economic status and health. Despite a reduction in these racial inequalities on both of these indicators during and immediately after the Civil Rights movement (the mid-1960s to the mid-1970s), they have remained large or have widened ever since 3. In the US, adjustment for socio-econom- ic status (SES) always reduces and sometimes even eliminates racial disparities in health. A recent publication of the Institute of Medicine

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 661

also documented that there are large racial dif- and problems related to the use of this catego- ferences in the quality and intensity of medical ry in public health. treatment in the US, even after adjustment for access factors, SES, and severity of illness 4. The race concept In Brazil, there are fewer studies of racial in- equalities in health. Batista 5, using data from Numerous authors have argued recently that death certificates, has shown that “Black” men race is a social construct, with only limited (if and women had the highest crude mortality any) biological meaning. Human race is viewed rates in 1999 in the State of São Paulo. Data as a product of our history and culture and not based on census and national household sur- as a marker of our genes. As J. Kaufman 11 (p. veys show that aggregate infant mortality in 101) argues, race is a product of our cultural Brazil in the years 1977, 1987, and 1993 was imagination: “…race is not in our heads be- higher for “Blacks and “Pardos” (“Browns”) and cause it is real, but rather it is real because it is that it declined at a lower rate when compared in our heads”. Thomas LaVeist 12 reviewed the with “Whites” 6. Martins & Tanaka 7, using data definitions of race in various medical dictio- from the Committee on Maternal Mortality, naries and showed a lack of scientific rigor in have also shown large differences in the risk of such definitions. David Williams 13 reviewed dying due to maternal causes in the State of the changes over time in the definition of race Paraná in the years 1993 and 1997, which dis- in biomedical and social sciences dictionaries proportionately affected “Black” and “Yellow” (anthropology, psychology, and sociology). In (Asian) women. Maternal mortality did not dif- most but not all of the social science dictionar- fer between “Parda” (“Brown”) and “White” ies, race is viewed primarily as a sociopolitical women. Dachs 8, using data from the 1998 Na- construct with strong cultural components. In tional Household Survey (PNAD), found no sta- contrast, all biomedical dictionaries except tistically significant difference by “skin color/ one defined race as reflecting genetic traits and race” in self-assessed health status after adjust- thus ascribed a dominantly biological meaning ing for education and income level. Barros et to race 13. al. 9, based on longitudinal data, have shown In the early 19th century, the various fields worse health outcomes for “Black” children in of science had adopted the ideological public Southern Brazil, which is reduced after adjust- view of human differences based on race. This ment for SES and various other variables (mar- ideology was linked to colonialism and slavery, ital status, maternal age, parity, planned preg- which (especially in the United States) estab- nancy, social support, smoking, work during lished a mode of classification based on a rigid pregnancy, and antenatal care). The study re- hierarchy of socially exclusive categories that sults also suggest that “Black” mothers receive divided Europeans, Africans, and Indians (http: lower quality of care as compared to “White” //www.aaanet.org/stmts/racepp.htm, accessed ones. There are also indications that in Brazil on 13/Aug/2002). racial inequalities are more common in treat- The concept of race began to be questioned ment than in access to health care services 10. in the late 1930s by Ashley Montagu, culminat- The objective of this article is to review the ing with his book Man’s Most Dangerous Myth: literature related to the concept and measure- the Fallacy of Race, published in 1942, in which ment of race with a focus on the US and Brazil. he claimed that race was a biological myth 1. We will discuss both the measurement of race After World War II, Montagu wrote the final text in these two multiracial societies and data qual- of the UNESCO Statement on Race that denied ity problems. We also make recommendations any scientific meaning for the concept of race. about the measurement of race in medical Many physical anthropologists and biomedical records and public health research. Although scientists strongly opposed this document 14. the use of race in public health research has However, a study of the definitions of race in been discussed in relation to definitional and physical anthropology textbooks published be- methodological problems in the United States, tween 1932 and 1979 clearly showed that the the Brazilian public health literature has not entire field moved towards Montagu’s view discussed in detail how such problems apply to over time 15. By the 1970s the modal position of Brazil. This article is intended to review the lit- these texts was that human races do not exist. erature and introduce a discussion regarding However, a few physical anthropologists still broader as well as country-specific questions hold the traditional view. For example, in the

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 662 Travassos C, Williams DR

1985 edition of the Anthropological Glossary 16 but fits well to a restricted gene flow model race was defined as “a genetically distinct in- (Trellis Model). Therefore, genetic distance does breeding division within species”. Howe ve r, a t not validate the concept of human races as evo- the end of the 20th century the Executive Board lutionary distinct lineages. By using modern of the North-American Anthropological Asso- genetics, Templeton shows that human beings ciation, based on recent scientific evidence, but are not biologically distinct groups. without consensus from all members, officially Human classifications based on physical stated: “physical variations in the human species traits are also considered not to have any evo- have no meaning except the social ones that hu- lutionary validity. They represent adaptive traits mans put on them” (http://www.aaanet.org/ of human population to the environment. stmts/racepp.htm, accessed on 13/Aug/2002). Physical traits such as skin pigmentation, hair In the evolutionary field, race is assumed as color and texture, and the shape of one’s nose a synonym for subspecies. Similar to race, sub- or lips are erroneously used as markers of race. species is also an imprecise concept. The tradi- They are assumed to be adaptations to geo- tional definition relates subspecies to a geo- graphic factors such as solar radiation and graphically circumscribed and genetically dif- temperature 19. Melanesian and African popu- ferentiated population, but from an evolution- lations share traits such as dark skin, hair tex- ary genetic point of view the problem is “that ture, and cranial-facial morphology, used to many traits and their underlying polymorphic classify people into “races”, but they have great genes show independent patterns of geographi- genetic diversity. Europeans are genetically clos- cal variation” 17 (p. 632). One argument, based er to Melanesians and Africans than Africans on the traditional concept of subspecies, fre- are to Melanesians 17. quently present in the debate about the subdi- Contrary to the Mendelian tradition, which vision of the human population into “races”, is assumes that for any one gene there is only one that most of human genetic diversity exists as phenotypic outcome, modern genetics shows differences between individuals within popu- that the phenotypic expression of any gene may lations. Amongst other studies, Rosenberg et vary expressively depending on the environ- al. 18 have shown that only 3 to 5% of the exist- ment (the organism’s genome as well as exter- ing genetic diversity can be used to distinguish nal factors). Mendelian ratios (three-to-one ra- what could be called as human “races”. The ge- tio) are the result of very special cases, in which netic diversity between groups of humans is the phenotypic expressions of enzyme path- the lower than that of several other mammalian ways are little influenced by the environment. species and is considered to be too small to al- This is because they might express very trivial low a distinction of humans subspecies under traits of the phenotype 20. the traditional concept 17. The complexity of the expression of most Nonetheless, methods traditionally used genes, which relates to the other genes in the (Fst and related statistics) to measure levels of organism’s genome, the cellular and extra-cel- human genetic diversity are not sufficient to lular environment and the environment out- discriminate contrasting theories about human side the organism 20 imply that the number of evolution, because in modern evolutionary sci- diseases that follow the Mendelian tradition is ence, subspecies must be genetically differenti- also limited. One example is sickle-cell anemia, ated due to barriers to genetic exchange that but the majority of diseases do not represent a have persisted for long periods of time. Temple- single-gene disorder, and molecular genetics is ton 17 uses the concept of “genetic distance” to still far from being able to provide further un- test contrasting evolutionary models. The Can- derstanding of chronic illnesses. delabra Models (old and new) imply that major Despite the fact that race has been used as human geographical populations (Africans, a surrogate for genetic information until the Asians, and Europeans) are branches of a can- onset of molecular genetics, there is no scien- delabra and therefore valid “races”, while the tific support to continue using race in Public Trellis Model poses recurrent genetic inter- Health as a marker for genetic susceptibility 21. change among Old World human populations Parra et al. 19 have recently shown that skin col- in such a way that there was no separation into or in Brazilians cannot be used as a genetic evolutionary lineages and as a result there is no marker, because physical traits have been shown such thing as human subspecies or races. Both to be a poor predictor of African ancestry in models agree with the African origin of anatom- this population. In both the United States 21 ically modern humans. Templeton presents ev- and Brazil, although the risk of sickle-cell ane- idence that genetic distance models do not fit mia varies by race, race is not a reliable predic- treeness as required by the Candelabra Models, tor of sickle-cell anemia.

