Racial inequality in Latin America

Abstract

This study examines socioeconomic inequality in Latin America through the lens of race and ethnicity. Its primary source is census data, but it also draws on national survey data from the Latin American Public Opinion Project. Unlike national censuses, the Latin American Public Opinion Project uses more consistent measures of ethno-racial composition across countries and includes a unique interviewer-rated measure of skin color. The study finds that individuals with darker skin, as well as Black and indigenous populations, experience educational, income and occupational disadvantages, even when controlling for social origin. Nonetheless, the degree of inequality and the hierarchical order of Afro-descendants, indigenous peoples, mestizos, Whites and others varies across countries. Finally, the study explores possible policy solutions to address ethnoracial disadvantage, highlighting Brazil’s targeted anti-racism policies, the region’s most extensive case.

inequality, census, racial categories, skin color, education, income, occupation, racial quotas

Introduction

Latin America has some of the world’s highest levels of social inequality. However, the relationship between this inequality and the region’s significant racial and ethnic diversity is understudied and often overlooked. The populations of Latin America and the Caribbean, including Afro-descendants, indigenous peoples, Whites and mestizos (people of mixed ancestry), have their origins in European conquest and colonialism, which began in the late 15th century. The colonizers’ primary goals were territorial expansion and economic extractivism, leading to significant changes in the region’s demographics. Critical elements or by-products of colonial ‘world-making’ included catastrophic declines in indigenous populations (on the order of millions) and the collapse of countless indigenous peoples/nations. Indigenous populations precipitously declined from diseases introduced by colonists, wars, enslavement and punishment. Colonialism further transformed the human face of the vast region through the forced transportation of millions of Africans to the ‘New World’ and the subsequent enslavement of those who survived the passage. Latin America has ~550 million people today, with some 40 million identifying as indigenous and 114–137 million as Afro-descendants (Telles and PERLA 2014). In total, indigenous and Afro-descendant peoples make up ~30% of the region’s population, although the proportions vary from country to country. After the colonial period, geopolitics and nation-building projects further transformed Latin America’s populations through significant and often targeted European immigration and the arrival of Asian and other non-European migrants.

The region’s Black and Indigenous populations, though, are disproportionately among the poorest in contemporary Latin America, almost without exception, and are largely absent among the middle and upper classes. Although the statistical analysis of socioeconomic inequalities in Latin America has a long history, the detailed analysis of racial and ethnic inequality is surprisingly new (Oxhorn and Jouve-Martín 2017). The absence of analysis might be attributed to near-consensus thinking that race and ethnicity were relatively unimportant or transitory in the region and, relatedly, to the absence of race and ethnic data (Loveman 2014). Except for occasional studies of race in Brazil (Silva 1978; Hasenbalg, 1985; Telles 1994; Lovell and Wood 1998) and a multi-country study of Indigenous disadvantage (Psacharopoulos and Patrinos, 1994), one would be hard-pressed to find any systematic studies or racial or ethnic inequality in Latin America before 2000 (see e.g. Atal et al.’s (2009) extensive bibliography). Considering that standard measures of race are based on self-identification, which tends to be fluid in the region, analysts increasingly use a multidimensional approach with additional measures, such as skin color (Woo-Mora 2022; Telles et al., 2015; Bailey et al., 2014). In terms of policy, Lustig (2017) notes how government-led fiscal interventions throughout the region, despite reducing income inequality, have had relatively little effect on Black and indigenous disadvantage.

This study employs a comparative lens to examine multiple dimensions of race and ethnicity in 20 Latin American and Caribbean countries in three sections. We first describe three principal ethnoracial formation projects behind the historical production of official census statistics and understandings of racial dynamics in the region: whitening, mestizaje and multiculturalism. We then explore the region’s ethnoracial demography, drawing on three racial and ethnic measures: census categories, national survey categories and interviewer-rated skin color categories.¹ Lastly, we examine the region’s ethnoracial inequality in education, income and occupation, leveraging the multiple ethnoracial measures for unique insights into social stratification by race and ethnicity across Latin America.

We find that Latin America’s ethnoracial landscape is characterized by a mosaic of identities shaped by centuries of populational conflict and hierarchy, as well as admixture and commonalities. Broadly speaking, mestizos are prevalent across much of the region, representing a particular vision of individuals and populations of perceived mixed European and indigenous ancestry. Countries like Mexico, Ecuador and Peru have large mestizo majorities. Indigenous populations are particularly substantial in the Andean and Mesoamerican regions, with many maintaining distinct cultural identities and languages. Afro-descendant or Black populations are particularly salient and large in Brazil, Colombia and coastal areas of countries like Ecuador and Venezuela. In Brazil, Cuba and Venezuela, individuals of some Afro-descendency self-classify as pardo, mestizo, mulatto or moreno, analogous to the ‘intermediacy’ of the mestizo category. European-descended or White populations, notably Spanish, Portuguese and other European origins, figure prominently in Argentina, Chile and Uruguay.

Despite noted ethnoracial fluidity and salient population admixture, which Telles (2004) labeled the ‘horizontal’ dimension of ethnoracial dynamics in Latin America, our research uniquely illustrates ethnoraciality’s ‘vertical’ or hierarchical dimension as well. Most prominently, we show that the region’s White populations continue atop the region’s ethnoracial hierarchies while individuals with darker skin, as well as Black and indigenous populations, experience educational, income and occupational disadvantages, almost without exception. The region’s patterned inequality is complex but clear even when controlling for social class, a dimension often argued historically as ‘explaining’ Latin America’s ethnoracial inequalities. Nonetheless, we show how the degree of inequality and the hierarchical order of Afro-descendants, Indigenous peoples, mestizos, Whites and others can vary across countries while the overall negative impact of racial and ethnic socioeconomic divides is stubbornly persistent.

