Residential Segregation by Skin Color in Author(s): Edward E. Telles Source: American Sociological Review , Apr., 1992, Vol. 57, No. 2 (Apr., 1992), pp. 186- 197 Published by: American Sociological Association Stable URL: https://www.jstor.org/stable/2096204

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL*

EDWARD E. TELLES University of California, Los Angeles

I examine residential segregation by skin color in 35 of the largest metropolitan areas in Brazil, using census tract data from the 1980 Brazilian census. Residential dissimilarity among whites, mulattoes (browns) and blacks is only moderate by U.S. standards. White- black dissimilarity is the highest,followed by brown-black and then white-brown dissimilar- ity. Residential segregation within income groups is lowest for the low income groups and increases with increasing income level. For most of the white middle class, residential sep- aration is ensured by the concentration of blacks and mulattoes in low socioeconomic class- es and in distinct regions. Multivariate analysis reveals that an urban area's socioeconomic status and housing market are strong predictors of overall residential segregation.* segrega- tion is significantly higher in urban areas with high occupational inequality, low mean in- come, high levels of housing turnover, and high homeownership. Measures of industrializa- tion, immigrant influence and color heterogeneity were not significant. Implicationsfor Bra- zilian race relations are discussed.

R esidential segregation occupies a central role widespread racial inequalities. For example, in in the sociological literature on race rela- Brazil in 1976, the mean income of African-ori- tions and racial inequality. Although there is a gin males was 47 percent of that of white males growing recognition of the importance of com- (Silva 1985, p. 45).l Blacks and mulattoes are dis- parative studies, virtually nothing is known about proportionately represented among the lower so- racial residential segregation outside of the Unit- cial classes - they have considerably lower in- ed States. Brazil's African-origin population is comes than whites, experience less social mobil- the second largest in the world only after Nige- ity than whites and and are more likely than whites ria's. Unlike the United States, Brazil has had no to be in the urban informal labor market (Olivei- race-based laws that encourage residential segre- ra, Porcaro, and Araijo 1985; Silva 1985; Hasen- gation since Abolition in 1888, yet segregation balg 1985; Lovell 1989; Telles forthcoming a). by skin color is prevalent. I examine overall lev- Since the abolition of slavery in 1888, three els of white-black, white-brown and brown-black important features have distinguished Brazilian residential segregation as well as segregation pat- race relations: (1) a color continuum rather than terns among color groups within income groups. a color line, (2) its racial ideology, and (3) an I then analyze factors that may contribute to vari- avoidance by the Brazilian government of legis- ation among Brazilian metropolitan areas in res- lation that mentions race. idential segregation. An important difference between U.S. and Bra- zilian race relations is that conceive of race as a continuous color variable rather than a BACKGROUND categorical variable (Hoetink 1967; Skidmore A resurgent academic interest in race issues in 1972; Denton and Massey 1989). Throughout Brazil has been fueled by the recent availability it's colonial period, Brazil's white population had of national statistics that reflect persistent and a very high ratio of males to females. At the same time, white males had relatively unlimited ac- * Direct all correspondence to Edward E. Telles, Department of Sociology, University of California, anonymous reviewers. Data were gathered while I Los Angeles, 264 Haines Hall, Los Angeles, CA was a Rockefeller Foundation Fellow in Population 90024-1551. I thank Arlindo Mello do Nascimento, Sciences at the State University of Campinas (UNI- Luisa Pinheiro Texeira and Luis Armando de Medei- CAMP) in Brazil and with the support of a UCLA ros Frias of the Instituto Brasileiro de Geografia e Academic Senate Grant. Estatistica for providing the data. I benefited from I In the United States in 1979, mean income for valuable comments by Walter Allen, David Lopez, black males was 61 percent of white male mean in- Joe Potter, Donald Treiman, the ASR editor, and the come (Farley and Allen 1989, p. 315).

186 American Sociological Review, 1992, Vol. 57 (April:186-197)

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 187 cess to Indian slaves and later, African origin racism of the time caused European countries to slaves. These two facts led to high levels of mise- look down on Latin America because of its large genation (Degler [1971] 1986) and in Brazil to- African and Indian populations (Skidmore 1974). day racial intermarriage is not uncommon (Silva The Brazilian legislature decided to encourage 1987).2 Although some localities recognize many the immigration of European workers to increase color gradients, all Brazilians recognize at least a the proportion of whites in the population. These tripartite color distinction: white, mulatto (brown), white immigrants would also be expected to mar- and black (Hutchinson 1957; Degler 1986). Dur- ry with the nonwhite population to eventually ing slavery, the color distinction often became a create a "whiter" population (Degler 1986; Vain- social distinction in which the common practice ner 1990). The whitening ideology has subse- of manumitting the offspring of white males and quently been blamed for preventing the forma- black slave women made mulattoes predominant tion of ethnic identities based on race and for among an intermediate social category while defeating attempts to form a black consciousness "blacks" made up most of the slave population (Skidmore 1974; Nascimento 1982). (Harris 1964; Schwartz 1974). This process has- Another feature of Brazil's racial ideology that tened the acculturation of nonwhites and thus dates to the 1920s is its claim to being a racial forestalled the development of rigid and legal- democracy in which race makes no difference to ized color lines, such as emerged in the United opportunity or status. Gilberto Freyre popular- States (Schermerhorn 1978; Degler 1986). How- ized this idea in his historical accounts by extol- ever, findings based on recent national surveys ling the virtues of the Brazilian race system while and census data demonstrate that the earnings of overlooking its subtle forms of racism (Freyre mulattoes are much more similar to black earn- 1933).3 The racial democracy idea was grafted ings than to white earnings and the earnings "cost" into official ideology by the mostly conservative due to racial discrimination is often similar be- and military elite that had ruled Brazil since the tween mulattoes and blacks (Silva 1985; Lovell 1930s to forestall racial uprisings and to show- 1989). By contrast, findings for intermarriage case Brazil internationally as a positive example reveal a continuum where white-brown and of race relations (Skidmore 1972). brown-black marriages are far more common than With one exception, laws in post-Abolition white-black marriages (Silva 1987; Telles forth- Brazil have not been directed at racial segrega- coming b). tion or integration (Eccles 1991). Furthermore, The lack of clearly defined color lines has made there has never been a legal basis for racial cate- racial self-identification somewhat flexible in gorization, unlike the United States or South Af- Brazil and Latin America, and sometimes varies rica. Brazilian law has also never been used to depending on social context (Rodriguez and Cord- encourage racial equality, probably because this ero-Guzman forthcoming; Wood 1991). Persons would mean acknowledging the existence of rac- identifying as white or black often are not "ra- ism and racial inequalities (Skidmore 1972). The cially pure" but are "relatively white" or "rela- only exception to this generalization was the re- tively black" and there is a tendency for persons luctantly passed Alfonso Arinas law of 1951, on the border of a color category to "pass" into which made racial discrimination in public plac- the lighter category (Hutchinson 1957; Degler es a misdemeanor sanctioned by fine or impris- 1986). The high value placed on light skin color onment (Eccles 1991). The rare cases prosecuted and the continued importance of the mixed-race under this law have been of a blatantly racist category emerges largely out of the Brazilian ide- nature, leaving the subtler institutionalized forms ology of whitening. of racism - including clear cases of housing The goal of whitening was made explicit in discrimination - free of penalty or reparation nineteenth century debates among the Brazilian (Skidmore 1972; Eccles 1991). elite about how to replace the largely slave labor force that would soon be manumitted through I Freyre left his native Brazil to study in the U.S. Abolition. The elite were concerned about Bra- South between 1917 and 1920. It is easy to imagine zil's international status in which the scientific how the contrast between the two systems at that time may have led him to such conclusions. Donald Pier- 2Silva (1987) showed that in 1980, about 21 per- son, an American sociologist, later upheld Freyre's cent of currently-married couples reported being in findings and effectively promoted the racial democ- racially mixed unions: 16.1 percent were between racy thesis in the U.S. academic community (Pierson whites and browns, 3.2 percent between browns and 1942). Notably, both scholars had studied the mostly blacks, and 1.4 percent between whites and blacks. nonwhite Brazilian Northeast.