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 663

However, the controversy in the biomedical Race and class literature in relation to race remains alive. Neil Risch 22 recently argued in favor of the validity The salience of race in measuring social in- of using racial/ethnic self-classification in bio- equalities varies across countries. The United medical research. Their arguments are based States, a highly racialized society, has been on the acceptance of an evolutionary tree for measuring race since its first census in 1790 25. human races. Race is defined by these authors In this census, there was no information on so- as the person’s primary continent of origin cial class (occupational class), but people were based on the evolutionary tree, while ethnicity classified by race. Race has been more salient is viewed as a self-defined construct with geo- than class in the US, and in some ways this sup- graphical, social, cultural, and religious mean- ported an ideology of a classless society 2. ings. For these authors race is a biological con- In addition, the importance given to race in cept, while ethnicity is a social construct that US health statistics is not universal. Studies of could be used as a proxy for race/genetic varia- health disparities in European countries have tion. In opposition to the conclusions of a re- attributed much more salience to the concept cent study on drug response, which proposed of social class than to race. In the US, until re- that genetically inferred clusters are more in- cently, age, sex and race were the routinely formative than commonly used ethnic cate- used variables in reports on vital statistics. In gories 23, Neil Risch and colleagues claim that contrast, social class data have been linked to ethnicity or ancestry are actually more geneti- health information for more than a century in cally informative than clusters based on the England, although there is growing attention to analysis of random genetic markers. The argu- race with both the Census of England and Wales. ment in favor of ethnicity is that the number of In Portugal, information on race is not collect- loci required in genetic data to discriminate ed in the National Census. more recently separated populations (or those Differences in demography and ideology influenced by admixture or migration) is much are identified as important reasons behind the greater than that needed to discriminate popu- variables selected to be included in health data lations with ancient separation. However, these systems 2. Ancestry and therefore race had lim- arguments must be interpreted in relation to ited influence among public health profession- the validity of the evolutionary theory used als in the early 20th century in Great Britain, that assumes the existence of “races” in the hu- and preventive medicine looked for historical man population. Secondly, as will be discussed and social explanation of health, in contrast later in this paper, there are validity and relia- with what happened in the US 26. Only recent- bility problems with racial taxonomies. Method- ly, in 1991, was a measure of ethnicity included ological problems are likely to increase when in both the Health Survey of England and Wales the taxonomy is applied across countries and census 25, given the extensive migration of peo- to admixed populations. These are major limi- ple from former colonies in the second half of tations given that biomedical research should the 20th century, ethnic minority groups now generate information with the greatest possi- constitute 6 percent of the population of Eng- ble validity across populations. land and Wales. As this population continues Another recent article 24 based on a small to grow, there is increasing interest in racial/ sample of southern Brazilian women suggests ethnic variations in health in the UK, as is hap- that the self-referred number of “Black” ascen- pening in other countries. In Brazil, only in 1996 dants is a reliable marker of susceptibility to did it become mandatory to include data on certain heritable health conditions. “race/skin color” on the death certificate and the Despite existing controversies in the bio- Information System on Live Births (SINASC). medical literature, it is widely accepted that As for other areas (variables) of social life, racial/ethnic categories are imprecise and race is a particular dimension of social stratifi- changing measures that are historically, ad- cation which defines differences in access to ministratively, and politically constructed. The goods and services. Race may be related to so- salience given to race, as well as the meaning cial class but is different from it, even though and the measure of race itself in census and both concepts carry socially constructed mean- health data, varies across countries and across ings. Race is based on the physical (color/race) time. The history of race classification in the characteristics of individuals, while social class US and Brazil are good examples of these vari- is a product of social relations. ations as will be discussed later in this article. Marger 27 highlights that the unequal distri- bution of resources within a society creates a system of stratification, meaning that in mod-

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 664 Travassos C, Williams DR

ern societies stratification occurs along several of particular groups in society. Race was con- dimensions, with class stratification as the sidered by the “White” elite as a natural distinc- most prominent. Multiethnic societies are also tion in human identity 33. Following the US stratified in relation to ethnicity. However, as Constitution, “Blacks” were counted as three- this author argues, “the empirical relationship fifths of a person 25, and until 1850 “Blacks” in between class and ethnicity is complex and sub- the US census were categorized as either “slave” ject to frequent change” 27 (p. 63). According to or “free colored”. Early censuses did not count E. O. Wright 28, race represents non-class forms Indians unless they were “Civilized” (the latter of oppression and can reflect relations that are being those who did not live on reservations quite independent of class. Candace Nelson and who paid taxes). The 1870 census classified and Martha Tienda 29 argue that although eth- the indigenous population as “Pure Indians” nicity cannot be reduced to a class phenome- and “Half-breeds” 34. Only after 1924, when non, social position plays an important role in American Indians were given citizenship, they molding ethnicity over time and place. began to be classified in a single racial category Socio-economic status (SES) is typically mea- according to the US census 25. sured by income, education, and occupational The 1850 and 1860 censuses used the cate- status or some combination of these three mea- gories “Black” and “Mulatto” (tabulations would sures 30. Variations in health status across SES in aggregate under the term “colored”) for free the United States are in general larger than racial African descended people, and from the 1910 ones, but race tends to predict increased health to the 1930 census the term “Mulatto” was used risks independently of SES 25,31. Race and SES again together with “Negro” to classify African are correlated but not identical. For example, descendents 34. The temporary usage of a term “Blacks” are three times more likely to be poor [mulatto] to classify admixed people in the US than “Whites”, but two-thirds of blacks are not census supported the polygenist theory of the poor and two thirds of all poor Americans are superiority of “Whites”, which additionally white 30. Similarly, in Brazil, race and SES are not contended that hybrid racial species were less equivalent, and racial disparities are smaller than fertile and had shorter life spans than pure- SES ones. Data from the 1998 PNAD show that in- race persons 33. For this purpose, the class “Mu- come is unevenly distributed across race cate- latto” was defined as including anyone having gories: “Pardos” are largely concentrated in the any percentage of African blood. “Mulatto” was poorest quintile, followed by “Blacks”. On the perceived by the color of the skin by census other hand, Asians (“Yellows”) are largely con- enumerators and was not based on genealogi- centrated in the richest quintile, followed by cal history. It referred to people in whom the “Whites”. In the Northeast, the poorest region in mixture of “White” and “Black” was visible 35. the country, “Whites” are largely concentrated The 1890 census “refined” this admixed amongst the poor. Gender inequalities in the la- racial category: besides “Mulattos” it included bor market have also been shown to be greater the categories “Quadroon” (one Black grand- than racial inequalities in Brazil 32. parent or one mulatto parent and the other white) and “Octoroon” (a Black great-grand- parent or one Quadroon parent and the other Measures of race White) to further distinguish the level of Black blood. The “Mulatto” category remained in the The US experience 1910 and 1920 censuses, but was dropped in 1930 by census officials who claimed it was in- Race is central to the organization of life in the accurate 33. Consistent with racist laws, the ter- United States, which was the first country to minology for admixed populations was even- collect census data on race. Racial categories tually substituted by “Non-White” categories in census data in the US have changed regular- based on the “One Drop of non-White Blood” ly, such that no racial classification scheme has rule. This rule stated that a single Black ances- been used in more than two censuses 25. Prior tor would classify a person as Black, despite to 1960, race was assessed by interviewer ob- appearance. The 1930 US census stated that “a servation. Since then, the assessment of race person of mixed White and Negro blood should has been self-reported. A new classification was be returned [classified] a Negro, no matter how used in the 2000 census. The main characteris- small the percentage of Negro blood” 33 (p. 1741). tic of the US race classification is that it is main- The ideology behind the “One-Drop” rule was ly based on ancestry and not on phenotype. one against miscegenation 33 and shaped the From the very beginning, racial categoriza- “Black” and “White” racial divide of the US tion was an expression of social status (value) population. Many Southern states in the US

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 665

had laws that prohibited interracial (“Black- (“North-American Indian or Alaskan Native”, White”) marriage until the US Supreme Court “Asian or Pacific Islander”, “Black”, and “White”) ruled such laws unconstitutional in 1967. It and two ethnic categories (“Hispanic” and “Not was not that admixture did not exist, but that it Hispanic origin”). was seen as something to be prevented and re- “Hispanic” was defined as a person of duced. From then until the 2000 census, racial Cuban, Mexican, Puerto Rican, South or Cen- classification in the United States did not give tral American, or other Spanish culture or ori- any room for multiracial classification. Nonethe- gin, regardless of race. Race and ethnicity less, in earlier censuses, a small number of per- could be collected separately or in a combined sons checked the “other” race category and format. In the separate format “Hispanics” specified that they belonged to multiple-race were also to be classified as “Black” or “White”, categories. and in the combined format these two “races” Influenced by the 19th century notion of were to exclude “Hispanics” (http://www.white race in anthropology, race categories were house.gov/omb/fedreg/ombdir15.html). changed in the 1900 census to match what At the time of the establishment of OMB were considered the four major races. Discrete standards for racial classification, civil rights categories were created to express mainly the advocates accepted the definition of race as it continent of ancestry: “Caucasian” or “White”, was, arguing that since it was the basis for dis- “Negro” or “Black”, “Mongolian” or “Yellow”, crimination, it should be assessed to monitor and “Indian” or “Red” 34. As such, racial catego- the success of policies to reduce discrimina- rization gained scientific status. tion 33. As a result, the dichotomy between “Whites” are the large majority of the US “Blacks” and “Whites” based on the racist “One- population. In the 2000 US census “Whites” Drop” rule and the absence of multiracial cate- corresponded to 75% of the US population. The gories remained in the US official taxonomy. distinctiveness of “Whites” as the dominant In the 1990s the OMB reviewed its race race is reflected in the US census categories in standards and adopted a revised classification the fact that this group was always separated system in 1997 that was applied in the 2000 from “Non-Whites”. For a short period of time census. Under the new standards, as in earlier in the 19th century the US census term “Col- ones, the OMB does not make a clear concep- ored” included all races except “Whites” 36. Un- tual distinction between race and ethnicity or like the other racial categories, the term “Whites” completely rule out biology from them: “racial never changed, although its meaning did. In and ethnic categories should not be interpreted 1910 a category “All Others” was created that as being primarily biological or genetic in refer- included Mexicans, but in 1930 they became ence. Race and ethnicity may be thought of in counted as “Whites”. terms of social and cultural characteristics as The use of “Chinese” and “Japanese” in ear- well as ancestry” (http://www.whitehouse.gov/ ly racial classification in the US shows that the omb/fedreg/ombdir15.html, accessed on 23/ lack of distinction between the concept of race Aug/2002). and ethnicity was present in the measurement Various changes were introduced in the defi- of race in the US since its early times. “Chinese” nition of the categories, in question format (the was a separate racial category in the 1869 cen- use of separate questions to inquire about race sus. Later it was merged into the “Colored” and ethnicity is recommended, with the ethnici- group, but in 1900 it became a sub-category ty question coming first), and in terminology. among the “Yellows” or “Mongolians”, with the The new classification provides a minimum stan- “Japanese” as another sub-group. Other Asians dard with five race categories (“American Indian were counted but rarely presented in published or Alaska Native”; “Asian”; “Black or African Amer- tabulations 34. ican”; “Native Hawaiian or other Pacific Islander”; The Civil Rights Act of 1964 and 1968 re- and “White”) and two ethnic categories (“His- quired racial/ethnic data for monitoring poli- panic or Latino” and “Not Hispanic or Latino”). cy. In response to these needs, the US Office of The question of how to measure people Management and Budget (OMB) established a with multiracial ancestry was also discussed, race classification standard in 1977 to be used but a “multiracial” category was not included. by all the statistical agencies of the Federal gov- There was strong activism on both sides of the ernment, including the US Census. These guide- issue, but in the end the OMB decided to main- lines were used in the 1980 and 1990 censuses. tain race classification based on discrete race The OMB classification officially introduced categories. As race classification reflects peo- ethnicity in the US race classification. It con- ple’s consciousness about race, but also acts as tained a minimum set of four race categories reifying race in people’s consciousness, the