Context: ethnoracial formation projects

Whitening

Social distinctions based on perceived phenotype, ancestry and culture—markers associated with ‘race’ and ethnicity—have been prominent features of the New World throughout its 500 year history. These ethnoracial markers are important because they have come to signify and reflect power dynamics and determine innumerable social outcomes throughout the Americas.² The dominance of Europeans and their descendants over non-European ethnoracial populations, sometimes de jure but almost always de facto, continues today, as our findings will show. The region’s early colonial history of various iterations of subjugation and enslavement of Indigenous and Afro-descendant populations speaks for itself on that account.

However, ethnoracial markers as social constructions shift in meaning and salience over time and across contexts. For example, during much of the 17th and 18th centuries, until independence in the early 19th century, Spanish colonial authorities established a ‘castas’ (literally, castes) system defining individuals’ proportion of Spanish ancestry for taxation and the assignment of trades and offices. Authorities often used phenotype, especially skin color, as a proxy for ancestry. Over time, through various stages and forms of oppression, the association of European ancestry and ‘whiteness’ became deeply embedded across the region (Martínez-Alier 1974; Graham 1990; de los Angeles and López 1991; Shumway 2001; Martínez 2008). Whiteness, or proximity to it, conveyed material and symbolic privileges in Latin American society.

Local elites began to suspend ‘casta’ laws with Spain’s liberal 1812 constitution, and these laws would fall across the entire region with the independence movements of the 1820s. Moreover, the abolition of slavery in the 19th century shifted and challenged the meaning of ethnoracial markers. However, from the late 19th to the early 20th century, Latin American elites became increasingly concerned about their place among modern nation-states in Europe and the United States. Those countries’ race science stipulated the biological superiority of a ‘White race’, the inferiority of the ‘Black race’ and the degeneracy of ‘race-mixing’. Hence, elites were concerned that the region’s large non-White populations and extensive miscegenation would relegate the region’s countries to a perpetual second-class status among modern nation-states (Skidmore 1976; Helg 1990; Stepan 1991).

In response, Latin American elites banked on an alternative ‘race science’ in their attempts to portray themselves as largely White or near White. The alternative view stressed neo-Lamarckian ideas about the mutability of ‘race’. Racial formation projects thus promoted ‘constructive miscegenation’ to project population change in the direction of whiteness. It was believed that ‘White genes’ could predominate in the progeny of population admixtures, thereby encouraging gradual cross-generational whitening. Additionally, elites enacted discriminatory immigration policies—recruiting Europeans while constricting non-Europeans—toward conshtructing a whiter national population. Essentially, many Latin American countries believed that to modernise, they could whiten their populations. Through whitening, elites thus sought to push their burgeoning nations into the ranks of the modern nation-states with the US and Europe (Skidmore 1976; Stepan 1991; FitzGerald and Cook-Martín 2014; Loveman 2014). An ideology of racial whitening became a defining feature of the Latin American landscape and policy. It is also recognizable even today across diverse spheres, such as the marriage market and personal aesthetics (Wade 1997; Telles and PERLA 2014; Osuji 2019).

Mestizaje

By the 1930s, leading thinkers in many Latin American countries would turn the previous racialist thinking about whitening on its head through an overt embrace of ethnoracial population admixture or mestizaje. Earlier race science’s undoing was primarily based on ideas by anthropologist Franz Boas. He argued that so-called racial differences were not rooted in biology but in culture. Boas had a tremendous influence in Latin America as he trained two of the region’s influential thinkers in Brazil and Mexico: Gilberto Freyre and Manuel Gamio. Their writings supported novel progressive ideologies that viewed ethnoracial admixture as positive and an essential feature of these new nations. Elites sought to create visions of the nation as homogeneous; in these visions, national or mestizo identities would replace the previous ethnoracial originating populations (cf. Knight 1990; Wade 1993; Telles 2004). For example, mestizaje narratives presented Brazilians, Mexicans and other national subjects as meta races that fused White, Indigenous and (sometimes) Black lineages and cultures. Mestizos or ‘mixed-race’ persons came to be considered the ideal or prototypical citizens (cf. Skidmore 1976; Knight 1990; Mallon 1992; Whitten 2004), and identities alluding to ‘mixed-race’ came to represent large swaths of Latin Americans, even those who might identify, or be classified by others, as Indigenous, Black or White (de la Cadena 2004; Telles and PERLA 2014).

