Introduction
Domestic and international forces have combined to shape the ways states classify their populations along ethnic and racial lines over the last century. Following World War II, many countries dropped racial classification from censuses and official statistics. This trend was later reversed, and a majority of national censuses worldwide now classify their populations by race or ethnicity.Footnote 1 This pattern has been particularly striking in Latin America: while in 1980 only three countries in the region (Brazil, Cuba and Guatemala) had a direct question about racial or ethnic identification on the national census, in 2010 only one (Dominican Republic) did not include such a question.Footnote 2
Brazil presents a critical case for understanding the complex politics of racial categorisation. We explore the state's approach to racial classification from the exclusion (1970) and reintroduction (1980) of the colour category, to the most recent decennial census (2010). Brazil has historically presented itself internationally as a racial democracy, where race relations are defined by blurred and transmutable intergroup boundaries. Yet it stands with the United States and South Africa as one of only a few countries that continued to classify its population along racial lines throughout the twentieth century.Footnote 3 The parallel here is puzzling, as the United States and South Africa employed racial statistics to monitor and enforce official systems of racial hierarchy. Other Latin American countries, espousing visions of racial order more similar to Brazil's, generally excluded ethnoracial classification from national censuses during this period.Footnote 4
Also puzzling is the continuity in the Brazilian census's colour categories. In contrast to the United States, where census racial categories changed substantially over time,Footnote 5 Brazil's categories have remained essentially the same since their initial adoption in 1872. Domestically, Brazil experienced many of the same sweeping demographic shifts as those that helped drive change on the US census – the abolition of slavery, major waves of immigration from southern and eastern Europe as well as Asia, and internal migration from agricultural to industrial areas. Narratives about race in Brazil have also radically changed: from the emergence and crisis of eugenics as a science, the rise and fall of racial democracy narratives and the recent strengthening of black and indigenous activism and implementation of race-based affirmative action.
Brazil's census classification scheme has remained stable despite divergent approaches to racial categorisation in other domains of government, and society at large. The classification system used on the census since 1872 employs a graded colour spectrum (branco/pardo/preto, or white/brown/black)Footnote 6 whose categories are generally based on phenotype.Footnote 7 A different system, codified in the 2010 Racial Equality Statute, employs a binary (white/black; branco/negro) approach, combining pardos and pretos within the negro category. Neither approach maps neatly onto spontaneous self-identification patterns of the Brazilian population, which many scholars assert is ambiguous or situation-specific.Footnote 8
How has this continuity in racial categories on the census been possible in Brazil despite striking transformations in racial narratives and the racial order? Existing scholarship has focused on contestation between the state and civil society,Footnote 9 alongside other international factors.Footnote 10 This literature sheds important light on how Brazilian classification policy emerged and evolved. Yet it does not fully explain the trajectory that official policies of racial classification in Brazil have taken in recent decades. Complementing this literature, the focus of this article lies in the policy-making environment of the national census, and the technocratic actors it privileges. We offer a textured picture of policy-making by technocrats within the Brazilian census bureau (the Instituto Brasileiro de Geografia e Estatística – Brazilian Institute of Geography and Statistics, IBGE), and the context in which those decisions are made.
The core contention of this paper is that in modern states census politics are often bureaucratic politics.Footnote 11 Technocrats, trained as statisticians and social scientists, have direct access to the census design and implementation process. They play a key role in proposing and defending changes or continuities in the official classification system, even producing the data on which classification policy debates are based. Moreover, the interests and motivations of such actors are often distinct from the political leadership of the state, as well as from interested parties outside of government such as academics and social movements.Footnote 12
Technocrats are positioned to directly influence census policy, but they often do so in a context of technical and political uncertainty. We show that throughout the period of our study, Brazilian census officials enjoyed an important degree of autonomy to design, interpret and experiment with racial classification policies. But the appropriate policy path to pursue was almost always unclear or contested. Census officials debated amongst themselves, as well as with other branches of government, civil society, academia and the public at large over the form that official racial categories should take. We argue that Brazilian classification policy outcomes reflect what we call the ‘technocrats’ compromise’: policy choices that emerge as technocratic actors grapple with various professional, political and ideological motivations in a context of uncertainty.
We organise the remainder of the article as follows: the next section reviews the existing literature on the politics of ethnoracial classification and develops the concept of the technocrats’ compromise. We then present the literature on the Brazilian census. The case study follows, using archival material, survey and census data, as well as key informant interviews with senior IBGE technocrats responsible for racial classification policy design, implementation and analysis in recent decades.Footnote 13 We conclude by summarising our main findings and underscoring the importance of better understanding the key role of decision-making processes behind the choice and interpretation of social policy categories.
Explaining the Politics of Ethnoracial Classification
To explain the politics of ethnoracial classification, social scientists emphasise the authority and agency of powerful actors such as the state, civil society, or the international community. Classic studies begin with the state, and understand censuses as efforts by the state to bring order and intelligibility to disordered realities.Footnote 14 In this view, racial and ethnic classification systems reflect state efforts to understand, manage and even control a complex social terrain.Footnote 15 In turn, classification systems and the statistics censuses generate become the focus of ethnic and racial group mobilisation, as they create the bases for social and political inclusion or exclusion.
According to rational choice perspectives, ethnic and racial groups are instrumentalist in how they seek to define group boundaries.Footnote 16 In this view, groups are likely to adopt the strategy of the ‘minimum winning coalition’, pursuing the intergroup boundary which establishes a constituency large enough to secure available benefits, but not so large that payoffs must be distributed too broadly.Footnote 17
Other scholars emphasise the variability in states’ motives for classifying their populations along ethnoracial lines, and argue that the classification policies ethnoracial groups pursue vary accordingly as well.Footnote 18 When states enumerate groups for the purposes of monitoring and control, groups are likely to lobby against explicit classification or differentiation of their group to avoid repression. When states enumerate groups to monitor inequality or to facilitate the distribution of particular rights or recognition, groups will seek to be uniquely identified to access benefits.Footnote 19
Scholars also point to the role of international actors in shaping racial and ethnic classification policies, especially in Latin America.Footnote 20 This literature demonstrates how cross-national movements as well as international organisations such as the UN have been instrumental in incentivising states to classify their populations by race or ethnicity on national censuses.