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 666 Travassos C, Williams DR

OMB taxonomy has played a role in perpetuat- icas. The use of the adjective “American” to de- ing the idea of “pure” races in many people’s nominate people with origins from other con- minds. However, to deal with people of multi- tinents or countries (e.g., African American, ple racial heritage, respondents are now allowed Italian American, Mexican American, or Brazil- to check as many responses as are applicable ian American) disregards the fact that an Amer- in response to the race question 37. ican is anyone from the American continent, as The concept of race based on ancestry in a European is anyone from a European country the US racial taxonomy and the absence of a or an African is anyone from an African coun- multiracial category is likely to raise difficulties try. Strictly speaking, Brazilian American is for many admixed people. In the census ques- synonymous of Brazilian, not a US resident/ tion about race, the respondent can specify citizen of Brazilian origin. when he or she selects “some other race”. In the long form of the 2000 census questionnaire, The Brazilian experience applied to a sample of the US population, peo- ple also answer a question about their national Brazil is the country in the Americas that has origin. However, published data are mainly the largest population of African ancestry 39. based on the OMB racial/ethnic categories. Miscegenation is part of Brazil’s history, as in Some have also pointed to difficulties in data other Latin American countries 40. The official analysis that the option for selecting more than term for the admixed population in the census one race poses 38. is “Pardo”, literally meaning “Brown” or “Gray”. The use of race and ethnic categories in a The 2000 census showed that with a total pop- single racial/ethnic classification seems to be ulation of 169 million, 53.7% of Brazilians self- of common usage, as is the case in the UK cen- assessed themselves as “White”, 39.1% as “Par- sus and the Canadian census, denoting a com- do”, 6.2% as “Black”, 0.5% as “Yellow”, and 0.4% mon lack of clarity in the distinction between as “Indigenous” (http://www.ibge.com.br, ac- these two concepts. As presented, since 1980, cessed on 10/Feb/2002). the census has used the ethnic category “His- It is important to note that racial measure- panic” to classify people of Latin American ori- ment in the Brazilian census has always re- gin. Brazilians, who are of Portuguese rather ferred to skin color. It refers to phenotype (phys- than Spanish origin and culture, hardly identi- ical appearance) and not to ancestry (origin), fy with the term “Hispanic” The new OMB ter- as in the US. Racial categories were created minology added the term “Latino” to “Hispan- based on a combination of physical appear- ic”, but again, Brazilians rarely identify them- ance and social position 41. While the “One- selves as “Latinos”. The official Canadian clas- Drop” rule shaped the racial division in US so- sification, for instance, uses the term Latin ciety based on “pure” racial categories, the Americans, a more inclusive term, and of more Brazilian census categories have always includ- common usage in the American Continent out- ed a term (“Pardo”) for the admixed popula- side the USA. tion. Miscegenation was an early pattern in “Hispanics” or “Latinos” represent a very Brazilian society, and the first census in 1872 heterogeneous group in relation to nationality, showed that 38.3% of Brazilians were already race, and even ethnicity. Candace Nelson and mixed (“Pardo”) 42. Martha Tienda 29 argue that ethnicity is struc- Similar to the US, in the 19th century the tured by the relationship between a particular gathering of population data in the Brazilian national origin group and the organization of census distinguished free people from slaves society, and that this relationship is shaped over and various censuses applied different race/col- time by immigration history, reception factors or classifications. The first two general census- when arriving in the US, and race. They have es in the country (1872 and 1890) included shown large variations between Cubans, Mexi- questions on race. In these censuses, the ter- cans, and Puerto Ricans in the US in relation to minology used was race and not color. In the assimilation, residential segregation, and social 1872 census, race was presented in four cate- mobility. Therefore, the self-classification of US gories (“White”, “Black”, “Pardo”, and “Cablo- residents from Latin American countries de- co”). “Cabloco” referred to indigenous people pends on people’s SES background, time, histo- and their descendents. In 1890, “Mestiço” re- ry, country of migration, and even skin color. placed the term “Pardo”. From then on the vari- Although less a criticism of OMB standards, able was referred to as color until the 1990 cen- the use of “American” in standard terminology sus, when color/race was used. for race and ethnicity in the US can be viewed This information was absent from the 1900 as disrespectful to other countries of the Amer- and 1920 censuses and was reintroduced in

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 667

1940. There was no census in either 1910 or the patriarch with his illegitimate or natural 1930. In 1940, the interviewer, in an open ques- children with slaves or natives (a Moorish influ- tion, made a judgment about the respondent’s ence, according to Freyre). …We know that the color. The main options were “White”, “Black”, Portuguese, despite being intensely Christian and “Yellow”. All other designations, indicating (and even more, champions of the Christian admixed individuals (“Caboclo”,“Mulatto”,“Mo- cause against that of Islam), imitated the Arabs, reno”), were classified under “Pardo”. Howe ve r, or Moors, the Mohammedans in certain tech- official tabulations presented “Black” and “Par- niques and in certain costumes, assimilating do” in aggregate form. In the 1950 census, the countless cultural values from them. The Mo- major change was that color became self-as- hammedan concept of slavery, as a domestic sys- sessed. Besides, in the questionnaire it was ex- tem linked to the organization of the family, in- plicitly stated that the term “Moreno” could not cluding domestic activities, without being deci- be accepted as a skin color option. sively dominated by an economic-industrial pro- The 1960 census included the term “Índio” posal, was one of the Moorish or Mohammedan [Indian or Indigenous], which was previously values that the Portuguese applied to coloniza- counted as “Pardo”. In this case, “Índio” was tion, predominately, but not exclusively Christ- limited to individuals living on reservations. ian, of Brazil.” He argues that “because of the The question format changed to a five-fold one, emphasis placed on purity of origin in the Unit- but, the micro-data on race from this census ed States, this was not even a possibility there”. were never made available, and published data Nineteenth-century urbanization was an showed “Indians” collapsed into “Pardos”. Data opportunity for social mobility for mestizos, on race were again absent from the 1970 cen- people that where somehow outside the polar sus. This happened during the military dicta- relations established by the master/slave posi- torship that ruled the country for more than 20 tions. Miscegenation increased substantially in years and was responsible for violent repres- the 19th century, with the proportion of ad- sion of political groups and severe censorship. mixed individuals increasing from 10 to 41% of When race was reintroduced into the 1980 the population (http://www.iuperj.br/profes- census data, the self-classification adopted was sores/texto3jesse.htm). four-fold, which again excluded the category However, at the turn of the 19th century, “Indian”. Eventually, the 1991 and 2000 census- miscegenation became related to the notion of es readopted a five-fold classification, which “embranquecimento” or “whitening” 42. What included “Indigenous” as a separate group. became known as the “whitening ideal” aimed Strictly speaking, “Indigenous” is not a “skin to dilute “Black” blood in the “White” blood of color” category. “Indigenous” might represent European migrants to make the population some physical traits, ancestry/ethnicity, or even lighter-colored 42. The late 19th and early 20th- group identity. As such, the present Brazilian century Brazilian elites and intellectuals ap- census color/race classification also mixes skin proached “purification” of the population through color with ethnicity. The color/race question is racial miscegenation and migration, contrary only included in the census questionnaire ap- to the US approach to racial “purification” by plied to a sample of the Brazilian population. decreasing the proportion of admixed individ- Telles & Lim 43 point out that in Brazil, for vari- uals with “dark color” in the population, using ous reasons, interviewers do not always follow legal enforcement. In contrast, racial discrimi- the instructions, and census data on color/race nation in Brazil has been considered illegal since are a combination of self-classification and in- the beginning of the Republic in 1890 44. David terviewer classification. Cleary 45 (p. 6) points out that “the ‘whitening The characteristics of miscegenation in thesis’ was not unique to Brazil: it was a stan- Brazilian society are discussed on the basis of dard response of Latin American intellectuals to two dominant approaches. Gilberto Freyre, one eugenicist orthodoxy, with local variances in of the most important and controversial Brazil- Venezuela, Mexico, Argentina, and elsewhere. ian sociologists and anthropologists, regarded The contradiction here, that increasing diversity miscegenation as the core of Brazilian identity and miscegenation could result in racial unifor- 44. During colonial times there were numerous mity over time, was typical of the mental gym- admixed and illegitimate children of slave mas- nastics intellectuals of the period were obliged ters and priests. According to Jessé de Souza to perform in order to reconcile the tenets of sci- (http://www.iuperj.br/professores/texto3jesse. entific racism with Latin American patent mul- htm), interpreting Freyre’s ideas, the specificity tiracial reality”. of Brazilian miscegenation is represented “by Three to five million Europeans migrated the uncertain but real possibility of identifying to Brazil in the late 19th century. The “Euro-