Nevertheless, mestizaje ideologies varied within the region (Wade 2009; Telles and Garcia 2013). Mestizaje had particularly low resonance in the perceived White-majority countries of the Southern Cone (Argentina, Uruguay and, to some extent, Chile) and Costa Rica (Andrews 1980, 2010; Telles and Flores 2013). In contrast, elites in Mexico and Brazil were politically willing and able to promote powerful versions of mestizaje ideologies, partly because their states had strong capacities to disseminate these ideas through educational and cultural campaigns (Mallon 1992; Wade 2009; Telles and Garcia 2013). Moreover, mestizaje thinkers like Manuel Gamio were well positioned in the Mexican state apparatus. Freyre’s ideas gained wide popularity in Brazilian literary circles as his project fit well with nation-making and modernizing efforts. Andean countries tended to stress the binary racialized distinctions between the Spanish and the Indigenous; to the extent that mestizaje ideologies existed, they were weaker and shorter-lasting than in Mexico, the country to which they are sometimes compared (Mallon 1992; de la Cadena 2000; Larson 2004; Sulmont and Callirgos 2014). In Mexico, Colombia, the Dominican Republic and Peru, the mestizaje narratives stressed Indigenous and Spanish lineage while downplaying or ignoring African contributions, even though these countries had forcibly brought hundreds of thousands of enslaved Africans. This exclusion is arguably more severe in the Dominican Republic, where features associated with African origin are apparent among much of the population. Its capital, Santo Domingo, was a major slave port, yet Afro-Dominicans are conspicuously absent from the national mestizaje narratives, which emphasize Spanish-Taino mixture as the root of the Dominican nation (Howard 2001; Candelario 2007; Roth 2012). On the other hand, mestizaje ideologies in Brazil and Cuba stressed the inclusion of African elements, as well as those of Europeans and Amerindians (Telles 2004; Bailey 2009; de la Fuente 2001; de la Fuente and Bailey 2021).

Mestizaje became a widely shared Latin American experience (Wade 1997; Telles and PERLA 2014). Even though the ideas of mestizaje were considered progressive for much of the 20th century, they also had critics. These alternative voices and scholars argued that mestizaje ideologies erased the ethnoracial consciousness of the general population or even implied a cultural or statistical genocide of Black and Indigenous peoples (Do Nascimento 1979; Guillermo 1990).

The multicultural turn

The context for understanding the character and place of ethnoracial diversity in Latin America is rapidly changing. In challenges to ideologies of whitening and mestizaje, many movement, state, academic and international actors embrace and promote what is often referred to as the multicultural turn in Latin America (e.g. Hooker 2005; Fontaine 2012; Rahier 2020). This has occurred in the context of an economic transition and growing democratization. The domestically focused economic model of industrial growth of the 1980s, based on import substitution, had been in decline, replaced mainly by dominant Western economic models. Neoliberalism and globalization have exposed these countries to more significant external pressure and scrutiny on all sides. The opening of Latin American markets bolstered US and European exploits and extractivism in Latin America, and states often turned to austerity-conditioned loans and serial indebtedness to remain afloat (Main 2020). Conversely, the globalization experience included monitoring human rights norms by private international organizations and UN human rights committees, legislation and forums and internal pressure from Indigenous, Afro-descendant and other human rights movements (Telles 2004; Van Cott 2005).

In a rapid transition taking less than three decades, nearly all Latin American countries were considered representative democracies by the mid-2010s, in contrast to only four of the nineteen in the mid-1970s (Mainwaring and Pérez-Liñán, 2014). As part of their democratization process, more countries officially recognize the identities, dignity and rights of Afro-descendants and Indigenous people. Since the 1980s, Black and Indigenous movements have emerged as important new global and domestic actors, pointing out and challenging the region’s inequalities. Often backed by an international network of human rights supporters and institutions, including UN forums to promote human rights, Indigenous and Black movement activists are increasingly effective at pressing their governments to address persistent social exclusion (Yashar 1998; Paschel and Sawyer 2008). In some cases, new constitutions, laws and social policies have sought to respond to claims for greater racial, ethnic and gender justice. In sum, multiculturalism refers to a new stage of ethnoracial thinking in Latin America. It seeks to recognize, respect and endorse the region’s ethnoracial diversity and hence benefit national belonging and well-being for all. This includes well-known efforts to promote racial justice and diminish racial inequality in Brazil.

Censuses and other official efforts at data collection parallel these stages of ethnoracial formation projects. Early censuses often collected racial data to gauge national progress in whitening (Skidmore 1976; Loveman 2014). The inclusion of ethnoracial queries in some national censuses peaked in the 1920s. These queries were then dropped in many countries for ideological and political reasons tied to mestizaje and because the scientific consensus began to invalidate ‘race’ as a scientific concept and eschew its use as a population category (Loveman 2014).

By the 1990s, though, the collection of ethnoracial data intensified with the shift to multiculturalism and the demand for ethnoracial recognition (Nobles 2000; Loveman 2014). Moreover, there was growing social scientific documentation and elite recognition that ethnoracial classifications, though social constructs, were nonetheless associated with or proxied social stratification in the region. Mainly because of pressure from international human rights groups, international funding organizations, such as the World Bank, and international conventions, particularly the International Labor Organization’s Convention 169 (Indigenous and Tribal Peoples Convention), adopted in 1989 and ratified by most Latin American countries by 2000, many Latin American countries began collecting ethnoracial data for the first time in decades (Loveman 2014). In addition, Latin American activists at the 2001 UN Conference against Racism, Racial Discrimination and Xenophobia in Durban, South Africa, demanded that their governments collect data on ethnicity and race (Htun 2004). Having data to document ethnoracial inequities, governments would find it increasingly challenging to sustain national narratives of mestizaje, non-discrimination, racial harmony and racial equality. Collecting data on race and ethnicity was also considered an essential first step—arguably, the initial policy shift—in the transition to multiculturalism. Since 2010, almost all Latin American countries now have census data for ethnic minorities. Nonetheless, as shown below, variation in census coverage continues to create barriers to a fuller enumeration of the region’s ethnoracial diversity and the comparative study of the region’s ethnoracial stratification. We now proceed to describe our data before presenting our findings.