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms 188 AMERICAN SOCIOLOGICAL REVIEW

RESIDENTIAL SEGREGATION Two studies noted that racial residential segre- BY SKIN COLOR gation appears to be particularly high in the Bra- zilian South (Willems 1949; Turner 1985). Re- Work on racial residential segregation in Brazil gional and metropolitan area variations in race is scant. Pierson (1942) investigated residential relations in general were suggested by the segregation in his seminal study of Salvador, UNESCO scholars and others (Degler 1986). Bahia, a predominantly nonwhite city in Brazil's Scattered evidence suggests that the less devel- Northeast. Based on his impressions of various oped and mostly nonwhite Northeast has a par- parts of the city, he observed that the skin color ticularly amicable system of race relations, while of residents varied according to the quality of the the state of Sdo Paulo and the three southern neighborhood: The poorest and most overcrowd- states are particularly racist (Degler 1986). ed areas of the city were inhabited by blacks, dark- Such variation in race relations within Brazil skinned mulattoes, and a limited number of light- is said to be the result of local industrialization skinned mulattoes, while whites and occasional and/or immigrant influences. Bastide and Van light-skinned mulattoes lived in the middle-class den Berghe have argued that race relations are sectors of town. Thus, he painted a picture of con- more competitive and racism heightened in more siderable residential segregation by color although industrialized areas (Bastide and Van den Berghe he concluded that it existed only to the extent that 1957; Bastide 1965; Van den Berghe 1967). By color and class were coterminous. He noted that, contrast, Fernandes (1969) contended that with unlike the United States, there was no attempt to industrialization the salience of race in the labor segregate races and class distinctions were the market would diminish and consequently racism result of the Afro-Brazilians' recent emergence would diminish. The immigration hypothesis from a servile status. theorizes that European immigration since Ab- During the 1950s and early 1960s, research olition affected race relations in those areas sponsored by UNESCO reported substantial res- where immigrants settled. One variant argues idential segregation in Rio de Janeiro, a city with that the predominantly Italian and German im- a large minority population (roughly 30 percent migrants competed for jobs with the African- nonwhite) and in Florianopolis, a city with a small origin population, thus creating a high level of minority population (roughly 10 percent non- racial animosity (Andrews 1988). Another ver- white) but concluded that in both cases, segrega- sion suggests that these immigrants brought tion was highly conditioned by social class (Ian- with them ideologies that were more racist than ni and Cardoso 1959; Pinto 1953).4 However, those of the native Luzo-Brazilian culture (Fra- surveys of racial attitudes conducted during the zier 1942; Willems 1949). Fernandes (1955) 1950s (the same period), found considerable ra- disputed these theories with evidence that cit- cial prejudice, including resistance by many ies like Campinas that had relatively few im- whites to having blacks and mulattoes as neigh- migrants were clearly more racist than other bors (lanni and Cardoso 1959; Bastide and Van cities in the state of Sdo Paulo with large im- den Berghe 1957; Fernandes 1955). A recent his- migrant populations. torical study of the two largest metropolitan areas An alternative explanation derives from showed that the African origin population tends Blau's (1977) focus on the effects of racial to live in certain poor neighborhoods, near other heterogeneity. Blau hypothesized that social "co-ethnics" and generally near Afro-Brazilian group differences are more noticeable when cultural and religious institutions. These neigh- one group is numerically predominant and less borhoods often grew around the cores of former obvious when the population is most hetero- slave neighborhoods (Rolnik 1989). geneous. Where there is greater heterogene- ity, racial differences are less conspicuous and interracial exchange is less constrained. This 4 The flurry of UNESCO-supported research theory has particular appeal in Brazil, given showed widespread racial prejudice and inequality in its wide variation in racial composition. The Brazil, contrary to the earlier findings of scholars like Freyre and Pierson who believed that any racial prej- five urban areas with the highest white-brown udice could be attributed almost entirely to class prej- segregation (Salvador, Teresina, Florianopo- udice and that the social position of nonwhites was a lis, Joinville and Caxias do Sul), also have the result of their recent emergence from slavery in a least heterogeneity, where nonwhites are a society where upward mobility was very limited for large majority in the first two areas and whites persons of any race (Skidmore 1974; Degler 1986). are a large majority in the latter three areas.