The interplay of state and non-state actors, as well as the influence of international actors and institutions, are important for understanding the context in which official racial categories emerge, and how they change over time. However, as G. Cristina Mora notes, much of the literature on racial classification ‘stresses the workings of the state, or the agendas of ethnic political and market leaders, without paying much attention to how categories are negotiated and developed’.Footnote 21 We therefore propose to take a closer look at how classification policy is made and debated within the state institutions most directly responsible.
Our empirical focus is on the role of technocrats. We define technocrats as government actors with technical training or expertise that provides the rationale for their unique authority to make and implement policy. This authority rests on access to ‘scientific forms of knowledge’ which is presumed to be rational and thus free from bias or other ideological commitments.Footnote 22 Yet even early studies demonstrated that technocrats are never completely impervious to political pressure, normative commitments, or status and reputational concerns that may be at odds with a strictly technical approach to their professional duties.Footnote 23 In the United States, Jennifer Hochschild and Brenna Powell found that a range of motivations drove the policy choices of census technocrats (statisticians, demographers, anthropologists) responsible for producing demographic statistics: a desire for political control vis-à-vis other branches of government, a desire for scientific credibility and ideological convictions about the nature of racial and ethnic hierarchy. These goals in some cases conflicted, and in other cases reinforced one another.Footnote 24
Technical expertise alone fails to provide precise answers to many important policy questions, and even specialists may disagree about which policy choice is optimal. Thus in making decisions about how to design and implement policy, technocrats often grapple with a good deal of uncertainty regarding the appropriate path to pursue. In order to resolve uncertainty, technocrats rely on what Peter Hall has called ‘instrumental beliefs’, or ‘means–ends schema’, that help actors interpret how particular choices will ‘affect the likelihood of achieving various types of goals’.Footnote 25 Such uncertainty provides the context for what we term ‘technocratic compromises’.
We define technocratic compromise as policy choices that emerge as technocratic actors contend with various professional, political and ideological motivations in a context of policy uncertainty. The notion of compromise should not indicate reaching middle ground between opposing goals, but rather navigating a range of objectives that may in some cases work in harmony, and in other cases at cross purposes. Professional motivations include the desire for professional legitimacy, as well as the autonomy (and resources) to make technologically sound decisions. Political motivations range from commitments to one political vision or another to the pragmatic need to ensure that policy choices comply with the agendas of powerful political actors. Ideological motivations include normative commitments to particular views of social justice or social order.
The notion of technocratic compromise here is consistent with other scholarly accounts of Brazilian policy-making. As demonstrated by the literature on technocrats’ insulation in Latin America, even during the most repressive era of military dictatorship technocrats in Brazil enjoyed a significant degree of independence.Footnote 26 In addition, scholars have shown that ideologically motivated technocrats gained access to government and shaped policy in fields as diverse as health care, domestic high-tech manufacturing and economic policy, even in contexts where these technocrats’ agendas were squarely at odds with the political regime in power. However, there is variation in the degree to which such technocrats were motivated by coherent ideological commitments. Emanuel Adler dubs high-tech policy-makers ‘ideological guerrillas’, who function as a ‘subversive elite’.Footnote 27 Tulia Falleti similarly describes a group of health-policy technocrats who understood themselves as a movement with shared ideas and commitments, and actively tried to persuade the military regime to adopt their preferred policies.Footnote 28
In contrast, the technocratic actors who shaped racial classification policy on the Brazilian census during the period of our study (1970–2010) did not represent a movement, nor share a singular ideological agenda. However, the absence of a self-defined ‘guerrilla’ movement has not rendered technocrats unable to influence census policy. In fact, the institutional structure at IBGE has meant that a small number of officials have had a relatively high degree of influence over census classification policy.Footnote 29 Furthermore, the regularity of the decennial census and other monthly household surveys, together with persistent dissatisfaction with the existing classification scheme, have provided these officials with repeated opportunities to revisit and reinterpret classification policy.
Throughout the years our study covers, uncertainty about the validity of the racial categories used on the Brazilian census has been high, due to the discrepancies between how Brazilians identify their race or colour in everyday practice, and the constrained manner in which they are required to do so on the census. At the same time, confidence has been low in IBGE's ability to identify census categories which would simultaneously satisfy the technical objectives of, on the one hand, validity and, on the other, ease of statistical analysis. To this day, the appropriate classification scheme to be employed remains unclear and contested.
Ethnoracial Classification in Brazil
It is tempting to attribute the relative stability in Brazil's census categories to path dependency or bureaucratic preference for the status quo. Changing census categories is costly because it destroys the comparability of data over time, one of Stanley Lieberson's ‘devilish principles’ of census-taking.Footnote 30 However, a path dependency explanation is clearly insufficient. Over the last four decades, IBGE officials have experimented repeatedly with different modes of classification. As early as 1976, IBGE conducted surveys with open-ended questions to identify the racial categories used spontaneously in self-identification; the Institute conducted similar studies in 1998 and 2008.Footnote 31 At the same time, political movements have called for the reinstatement of racial classification on the census following its removal, and leading voices across the political spectrum have contested the legitimacy of the existing system.