Data

National Censuses

Our primary data sources are national censuses. We analyzed the latest census data in the International Public Use Micro Sample (IPUMS). Each country designs its own Census ethnoracial questions and response formats, but all national censuses use self-classification. Although most Latin American censuses now include questions and a variety of categories to gather data on their Afro-descendant and Indigenous populations, coverage of other ethnoracial populations remains inconsistent. Note that some countries still do not ask about indigenous people (Cuba and the Dominican Republic), and some do not ask about Afro-descendants (e.g. Chile and the Dominican Republic). Many began asking about ethnoracial categories only in the 2010 round, with Mexico (regarding Afro-descendants) only in 2020; the Dominican Republic has not collected such data since 1970. For our analysis of census data, we thus include only those countries with recent data in IPUMS and with ethnoracial variables, with a couple of exceptions: Peru and Venezuela are usually not included because their latest censuses in the IPUMS file do not include data for race and ethnicity.

Latin American Public Opinion Project (LAPOP)

For our inequality analyses by skin color and alternative race classification, we also utilize data from the LAPOP, conducting biennially random sample face-to-face or phone surveys in all Latin American countries except Cuba. LAPOP has included an interviewer-rated skin color item since 2010. Each survey included ~1500 respondents, with larger samples from countries such as Brazil and Colombia.

Project on Ethnicity and Race in Latin America (PERLA) surveys

We also draw on data from the Project on Ethnicity and Race in Latin America (PERLA) conducted in 2010 in Brazil, Colombia, Mexico and Peru. The data are nationally representative and include ~1500 respondents per country. These surveys focused on race and racism issues, exploring diverse race/ethnic classification questions, including the skin color measure, which was also introduced as a core element in the biannual LAPOP surveys since 2010.

Brazilian National Household Sample Survey (PNAD)

We explore shifts in ethnoracial makeup and university enrollment trends in Brazil from 1992 to 2021, utilizing data from the National Household Sample Survey (PNAD) conducted by the Brazilian Institute of Geography and Statistics (IBGE). This survey gathers household-level data on demographics (age, race and gender), education, employment and income through in-person interviews with participants aged 14 and above.

In the upcoming sections, we detail our findings regarding ethnoracial breakdowns, educational, income and occupational inequalities by race/ethnicity. We segment the findings by data source and ethnoracial measure, concluding with analyses on ‘race versus class’ effects, the unique Panamanian scenario and urban residency’s impact on ethnoracial stratification.

Findings: ethnoracial composition

Ethnoracial composition, census data

Official censuses are developed by each nation’s census office, selecting ethnoracial categories that reflect and shape the country’s racial/ethnic demographics. Categories and portrayals vary across recent censuses, with most including indigenous and Afro-descendant groups. Figure 1 provides a comparative overview of ethnoracial composition based on national censuses (using self-identification) across the region. It includes 16 Latin American countries plus the US territory of Puerto Rico. Although Fig. 1 mainly represents cases of those countries with ethnoracial census data found in IPUMS, we also included ethnoracial statistics for Argentina, Peru and Venezuela. These countries have ethnoracial data in their most recent censuses but are unavailable via IPUMS. For these cases, we used information on ethnoracial composition from the official websites of these countries.³

Figure 1

Ethnoracial composition by country census classification scheme. Source: Harmonized National Censuses for each ‘nation (year)’ accessed through IPUMS variable(s): ‘race’ in harmonized IPUMS data for each ‘nation (year)’ race categories across nation (year): White, Black (including Black African, Black Caribbean, Afro-Ecuadorian, Other Black), Indigenous (including American Indian and Latin American Indian), Asian (including Chinese, Japanese, Korean, Vietnamese, Filipino, Indian, Pakistani, Bangladeshi, other Asian), Mixed race (including Brown (Brazil), Mestizo (Indigenous and White) and Mulatto (Black and White)

Figure 1 is divided into two panels for ease of interpretation. From left to right, the first six countries (Uruguay to Ecuador) include ethnoracial composition questions using ‘full coverage’ classification schemes, i.e. they included White, mestizo/mixed, Indigenous and Black or Afro-descendant categories. These full-coverage countries are ordered by their White populations’ size (percentage), from highest to lowest. The remaining nine countries (Argentina to Guatemala) exclusively targeted Indigenous and Black populations. Thus, we created a residual category (R) to represent the non-Indigenous and non-Afro-descendant population, which are predominantly White and mestizo. These countries are ordered by percent residual, from highest to lowest.

In the initial group of countries, Uruguay has the highest White population at ~90%, while Ecuador reports <10% of its total population as White. That variation is striking and certainly illustrates the substantial variation in ethnoracial landscapes across Latin America. In the first set, Ecuador and El Salvador stand out as predominately mestizo countries with small White populations, in contrast to majority White countries like Uruguay. Other countries fall at various midway points between those two poles. One case, the US territory of Puerto Rico, is particularly interesting. A large majority of Puerto Ricans on the island identify as White (70%), but among Puerto Ricans on the U.S. mainland, less than 50% identified as White, which may suggest different conceptions of who is considered White (Duany 2005, see Loveman and Muniz 2007 for a history).