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 668 Travassos C, Williams DR

peanization” of Brazil, the respective modern journalists in Brazil to assume the term “Pardo” values attached to it, and the “whitening” poli- as “Black” to refer to the Brazilian population of cies defined the new basis of Brazilian social/ African ancestry. “Pardos” are aggregated with racial inequalities. In this new context, “White” “Blacks” under the latter designation. Brazilian was superior to “Non-White”, “but ‘White’ was Ministry of Health documents began to adopt (and continues to be) more an indicator of exis- the term “Black” to refer to Brazilians of African tence of a series of moral and cultural attributes ancestry 48. The same is becoming common- than skin color” (http://www.iuperj.br/profes- place in newspapers and scientific articles sores/texto3jesse.htm). about racial inequalities 49. Lovell & Wood 32 There is a strong ongoing debate regarding used this procedure, but justified it by the fact color/race classification and terminology used that the category “Non-White” is more stable by the Brazilian census. The Brazilian Census over time than “Pardo” and “Black”, and that the Bureau (IBGE) organized a Committee to ad- category “White” shows greater reliability. How- vise the design of the 2000 census, but despite ever, it is uncommon for authors to base this de- the existing opposition, the final decision by cision on technical grounds; in most cases it is a IBGE was not to modify the 1991 classification, political decision, and not always made explicit. maintaining a forced question with a five-fold An opposite view is expressed by other category. Brazilian and US scholars who recognize mis- Color/race classification has been criticized cegenation as part of Brazilian history and con- from contrasting viewpoints. The use of dis- sider Brazilian racial identity intrinsically am- crete race/color classification in a country with biguous 50 In this case, racial classification a large admixed population has been ques- does not fit into a discrete “Black”/“White” di- tioned by various scholars, along with pressure chotomy. from some scholars and activists 44 for the Ambiguity is reflected in the fact that Brazil- IBGE to limit its classification to the dichoto- ians, when inquired about their color/race in mous “Black” and “White” US racial approach, an open-ended question, may answer with 135 by collapsing the category “Pardo” with “Black”. to 500 different race-color terms 51,52,53. Difícil Influenced by the experience of other coun- dizer [hard to say]; tostada [toasted]; leite [milk] tries, in particular the US, Brazilian scholars are some examples, besides the many different and activists of the Black Movement argue in terms people use to specify very fine color nu- favor of using the dichotomous “Black”/“White” ances such as morena café [coffee-colored mor- classification. An increasing number of Brazil- ena or tan], morena clara [light morena], or ian activists and scholars, influenced by racial morena canela [cinnamon morena]. Ambiguity movements in the United States, do not accept also refers to the fact that skin color is a very that Brazilians be seen as a continuum of col- fluid measure that varies greatly with the con- ors 44. The main argument in this case is politi- text 52 Lovell 42 attributes the influence of so- cal: that the emancipation of people with cial class to the fluidity of color/race identifica- African ancestry requires greater racial polar- tion in Brazil. Some authors say that in Brazil, ization than exists in Brazil 46. Moreover, it is “money whitens”. Wealthier people with darker argued that affirmative action policies require phenotypes tend to classify themselves and be identifying who will benefit from them and classified by others in lighter categories 43. Oth- that the US notion of race as ancestry is more er contextual circumstances, such as dressing appropriate. However, even in the US, other and social status, can also influence color/racial criteria have been used for affirmative action labeling 51. Given this ambiguity and fluidity of purposes. Most of the initial affirmative action color/race for Brazilians, some have argued polices in the US included women as well as that to restrict respondents’ choice to a few racial minorities. And women, most of whom color/racial categories in the Brazilian census were white, experienced greater economic im- is a violation of the principle of self-identity 51. provement than minorities in the last three Another aspect of skin-color ambiguity is il- decades 47. Moreover, US affirmative action lustrated by the fact that terms such as “More- policies covered multiple racial/ethnic groups, no” or “Pardo” refer to a wide spectrum of phe- including “Blacks”, “Hispanics”, and “American notypes. According to the standard Brazilian Indians”. Thus, affirmative action polices per se Portuguese dictionary Aurélio, “Moreno” de- do not require any particular racial categoriza- rives from Spanish and refers to Moorish skin tion. What is needed is a political decision about color. Variation in phenotype is even wider for which groups should be covered by the program. “Moreno” than for “Pardo” 51. It is now becoming common practice among The usage of the term “Pardo” is also highly many scholars, activists, policy-makers, and questioned. Despite the fact that it is adopted

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 669

in the census, the word “Pardo” is rarely used Applying an index for African ancestry, they did by Brazilians to refer to skin color. Before the not find statistically significant differences in reintroduction of color/race category in the their sample in the index for “Blacks” and 1980 census, IBGE included in the 1976 PNAD “Whites”. These authors emphasize the risks of both an open-ended question and a forced- equating color with geographic ancestry and of choice question with the color/race categories interchangeably using the terms “White” and used in the previous census, to test new ways “European” or “Black” and “African” in the of inquiring about color/race. In the open-end- Brazilian context. ed question, only a few respondents (6%) rec- Besides inaccuracies regarding ancestry, ognized themselves as “Pardo” 53. Various stud- Brazilian census classification never included a ies have found that the most preferred skin col- specific category for admixed people with In- or term in the country is “Moreno” 51,53. In digenous ancestry and, as previously shown, agreement with other surveys 51,52, a study pub- until 1980 census data classified all admixed lished by the Folha de São Paulo newspaper 53 people as “Pardo”. In the present color/race showed “Moreno” to be the term Brazilians pre- classification, people of indigenous ancestry fer to identify themselves. However, for some 54 can choose to identify themselves as “Pardo”, “Moreno” means both a color and the absence “White”, or “Indigenous”. People self-classified of color. as “Indigenous” or “Pardo” in the census forced- Nelson do Valle Silva 55 has argued in favor choice question tend to classify themselves in of the validity of the official classification. The similar color/race categories when given the core of Silva’s argument in favor of the official choice to do so. In a recent study 52, comparing category is that “Pardo” refers to an objective, census categories with an open-ended ques- more precise, “demographic” characteristic, tion, 61.73% who identified themselves as “In- while “Moreno” is related to skin-color identity. digenous” in the forced question answered By doing so, he assumes the variable color/race “Moreno” in the open-ended question. Similar- as being an objective physical trait, not a social ly, 54% of “Pardos” also identified themselves construct. According to Silva, for census pur- as “Morenos”. On the other hand, only 12.8% of poses the demographic trait “color of the skin” the people who identified themselves as “In- is a more appropriate measure than color iden- digenous” in the closed question chose the tity. However, “skin color” matters if it affects same option in the open-ended one. peoples’ identity in a different manner. It mat- Data from the 1998 National Household ters when it serves as a determinant of people’s Survey (PNAD) signal significant demographic, choices and opportunities due to discrimina- socioeconomic, and geographic variations tion in relation to their phenotype. As a social across the categories “Pardo” and “Black”, ar- construct, regardless of the terminology ap- guing against aggregation of these two groups. plied or the way the color/race question is con- In the North and Northeast, the poorest re- structed or assessed (self-assessed or assessed gions in the country, the majority of the popu- by the interviewer), skin color will reflect re- lation are “Pardo” (68.49% and 64.3% respec- spondent/societal values. It is not a character- tively), while in the Southeast and South (the istic like age, which despite having social mean- country’s wealthiest regions) the majority are ing, is an objective and meaningful demograph- “White” (64.03% and 82.91% respectively). The ic trait in itself. Moreover, respect for individual proportion of “Blacks” is higher in the South- dignity is one principle that has to be used in east (7.31%), followed by the Northeast (5.74%). the assessment of color/race. This means that The North is the region with the lowest concen- when collecting racial data the most preferred tration of “Blacks” (2.2%). People that declared terms that the group uses should be adopted 13. themselves “Yellow” are on average the oldest It should also be noted that the definition amongst Brazilians (37.44 years), followed by of the Brazilian census color/race categories “Whites” and “Blacks” (29.75 and 30.79 years, does not support the assumption that “Pardo” respectively). “Indians” and “Pardos” are the refers strictly to African descendents. Parra et youngest in all regions (25.51 and 26.16 years, al. 19 using a classification of color (“White”, respectively). Among individuals 10 years old “Intermediate”, and “Black”), based on multi- or greater, “Blacks” have the worse educational variate evaluation using skin color on the me- achievement, especially in the Northeast (4.74 dial surface of the arm, hair color, and texture years on average; median = 3 years). However, and the shape of the lips and nose, indicated the concentration of “Pardos” among the poor that for Brazil these phenotypical traits are is greater than “Blacks”: the poorest quintile of weak individual predictors of African ancestry per capita family income consists of 23.7% estimated on the basis of molecular markers. “Blacks” and 30.7% “Pardos”

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 670 Travassos C, Williams DR