Attention to the strategic incentives of various actors provides some leverage in explaining the trajectory of racial and ethnic classification in Brazil, as the existing literature would suggest. In the late nineteenth and early twentieth centuries, elite interests in ‘whitening’ the population provided the logic for the graded colour scheme on the census – even if through an unorthodox understanding of eugenics.Footnote 32 In the introduction to the 1940 census, social scientist Fernando de Azevedo explained that the census categories would help to monitor progress towards the government's goal of achieving an increasingly white population. Melissa Nobles argues that the census categories have remained stable over time due to a racially exclusive political elite, which was challenged only by an ineffective or weak black political movement. In her account, even latter-day officials are loath to wrestle with the social significance of the existing classification system.Footnote 33
However, Nobles’ narrative of conservative elite consensus stands in contrast to the debate and experimentation that has taken place within IBGE since the 1970s. Moreover, the notion that continuity in the colour categories reflects hostility or indifference to addressing racial inequality presumes that there has been agreement as to how greater racial equality might be achieved. In general, scholars have underplayed the significance of debates over classification policy within government institutions, and overlooked developments within IBGE.Footnote 34 Our research indicates that, to the contrary, those officials responsible for designing the Brazilian census and producing the nation's racial statistics have remained dissatisfied with the official classification scheme. At the same time, there has been technical and normative uncertainty about the superior policy path to pursue. As we will see, the agency and interests of such actors are central to understanding the trajectory of Brazilian classification policy. José Luis Petruccelli, a leading analyst for racial statistics at IBGE, sums up Brazil's history of racial classification in the following way:
There is the good news and the bad news. The good news is that we've got more than a hundred years with racial classification [on the census]. The bad news is we've got more than a hundred years of the same racial classification.Footnote 35
Case Study: Brazil 1970–2010
Debate under Dictatorship: Understanding the Exclusion of the Race Question in 1970
The military coup of 1964 ushered in an era of violent repression, suppressing discussion of race and inequality along with many other social and political causes. The dictatorship initiated a programme of economic stabilisation and bureaucratic professionalisation, which ultimately brought unintended and long-term consequences for racial politics.
In this period the only significant policy instrument the government used to classify the population by race or colour was the inquiry on the decennial census. The 1960 census had asked Brazilians to self-identifyFootnote 36 according to the traditional census categories preto, pardo, branco and amarelo (yellow, intended for Asian Brazilians). Expert advisors preparing for the 1970 census raised questions about whether or not the colour inquiry should be included and what the categories actually measured, given the nation's history of racial intermixture and regional variation in beliefs about who belonged in which category.
According to government records, the terms of the debate were social-scientific and centred on how race should be defined.Footnote 37 If race was understood in anthropological terms – caucasoid, mongoloid, negroid – then the classification of the Brazilian population into racial categories was unlikely to generate racial data of any value. Intermixture had rendered the lines between these ‘pure types’ inscrutable to anyone without an anthropologist's technical expertise. A sociological perspective defined race through social interactions rather than biology. By this logic, including the colour question on the census could provide useful data as well as providing the opportunity to analyse patterns of racial classification over time. In the end, the expert advisors could not agree on a valid mode of classification and voted to remove the colour inquiry. They did note that excluding the colour inquiry might provoke the protest of some sociologists, but concluded this was perhaps worth the risk.Footnote 38
Scholars have commonly attributed the exclusion of the colour question in 1970 to the broader racial ideology of the military regime, which was closely aligned to Gilberto Freyre's ideas of racial democracy.Footnote 39 These ideological alignments make it tempting to dismiss the official justification for data quality as a cover-up for the regime's desire to present Brazil as a racially homogeneous and united nation. Jane Souto de Oliveira, a top official at IBGE in the 1970s, acknowledged that while the dictatorship bolstered IBGE's funding and mandate, the repressive climate had a chilling effect at the Institute:
the instruments of control and censure also managed to reach the information produced and divulged in the country. It was a time in which it was not possible to speak of ‘cats’ and when they were spoken about, a metaphorical language was resorted to in which they became instead ‘felines with four paws’.Footnote 40
Nevertheless, there is no conclusive evidence that the 1970 exclusion followed directly from regime intervention. In fact, in abandoning the colour question, Brazil joined other Latin American countries which dropped ethnoracial classification from their censuses after the 1950s. Stressing the impact of transnational processes, as well as intellectual criticism of racial determinism and the new cultural conceptualisations of race, Mara Loveman argues that ‘(t)echnical, scientific, political and ideological reasons […] worked in combination to lead to the omission of race queries from nearly all Latin American census questionnaires in the latter part of the twentieth century’.Footnote 41
If regime pressure on the colour question did occur, there is no official record of it, which is not surprising. More surprising is that no accounts provided by IBGE officials support the idea. Tereza Cristina Costa, who led the Institute's social indicators work in the 1970s, attributed the decision to debates over the concept's scientific validity.Footnote 42 Other long-standing IBGE analysts interviewed for this study and familiar with the period maintained that political intervention was not responsible for the exclusion.Footnote 43
Even if some scepticism is warranted regarding the scientific justifications and the actual leverage of IBGE officials vis-à-vis the military regime, the importance of census officials’ concerns about scientific validity at that time should not be underplayed. These concerns are critical for understanding the experimentation that ensued in the 1970s, as well as the return of the colour question to the census in 1980.
Bringing Race back in: Changes from within IBGE
The return of the race question in 1980 is commonly understood as a consequence of the democratic transition and the strong pressure of activist groups. In explaining Brazilian exceptionalism, Mara Loveman argues (largely based on Melissa Nobles) that ‘absent the organized pressure from academics and activists, Brazil's 1980 census would have been fielded without a color question’.Footnote 44
Although public pressure did emerge, the groundwork for the return of the colour question was laid during the previous decade, during the most repressive years of the Brazilian dictatorship when little room for public advocacy or pressure on matters of racial representation existed.Footnote 45 During this period, IBGE officials were uniquely positioned to shape census policy.