Brazil and Cuba have consistently documented their Afro-descendant populations since the 19th century, a practice uncommon among other regional countries. Brazil’s 2010 census documents a substantial ‘pardo’ population of almost equal size to the White population. However, although ‘pardo’ is essentially a proxy category for ethnoracial population admixture, the Brazilian state and many Brazilians understand the ‘pardo’ category as belonging to a ‘collective Black’ or Afro-descendant population (Bailey and Telles 2006). The Afro-descendant component of ‘pardo’ is thus prioritized, and it is standard practice for the Brazilian state to consider the ‘pardo’ and ‘preto’ categories as proxies for Afro-descendancy that can be collapsed into a ‘negro’ category. When ‘pardo’ and ‘preto’ are counted together in Brazil’s 2010 census, the ‘collective Black’ or ‘negro’ population segment was a numerical majority population for the first time in the Republic’s history, the only majority ‘Black’ country among all of those examined here.

Cuba uses a three-option question to enumerate four categories of ‘skin color’ in its Census: (1) ‘negro’ (Black), (2) ‘blanco’ (White) and (3) ‘mestizo o mulato’ (mestizo or mulatto). The size of the Mestizo (i.e. individuals self-classifying as either ‘mestizo o mulato’) category in Cuba also stands out in its relative size, at 27% of the population according to the 2012 census, as does its Black population at just under 10%. In contrast to the Brazilian state in recent decades, however, the Cuban state’s ethnoracial formation project does not generally reflect a collective blackness perspective of its ethnoracial diversity.

Cuba and Puerto Rico (via U.S.) censuses do not include Indigenous categories. From an official census lens, these regions’ populations do not possess the markers of indigeneity often found in other Latin American countries, such as language and recognized communities/territories. In other countries that include the White category in their census, Indigenous populations generally represent less than3% of national ethnoracial compositions. The exception is Ecuador, where it is 7%.

The second set of countries does not include White and Mestizo categories in their censuses’ enumeration of ethnoracial populations. Instead, these countries only target Indigenous and Afro-descendants for census enumeration, with few exceptions. The result of the decision not to enumerate Whites and Mestizos is that most countries’ populations of this set are represented as undifferentiated masses, devoid of salient or impactful ethnoracial markers. Among these countries, Guatemala, Bolivia, Mexico and Peru stand out because of the relatively large size of their Indigenous populations when using the census lens. The set’s largest Black populations are in Colombia (11%) and Panama (9%).

Figure 1 presents those countries’ official ethnoracial composition. The ethnoracial categories used by each census represent the official portrait of their ethnoracial composition and are artifacts of their ethnoracial formation projects. The countries of the first set include White and mestizo categories, providing fuller information for studying ethnoracial inequality. The countries of the second set exclude White and mestizo categories and hence are arguably incomplete and complicate research on racial stratification. Both sets of countries, however, arguably represent national ethnoracial projects of ‘mestizaje’, albeit differently. Those national schemes in the first set of countries contain ‘mestizo’ categories and reflect ‘mestizaje’ formation projects more directly. El Salvador and Ecuador are illustrative cases. Research on the first set of countries often emphasizes the fluidity dynamics between the mestizo/mulato/pardo categories, on the one hand, and the indigenous and Black categories, on the other. They may also represent a gradual de-ethnicization process and its intersection with whitening ideologies. Brazilian social movement actors, e.g. have long seen the pardo category in Brazil as a reflection of state whitening projects but also tied to the country’s version of the ideology of mestizaje, which is known as the ideology of ‘racial democracy’.

Nonetheless, the second set suggests that ‘mestizaje’ as a state ethnoracial project does not require a separate mestizo census category; instead, only Black and Indigenous populations are counted as ethnic and racially distinct compared to all (undifferentiated) others. Moreover, public and informal discourse regarding population admixture as defining national identity in these countries is common, which differs in most Southern Cone countries. That is, to be non-indigenous or non-Black in Chile and Argentina, e.g. is to be simply Chilean and Argentinian: sometimes ‘mixed’, but unremarkably so from the census perspective.

Ethnoracial composition, LAPOP

We turn now to an alternative lens on these countries’ ethnoracial compositions. The classification schemes and data from LAPOP are often fundamentally distinct from the census data. LAPOP does not replicate national classification schemes in any of the countries. LAPOP surveys generally standardize the region’s ethnoracial questions using, in most cases, six ethnoracial categories: White, Mestizo, Mulatto, Black, Indigenous and Asian. A few exceptions exist regarding those six standard categories, most notably in Guatemala, Brazil and the Dominican Republic. LAPOP uses Ladino and Indígena in Guatemala; branco, pardo and preto for Brazil and blanco, Índio, mulato and negro in the Dominican Republic. LAPOP data provide more detailed information for analyzing ethnoracial stratification, though their sample sizes and coverage are far less than those of national censuses.

The alternative LAPOP source allows us to deconstruct and disaggregate most countries’ ‘R’ or residual categories shown in Fig. 1. For example, the LAPOP data in Fig. 2 reveal that White populations dominate the ethnoracial demography in Argentina, Chile and Costa Rica, which Fig. 1 somewhat obscures. In Honduras, Nicaragua, Panama, Mexico and Colombia, LAPOP reveals that ‘mestizo’ populations dominate, which was also obscured by the R category. In addition, LAPOP’s inclusion of the ‘mestizo’ category in Uruguay considerably changes the portrait of that country: its White segment shrinks from 90 to 65%, and fully 20% of Uruguayans opt for ‘mestizo’ self-identification. In Brazil, the nuanced differences in LAPOP’s question, which uses White, pardo (mixed ancestry), negro or Afro-Brazilian (Black), Asian and Indigenous, result in ‘pardos’ overtaking Whites as the largest population segment compared to Brazil’s 2010 census, as well as a near doubling of the overall percentage of individuals self-classifying as Black.⁴

Figure 2

Percent ethnoracial category composition. Source: Latin American Public Opinion Project (LAPOP) 2010–2021. Variable(s): ‘Etid’ in merged LAPOP data for each nation 2010–2021. Race categories across nations: (1) Blanca, (2) mestiza, (3) Indigena, (4) Negra, (5) Mulata and (6) Otra

In addition, LAPOP provides information on country cases (i.e. Dominican Republic, Peru and Paraguay) for which we could not obtain ethnoracial data via IPUMS. In contrast, Cuba is the only country case for which we have census data but no LAPOP.