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 189

METHODS AND DATA and blacks suggests a system of race relations distinct from most of Brazil in which there is a I focus on the traditional conception of residen- continuum from European to African. The 35 tial segregation - evenness, i.e., the extent to metropolitan areas include 36.1 percent of Bra- which minority and majority group members are zil's total population in 1980 and 53.5 percent of evenly distributed across an urban area (Massey its urban population. and Denton 1988). Evenness is measured with The study of residential segregation in Brazil the traditional index of dissimilarity, D, defined has been impeded by the lack of data available at as the proportion of one group that would have to the neighborhood level. Because data had been change census tracts to achieve the same spatial available only for large districts, the study of seg- distribution as the other group (Duncan and Dun- regation in a context in which squatter settle- can 1955). D is mathematically independent of a ments (favelas) are often located amidst formal- metropolitan area's racial composition (Massey ly-zoned housing was highly problematic. How- and Denton 1988). ever, an agreement by the IBGE to compute seg- The value of D varies from 0, when two racial regation indexes using census tract level data groups are evenly distributed throughout the ur- overcame this problem. ban area, to 1, when there is complete segrega- In Brazil, census tracts in urban areas contain tion. about 250 to 300 households. At each decennial census since 1940, the IBGE has updated census D = .5 x >I(xIX) - (y,/Y)l, (1) tract boundaries to conform to this size specifi- where X and Y are the metropolitan area popula- cation. Although a strict policy about disclosure tions of the groups being compared and x, and y, of census tract level data prevents the IBGE from are their respective populations in census tract i. making such data publicly available, the IBGE Residential dissimilarity indices among whites, agreed to compute the segregation indices them- browns and blacks in 35 metropolitan areas were selves while allowing me to closely monitor the computed for the total population and within fam- process. When comparing these indices with those ily income groups. The 1980 Population Census of other countries it should be kept in mind that of Brazil, produced by the Instituto Brasileiro de the average census tract population in Brazilian Geografia e Estatistica (IBGE) is the only source urban areas is about 1150, because "D" tends to for computing segregation indexes based on col- mathematically increase as the size of the aerial or.5 Segregation indexes are based on census tract unit decreases (Massey and Denton 1988). data from the 25 percent sample of households The 1980 census asked respondents to identify that answered the long form of the census ques- themselves as to their color. Possible responses tionnaire. The independent variables in the re- were white brancho), brown (pardo), black (pre- gression analysis were also calculated from this to), yellow (amarello), and other. Fully 99 per- sample. cent of the responses were in the first three cate- Metropolitan areas are defined as contiguous gories. Brown refers to both Indians and mixed urbanized areas in one or more municipalities race persons, i.e., those whose skin color is be- and were defined in a previous IBGE publication tween black and white. (Vetter 1988). I chose 35 of the 40 metropolitan Although the native Indian population is cate- areas with populations over 200,000 in 1980. The gorized as brown in this census, a survey showed five excluded areas are in the North and Central- that persons who would identify themselves as West regions (the Brazilian frontier) where a high- "Indian" comprise a very small proportion of the er concentration of Indians and their descendants national population although they are a signifi- among persons identifying as whites, browns, cant proportion of the population in the Amazon (North) region and in the Central-West (Oliveira IThe 1970 Census of Brazil, the only other avail- Filho 1987). Since the beginnings of European able census with machine readable files, omitted the colonization, disease and war have led to huge race question. The Census scheduled for 1990 was declines in the size of the Brazilian Indian popu- actually taken in September of 1991. Skidmore (1974) lation (Denevan 1976; Hemming 1978). Howev- interprets the lack of data on race in the 1970 Census er, many Brazilians today, including some iden- as reflecting the military government's concern that such data that would point to racial inequalities and tifying themselves as white, have Indian slave thus encourage racial opposition. Officially, they ancestors (Monteiro 1985). The extent to which claimed that race was not a statistically meaningful the Brazilian population has Indian ancestry is category. difficult to ascertain except by self-identification.

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms 190 AMERICAN SOCIOLOGICAL REVIEW

Table 1. Regional Distribution of Total Population by Re- regionally segregated from each other (D = .447) gion for Color Groups: Brazil, 1980 than are blacks from whites (D = .286) or blacks from browns (D = .268). Color