The military regime took seriously the importance of bureaucratic insulation, or the idea that key posts of the administration should be occupied based on technical expertise (even if still depending on personal and informal connections with the government).Footnote 46 IBGE was placed under the Planning Ministry and became responsible not only for mapping and collecting demographic information but also for creating an input–output matrix to guide state investments and economic planning. Of course, the Brazilian economy was never truly directed by the Planning Ministry, nor by IBGE data.Footnote 47 Nevertheless, the image of scientific planning gave a new outlook to the Institute and provided some technical autonomy and legitimacy.
The dictatorship's top priority was economic development, and it became clear that development could not progress without addressing social inequalities. Embarrassing economic data from the 1970 census revealed that poverty and inequality had actually worsened after the military take-over, contradicting official doctrine that economic growth would answer all the nation's ills. The government was forced to rethink its approach to poverty, and ‘in the official discourse, the poor came to be defined as privileged subjects for public action’, whereas prior political agendas had rendered them invisible.Footnote 48
Taking action required good data. The government appointed distinguished economist Isaac Kerstenetzky as President of IBGE in 1970. Kerstenetzky oversaw fundamental reform at IBGE, modernising its methodology and prioritising rigorous, multidisciplinary research with special attention to problems of poverty and inequality.Footnote 49 Our interviewees and archival research point to Kerstenetzky's leadership as critical for providing the space for sensitive topics such as race to be discussed.Footnote 50
In this period, IBGE embraced a new logic for data collection grounded in the concept of ‘social indicators’.Footnote 51 Social indicators stood in contrast to standard economic indicators, and promised to shed light on social conditions (notably social conflict and inequality) increasingly regarded as central to economic development. Anthropologist Tereza Cristina Costa came to IBGE in the early 1970s and created the Social Indicators Project.Footnote 52 She described social indicators as ‘normative’ concepts that would structure ‘social objectives and priorities’.Footnote 53 Data generated on social indicators would make it possible for the government to take action over important social conditions.Footnote 54
Developing socially relevant concepts for measurement formed part of a broader and at times radical process of social engagement at IBGE. Moema Poli Teixeira came to the Institute in 1979 as an anthropologist. She recalls that the head of the department charged with producing statistics about favelas (informal housing settlements) sought a total overhaul of the data collected – so much so that he took the radical step of living in a favela in order to understand the important dimensions of favela life.
The embrace of social indicators provided a rationale for collecting official statistics on race and colour that was nevertheless consistent with the broad objectives of the dictatorship. The logic of social indicators meant that the imprecision or unwieldiness of race or colour could no longer justify its exclusion from government statistics. Costa explained:
the concept of Social Indicators should not be tied to a quantitative perspective that limits sociological research to only that which is measureable, but to what is relevant … many times we will have to opt between a sophisticated measure of a less fruitful concept, and a crude measure of a relevant concept.Footnote 55
Consistent with the transnational transformations discussed by Mara Loveman, the idea of race as a social indicator followed the new science of race, which held that race and racial differentiation were social constructions and rejected the biological foundations of racial groups. Racial classification was not about making objective observations about the natural world, but participating in a fundamentally social process.Footnote 56
Measuring race as a social indicator required identifying socially meaningful categories, or the words that people used to describe themselves in everyday life. Costa proposed a study that would identify ‘racial vocabulary’, and thus ‘make more intelligible the ambiguity that seems to characterise ethnic identification in Brazilian society’.Footnote 57 Accordingly, IBGE administered the 1976 Pesquisa Nacional por Amostra de Domicílios (National Household Survey, PNAD) with a special supplement that included two questions about colour. One was open-ended, asking respondents to identify their colour spontaneously. The other asked respondents to select their colour from the categories used on the census (branco, pardo, preto, amarelo). Table 1 below lists the results of the two questions.
Table 1. Results from the 1976 PNAD: a Open-ended and Pre-codified Colour Questions
Categories in bold are the official categories of the Brazilian census.
a Pesquisa Nacional por Amostra de Domicílios (National Household Survey).
b The adjectives in the column below are in the feminine form as they modify the Portuguese noun cor, ‘colour’.
c Petruccelli (see source) included the figures for those selecting amarela in the ‘others’ category.
Source: Adapted from Petruccelli, A cor denominada, pp. 27–8.
The PNAD allowed researchers to investigate two new sets of questions. First, it allowed analysts to compare the economic well-being of Brazilians under the dictatorship by colour, showing the persistence of racial inequalities. It had not been possible to do so previously because racial statistics had not been collected since 1960. Secondly, it allowed researchers to learn how the census colour categories related to the categories people used spontaneously. Respondents provided a huge variety of responses to the open-ended question (136 in total), suggesting no consensus on the meaningful categories in Brazilian society. At the same time, a majority of respondents spontaneously used one of the census categories. Variants of moreno also proved popular. The vast majority of the other categories were used by only one person each.
The fact that explicit research on racial classification was planned and conceptualised during the most repressive period of the military dictatorship illustrates the degree of bureaucratic insulation IBGE enjoyed. Nevertheless, the results of the 1976 PNAD were controversial, and IBGE did not publish them for nearly a decade. Our interviewees reported that despite the controversy, officials quietly shared the results with researchers and black movement activists. ‘When I came [to IBGE]’, Moema Teixeira recalls, ‘the [1976] PNAD was being studied together with a lot of dialogue with the black movement – with researchers and intellectuals of the black movement’.Footnote 58
Officially, draft versions of the 1980 census followed the 1970 format, without a colour inquiry. Yet by the late 1970s the political climate had radically changed and the omission prompted an outcry from scholars, civil society and political opposition members. Political opposition leader Marcello Cerqueira introduced a bill to return the colour question to the census in the name of generating data to enact better social policy.Footnote 59 Activists used their connections at IBGE to press for the colour question's inclusion, and met with new IBGE president Jessé Montello to voice their concerns. According to two of our informants, dialogue between IBGE and movement activists was directly related to the return of the race question to the 1980 census. Unsurprisingly for the dictatorship period, we could not find any formal documentation to independently verify this claim. However, such a dialogue seems to be in line with the later adoption by IBGE officials of the term ‘negro’ to report on racial inequalities from the early 1980s, even before it was widespread in academia.