According to Elza Berquó 56, “Pardos” and advantage and their social and geographic iso- “Blacks” varied in relation to infant mortality, lation from Whites 14. marriage, and fertility. In relation to the above These are controversial and politically diffi- indicators “Pardos” were becoming more simi- cult challenges. Despite the pressure for a more lar to “Whites”. Compared to “Whites” and “Par- clear-cut classification, Brazilian racial history dos”, “Blacks” presented higher mortality, later does not perfectly fit the US racial divide, and marriage, and higher single marital status, espe- many Brazilians apparently resist being classi- cially among women. From 1960 to 1980, “Black” fied in discrete and finite “Black” and “”White” women presented the highest fertility rate. racial categories. On the contrary, the US pop- In short, collapsing “Pardos” into the “Black” ulation does not easily absorb the mixed-races category is a questionable approach for mea- concept because of the “One-Drop of non- suring people with African ancestry. As shown White Blood” rule which constrained the cre- previously, color is not a reliable marker of ge- ation of a “multiracial” consciousness. A recent netic variation in the Brazilian population and survey showed that only 2% of the US popula- “Pardo” is not a valid category for admixed tion selected two or more races when given the Brazilians with African ancestry. The simple ag- option to do so 37, which is most probably an gregation of “Blacks” and “Pardos” is likely to underestimation of the multiracial nature of misestimate the size of the country’s African the US population 57; meanwhile, the term descendent population. Moreover, relevant vari- “multiracial” was not clearly understood by all ations in the demographic and socio-econom- respondents, many of whom preferred not to ic characteristics of color/race groups show identify themselves with a multiracial category that these may not be comparable groups and 37. However, there are signs that racial con- that aggregation may not be the appropriate sciousness might be changing in the United analytical procedure. Finally, this approach States. A recent study of racial self-identifica- does not respect the color/race category indi- tion by multiracial adolescents showed that viduals have chosen to self-classify. this group has a new understanding of race Ethnicity is unlikely to be a reliable alterna- based on the acceptance of diversity and mul- tive for skin color/race classification in Brazil. tiracial admixture that does not reflect the in- A recent survey conducted by IBGE and repre- fluence of the “One-Drop” rule, still prevalent sentative of the population from six greater among their parents 58. metropolitan areas in Brazil allowed the analy- In summary, the comparison of race mea- sis of how Brazilians respond when inquired surement in Brazil and the US shows that the about their ethnicity. Despite being a society of meaning of this concept differs between the migrants, Brazil does not show clear cultural two countries. Differences reflect the countries’ divides. When exposed to an ethnicity question history, culture, bureaucracy, and political forces with 12 categories of origin, 86% of the respon- that shape color/racial identity. Racial identity dents identified themselves as Brazilians. Brazil- is mainly related to how people perceive them- ians do not understand the term “origin” in a selves (and perceive how others perceive them) homogenous fashion. It can mean city of ori- in relation to their color/race. The concept of gin, race, or nationality. Responses also varied race (ancestry) is not the same as skin color. in relation to age, country/ethnic origin, and Race in the US originated in the old notion of time of migration 52. race as biology, while skin color, despite being Given their differences in history and the associated with racial ancestry, absorbs the so- social meaning of race, in the early 21st century cial context more and is more fluid. Racial clas- Brazil and the US continue to represent con- sification is not directly comparable or trans- trasting experiences in relation to racial/ethnic posable from one country to another. To mea- classification. However, the current contrast sure and interpret race relations in Brazil has a different basis from that of the 19th cen- through the prism of the US experience, or vice tury. In Brazil, it is represented by the move- versa, is to blind one’s analysis of the existing ment to change the racial/skin color classifica- differences, to separate the measurement from tion to an approach based on ancestry, which its meaning, and to ignore the singularities of does not allow for mixed categories. Mean- each society’s historical, cultural, and political while, the increasing intermarriage and migra- experience, which shape people’s perceptions, tion in the US may soon lead this country to societal values, and the measurement of race adjust its racial/ethnic classification to a more itself. admixed society for Whites and non-Black groups. Blacks remain distinctive in the US with regard to the persistence of socioeconomic dis-

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 671

Quality of data ists and seems to be more common at younger ages, but does not vary by gender. Reasons for The definition of race is not consistent across changes in self-assessment may be optional or societies, as seen in the US and Brazilian exam- contextual “depending on the form of the race ples, nor is it consistent within societies or over question, economic incentives for being Ameri- time. The difficulties linked to defining race in can Indian in some states, reduced discrimina- a precise and unchanging manner directly af- tion against American Indians, an increased fect the validity and reliability of this concept. willingness to self-identify as American Indians, Given the lack of scientific support for the and the increased use of self-enumeration in the claim that the human species is subdivided in- Census” 25 (p. 174). The same can be true for to different lineages (races), it is neither possi- any other racial category, e.g., the case of US ble nor desirable to establish a valid measure- residents from Latin America, as discussed ear- ment of race in biological terms. At the same lier. The proportion of the Brazilian population time, race is a very meaningful social category in each color/racial category also depends on that can determine differential access to a how such categories are presented and inter- broad range of societal resources. Measuring preted. The meaning of color/racial terminolo- race is not simple or easy because the concepts gy apparently varies across geographical re- of race, skin color, ethnicity, origin, ancestry, gions in Brazil 56. An increasing willingness by nationality, and identity overlap. In Canadian some groups of Brazilians to self-identify as statistics there is a preference to use the term “Black” may be the explanation for the relative ethnicity, but its definitions and measurement increase in the proportion of “Blacks” in the represent a clear example of the lack of bound- 2000 Brazilian census. “Blacks” in the 1991 cen- aries between these concepts. sus 42 represented 5.0% of the Brazilian popu- Definitions of racial categories vary across lation and increased to 6.2% in 2000. countries and time. Some people that refer to As noted previously, the skin-color concept themselves as “Whites” in the US are not simi- is more influenced by socioeconomic position. lar (in regard to ancestry or skin color) to Unlike race as ancestry, which bases its defini- “Whites” in Brazil. For example, because of the tion on the idea of human subspecies, skin col- “One-Drop” rule, individuals with “White” skin or (influenced by the 19th century idea of race) color but African ancestry are likely to identify is less rigid and more influenced by context. themselves as “Black” in the US, but might re- Labeling and self-assessment of race/color in gard themselves as “White” in Brazil. Even for Brazil has been shown to vary with socioeco- countries that base racial classification on an- nomic position. Studies on socio-economic dis- cestry or ethnicity (origin) the definition of tribution by color/race must be mindful of the racial categories is not comparable. “Whites” fact that color/racial discrepancies may be par- in the US are not comparable to “Whites” in the tially due to the fact that SES shapes color/racial UK. The term “White” in the UK never consid- identity and can lead to errors in association ered Asian Indians, Middle Eastern, and North when data on race and SES are collected at the Africans, but until recently people from India same point in time. were considered “Whites” by the official US It has been suggested that skin color desig- racial classification, and Middle Easterners still nation in the US is also influenced by socioe- are 59. There are large variations in racial tax- conomic position 35 and that skin color within onomy across countries. Some taxonomies em- the existing racial categories affects people’s phasize ancestry (e.g., US), others ethnicity life chances (for example, lighter-skinned (UK and Canada), and still others skin color “Blacks” and “Latinos” experiencing better life (Brazil). Strictly speaking, these taxonomies, chances than those with darker skin). In a na- although they overlap in some cases, are not tional study of “Blacks” in the US, lighter skin identical. color was a stronger determinant of adult in- The understanding of the race concept also come and occupational status than was parental varies across respondents and for the same SES. This association was stronger for women person (on different occasions); the impact of than for men 59,60. As in Brazil, during slavery this variation is not the same across racial cat- in the US, “Mulatto” slaves enjoyed more pres- egories. Large changes in the number of “Amer- tige than darker ones 60. However, unlike a study ican Indians” between the 1960 and the 1990 in which skin color was associated with racial US censuses may be due to improvement in discrimination 60, Krieger et al. 61 failed to find the quality of data, but most likely represent a any association between skin color and 5 of 7 shift in self-identification. Evidence of a shift specified situations of racial discrimination in favor of “American Indian” identification ex- (getting a job, at work, purchasing a home, get-

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 672 Travassos C, Williams DR