Region White Brown Black Total Metropolitan Areas Northeast 14.5 49.6 33.2 29.3 Central-East 24.0 16.8 33.6 22.5 The final column of Table 2 shows that the racial Sdo Paulo 29.0 11.5 17.9 21.1 composition of individual metropolitan areas var- South 24.8 5.0 8.5 16.0 ies widely, from Joinville where whites are 95.2 North 1.8 9.7 2.2 4.9 percent of the local population to Teresina, where Central-West 5.7 7.6 4.5 6.3 they are 15.0 percent. Whites are a minority in all Total 99.8 100.2 99.9 100.1 13 urban areas of the less developed Northeast, but make up more than 80 percent of the popula- Source: Instituto Brasileiro de Geografia e Estatistica tion in 10 of the 14 areas in Sdo Paulo and the 1983, Table 1.11. South. Brazil's two giant cities, Sdo Paulo and Rio de Janeiro, both have white majorities (71.5 Perhaps most Brazilians identify as white, black, and 60.0 percent) but have African origin popu- or mulatto rather than Indian or part-Indian be- lations surpassing three million. cause extensive miscegenation with Indians oc- By U.S. standards, values of D between .3 and curred mostly in the sixteenth and seventeenth .6 are considered moderate (Massey and Denton centuries as compared to more recent miscege- 1987). Table 2 thus shows that residential segre- nation with blacks (Monteiro 1985) and because gation by color in Brazil is moderate - all but of a racial ideology that emphasizes a continuum one of the dissimilarity indexes are below .6. from white to black. The concentration of Indi- One factor that may depress these indexes is the ans and their descendants in the two regions of tendency for domestic workers and building the Brazilian frontier along with current strug- guards and their families, who are predominately gles for indigenous land rights (Oliveira Filho nonwhite, to live in the same household or build- 1987) contributed to a system of race relations ing as their employers. On the other hand, the and concerns in those areas that differs from the smaller average population size of Brazilian cen- eastern half of the country where about 90 per- sus tracts (1,150 compared to 4,000 in the United cent of Brazilians reside. Thus, the problems of States) may operate to inflate Brazilian segrega- including large numbers of Indians with the mu- tion indexes. latto population and of classifying persons with The mean segregation of blacks versus whites Indian admixture as "white" are attenuated by (.450) for these 35 urban areas is only slightly eliminating North and Central-West metropoli- greater than that for browns versus whites (.397), tan areas from the analysiL. a pattern typified in 29 of the 35 urban areas. In the remaining six urban areas, four of which are in the Central-East region, the black-white segregation RESULTS index was lower than the brown-white index. White-black dissimilarity was highest in all 13 Regions metropolitan areas of the Northeast. In Brazil, nonwhites are clearly a numerical mi- Interestingly, the mean index of brown-black nority in the more developed areas of the country segregation (.407) was higher than that for white- while they are a clear majority in the less devel- brown segregation (.397). The brown-black seg- oped Northeast, which was the center of the slave regation index was the highest of the three index- based economy prior to 1888. Table 1 shows the es in five metropolitan areas, including three in distribution of color groups by region. The ma- the South, and the lowest in 15 areas. In the North- jority (53.8 percent) of whites live in the indus- east-, brown-black dissimilarity was between trialized, highly populated and relatively wealthy white-black and white-brown dissimilarity in eight state of Sdo Paulo and the South, a region com- of the 13 areas and in Sdo Paulo state it was the prising three slightly less developed states. Con- lowest of the three indexes in five of the seven versely, about one-half of browns and one-third metropolitan areas. However, white-brown and of blacks live in the poor and underdeveloped white-black segregation indexes may be mislead- Northeast. Indexes of dissimilarity computed by ingly low and thus not directly comparable to region reveal that browns and whites are more brown-black segregation indexes because of the

This content downloaded from 68.8.165.ff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 191 tendency for many nonwhite domestic workers to Table 2. Indexes of Dissimilarity Among Whites, Browns, reside with their white middle-class employers. and Blacks by Metropolitan Areas: Brazil, 1980 Nevertheless, the moderate level of segregation Index of Dissimilarity between browns and blacks, who are of similar White White Brown socioeconomic status, demonstrates the impor- Region and vs. vs. vs. Percent tance of the color continuum. Metropolitan Area Black Brown Black White Average residential dissimilarity by region is Northeast .509 .405 .434 30.1 highest in the Northeast and South, followed by Recife .499 .389 .424 37.1 Sdo Paulo and finally, the Central-East. Previous studies, based on impressionistic evidence, sug- Salvador .534 .488 .353 22.8 gested that the Northeast was the least segregat- Fortaleza .563 .409 .562 30.6 ed of all regions (Willems 1949; Turner 1985). It Belem .495 .376 .431 27.1 is clearly the most segregated in terms of white- Joao Pessoa .541 .439 .482 33.0 black dissimilarity. As long as segregation is not Natal .596 .393 .585 47.2 extreme, there will be a noticeable presence of Teresina .500 .486 .469 15.0 nonwhites in areas of the Northeast where whites Maceio .495 .379 .450 36.5 live because nonwhites are a majority in the Aracaju .467 .381 .398 27.0 Northeast. This gives the impression to a casual observer that white segregation is lowest in the Sao Luis .471 .362 .320 28.0 Northeast. Feira de Santana .504 .409 .375 20.5 Campina Grande .468 .356 .444 45.0

Socioeconomic Segregation Itabuna .488 .395 .343 21.1

To examine the extent to which color segrega- Centr-al-East .381 .367 .344 61.3 tion is independent of social class, I computed Rio de Janeiro .427 .383 .312 60.0 residential dissimilarity indexes among the three Belo Horizonte .419 .427 .341 54.1 color groups within household income brackets. Vitoria .455 .356 .371 48.6

These indexes provide a reasonable control for Barra Mansa- .297 .318 .324 65.2 the "money whitens" effect - if color identifi- Volta Redonda cation changes when individuals move up the Juiz de Fora .364 .358 .327 71.1 economic ladder, then overall segregation may Ipatinga .356 .396 .393 50.0 reflect not only inequality, but the misreporting Uberlandia .381 .388 .363 75.5 of color by high income nonwhites. Campos .346 .312 .319 66.0 Dissimilarity indexes within income levels were calculated using households as the unit of Sdo Paulo .402 .380 .364 78.3 analysis and defining color as the color of the Sao Paulo .412 .394 .364 71.5 family head. Results are presented in Table 3. Santos .441 .424 .369 67.4 Indexes were not calculated if either of the two Campinas .429 .412 .359 78.6 color groups in an income bracket averaged less Sao Jose dos .370 .332 .389 80.1 than three households per tract. This was done to Campos avoid random departures from evenness that Sorocaba .377 .305 .358 85.5 might otherwise occur by using very small popu- Ribeirao Preto .406 .387 .337 82.1 lations in the calculation of D. This criterion meant that indexes could not be calculated for many Jundiai .378 .405 .369 83.0 income groups in many of the urban areas, which South .467 .433 .476 86.4 reflects the concentration of nonwhites in low- Porto Alegre .422 .407 .421 85.3 income categories. Thus, residential isolation of .477 .417 .467 84.1 whites is virtually assured by the absence of a Pelotas-Rio Grande .386 .352 .440 84.6 significant nonwhite middle class. Household income is measured as the the num- Florian6polis .461 .492 .516 91.8 ber of "minimum salaries," a common procedure Londrina .449 .397 .404 74.3 in Brazil where high inflation makes actual dol- Joinville .526 .470 .624 95.2 lar or cruzeiro values less meaningful. In Sep- Caxias do Sul .546 .495 .457 89.2 tember 1980 when the census was taken, one minimum salary was worth about 75 U.S. dollars Mean .450 .397 .407 58.1