Democratic Opening: The 1980s and the Push for New Racial Categories
After the reintroduction of the colour question in 1980, IBGE expanded the collection, analysis and publication of racial statistics.Footnote 60 New debates followed as well. For years, scholars and IBGE officials like Costa had been concerned that the census categories did not correspond to the categories people commonly used to describe their racial identity. The purpose of census statistics was to reflect reality, but there were a number of realities to choose from – the reality of how people chose to self-identify, the reality of socio-economic stratification along racial lines, or the reality of increasing black political mobilisation. Different categories reflected different realities. Moreno, for example, is a popular colour category that never appeared on the census. It is associated with the ideal of racial democracy: Brazilians as one people united by mixture rather than divided by race. In popular use, it can de-emphasise or obscure racial differentiation. Fully 28 per cent of respondents on the 1976 PNAD used moreno or some modification (e.g. moreno claro, moreno escuro) in the open-ended colour question.Footnote 61
At the same time, PNAD data demonstrated that a binary concept of race, white vs. non-white, did reflect a powerful socio-economic reality. In separate studies, Brazilian sociologists Nelson do Valle e Silva and Carlos Hasenbalg used PNAD data to demonstrate that pardos and pretos together fared worse economically than brancos, and that current inequalities could only be the result of ongoing racial discrimination.Footnote 62 The work was controversial at the time, directly contradicting the prevailing notion, supported by the military regime, that neither race nor racism existed in Brazil.Footnote 63
In 1985 Costa and her co-authors published their analysis of the 1976 PNAD data, and combined pardos and pretos together into a single negro category.Footnote 64 The move was significant as the term negro was associated with black activism, and was not widely used by mainstream academics. In fact, the PNAD data indicated that only a tiny percentage of Brazilians spontaneously self-identified as negro, despite the consciousness-raising efforts by activists. Yet defined as the sum of pretos and pardos, the negro category represented an observable reality of socio-economic inequality, and thus, for IBGE officials, a scientifically valid concept. While moreno was a popular category for self-identification, it was not reflective of the socio-economic realities IBGE officials believed it was important to document. Negro was the opposite.
Moving from Colour to Race: The 1991 Census
In the lead-up to the 1990 census, black activists empowered by the democratic transition lobbied to replace preto and pardo with a single negro category. They also sought to change how individuals classified themselves. Activists launched a campaign urging Brazilians with African ancestry not to identify as white. A poster showing the bare torsos of three people with varying skin tones became the iconic image of the campaign.Footnote 65 It urged ‘Não deixa sua cor passar em branco. Responda com bom (C)senso’. Or, ‘Don't let your colour pass into white. Respond with good sense.’ By overwriting the ‘S’ with a ‘C’, the poster cleverly urged Brazilians to use good sense on the census. Senso and censo have the same sound in Portuguese, the latter meaning ‘census’ and the former ‘sense’.
Ultimately, the census categories remained essentially the same. Amidst political and economic upheaval, President Fernando Collor de Mello did not release the funds to carry out the census until 1991. We could find little institutional memory or documentation about the colour inquiry. However, in 1991 there was one notable addition: the indigenous category was included as an optional response. Because indigenous is not a colour distinction but an ethnic or racial one, the question itself was changed from ‘What is your colour?’ to ‘What is your colour or race?’
There is no official record of the decision to modify the question wording, and none of the current or former IBGE analysts and officials we interviewed could account for the origins of the indigenous category – not even Nilza Pereira, a long-serving IBGE official and later the chief analyst for indigenous statistics.Footnote 66 According to Melissa Nobles, some black movement activists speculated that the category was included at the behest of the World Bank.Footnote 67 IBGE officials we interviewed acknowledged it is possible the Bank made a request for indigenous statistics but, perhaps consistent with a desire to defend IBGE autonomy, rejected the idea that a census category would be included simply because the Bank demanded it.
The important aspect of the indigenous category is that its introduction was a non-event. In contrast to the controversy surrounding the enumeration of blacks, the enumeration of indigenous peoples does not appear to have been politically or technically problematic. ‘Indigenous’ was an administrative category with a long history in Brazil, and the concept was also enshrined in the new democratic constitution.Footnote 68 The relatively tiny numbers of indigenous Brazilians (less than 1 per cent of the total population) also lowered the political stakes.