ting medical care, in a public setting) in a US population only, 2.1% identified their origin as sample of young “Black” men and women. “African” and 6.7% as “Native Indians” 52. Terminology is another source of disagree- Ethnicity may also not be perceived as im- ment between respondents, who vary in their portant when people feel themselves as part of preferred racial terminology. For instance, in a another group. However, when individuals have national study in the US, 44% of “Blacks” pre- multiple ethnicities, the choice of one may be ferred “Black”, while 28% preferred “African situational. Not surprisingly, ethnicity has been American” 25. Individuals may also experience found to be quite fluid, with only 58% of peo- difficulties in understanding the terminology ple reporting the same ethnicity between two of racial classification 62. This influence is im- surveys in the US 31. portant because variation in terminology across Racial categories are also very heteroge- data sources or in time impact the size of the neous 13. Particularly due to recent migration, racial groups. As indicated earlier, Brazilians “Black” in the US is increasingly diverse in prefer the term “Moreno” to “Pardo”. Harris et al. terms of “ethnic” origin and needs 65 There is 52 have shown that the proportion of “Whites” also considerable regional variation within the and “Blacks” diminishes significantly when Black population. Scottish-born and Irish-born “Moreno” is used instead of “Pardo” in Brazil. people living in England and Wales are sub- Self-assessment, interviewer’s observation, groups of the “White” population with greater health provider reporting, and other factors needs than some racial/ethnic minorities 59. represent different ways of inquiring about Data from the 1998 PNAD show that “Whites” race, and each method may influence the re- in Brazil are the skin color/race group with the sponse in a different way. It has thus been rec- greatest income inequality. Poor “Whites” are ommended 13 that researchers specify the way far from the rich “Whites” in their needs and in which data on race by obtained. A study by lack of opportunities. the US Bureau of the Census showed high The optimal measurement of race depends agreement between observer and self-assess- on the purpose for which race is being used. ment responses for “Whites” and “Blacks”, but Within and between countries differences in low agreement for the other racial categories racial measurement may arise due to diversity 63. Telles & Lim 43, using 1995 data from a na- in the race concept, data collection methods, tional survey to compare self-assessment of question format, terminology, or classification skin color with interviewer classification in system 60. Validity and reliability are major Brazil, found that the interviewer classification problems when measuring race, and measure- darkens the skin color of low-SES individuals ment error is likely to increase in more interra- and whitens it for high-SES individuals. cially or inter-ethnically admixed populations. It is generally recommended that self-as- sessment is the appropriate way to inquire Measuring race in admixed populations about race. However, the method should be se- lected in a more flexible way in relation to the The question of whether populations of mixed research question. In public health research, origins can be categorized into any simple, fi- observer assessment may be the best option nite, discrete categories is becoming central to when investigating racial disparities in health racial/ethnic taxonomy. Some societies have care (treatment), because it captures how oth- large proportions of admixed people and many ers perceive the health system client. In con- others are increasingly becoming admixed. Im- trast, self-assessment may be a more appropri- migration in the US, especially from Latin ate method for studies on health disparities American countries, increased in the last few and disparities in access to health services. decades, making its population much more Respondents also show large disagreement heterogeneous. The projection of the US Cen- between “color/racial” terminology and “eth- sus Bureau is that by 2050 one half of the US nic” categories. In the 1980 US census, 26.5 population will be “Non-White” 63 and 21% of million individuals in the US self-identified the population will be of multiple ancestry 57. themselves as “Black or Negro”, but only 21 Despite the possibility of answering ques- million reported having African ancestry. Dis- tions with multiple races, the new OMB classi- agreement varies across population groups fication in the US is not a good solution for and was greater for “American Indians” 64 In classifying admixed people. For miscegenation Brazil, discrepancies are even sharper than in that goes back many generations, individuals the US, mostly because the majority of Brazil- simply do not know about their ancestry. When- ians tend not to identify themselves with dis- ever people’s parents, grandparents, and great- tinct origins. In a country with a large admixed grandparents descend from intermarriages of

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 673

admixed people, “pure” ancestry becomes very Racial measurement difficult to trace. In Latin American countries in health-related documents such as Brazil where miscegenation occurred at very early stages, it is difficult for a large Validity and reliability problems with racial number of people to answer questions about measurement are of major concern in public their origins. health. The various quality problems associat- It can also be argued that people do not ed with race data lead to bias in health indica- know their ancestry because origin played a tors. Bias can happen whenever individuals in distinct role in societies with early miscegena- the numerator do not come from the same tion. As a result, many people may not find a population as in the denominator. Inconsis- place in any of the selected discrete “races” cat- tencies also vary between indicators and are egories. In the 2000 US census, 43% of people expected to be greater for admixed popula- that identified themselves as “Hispanic or Lati- tions. Problems of comparability increase when no” chose, in the race question, to answer “some rates are used for the purpose of between-coun- other race” (http://www.census.gov/mso/www/ try comparisons, given differences in racial, rsf/racedata/sld008.htm, accessed on 10/Oct/ conceptual, classificatory, terminological, and 2002). And they usually inserted their country data-collection methodology. of origin or an alternative term for their His- Racial classification, question content and panic ethnicity for their race. format, and data collection methods, which When assessing their own race, recent im- impact racial estimates, are not and cannot al- migrants from countries where race is not as ways be the same across data sources. In gen- central in social structure as in the US may ap- eral, health indicators depend both on cen- ply criteria adopted in their original country. sus/survey and vital statistics data. In routine On the other hand, descendents of migrants health indicators, censuses and surveys fre- are more likely to respond to the race question quently provide data for the denominator, while using different criteria from the ones used by data for the numerator come from a different their parents. The fact that this classification is source. In this case, data are collected at differ- based on pure-race categories of ancestry and ent points in time, potentially affecting the clas- the absence of a multiracial category increases sification. As discussed before, people change the chance of misclassification or non-specifi- their racial (ethnic or color) self-assessment cation for admixed people. On the other hand, over time and in relation to context, introduc- multiracial categories tend to be very hetero- ing variation in health indicators that may not geneous, and the greater the admixture in a be related to variation in frequency of the event population, the lower the discriminatory pow- under study, but to the size of the population er of racial classifications. in the race category due to changes in the way Therefore, fluidity and ambiguity of racial people identify themselves. Although there is measurement increases as the population be- little evidence that this currently distorts US comes more multicultural and admixed. The racial data on health, it is likely to be of in- more admixed a society, the greater the mis- creasing importance in the future due to in- specification and heterogeneity of racial cate- creasing migration and intermarriage. This gories based on ancestry. Bias will also affect variation in assessment will also affect time- classifications that allow people to be classified trend analysis. Data collection methods can al- in more than one pure-race category, as in the so vary. Self-assessment is generally used in new US classification. Multiracial categories al- census and household surveys, but it is not al- so tend to be very heterogeneous. At the same ways possible to be ascertained on death cer- time, US data on children born to Black/White tificates, for example. Data on race based on unions indicate that infants with a Black moth- information from medical records (which are er and White father consistently have higher more likely to be obtained by hospital staff or health risks than those with a White mother physicians) are likely to vary from data obtained and Black father 65, suggesting that in at least by census interviewers. some situations there may be health risks linked As noted previously, in Brazil data on col- to the specific pattern of multiracial status. or/race only became mandatory on death cer- The use of skin color may be a more ade- tificates and in the SINASC in 1986 and were al- quate proxy for racial/ethnic discrimination in so included more recently on the forms for admixed societies than racial measurement compulsory notification of diseases (SINAN) based on ancestry. Ethnicity or nationality may and in protocols for research involving human also be more meaningful in societies with re- beings. Reporting of this variable increased year cent migrants. by year, with about 12% of missing data in both

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 674 Travassos C, Williams DR

mortality and SINASC data in 2002. Health ser- ta for the denominator) for this population vices data systems do not include information group. It is also assumed that underestimation on race. The color/racial classification is the on the census data of this group is higher in same used in the census, but data collection some urban areas 65. A recent study that re- methods are not standardized. On the death viewed race and ethnicity data in US Public certificate, the physician generally informs the Health Data Systems in New England conclud- color/race of the deceased, but in case of in- ed that data from these systems are inconsis- hospital deaths some other hospital staff mem- tent, unclear, and not comparable. Moreover, bers can provide this information. The same some classifications presented error in relation procedure is adopted for deaths of children. In to geography and social/cultural realities of the SINASC, the mother reports the color/race race/ethnicity categories 67. The New England of the newborn. Estimates of infant mortality area has a large concentration of Brazilian resi- rates by color/race based on SINASC data may dents and descendents, but the classifications be particularly affected by differences in data used are not appropriate to classify Brazilians in collection methods between the death certifi- a single race, given the differences in racial iden- cate and SINASC. This can also have an impact tity and the large number of admixed people. on life expectancy estimates based on color/ In short, lack of accuracy on race-specific race. health indicators is directly related to reliabili- The US experience in measuring race on ty problems in racial measurement, and there birth and death certificates demonstrates some are no easy solutions for dealing with many of of the problems with racial measurement on the possible errors. Errors lead to over- or un- health documents. Until 1988, the National derestimates, and the direction and magnitude Center for Health Statistics used a complex al- of these miscalculations vary across race cate- gorithm for birth certificates which assumed gories, health indicators, and time. Differences “White” as the only race that required both par- in the definitions of race and reliability prob- ents to be “White”. Based on the “One-Drop” lems also point to low internal and external va- rule, these criteria implied that when one of lidity of race-related studies in public health. the parents was “White” but the other was not, Given these limitations and to avoid the the child would receive the race of the “non- misuse of race, ethnicity, and nationality as bi- White” parent. This changed in 1989, and the ological variables, various scientific journals, United States now reports infant statistics by including Paediatric and Perinatal Epidemiolo- the mother’s race. However, the definition on gy 68 and Nature Genetics 69 recently issued the birth certificate is not always the same as rules for articles to be published using any of on the death certificate. On the death certifi- these variables. In a recent Editorial in The New cate, funeral directors (or other certifiers) are England Journal of Medicine 70 (p. 1393), the au- supposed to ask the next-of-kin to inform the thor makes similar claims stating “as for med- deceased’s “self-assessed” race. Most funeral ical research, any investigator that entails so- home directors do not follow this procedure. called racial distinctions, whether a clinical tri- Accordingly, race on the death certificate has al or a laboratory study, should begin with a significant inconsistencies with racial data in plausible, clearly defined, and testable hypothe- the census for groups other than Black or White, sis… but, tax-supported trolling of data bases to thereby affecting death rates. Inconsistency in find racial distinctions in human biology must racial measurement on birth and death certifi- end”. cates for children born between 1983 and 1985 in the US and who died within one year was Why keep on measuring race? 43.2% for all races 66. This problem also exists on death certificates of adults, with many indi- Race is a 19th century biological concept that viduals who would have self-identified as lost its once-claimed scientific support and “American Indian”, “Asian”, or “Hispanic” being that is no longer seen as the appropriate way to misclassified as non-Hispanic White on the approach genetic differences across popula- death certificate. This numerator problem un- tion subgroups. As a social construct, its defin- derstates the nationally reported mortality rates ition varies across societies, groups, and indi- for these groups. viduals, and it is challenging to measure race In the US, problems have also been docu- in a precise and reliable manner. Increasing mented with the denominator. Estimates of an migration and multiracial marriages pose new- under-count of middle-aged Black men in the er and greater difficulties for racial taxonomy. census lead to the overestimation of mortality Some scholars and users in the US, Canada, rates (and other health data that use census da- and other countries claim that race should be