Nonetheless, as shown below, LAPOP presents a different portrait of Latin America’s ethnoracial demography and provides a contrasting lens on its ethnoracial inequality. The differences in racial composition between Figs 1 and 2 also reflect how the questions and categories are created differently in LAPOP and the Censuses. For example, an Indigenous category is collected in various ways in the Censuses, with some referring to particular Indigenous groups. In contrast, others refer to the general or pan-ethnic Indigenous category. However, LAPOP uses a uniform version of the pan-ethnic ‘Indigenous’ format, which has implications for which persons are considered in the Indigenous category (Telles and Torche, 2019).

As Bailey et al. (2016) discussed, determining the most effective method to capture Latin America’s ethnoracial stratification dynamics through census or LAPOP surveys is a complex issue. While IPUMS’s extensive census sample sizes and sampling methods are advantageous for broader ethnoracial categories, national census classifications may better align with how individuals in each country perceive these categories (Telles 2014). However, for countries primarily focusing on specific minority ethnoracial groups, such as those toward the right of Fig. 1, LAPOP data can provide crucial insights into ethnoracial dynamics. This study examines ethnoracial stratification based on educational attainment, income and occupational prestige using data from national censuses and LAPOP surveys.

Findings: ethnoracial inequality

Educational inequality, censuses

Figure 3 presents the same countries as Fig. 1 according to the percentage of the population that completed a university education by ethnoracial category. A first glance reveals significant variation. In the first set of countries (i.e. those including White and mestizo categories in their census), Puerto Rico has a higher percentage of university completion overall, while the lowest overall levels of university completion are in Honduras, Nicaragua and Guatemala. In this set, Cuba shows the lowest level of ethnoracial inequality in university completion, as revealed by the close clustering of White, Black and mestizo/mulatto categories.

Figure 3

Percent university completed, ages 25–50. Source: Harmonized National Censuses for each ‘nation (year)’ accessed through IPUMS. Variable(s): ‘race’ and ‘edattain’ in harmonized IPUMS data for each ‘nation (year)’; race categories across nation (year): White, Black (including Black African, Black Caribbean, Afro-Ecuadorian, other Black), Indigenous (including American Indian and Latin American Indian), Asian (including Chinese, Japanese, Korean, Vietnamese, Filipino, Indian, Pakistani, Bangladeshi, other Asian), mixed race (including Brown (Brazil), Mestizo (Indigenous and White), Mulatto (Black and White), educational attainment categories across nation (year): (1) less than primary completed, (2) Primary completed, (3) Secondary completed and (4) University completed

Of further note, the Asian population in Costa Rica is clearly an extreme outlier. Asian populations also occupy the top of the educational achievement hierarchy in Uruguay and Brazil. The top position of the Black population in Panama is particularly striking, although again, the R category of the Panamanian Census disallows a complete accounting of ethnoracial effects. The case of Ecuador stands out among those countries that include ‘mestizo’ categories in their Census. It is the only country case where the mestizo population tops the ethnoracial hierarchy regarding university completion. In all the others in that set, Whites stand in privileged positions in relation to mestizo, Black and indigenous populations.

In contrast to parsing the highest level of educational attainment, in Fig. 4, we present a view of ethnoracial stratification among the lowest educational level segment of the population: those with ‘no schooling’ or only ‘some primary’ education among 25- to 60-year-olds. The figure demonstrates the wide variation in the universalization of education across countries in the region. Uruguay, Puerto Rico, Cuba, Costa Rica and Chile show comparatively lower percentages of any ethnoracial group at the lowest education levels. Ethnoracial stratification is less pronounced at this lowest level in these five countries. However, significant stratification exists at the highest educational level, with Cuba being an exception. Guatemala stands out for its low levels of educational achievement, particularly among the majority of its population. At the lowest education level, there is a stark contrast between indigenous populations (~90%) and Ladinos (~50%). This disparity is universal across the region, with indigenous populations consistently lagging behind non-indigenous populations at the lowest education level (see Fig. 5). Black populations experience notable disadvantages compared to Whites, especially in Brazil, El Salvador and Ecuador.