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms 192 AMERICAN SOCIOLOGICAL REVIEW

Table 3. Indexes of Dissimilarity Among Whites, Browns, criterion. Table 3 shows that in metropolitan ar- and Blacks by Family Income Group: Seven Met- eas in which blacks and mulattoes are represent- ropolitan Areas in Brazil, 1980 ed in the various income groups, moderate seg-

Dissimilarity Between: regation persists independent of income. In fact, segregation indexes tend to increase with income. Metropolitan Area and White White Brown Income Group (Numbers vs. vs. vs. This pattern is unmistakable for white-black dis- of Minimum Salaries) Black Brown Black similarity and the general pattern also holds for

Salvador brown-black and white-brown segregation. In 1.01 to2 .519 .499 .391 addition, the increase in white-black segregation 2.01 to 3 .547 .519 .411 over the income groups is greater in all seven 3.01 to 5 .547 .511 .418 areas than that for white-brown segregation while 5.01 to 10 .618 .497 .483 brown-black segregation increases are interme- 10.01 to 20 .509 diate. Salvador, the only metropolitan area with 20.01 and over .507 a significant nonwhite elite, is typical of the first Feira de Santana four cities in Table 3. Although white-black seg- 1.01 to 2 .514 .468 .348 regation in the lowest income group (.519) is 2.01 to 3 .572 .464 .439 clearly higher than brown-black segregation 3.01 to 5 .585 .481 .445 5.01 to 10 .615 .465 .549 (.391), both increase by similar amounts between 10.01 to 20 .518 the first and fourth income categories (.092 and

Rio de Janeiro .099). On the other hand, white-brown segrega- 1.01 to 2 .419 .382 .399 tion in the lowest income category is only slight- 2.01 to 3 .456 .389 .424 ly less than black-white segregation but is little 3.01 to 5 .452 .387 .409 changed by the fourth category (.497) and in- 5.01 to 10 .543 .421 .479 creases only slightly in the two highest income 10.01 to 20 .546 groups. In Barra Mansa-Volta Redonda, a highly Belo Horizonte industrialized area, white-brown segregation ac- 1.01 to 2 .435 .412 .405 tually decreases slightly with increasing income 2.01 to 3 .450 .419 .437 until the highest category where it jumps to a 3.01 to 5 .462 .416 .450 relatively high level (.549). These cases demon- 5.01 to 10 .568 .454 .518 10.01 to 20 .554 strate the importance of the color continuum in which black is perceived as lower status than Barra Mansa-Volta Redonda 1.01 to 2 .355 .387 .423 brown and brown-black distinctions are made 2.01 to 3 .360 .358 .406 even by nonwhites. The pattern becomes clearer 3.01 to 5 .377 .362 .422 as income increases - segregation is greater 5.01 to 10 .421 . .361 .439 among high income households where housing 10.01 to 20 .549 options are greater. Juiz de Fora Two urban areas, Juiz de Fora and Campos, 1.01 to 2 .318 .337 .337 show a different pattern - segregation indexes 2.01 to 3 .380 .415 .408 for all color comparisons increase sharply as in- 3.01 to 5 .410 .436 .449 come rises. However, the difference between the 5.01 to 10 .568 .465 .556 lowest and the highest income group is greatest Campos for white-black segregation (.250 and .241) com- 1.01 to 2 .377 .335 .376 pared to those for brown-black segregation (.219 2.01 to3 .411 .374 .451 and .228) and white-brown segregation (.128 and 3.01 to 5 .448 .384 .461 5.01 to 10 .618 .450 .604 .115). Segregation in these two areas is particu- larly low among the lowest income groups where, for example, white-black segregation resembles per month. Indexes were not presented for the white-brown segregation. However, white-black "less than one minimum salary" category because segregation, which in the other five areas is the results were inconsistent with those for the other highest of the three indexes at each income level, income categories, suggesting substantial misre- is the highest of the three income groups only at porting at the low end of the income spectrum. the highest income group (5.01 to 10 minimum The seven metropolitan areas presented in Ta- salaries) in Juiz de Fora and Campos. ble 3 are those for which the greatest number of The fact that residential segregation by color is indexes could be calculated based on the above lowest among the very poor may help explain

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 193 why race relations among the poor have been ity data, this admittedly approximates the rela- comparatively amicable - historically there has tive immigrant influence. Heterogeneity is cal- been little race-related violence (Skidmore 1974). culated using the following formula: The low levels of segregation among the poor H= I -(w 2 + M 2+b 2) , (2) may also reflect the precarious living conditions of the poorest sectors of the Brazilian popula- where w is the percent white, m is the percent tion. Residential choices are extremely limited brown and b is the percent black (Gibbs and and must often be made on the basis of survival Martin 1962). chances. For the poor, the decision about where Socioeconomic status is measured using two to live is especially likely to be based on criteria variables: (1) male occupational dissimilarity in- like walking distance to work (because bus fares dexes between paired color groups in each met- are often not affordable) or where friends live ropolitan area and (2) mean household income who can assist in tasks such as child care or hous- for the entire population of each area. Color ine- ing construction. This often means living illega- quality is commonly postulated as the sole source ly in makeshift housing. A context of extreme of residential segregation in Brazil where hous- poverty coupled with an absence of overwhelm- ing markets merely reflect a social division of ing social pressures to segregate by color make labor; income provides the economic resources concerns about color a low priority. If segrega- that permit residential mobility.- tion is to occur in a context where extremely Any analysis of residential segregation must segregated residential patterns do not already exist consider the local housing market. I control for or segregation is delegalized, it is most likely to three housing related factors: residential turnover, occur at income levels where housing options homeownership, and density. High housing turn- are greater and color can become a criterion in over in an area indicates a high level of internal neighborhood selection. movement and migration. Turnover is measured Salvador is significant in that it is one city as the percent of households living less than two which has a significant mulatto middle class (Aze- years in their current dwelling. Persons in high vedo 1953). Tables 2 and 3 show that Salvador turnover areas are freer to choose their residences has high white-brown and white-black segrega- than are those in the tight markets typical of some tion but relatively low brown-black segregation. large Brazilian metropolitan areas. Movers are The relatively high segregation between whites often channeled into relatively homogeneous and nonwhites may promote social solidarity, as neighborhoods through social networks. This may manifested spatially, between browns and blacks either increase segregation or contribute to ex- in Salvador. Also important in Salvador is the panding areas of transition, thus reducing segre- strong presence of African-origin culture and gation. Homeownership is the percent of house- institutions which further differentiates the so- holds that are owner-occupied. Homeowners see cial lives of Afro-Salvadorans from Salvadorans their homes as long-term investments so they are of European origin. more likely than renters to be concerned about who their neighbors are. If neighborhood homo- geneity is preferred, then segregation is likely to Explaining Inter-Urban Patterns be higher where homeownership is high. Finally, The causes of differential segregation are tested a given level of segregation is entirely different in using an OLS regression model that considers a dense urban area where persons live in close urban segregation indexes as a function of social proximity to each other compared to areas that context, socioeconomic status, and housing mar- are more spread out. If whites want to avoid non- ket variables. Dissimilarity indexes (presented in whites, they would be physically closer to Afri- Table 2) are transformed into logits because of can-origin neighbors in a dense neighborhood, their limited range. Social context variables in- thus making segregation more likely in dense ar- clude industrialization, immigrant influence, and eas. Density is operationalized as the percent of heterogeneity. Industrialization is the percent of households living in apartment buildings. the total labor force in manufacturing. Immigrant Table 4 shows the results of the regression influence is estimated by -the percent Italian- and analysis. The coefficients are fairly consistent German-born individuals among the white pop- across the three models although levels of signif- ulation aged 60 and over in 1980 because such icance vary. Overall, variables representing so- immigration continued only until the 1930s (Mer- cioeconomic status and the housing market are rick and Graham 1979). In the absence of ethnic- better predictors of segregation than are the so-