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 675

abandoned because it is an ambiguous and causal relationship. Race must be studied and vague concept with enormous problems of re- understood within the context of other relevant liability and validity. During the discussion for social inequalities. Social position and discrim- review of OMB categories, proposals were made ination have been shown to affect health, ac- for the complete elimination of racial classifi- cess, and quality of care in a complex and cu- cation in the US. On the other hand, many au- mulative way 74, implying that the analysis of thors, despite recognizing these limitations, ar- racial disparities in health and health care gue in favor of the use of race. They argue that should take these relationships into considera- because race has historically and currently re- tion to avoid spurious association between flected differences in power and desirable access health and race. When comparing racial differ- in society, as long as these inequalities exist, race ences, one has to adjust for all relevant SES should be measured to monitor them and build measures assumed as potential confounders, societal support to eliminate them 2,14,65. but residual confounding may lead to a spuri- Racial disparities are the expression of so- ous independent effect of race 75. cial stratification based on racial discrimina- Finally, since race is a proxy measure for tion and racial stigma. Glenn Loury 71 (p. 167) other factors besides biology, some authors, points to differences between racism and stig- such as LaVeist 12, have pointed to the need for ma: racism is related to discrimination in treat- finding more creative ways of measuring such ment, while stigma “is about who, at the deep- factors. est level, they [Afro-Americans] are understood to be”. According to Loury, racial discrimination and racial stigma are distinct phenomena, both Recommendations rooted in the social context. As a social con- struct, racial measurement captures social sta- • The use of race in public health data should tus related to the social history of each society. be justified and racial measurement clearly con- Racial discrimination is assumed as one ceptualized. It is also important to shed light important pathway associated with poorer on what race is a proxy for and to provide indi- health in particular racial groups 12. The other cations on how findings should be interpreted. aspect of racial discrimination relates to in- • Health documents should include at least equalities in treatment (quality of care), a re- one social variable to allow for the monitoring flection of health professionals’ discrimina- of social inequalities. However, whenever data tion. Existing evidence indicates that discrimi- on race are included they should be accompa- nation is not uncommon in health services 4, nied by one or more social stratification vari- and it occurs across societies and is not only ables to avoid misspecification of complex focused on race. Various patients’ characteris- health risks or harmful stereotyping. The use of tics besides race, such as gender, income, and race in public health needs to avoid simplistic physical appearance have been shown to be as- conclusions and interpretations, which can sociated with differences in physician deci- lead to a spurious salience of race in the expla- sion-making and behavior 72. Discrimination nation of health and health-care inequalities. and stigma are aspects of human behavior that • The decision to incorporate race in routine- deleteriously affect the health and quality of ly collected health statistics should consider care received by socially disadvantaged groups. validity and reliability problems. To increase Understanding discrimination and the rela- the quality of data, data collection methods tionship between discrimination and health, (self-report, interviewer assessment, other) for access to services, and treatment in a broader measuring race should be clearly stated in each way is important. These represent a complex document, and the data collection process phenomenon that requires clear understand- should be standardized. The data collection ing of the specificities of stigma and discrimi- method may vary in relation to the objective of nation for allowing interventions to reduce in- the study. equalities that are not narrowly focused. • Race-specific indicators based on data from Crude measures of racial disparities in a single source, like census or survey data, are health and health care cannot be assumed as likely to display greater reliability than racial the existence of a causal relationship between indicators based on different data sources. Da- race and health. Race as a risk marker is not ta collection forms should be standardized to synonymous with race as a risk factor 73. In- ensure consistent classification across data creased frequency of a health indicator in a giv- systems. en racial group is informative for policy-mak- • The publication of race-specific indicators ing, but it should not be directly assumed as a should be accompanied by information on the

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 676 Travassos C, Williams DR

concept of race being used, data limitations, identifying people’s geographic origin or as a and potential bias to avoid misinterpretation. genetic marker. Such data may be useful as a • Publication of variations in frequencies in marker of the risk of discrimination or other race-related health indicators should specify social exposures. that they cannot be assumed as evidence for • Research is needed to test for more accept- racial causality. Within-group variations should able and adequate measurement of skin color also be published whenever possible. Ideally, in Brazil. This should be associated with the re- publications should include possible explana- search directed at providing a clearer under- tions for observed variations. standing of the meanings of skin color in the • Aggregation of racial categories, for exam- country. ple “Pardo” under “Black”, unless justified on • Since the concept and classification of race the grounds of data homogeneity in relation to are not uniform within datasets, between coun- other relevant demographic and social vari- tries, or over time, any comparison of study re- ables, should not be an acceptable public sults concerning the effects of race on health health practice. and health care must take into consideration • The use of skin color measurement in pub- variations in racial measurement use and data lic health should be avoided as a means of quality issues.

Resumo Contributors

Raça tem sido amplamente utilizada em estudos sobre C. Travassos and D. R. Willians participated in the con- desigualdades em saúde e assistência, principalmente cept, design, elaboration, and drafting of the article. nos Estados Unidos. Diversos problemas relativos à validade e confiabilidade da mensuração de raça são relevantes para o campo da saúde pública. Este artigo faz uma revisão da literatura sobre o conceito e men- suração de raça e compara como os achados se apli- cam aos Estados Unidos e ao Brasil. Os autores dis- Acknowledgements cutem em detalhes diversas questões da qualidade dos dados relacionadas à mensuração de raça e os proble- The preparation of this paper was financed by the Co- mas levantados pela mensuração de raça em socieda- ordinating Body for Training University Level Person- des multirraciais como o Brasil. Além disso, analisam nel (CAPES, Brazil) and made possible by support como essas questões se aplicam à saúde pública, e for- from the Survey Research Center, Institute for Social mulam recomendações sobre a mensuração de raça Research, University of Michigan (United States). We em registros médicos e pesquisa em saúde pública. wish to thank Scott Wyatt and Priscilla Mouta Mar- ques for their contributions to the preparation of this Raça; Sociologia Médica; Eqüidade; Literatura de Re- manuscript. visão; Cor da Pele

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 RACE CONCEPT AND MEASUREMENT 677