Figure 4

Percent lowest educational achievement, ages 25–60. Source: Harmonized National Censuses for each ‘nation (year)’ accessed through IPUMS. Variable(s): ‘race’ and ‘edattain’ in harmonized IPUMS data for each ‘nation (year)’; race categories across nation (year): White, Black (including Black African, Black Caribbean, Afro-Ecuadorian, other Black), Indigenous (including American Indian and Latin American Indian), Asian (including Chinese, Japanese, Korean, Vietnamese, Filipino, Indian, Pakistani, Bangladeshi, other Asian), mixed race (including Brown (Brazil), Mestizo (Indigenous and White), Mulatto (Black and White). Educational attainment categories across nation (year): (1) less than primary completed, (2) primary completed, (3) secondary completed and (4) university completed

Figure 5

Mean years of schooling by ethnoracial categories. Source: Latin American Public Opinion Project (LAPOP) 2010–2021. Variable(s): ‘Etid’ and ‘ed’ in merged LAPOP data for each nation 2010–2021. Race categories across nations: (1) Blanca, (2) Mestiza, (3) Indigena, (4) Negra, (5) Mulata and (6) Otra. Educational attainment categories across nations: (1) none, (2) primary, (3) secondary, (4) university and (5) post-secondary, not university

Educational inequality, LAPOP

Figure 5 depicts a comparative analysis of ethnoracial stratification in educational achievement using LAPOP data, offering insights into educational inequality across 18 Latin American countries.⁵ Due to sample size considerations within the LAPOP dataset, the study employs mean years of schooling instead of the percentage attending university.

The LAPOP perspective on ethnoracial inequality in educational achievement varies broadly. We highlight a few patterns. The results in Fig. 5 suggest that White populations are at the top of the ethnoracial hierarchy in mean years of schooling in only 4 of the 18 country cases: Uruguay, Brazil, Argentina and Panama. In the lion’s share of countries, the mestizo category ranks highest.⁶ This overall view on ethnoracial stratification revealing the high ranking of ‘Mestizo’ populations is counterintuitive. However, as discussed below, mestizo self-identification is often selective of more educated persons and may not necessarily align well with skin-color cleavages (Telles and PERLA 2014). Research suggests, e.g. that the symbolic value of the mestizo category as derived from national mixing ideologies makes it attractive to higher-status individuals; at the same time, self-identifying as White may be more attractive to low-status persons (Telles and Flores 2013). Hence, as we discuss further below, the mestizo category may be especially poor for capturing the type of ethnoracial dynamics and markers that scholars commonly associate with social stratification.

Despite the counterintuitive ordering of White and mestizo categories in years of schooling, the results in Fig. 5 generally suggest substantial indigenous and Black disadvantage in educational achievement. Seventeen of the eighteen countries have large enough Black populations for our analysis in LAPOP samples. In 8 of these 17, the Black category sits at the bottom of the hierarchy. The mean years of schooling for the Black category in Honduras, El Salvador, Nicaragua and Paraguay are <7 years. In contrast, our results in Fig. 5 show that those self-classifying in the Black category in Panama have the highest mean years of education than any other country’s Black category in our analysis, at 11 years.

Latin America’s Indigenous populations occupy the bottom of the ethnoracial hierarchy in mean school years in 9 of the 17 countries. Nonetheless, the overall situation of this population category across the region shows relative disadvantage in all countries vis-à-vis White and mestizo categories. Indigenous disadvantage in educational attainment is most apparent in Colombia and Guatemala, where these two populations appear as extreme outliers relative to other categories.

Overall, our analysis of categorical ethnoracial inequality in educational achievement using multiple measures and census and LAPOP data are mixed. The privileged position of the White category is clearest through the lens of the percent university completion in those countries with full classification schemes in their national censuses (Fig. 3, left side). The countries without full classification schemes in their national censuses (Fig. 3, right side) substantially obscure this measure. Nonetheless, Indigenous disadvantage at this elite level of educational attainment is most apparent there. In contrast, our analysis of LAPOP data and mean years of schooling (Fig. 5) is most telling regarding the relative location of mestizos in the region’s ethnoracial hierarchy: they hold top positions in 10 of the 16 countries where mestizo was a category option. Again, previous results using LAPOP data to study the ethnoracial category stratification in educational achievement support the robustness of these results (Telles and Steele 2012; Telles et al., 2015).

Educational inequality by skin color, LAPOP

We now turn to a third measure of ethnoracial boundary dynamics—skin color—to examine its impact on educational inequality. Racial and ethnic formations are multidimensional social constructs that vary across time and context (e.g. Roth 2016). LAPOP provides a unique opportunity to explore skin color and its impact on social inequality, a dimension that Roth (2016) labels ‘phenotype’ in her typology of racial dimensions. Importantly, there is a significant scholarly discussion around which measure or measures of ethnoracial boundaries (or dimensions) best capture inequality dynamics in Latin America (Telles and Lim 1998; Telles and PERLA 2014; Telles et al., 2015; Bailey et al., 2013; Bailey et al., 2014; Bailey et al., 2016). As a marker of ethnoraciality or ethnoracial boundary cleavages, does skin color provide leverage for examining ethnoracial stratification in mean years of schooling in Latin America?

Figures 6 and 7 present country comparisons of mean years of schooling by interviewer-evaluation of respondent skin color. The results are striking. Whereas the results in Fig. 5, which used ethnoracial categories as its measure, place mestizos at the top of the educational achievement hierarchy in mean years of schooling, Figs 6 and 7 reveal that light skin color persons are the most advantaged, almost without exception. That is, although there is variation in the mean years of schooling for the lightest categories across the region, its relative hierarchical position in each country is generally on the top. Moreover, dark skin color persons tend to be at the bottom, with medium skin color persons in the intermediate position. While Fig. 6 shows this for the light, medium and dark color categories, Fig. 7 provides a complementary perspective using a more detailed color continuum. Generally, in line with Fig. 6, results in Fig. 7, though, complicate its findings in a minority of cases regarding the expected positive association of lighter skin color with higher mean years of schooling, such as in Panama and Honduras.⁷