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms 194 AMERICAN SOCIOLOGICAL REVIEW

Table 4. Regression Coefficients for Residential Dissimi- tion between blacks and both whites and browns larity Indexes on Selected Independent Variables: is especially high in low income urban areas. Metropolitan Areas in Brazil, 1980 Urban areas with high residential turnover White White Brown clearly have higher segregation as indicated by vs. vs. vs. the significant positive coefficients in all three Independent Variable Brown Black Black models. Urban areas with high levels of home- Social context ownership are significantly associated with great-

Percent in manufacturing .009 .006 .008 er white-brown and white-black segregation. (.006) (.004) (.006)

Percent of whites age 2 60 .073 .006 .040 DISCUSSION AND CONCLUSION born in Italy/Germany (.056) (.039) (.061)

Color heterogeneity -.004 -.004 -.012 By U.S. standards, residential segregation among (.005) (.003) (.006) color groups in Brazil is generally moderate. Al-

Socioeconomic status though an absence of legal segregation through- out its modem history may explain the absence Occupational dissimilarity .026* .027* .013 (.012) (.010) (.022) of extreme segregation such as is found in other parts of the world, moderate segregation is a fact Mean household income -.032* -.007 -.039* -(.016) (.010) (.017) of urban Brazilian life. Results show that residential segregation Housing market among color groups cannot be accounted for by Percent < 2 years in .026* .020* .028* socioeconomic status, as previous literature has dwelling (.009) (.006) (.010) claimed - moderate residential segregation along Percent homeowners .019* .020* .013 color lines occurs among members of the same (.008) (.005) (.009) income group. Furthermore, segregation levels Percent living in .012 .009 .010 increase with rising income in those metropoli- apartments (.007) (.004) (.008) tan areas with substantial numbers of whites, Intercept -2.425 -2.773 -1.196 blacks and mulattoes in the income categories. R2 .496 .577 .537 Thus, the Afro-Brazilian middle class is more Adjusted R2 .341 .447 .394 spatially dissimilar from middle-class whites than Number of cases 35 35 35 poor Afro-Brazilians are from poor whites. For the white middle class in most metropolitan ar- *p ?.05 eas, though, the mere absence of a significant Note: Numbers in parentheses are standard errors. Dis- African origin middle class ensures that middle- similarity indexes were transformed into logits. class neighborhoods remain predominately white. The relatively low color segregation among the cial context variables. None of the social context large urban poor population suggests that the ex- variables are significant. Areas with high indus- tremely limited housing options make color a trialization, greater immigrant influence and lower low priority when choosing a place to live. racial heterogeneity tend to have higher segrega- Despite the suggestion in the literature that in- tion but these relationships are not significant.6 dustrialization and immigration in Brazil are cru- As expected, occupational dissimilarity has a cial to understanding regional differences in Bra- significant positive relationship with white-brown zilian race relations (Bastide 1965; Van den segregation and white-black segregation, but its Berghe 1967; Fernandes 1969), my analysis relationship is not significant for brown-black shows that these factors have weak effects at segregation. The lack of a relationship for brown- best. On the other hand, an area's socioeconomic black segregation is not surprising given the small status and its housing market are significantly differences in status between browns and blacks. associated with patterns of segregation. Urban Mean household income has a significant nega- areas with relatively high occupational inequali- tive relationship for white-black segregation and ty tend to be more segregated and high income brown-black segregation - residential segrega- urban areas tend to be more segregated. Urban areas with high housing turnover or a high rate of 6There is evidence of regional subcultures of rac- ism that may not be captured by by industrialization, homeownership also tend to have higher segre- immigration or racial composition (Telles forthcom- gation among color groups in Brazil. ing b). However, dummy variables for regions added The existence of a color continuum in Brazil to the multivariate models were not significant. from white to brown to black is thus supported in