In her cross-national comparison of Latin American censuses, Mara Loveman shows that indigenous classification returned to a number of countries at the same time, suggesting the role of transnational forces, or what Tianna Paschel calls a ‘multicultural alignment’ in the region.Footnote 69 In Brazil, enumerating indigenous peoples was not only in line with census policy recommendations of the UN and other bodies, but further added to the precision of the pardo category (wherein most people of indigenous descent were classified prior to the category's inclusion). It did not become clear until much later that indigeneity was as much a fluid social construction in Brazil as blackness, triggering new studies to explain the emergence of urban indigenous identification, as well as new ethnic classificatory schema to differentiate the urban indigenous population from ethnic-native-language-speaking indigenous peoples.Footnote 70
Expanding the Racial Classification Regime: The 1990s to 2000
Fernando Henrique Cardoso, a sociologist who had written on Brazilian race relations, was the first Brazilian president to publicly acknowledge that racial inequality was a problem in Brazil. He talked openly about the problem of discrimination, demanded equality for negros, and spoke of the cherished racial democracy as a myth. At the time simply using the term negro was significant. Elisa Reis, a prominent Brazilian sociologist who sat on the IBGE Census Advisory CommitteeFootnote 71 in preparations for the 2000 census, recalls Cardoso's leadership as critical in changing the national conversation about race. Up to this point, race
was even a forbidden issue … everybody felt uncomfortable mentioning it. And once [Cardoso] made it public, it became no longer incorrect … as a former academic, he was aware of the importance of the issue and he had the legitimacy to bring it to the fore without being perceived as incorrect.Footnote 72
Some scholars contend that political pressure from interest groups alone was not sufficient to explain Cardoso's initiative on issues of racial justice. Rather, under Cardoso's leadership ‘policymakers became convinced that combating inequalities was the right thing to do’.Footnote 73 Others have been more critical, contending that Cardoso failed to move beyond symbolic politics and never implemented concrete policies targeting racial inequalities.Footnote 74
Regardless, concurrent changes also took place within IBGE. Referred by Cardoso, the sociologist Simon Schwartzman served as President of IBGE between May 1994 and December 1998.Footnote 75 IBGE official Moema Teixeira compares Schwartzman's leadership to that of Isaac Kerstenetzky in the 1970s,Footnote 76 guiding the Institute through a period of technical reorganisation and professionalisation. Schwartzman was responsible for making IBGE datasets, including census and household surveys, publicly available – a significant move as previously data was only available for purchase, usually at a high price. Also like his predecessor, Schwartzman established a space for rigorous inquiry about race and ethnicity.
The debate over classification policy accelerated with preparations for the 2000 census. President Cardoso himself weighed in, suggesting that the preto and pardo categories should be replaced with a single negro category.Footnote 77 Schwartzman remained sceptical, and decided that more research was needed. It had been 20 years since Costa's PNAD study, and the data were no longer regarded as valid.Footnote 78 Schwartzman was also cognisant that changing the census categories could have significant political ramifications, particularly in the context of the new affirmative action programmes being proposed. ‘In a survey you can experiment with different ways of doing it, you have a better understanding’, he argued; ‘if you create that on the census, you are creating entitlements, and this is very different’.Footnote 79 Surveys were a lower-profile means to gather information, whereas censuses produced the nation's reality.
Schwartzman included a set of questions about race and national origins on the July 1998 Pesquisa Mensal de Emprego (Monthly Employment Survey, PME).Footnote 80 Like the 1976 PNAD, the PME asked respondents one open-ended question about their race or colour, and another pre-codified question that provided the census categories as options. In addition, the PME included an ambiguously-worded question about national origins: ‘Qual a origem que o(a) senhor(a) considera ter?’, or ‘What do you consider your origins to be?’ The question was asked in two ways, one open-ended, and the other with a list of nationality/regional options provided. The goal was to provide more data about the language Brazilians used to identify their heritage, and whether there was a better methodology than the existing racial classification system for capturing social identity.
PME results demonstrated that patterns of racial classification had not changed dramatically in the last 20 years.Footnote 81 As in 1976, the open-ended question generated a huge variety of responses (200 total), but most respondents chose one of the census categories, or some modification of moreno. The results also indicated that the negro category had gained some popularity but remained less popular than preto. In 1976 fewer than 1 per cent of respondents had self-classified as negro; in 1998 just over 3 per cent did. Table 2 presents the results from the 1998 PME.
Table 2. Results from the 1998 PME: a Open-ended and Pre-codified Colour Question
Categories in bold are the official categories of the Brazilian census.
a Pesquisa Mensal de Emprego (Monthly Employment Survey).
b The adjectives in the column below are in the feminine form as they modify the Portuguese noun cor, ‘colour’.
Source: Adapted from Schwartzman, ‘Fora de foco’.
The preto, pardo, and indígena categories demonstrated the most inconsistencies. Two-thirds of the individuals who identified themselves as pardo in the pre-codified question chose some other category in the open-ended question (usually moreno or some modification of it). Over half of the people who chose preto in the pre-codified question chose something else in the open-ended question (most often negro, though moreno was also popular). Only a small fraction (13 per cent) of those who chose indígena in the pre-codified question did so in the open-ended question. Most chose moreno instead. Clearly, for a large number of Brazilians, the census classification system did not provide a satisfactory language with which to identify themselves.
The results for the origins question indicated that Brazilians understood the concept in nationalistic terms: in both formulations of the question, the majority simply responded ‘Brazilian’.Footnote 82 The Census Advisory Committee debated the appropriate design for the 2000 census in light of the results, ultimately voting by a majority to keep the colour question in its current form and not to include any inquiry into origins.Footnote 83
The changing national conversation around race meant IBGE was no longer the only government body struggling to determine what the meaningful race and colour categories were, and who belonged in each. In 1998, black activist and Partido dos Trabalhadores (Workers’ Party, PT) Senator Benedita da Silva introduced national legislation that would require every Brazilian to register for an identification card with their race or colour on it. According to her, this ID card would provide the basis for affirmative action programmes in health, education and other arenas. Da Silva argued that changing the official racial categories was important because categories based on colour rather than descent were more ‘ideological’ than ‘anthropological’.Footnote 84 Most importantly, the census's graded colour-based system obscured the true size of the black population. Put simply, ‘the non-whites in Brazil are a majority and they become minorities with [the census] classification’, as she argued in an interview with the authors. ‘We fought a lot with IBGE’, da Silva recalls; ‘those debates were terrible for us’.Footnote 85
As in the past, consistency in the census categories belied a great deal of debate, research and public engagement. During the 1990s pressure came from government and civil society sources to change the colour categories. IBGE President Schwartzman was publicly sceptical about the idea of race as a key political category.Footnote 86 Yet his decision to collect new data on racial identification and to open IBGE datasets to researchers and the wider public increased the statistical visibility of racial inequalities, which proved key in the development of racial policies during the next decade.