References

1. Goodman AH. Why genes don’t count (for racial try in Brazilians. Proc Natl Acad Sci USA 2003; differences in health). Am J Public Health 2000; 100:177-82. 90:1699-702. 20. Rose S. Lifelines – Biology beyond determinism. 2. Krieger N, Rowley DL, Herman AA, Avery B, New York: Oxford University Press; 1998. Phillips MT. Racism, sexism, and social class: im- 21. Cooper RS. Gene-environment interactions and plications for studies of health, disease, and well- the etiology of common complex disease. Ann In- being. Am J Prev Med 1993; 9:82-122. tern Med 2003; 139:437-40. 3. Williams DR, Collins C. US socioeconomic and 22. Risch N, Burchard E, Ziv E, Tang H. Categoriza- racial differences in health: patterns and expla- tion of humans in biomedical research: genes, nations. Annu Rev Sociol 1995; 21:349-86. race and disease. Genome Biol 2002; 3. 4. Smedley BD, Stith AY, Nelson AR, editors. Un- 23. Wilson JF, Weale ME, Smith AC, Gratrix F, Fletcher equal treatment: confronting racial and ethnic B, Thomas MG, et al. Population genetic structure disparities in health care. Washington, DC: Na- of variable drug response. Nat Genet 2001; 29:265- tional Academies Press; 2003. 9. 5. Batista LE. Pode o estudo da mortalidade denun- 24. Fuchs SC, Guimarães SM, Sortica C, Wainberg F, ciar as desigualdades raciais? In: Barbosa LMA, Dias KO, Ughini M, et al. Reliability of race assess- Silva PBG, Silvério VR, organizadores. De preto a ment based on the race of the ascendants: a cross- afro-descendente: trajetos de pesquisa sobre rela- sectional study. BMC Public Health 2002; 2:1. ções étnico-raciais no Brasil. São Carlos: EdUFSCar; 25. Williams DR. Race, socioeconomic status, and 2003. p. 243-60. health: the added effects of racism and discrimi- 6. Cunha EMGP. Mortalidade infantil e raça: as dife- nation. Ann NY Acad Sci 1999; 896:173-88. renças da desigualdade. Jornal da Rede Feminista 26. Porter D. “Enemies of the race”: biologism, envi- de Saúde 2001; 23. http://www.redesaude.org.br. ronmentalism, and public health in Edwardian 7. Martins AL, Tanaka ACD. Mulheres negras e mor- England. Vic Stud 1991; 34:159-78. talidade materna no estado do Paraná, Brasil, de 27. Marger MN. Race and ethnic relations: Ameri- 1993 a 1998. Revista Brasileira de Crescimento e can and global perspectives. 6th Ed. Belmont: Desenvolvimento Humano 2000; 10:27-38. Wadsworth; 2003. 8. Dachs NW. Determinantes das desigualdades na 28. Wright EO. Class counts: comparative studies in auto-avaliação do estado de saúde no Brasil: aná- class analysis. Cambridge: Cambridge University lise dos dados da PNAD/1998. Ciênc Saúde Cole- Press; 1997. tiva 2002; 7:641-57. 29. Nelson C, Tienda M. The structuring of Hispanic 9. Barros FC, Victora CG, Horta BL. Ethnicity and in- ethnicity: historical and contemporary perspec- fant health in Southern Brazil. A birth cohort tives. In: Alba RD, editor. Ethnicity and race in the study. Int J Epidemiol 2001; 30:1001-8. U.S.A. London: Routledge and Kegan Paul; 1985. 10. Castro MSM, Travassos C, Carvalho MS. Fatores p. 49-74. associados às internações hospitalares no Brasil. 30. Williams DR, Lavizzo-Mourey R, Warren RC. The Ciênc Saúde Coletiva 2002; 7:795-811. concept of race and health status in America. 11. Kaufman JS. How inconsistencies in racial classi- Public Health Rep 1994; 109:26-41. fication demystify the race construct in public 31. Lin SS, Kelsey JL. Use of race and ethnicity in epi- health statistics. Epidemiology 1999; 10:101-3. demiologic research: concepts, methodological 12. LaVeist TA. Beyond dummy variables and sample issues, and suggestions for research. Epidemiol selection: what health services researchers ought Rev 2000; 22:187-202. to know about race as a variable. Health Serv Res 32. Lovell PA, Wood CH. Skin color, racial identity, 1994; 29:1-16. and life chances in Brazil. Lat Am Perspect 1998; 13. Williams DR. Race and health: basic questions, 25:90-109. emerging directions. Ann Epidemiol 1997; 7:322- 33. Nobles M. History counts: a comparative analysis 33. of racial/color categorization in US and Brazilian 14. Maio MC. O Brasil no concerto das nações: a luta censuses. Am J Public Health 2000; 90:1738-45. contra o racismo nos primórdios da UNESCO. 34. Byrd WM, Clayton LA. Racial and ethnic dispari- Hist Ciênc Saúde 1998; 5:375-413. ties in health care: a background and history. In: 15. Littlefield A, Liberman L, Reynolds LT. Redefining Smedley BD, Stith AY, Nelson AR, editors. Un- race: the potential demise of a concept in physi- equal treatment: confronting racial and ethnic cal anthropology. Curr Anthropol 1982; 23:641- disparities in health care. Washington, DC: Na- 55. tional Academy Press; 2002. p. 248-72. 16. Malabar FL, Robert E. Anthropological glossary. 35. Hill ME. Color differences in the socioeconomic Melbourne: Krieger Publishing; 1995. status of African American men: results of a lon- 17. Templeton AR. Human races: a genetic and evo- gitudinal study. Soc Forces 2000; 78:1437-60. lutionary perspective. Am Anthopol 1998; 100: 36. Byrd WM, Clayton LA. Changing concepts of race 632-50. and ethnicity. In: Smedley BD, Stith AY, Nelson 18. Rosenberg NA, Pritchard JK, Weber JL, Cann HM, AR, editors. Unequal treatment: confronting racial Kidd KK, Zhivotovsky LA, et al. Genetic structure and ethnic disparities in health care, Washington of human populations. Science 2002; 298:2381-5. DC: National Academy Press; 2002. p. 273-8. 19. Parra FC, Amado RC, Lambertucci JR, Rocha J, 37. Wallman KK, Evinger S, Schechter S. Measuring Antunes CM, Pena SD. Color and genomic ances- our nation’s diversity: developing a common lan-

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004 678 Travassos C, Williams DR

guage for data on race/ethnicity. Am J Public 57. Waters MC. Immigration, intermarriage, and the Health 2000; 90:1704-8. challenges of measuring racial/ethnic identities. 38. Harrison RJ. Inadequacies of multiple response Am J Public Health 2000; 90:1735-7. race data in the Federal Statistical System. In: 58. Harris DR, Sim JJ. Who is multiracial? Assessing Multiraciality: how will the new census data be the complexity of lived race. American Sociologi- used? New York: Levy Institute. cal Review 2002; 67:614-27. 39. Torres C. Descendientes de africanos en la Region 59. Bhopal R, Donaldson L. White, European, West- de las Américas y equidad en materia de salud. ern, Caucasian, or what? Inappropriate labeling Rev Panam Salud Publica 2002; 11:471-9. in research on race, ethnicity, and health. Am J 40. Sansone L. Racismo sem etnicidade. Políticas pú- Public Health 1998; 88:1303-7. blicas e discriminação racial em perspectiva com- 60. Keith VM, Herring C. Skin tone and stratification parada. Dados 1998; 41:751-83. in the Black community. AJS 1991; 97:760-78. 41. Agier M. Racism, culture and black identity in 61. Krieger N, Stephen S, Coakley E. Racial discrimi- Brazil. Bull Lat Am Res 1995; 14:245-64. nation and skin color in the CARDIA study: impli- 42. Lovell PA. Development and the persistence of cations for public health research. Am J Public racial inequality in Brazil: 1950-1991. J Dev Areas Health 1998; 88:1308-13. 1999; 33:395-418. 62. Sondik EJ, Lucas JW, Madans JH, Smith SS. Race/ 43. Telles EE, Lim N. Does it matter who answers the ethnicity and the 2000 census: implications for race question? Racial classification and income public health. Am J Public Health 2000; 90:1709- inequality in Brazil. Demography 1998; 35:465-74. 13. 44. Fry P. Politics, nationality, and the meaning of 63. Mckenney NR, Bennett CE. Issues regarding data “race” in Brazil. Daedalus 2000; 129:83-116. on race and ethnicity: the Census Bureau experi- 45. Cleary D. Race, nationalism and social theory in ence. Public Health Rep 1994; 109:16-25. Brazil: rethinking Gilberto Freyre. Cambridge: 64. Williams DR. Race/ethnicity and socioeconomic David Rockefeller Center for Latin American status: measurement and methodological issues. Studies, Harvard University; 1999. Int J Health Serv 1996; 26:483-505. 46. Hanchard M. Orpheus and power: the Movimen- 65. Williams DR, Jackson JS. Race/ethnicity and the to Negro of Rio de Janeiro and São Paulo, Brazil, 2000 census: recommendations for African Amer- 1945-1988. Princeton: Princeton University Press; ican and other black populations in the United 1994. States. Am J Public Health 2000; 90:1728-30. 47. Changing America: indicators of social and eco- 66. Hahn RA, Mulinare J, Teutsch SM. Inconsisten- nomic well-being by race and Hispanic origin. cies in coding of race and ethnicity between birth Washington, DC: Government Printing Office; and death in U.S. infants: a new look at infant 1998. mortality, 1983 through 1985. JAMA 1992; 267:259- 48. Ministério da Saúde. Manual de doenças mais 63. importantes, por razões étnicas, na população 67. Laws MB, Heckscher RA. Racial and ethnic identi- Brasileira afro-descendente. Brasília: Ministério fication practices in public health data systems in da Saúde; 2001. New England. Public Health Rep 2002; 117:50-61. 49. Henriques R. Desigualdade racial no Brasil: evo- 68. Editorial. Race, ethnicity and nationality. Paedia- lução das condições de vida na década de 90. Bra- tr Perinat Epidemiol 2000; 14:4-13. sília: Instituto de Pesquisa Econômica Aplicada; 69. Editorial. Census, race and science. Nat Genet 2001. (Texto para discussão; 807). 2000; 24:97-8. 50. Motta R. Paradigmas de interpretação das rela- 70. Schwartz RS. Racial profiling in medical research. ções raciais no Brasil. Estudos Afro-asiáticos 2000; N Engl J Med 2001; 344:1392-3. 38:1-17. 71. Loury GC. The anatomy of racial inequality. Cam- 51. Harris M, Consorte JG, Lang J, Byrne B. Who are bridge/London: Harvard University Press; 2002. the whites? Imposed census categories and the 72. Eisenberg JM. Physician utilization: the state of racial demography of Brazil. Social Forces 1993; research about physicians’ practice patterns. 72:451-62. Med Care 1985; 23:461-83. 52. Schwartzman S. Fora de foco: diversidade e iden- 73. Osborne NG, Feir MD. The use of race in medical tidades étnicas no Brasil. Cadernos de Pesquisa research. JAMA 1992; 267:275-9. 1999: 106:83-96. 74. House JS, Williams DR. Understanding and re- 53. Turra C, Venturini G. Racismo cordial. São Paulo: ducing socioeconomic and racial / ethnic dispar- Ática, 1995. ities in health. In: Smedly BD, Syme SL, editors. 54. Maggie Y. Aqueles a quem foi negada a cor do dia: Promoting health: intervention strategies from as categorias cor e raça na cultura brasileira. In: social and behavioral research. Washington, DC: Maio MC, Santos RV, organizadores. Raça, ciência National Academy Press; 2000. p. 81-124. e sociedade. Rio de Janeiro: Editora Fiocruz; 1996. 75. Kaufman JS, Cooper RS, McGee DL. Socioeco- p. 225-34. nomic status and health in blacks and whites: the 55. Silva NV. Morenidade: modo de usar. Estudos problem of residual confounding and the resilien- Afro-asiáticos 2002; 30:79-95. cy of race. Epidemiology 1997; 8:621-8. 56. Berquó E. The demography of inequality: consid- erations on Black people in Brazil. Brazilian Jour- Submitted on 13/Jan/2004 nal of Population Studies 1997; 1:165-77. Approved on 16/Feb/2004

Cad. Saúde Pública, Rio de Janeiro, 20(3):660-678, mai-jun, 2004