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 195 the case of residential segregation. Whites are high levels of interracial interaction. Whereas generally more segregated from blacks than from extreme segregation in South Africa and the Unit- mulattoes, both overall and within income groups. ed States has led to a high degree of race con- There is also significant residential segregation sciousness and corporate organization, moderate between blacks and mulattoes, often higher than segregation and the absence of clearly-defined white-brown segregation (but lower than white- racial categories have led to their relative ab- black segregation), suggesting that mulattoes also sence in Brazil. Along with an ideology that de- disdain blacks, presumably encouraged by an ide- nies racism, the fact that blacks, mulattoes, and ology that stresses whitening. Thus, secondary whites live in similar neighborhoods may race relations, including residential segregation strengthen a general perception that race has lit- and intermarriage (Telles forthcoming b), tend to tle or no effect on life chances, at least for indi- reflect a color continuum while primary relations viduals of the same social class. In this sense, the (labor market) are more indicative of a color line pervasive racism and racial inequality of Brazil- between white and nonwhite (Silva 1985; Lovell ian society is less conspicuous. Also, the absence 1989). Alsorthe fact that segregation increases at of extreme segregation in Brazil has precluded a faster rate with increasing income for both the formation of parallel institutions, a condition white-black and brown-black segregation than that was fundamental to the rise of black social for white-brown segregation suggests that black movements in South Africa and the United States. skin color is especially salient for higher income Finally, Brazil's less extreme segregation has groups in the same area. Both bivariate and mul- meant greater interracial exposure as evidenced tivariate findings also indicate that segregation by far greater prevalence of interracial friendship of blacks from both whites and browns is clearly and intermarriage, at least among Brazil's large greater than white-brown segregation in poorer poor population. and less developed areas. Moderate segregation has widespread impli- EDWARD E. TELLES is Assistant Professor of Sociol- ogy at the University of California, Los Angeles. His cations for other features of Brazilian race rela- current research focuses on Brazil, particularly so- tions, especially when compared to other coun- cioeconomic segregation, race relations, and income tries with large populations of African and Euro- inequality. His ongoing research also includes the ef- pean descent. These are: (1) a relatively low lev- fect of skin colorlphenotype on life chances among el of race consciousness and organization among Chicanos and the labor market effects of immigration the African origin population, and (2) relatively to the United States.

Appendix A: Correlation Matrix for Variables in Table 4

1 2 3 4 5 6 7 8 9 10 11 12 13

1. White-brown dissimilarity 2. White-black .616* dissimilarity 3. Brown-black .446* .680* dissimilarity 4. Percent of whites age .162 -.158 -.148 > 60 born in Italy/Ger 5. Percent in -.009 -.245 .096 .548* manufacturing 6. Color heterogeneity -.304 .012 -.393* -.345* -.520* 7. Mean income .085 -.369* -.286 .795* .531* -.352* 8. White-brown occupa- .225 .271 -.303 .111 -.450 .431* .042 tional dissimilarity 9. White-black occupa- .101 .344* -.149 -.470* -.792* .662* -.554* .702* tional dissimilarity 10. Brown-black occupa- .056 .371 * .055 -.717* -.702* .592* -.773* .269 .839* tional dissimilarity 11. Percent < 2 years .125 .148 .233 -.040 -.027 .069 .025 -.026 -.048 -.031 in dwelling 12. Percent living .140 -.134 -.296 .397* .002 .051 .538* .284 -.123 -.384* -.233 - in apartments 13. Percent homeowners .380* .401 * .374* -.275 -.241 -.273 -.386* -.161 .165 .347* -.403* -.206

p ?.05

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms 196 AMERICAN SOCIOLOGICAL REVIEW