From Census Classification to Redistributional Policies: The 2000s
The 2001 Durban UN Conference against Racism, Discrimination and Xenophobia marked another turning point in Brazilian racial politics. The election of the PT to the federal government accelerated the policy activity surrounding Durban. The PT's emphasis on redistributional policies and increased social spending, alongside collaboration with the movimento negro, saw expanded debate about racial inequalities.Footnote 87 Most importantly, race-based affirmative action policies officially linked the redistribution of resources to racial group membership. Notably, the relevant groups were conceived in binary terms. The first affirmative action policies, in the form of quotas for entrance into prestigious state universities, were for non-whites only (either indigenous peoples or negros, those self-classifying as preto or pardo).Footnote 88
The emergence of affirmative action policies and of other black movement policy demands was directly linked to the availability of IBGE racial statistics. Tianna Paschel argues that the much earlier implementation in Brazil of race-based affirmative action policies than in other Latin American countries can be largely explained by the presence of national statistics on racial inequality.Footnote 89 While black activists from other Latin American countries arrived in Durban with political demands and denunciations, Brazilian black activists had tables and statistics to prove their claims regarding the strength and persistence of racial inequality.
Ironically, IBGE race data proved as critical for the development of conservative arguments against affirmative action policies as it had for progressive arguments in favour of them. In 2006, Brazilian journalist Ali Kamel published a book entitled ‘We are not Racists: A Reaction to those who Want to Transform Us into a Bicolour Nation’.Footnote 90 Using IBGE data, Kamel challenged the idea that pardos were statistically indistinct from pretos, and rejected the logic of the negro category. If pardos were more likely to identify as moreno than negro in an open-ended question, why should they be combined together with pretos in affirmative action programmes as part of a negro category?
José Luis Petruccelli and Moema Teixeira responded to Kamel in newspaper op-eds, largely rejecting Kamel's conclusions. Petruccelli attempted to replicate Kamel's results and could not. ‘I think in one way or another he cheats [with the data]’, Petruccelli said in an interview with the authors, ‘and he cheats as well with his arguments’.Footnote 91 Petruccelli argued that the evidence for ongoing discrimination lay instead in the numbers, which demonstrated persistent inequality between whites and non-whites.
At the same time that public criticism over IBGE's practice of combining pretos and pardos as negros was ramping up, movimento negro activists were quietly withdrawing their demands to change the census categories. It was becoming clear that constructing a negro category by retroactively combining pretos and pardos would generate a far larger constituency than leaving it up to individuals to self-identify as negro.Footnote 92 According to the IBGE officials we interviewed, black activists now seemed content to maintain the census system and expressed their interest in leaving the existing categories as they were for the 2010 census.Footnote 93
If anything, increasing public debate over racial statistics underscored rather than resolved ongoing uncertainty at IBGE over the census classification scheme. José Luis Petruccelli and Moema Teixeira began work on another PNAD supplement, with an eye to implementing changes on the 2010 census. Petruccelli's concerns remained similar to those of Costa in the 1970s. The census system lacked both precision and accuracy, because it did not reflect the full complexity with which Brazilians described their race, ethnicity and colour. Too often the census system forced people into categories that were not socially meaningful. Moreover, the meaning of racial categories appeared to be changing. As a result, the census categories did not provide the best platform from which to construct effective social policy.
After years of consultation with scholars, activists and government officials, in 2008 IBGE implemented its first survey developed exclusively to address questions of race. The Pesquisa das Características Étnico-Raciais da População (Study on the Ethnoracial Characteristics of the Population, PCERP) included questions about racial identity, as well as perceptions of race and racial discrimination.Footnote 94 Ultimately, it produced results similar to those of the 1976 PNAD and 1998 PME, with the data indicating that Brazilians identify with numerous racial categories, but a majority use the census categories or the moreno category. It also revealed a small but growing identification with the negro category. The PCERP results came too late to inform the 2010 census design, and the classification scheme remained the same as in 2000. However, in an indication of the greater centrality of racial inequalities in the public debate, the 2010 census now included the colour or race question on the basic questionnaire, or short form.
Overall, the 2000s saw tremendous debate around racial classification among officials, scholars and policy-makers alike. IBGE policies and practices directly informed the ways in which political actors and activists developed new policy agendas. Both the raw data itself, and the organisation of the data into preto and pardo categories that could be easily combined into a sizeable negro constituency, shaped the kinds of claims scholars and activists made, as well as the legislation policy-makers could successfully draft and implement. The clearest example is the Racial Equality Statute, ratified by the Brazilian Congress in 2010.Footnote 95 Differently from the early twentieth-century immigration laws, the last official codification of racial boundaries in Brazil to exclude non-white immigrants, the Statute codified racial boundaries to make racially inclusive redistributive policies possible.Footnote 96 Section IV of Article 1 explicitly legislates the definition of the ‘negro population’ as ‘the group of people that self-classify as preto and pardo, according to the colour or race question used by IBGE, or who adopt an analogous self-definition’.Footnote 97 The new law both incorporated and superseded IBGE's racial classification scheme – leaving IBGE's census categories intact, but officially changing their meaning in the public policy domain. The statute's impact on the everyday meaning of these categories, however, is much more uncertain.
Conclusion
This article has sought to explain the Brazilian state's approach to racial classification on the census over the last four decades. We have shown how the apparent stability in Brazil's census categories belies a great deal of debate amongst census officials, and between these actors and other branches of government as well as the public, about which racial categories best capture the complex realities of racial identification in Brazil and how these categories should be interpreted. Crucially, technocrats’ policy choices in this domain have been defined as much by technical uncertainty as by certainty. The trajectory of racial classification policy in Brazil thus reflects what we have called technocrats’ compromises: policy choices that result as technocrats navigate various professional, political and ideological motivations in a context of uncertainty.
In consistently classifying its population along racial lines, Brazil has appeared less like other Latin American countries and more like the United States or South Africa, which developed official classification regimes to manage systems of racial hierarchy. Brazil has also diverged from other countries with long histories of census classification in the degree of discrepancy between popular approaches to self-identification, and the racial categories used by the state. The fact that patterns of self-identification have not followed official categories could suggest that Brazil's racial classification policies have simply not been consequential for everyday life. Nevertheless, recent studies comparing Brazil to other Latin American countries that do not have racial classification on the census show how Brazil's history of racial classification has enabled ‘the statistical visibility of racial inequality in Brazil’, which scholars argue has been key to informing recent changes in the racial order such as the implementation of affirmative action.Footnote 98 Similarly, Tianna Paschel shows how having census statistics for race was essential for the mobilisation of Brazilian black movements both domestically and abroad.Footnote 99
To explain Brazil's approach to racial classification, we have shown the central importance of technocratic actors working within the statistical bureaucracy in generating classification policies over the last four decades. Beginning in the 1970s, technocrats within IBGE pursued very different approaches to racial classification and statistics from those espoused by the military regime in power. Our study suggests that Brazilian census officials have enjoyed notable bureaucratic insulation – or what has been defined as the ‘opportunity for officials to pursue preferences and formulate policies independently’.Footnote 100 Ben Ross Schneider, analysing the insulation of bureaucracy from business demands, argues that when compared to more developed countries, insulation in Brazil (and Mexico) is contingent and political: ‘their insulation is fundamentally contingent on how presidents wield their appointment power’.Footnote 101 Clearly the capacity of technocrats at IBGE to establish census policy has not been unilateral: they have acted within domestic and international political contexts that limited the set of choices available, and privileged some options over others.Footnote 102 However, IBGE officials’ choices during the period of our study often stood in tension with the broader ideology of the federal governments in place.
The most significant challenge for Brazilian census officials has been a considerable degree of uncertainty about which racial categories represented the most technically valid approach to measurement. As we have shown, following the move towards social indicators in the 1970s, the question about whether to classify the population by race or by colour has largely been settled among the IBGE officials responsible for producing racial statistics. However, which categories to use has remained a constant source of debate and experimentation. In addition to internal questioning, external actors across the political spectrum have challenged IBGE's approach to classification on normative as well as technical grounds. IBGE officials have waded into very public debates to defend the census categories, as well as their statistical analysis and interpretation of those categories.
Our findings show that technical expertise alone has not resolved uncertainty about which racial categories should be used, and how racial data should be analysed. Other factors have also proven important as IBGE technocrats grappled with the best policies to pursue – from the desire for professional integrity and bureaucratic independence for IBGE, to normative commitments regarding particular visions of social equality in Brazilian society. The concept of technocratic compromise captures the ways in which technocrats navigate policy decisions when the appropriate or technically sound path to pursue is unclear. The clearest examples of technocrats’ compromise include IBGE's early choice to apply the negro category to the non-white population as a whole (excluding Asians, in the amarelo category), and the repeated rejection of the moreno category. Tereza Cristina Costa and her colleagues chose to use the negro category to describe pretos and pardos together, well before it became common practice in academia. They did so to capture a particular reality, but this approach to tabulating and reporting racial statistics was not the obvious choice. Over a decade later, even scholars who embraced the term conceded that it ‘may be more a classification attributed to nonwhites by movement actors than a real social group embraced by the general nonwhite population’.Footnote 103 Despite increasing educational levels, and the association between education and preference for the term negro,Footnote 104 only a small fraction of Brazilians choose to identify as negro, as shown in Tables 1 and 2. Yet the binary approach to race embraced by Costa and others at IBGE renders visible a specific racial reality – one in which inequality between whites and non-whites is a persistent fact of life.
In the case of moreno, the sizeable proportion of Brazilians who routinely identify as moreno in open-ended questions would seem to argue for its inclusion on the census.Footnote 105 Yet in our interviews and in the archival documents we analysed, technocrats within IBGE never took moreno seriously as potential census category. The rejection of racial democracy narratives appears to have played a key role. For example, Moema Teixeira asserts the purpose of census categories should be to ‘reveal the mechanisms of social and economic differentiation between groups’. She argues: ‘You could be moreno sometimes like, for example, at the time that IBGE comes to ask, perhaps at the time when you're in a more cultural space … but at other moments other things appear.’ Those other important moments are times like applying for a job, or ‘the moment of choosing who is going to marry my daughter’: in short, those moments which organise inequality and stratification.Footnote 106 Despite consistent evidence of the popular appeal of the moreno category, including it on the census would not render visible an important social reality.
The debate about Brazil's racial categories is ongoing. Recent studies have shown, for example, that the particular categories which statistical analyses elect to use affects the apparent size and strength of racial inequality, but in ways different from previous time periods. In a 2012 study Mara Loveman and colleagues show that collapsing pretos and pardos into a single negro group actually decreases the magnitude of racial inequality in Brazil.Footnote 107 Nevertheless, Jerônimo Oliveira Muniz has shown that despite the situational nature of many Brazilians’ racial identity, the gap between whites and non-whites remains consistent regardless of the categories used.Footnote 108
Beyond the debates about racial categories, our results also underscore the importance of technocrats in shaping the key categories that provide the concepts, data and analyses that inform social policy debates more generally. Technocrats make policy choices shaped by the political, institutional and ideological contexts in which they operate, but in modern bureaucratic states their institutional position and their ability to frame choices in technical terms provides them with leverage to influence the selection and creation of social categories. To understand states’ approaches to these categories, we must look beyond the strategic interactions of major actors (states, civil society, transnational organisations), to explore the internal debates and disagreements, as well as decision-making processes, within the institutions that design, implement and evaluate policies.