REFERENCES Hemming, John. 1978. Red Gold: The Conquest of the Brazilian Indians. Cambridge: Harvard Uni- Andrews, George Reid. 1988. "Black and White versity Press. Workers: Sdo Paulo, Brazil, 1888-1928." Hispan- Hoetink, H. 1967. The Two Variants in Caribbean ic American Historical Review 68:491-524. Race Relations. , England: Oxford Univer- Azevedo, Thales de. 1953. Les Elites de Couleur dans sity Press. une Ville Bre'sillienne (Elites of Color in a Brazil- Hutchinson, Harry William. 1957. Village and Plan- ian City). Paris: UNESCO. tation Life in Northeastern Brazil. Seattle: Univer- Bastide, Roger and Pierre Van den Berghe. 1957. sity of Washington Press. "Stereotypes, Norms and Interracial Behavior in Ianni, Octavio and Fernando Henrique Cardoso. 1959. Sdo Paulo, Brazil." American Sociological Review Cor e Mobilidade Social em Florian polis. (Color 22:689-94. and Social Mobility in Florian'polis). Sdo Paulo: Bastide, Roger. 1965. "The Development of Race Companhia Editora Nacional. Relations in Brazil." Pp. 9-29 in Industrialization Instituto Brasileiro de Estatistica e Geografia. 1983. and Race Relations, edited by G. Hunter. London, Censo Demografico: Dados Gerais-Migragdo-In- England: Oxford University Press. struqdo-Fecundidade-Mortalidade (General Infor- Blau, Peter. 1977. Heterogeneity and Inequality. New mation-Migration-Education-Fertility-Mortality). York: The Free Press. Vol. 1, ser. 4, no. 1. Rio de Janeiro: Instituto Brasil- Degler, Carl N. [1971] 1986. Neither Black nor White: eiro de Estatistica e Geografia. Slavery and Race Relations in Brazil and the Unit- Lovell, Peggy. 1989. Income and Racial Inequality in ed States. Madison: University of Wisconsin Press. Brazil. Ph.D. dissertation. Department of Sociol- Denavan, William M. 1976. "The Aboriginal Popula- ogy, University of Florida, Gainesville. tion of Amazonia." Pp. 205-34 in The Native Pop- Massey, Douglas S. and Nancy A. Denton. 1987. ulation of the Americas in 1492, edited by W. M. "Trends in the Residential Segregation of Hispan- Denevan. Madison: University of Wisconsin Press. ics, Blacks and Asians: 1970-1980."American So- Denton, Nancy and Douglas Massey, 1989. "Racial ciological Review 52:802-24. Identity Among Caribbean Hispanics: The Effect . 1988. "The Dimensions of Residential of Double Minority Status on Residential Segrega- Segregation." Social Forces. 67:281-315. tion." American Sociological Review 54:790-808. Merrick, Thomas W. and Douglas H. Graham. 1979. Duncan, Otis D. and Beverly Duncan. 1955. "A Meth- Population and Economic Development in Brazil: odological Analysis of Segregation Indices." Amer- 1800 to the Present. Baltimore: Johns Hopkins Uni- ican Sociological Review 20:210-17. versity. Eccles, Peter R. 1991. "Culpados Ate Prova em Con- Monteiro, John. 1985. Sdo Paulo in the 17th Century: trario: Os Negros, A Lei e os Direitos Humanos no Economy and Society. Ph.D. dissertation. Depart- Brasil" (Guilty Until Proven Innocent: Blacks, the ment of History, University of Chicago. Law and ). Estudos Afro- Nascimento, Abdias do. 1982. 0 Negro Revoltado Asiaticos 20:135-63. (The Insurgent Black Man). Rio de Janeiro: Nova Farley, Reynolds and Walter Allen. 1989. The Color Fronteira. Line and the Quality of Life in America. New York: Oliveira Filho, Joao Pacheco. 1987. Terras Indigenas Oxford University Press. no Brasil (Indian Lands in Brazil). Sdo Paulo: Cen- Fernandes, Florestan. 1955. Relag6es Raciais entre tro Ecumenico de Documentaqdo e Informaqdo. Brancos e Negros em Sdo Paulo (Black-White Race Oliveira, Lucia Elena Garcia de, Rosa Maria Porcaro Relations in Sao Paulo). Sao Paulo: Editora An- and Tereza Cristina N. Ara jo. 1985. 0 Lugar do hembi Limitada. Negro na Forga de Trabalho (The Position of .1969. The Negro in Brazilian Society. New Blacks in the Labor Market). Rio de Janeiro: Insti- York: Columbia University Press. tuto Brasileiro de Estatistica e Geografia. Frazier, E. Franklin. 1942. "Some Aspects of Race Pierson, Donald. 1942. Negroes in Brazil: A Study of Relations in Brazil." Phylon 3:287-295. Race Contact atBahia. Chicago: University of Chi- Freyre, Gilberto. 1933. Casa Grande e Senzala (The cago Press. Masters and the Slaves). Rio de Janeiro: Jose Pinto, L. A. de Costa. 1953. 0 Negro no Rio de Janei- Olimpio. ro (Blacks in Rio de Janeiro). Sao Paulo: Compan- Gibbs, Jack P. and Walter T. Martin. 1962. "Urban- hia Editora Nacional. ization, Technology and the Division of Labor." Rodriguez, Clara E. and Hector Cordero-Guzman. American Sociological Review 26:667-77. Forthcoming. "Placing Race in Context." Ethnic Harris, Marvin. 1964. Patterns of Race in the Ameri- and Racial Studies. cas. New York: Walker and Company. Rolnik, Raquel. 1989. "Territorios Negros nas Ci- Hasenbalg, Carlos A. 1985. "Race and Socioeconom- dades Brasileiras: Etnicidade e Cidade em Sdo Paulo ic Inequalities in Brazil." Chap. 3 in Race, Class e no Rio de Janeiro." (Black Territories in Brazil- and Power in Brazil, edited by P. M. Fontaine. Los ian Cities: Ethnicity and City in Sdo Paulo and Rio Angeles: UCLA Center for Afro-American Studies. de Janeiro). Estudos Afro-Asiaticos 17:29-41.

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms RESIDENTIAL SEGREGATION BY SKIN COLOR IN BRAZIL 197

Schermerhorn, R. A. 1978. Comparative Ethnic Re- Groups." Latin American Research Review. lations: A Framework for Theory and Research. Turner, J. Michael. 1985. "Brown into Black." Chap. Chicago: University of Chicago Press. 6 in Race, Class and Power in Brazil, edited by P. Schwartz, Stuart. 1974. "The Manumission of Slaves M. Fontaine. Los Angeles: UCLA Center for Afro- in ." Hispanic American Historical American Studies. Review 54:603-645. Vainer, Carlos. 1990. "Estado e Raqa no Brasil: No- Silva, Nelson do Valle. 1985. "Updating the Cost of tas Exploratorias" (State and Race in Brazil: Pre- Not Being White in Brazil." Chap. 4 in Race, Class liminary Notes). Estudos Afro-Asiaticos 18:103- and Power in Brazil, edited by P. M. Fontaine. Los 118. Angeles: UCLA Center for Afro-American Studies. Van den Berghe, Pierre. 1967. Race and Racism. New _ . 1987. "Distancia Social e Casamento In- York: Wiley. ter-Racial no Brasil" (Social Distance and Interra- Vetter, David. 1988. Indicadores Sociais: Regi6es cial Marriage in Brazil). Estudos Afro-Asiaticos Metropolitanas, Aglomeraqv5es Urbanas e Muni- 14:54-84. cipios com mais de 100 Mil Habitantes (Social Skidmore, Thomas E. 1972. "Toward a Comparative Indicators: Metropolitan Areas, Urban Agglomer- Analysis of Race Relations Since Abolition in Bra- ations and Municipalities With Populations Over zil and the U.S." Journal of Latin American Stud- 100,000). Rio de Janeiro: Instituto Brasileiro de ies 4:1-28. Estatistica e Geografia. _ . 1974. Black Into White: Race and Nation- Willems, Emilio. 1949. "Racial Attitudes in Brazil." ality in Brazilian Thought. New York: Oxford Uni- American Journal of Sociology 54:402-408. versity Press. Wood, Charles. 1991. "Categorias Censitarias e Clas- Telles, Edward E. Forthcoming a. "Who Gets Formal sificac~es Subjetivas de Raqa no Brasil" (Census Sector Jobs: Determinants of Formal-Informal Sec- Categories and Subjective Classifications of Race tor Participation in Brazilian Metropolitan Areas." in Brazil). Pp. 93-113 in Desigualdade Racial no Work and Occupations. Brasil Contemporaneo, edited by P. A. Lovell. Belo . Forthcoming b. "Racial Distance and Re- Horizonte: Universidade Federal de Minas Gerais/ gion in Brazil: The Case of Marriage Among Color Centro de Demografia e Planejamento Regional.

This content downloaded from 68.8.165.204 on Fri, 19 Feb 2021 07:02:54 UTC All use subject to https://about.jstor.org/terms