Figure 6

Mean years of schooling by major skin color categories. Source: Latin American Public Opinion Project (LAPOP) 2010–2021. Variable(s): ‘Etid’ and ‘colorr’ in merged LAPOP data for each nation 2010–2021. Race categories across nations: (1) Blanca, (2) mestiza, (3) Indigena, (4) Negra, (5) Mulata and (6) Otra. Color categories were recorded in three categories across nations: (1) 1–3, (2) 4–5 and (3) 6–11

Figure 7

Mean years of schooling by detailed skin color categories. Source: Latin American Public Opinion Project (LAPOP) 2010–2021. Variable(s): ‘Etid’ and ‘colorr’ in merged LAPOP data for each nation 2010–2021. Race categories across nations: (1) Blanca, (2) Mestiza, (3) Indigena, (4) Negra, (5) Mulata and (6) Otra. Color variable include 11 color categories coded 1–11 across all nations

Race versus class in educational inequality

The examination of ethnoracial inequality in Latin America has evolved. Earlier studies by scholars such as Wagley (1963) and González Casanova (1965) predominantly attributed ethnoracial disparities to the entrenched class stratification within the region. This perspective suggested that the observed racial inequalities were largely a byproduct of the broader class inequalities. Challenging these foundational views, later scholars, such as Silva (1985), began to untangle the complexities of ethnoracial dynamics, shedding light on their distinct influence on social inequality, independent of class factors. This perspective acknowledges the profound class divides but also recognizes that racial and ethnic boundary dynamics can have separate and substantial impacts on social outcomes.

Supporting Silva’s foundational work, the research literature suggests that the effect of skin color on educational achievement is sustained even with controls for social origin and a host of other variables (Telles et al., 2015; Telles and Steele 2012). We engage this debate further with results from an analysis presented in Fig. 8, showing differences in predicted education years by parental occupation, ethnoracial self-identification using Census categories and skin color (see Telles et al., 2015). The predicted probabilities are based on a multivariate regression model using data from LAPOP and the PERLA for eight countries representing nearly 80% of the region’s population.⁸  Figure 8 shows that even with controls, parental occupation, a leading measure of social origin or class, and skin color independently predict years of schooling.

Figure 8

Predicted years of schooling by parental occupation, ethnoracial and skin color. Note: The ‘lighter’ and ‘darker’ skin color bars indicate the predicted years of education of respondents with skin tones one standard deviation below or above the mean, respectively, for each country. The parental occupation bars also represent the estimated years of education for respondents one standard deviation below or above their country average occupational status, respectively

The first two bars in each country panel show that respondents with parents in high occupational categories (professionals and administrators) have 2–3 more years of education than those in low categories (unskilled and semiskilled occupations). The rightmost bars in each country panel represent skin color differences. These bars show that the lighter-skinned persons have 1–2 more years of schooling than the darkest respondents, net of controls for parental occupation and the other controls. On the other hand, the results using ethnoracial categories produce mixed results, both with and without controls. Considering that parental occupation may be the best proxy measure for class, as it is commonly conceived, the results in Fig. 8 suggest that both class and race, especially when measured by skin color, independently shape educational attainment in these eight Latin American countries.⁹

Income inequality, LAPOP

To delve deeper into ethnoracial stratification in Latin America, we focus on a second measure of socioeconomic status (SES): household income. Drawing on data from the LAPOP surveys conducted across 18 countries in 2012, our analysis examines per capita household income across ethnoracial categories and skin color. The income assessment relies on self-reported data, utilizing 16 intervals tailored to each country’s currency. Respondents were prompted to report the total monthly household income, encompassing earnings from working adults and children and remittances from abroad. To ensure robust analysis, we focus solely on color points and racial categories with a minimum of 30 cases, thereby excluding smaller racial populations and truncated color diversity in some countries (see Bailey et al., 2014).

Figure 9 visually represents the relationship between per capita household income, skin color and ethnoracial categories across the included countries. The x-axis arranges countries based on the percentage of the population identified as White, with the highest percentages on the left and the lowest on the right. The mean income distribution is depicted on the y-axis.

Figure 9

Mean per capita household income by ethnoracial category and skin color, LAPOP 2012. Source: LAPOP 2012, adapted from Bailey et al. (2014)

Notes: The mean per capita household income of skin color category five is the reference (0%) for each country. Skin color points are shaded to match the category number on the color scales. Racial categories are denoted by letters—W = White/Blanca, B = black/Negra, A = Asian/amarela, L = Ladina, Me = Mestiza, Mo = Morena, Mu = Mulata, I = Indígena.

In our analysis, we utilize the mean income of the mid-range (or fifth) skin color point as the reference category within each country. This choice allows for a consistent benchmark across nations. Mean income values for other skin color points and racial categories are graphed relative to this mid-range value as a percent difference. Specifically, the mean income of the skin color category 5 serves as the baseline at 0%; thus, mean income values above this benchmark are expressed as a positive percent difference, while those below it are represented as negative percent differences. This approach enables a comparative assessment of income disparities across different skin color points and racial categories within and across countries. It provides valuable insights into ethnoracial stratification and its impact on household income distribution.

To illustrate the interpretation of the data presented in Fig. 9, consider Uruguay. The data indicates a significant disparity in income based on perceived skin tone. Specifically, individuals with the lightest skin tone reported a per capita household income 60% higher than their counterparts in the median skin tone category, labeled as category 5. Conversely, those perceived as having the darkest skin tone earned ~18% less than those in the median category. Moreover, when examining ethnoracial classifications, the trend persists. Individuals in Uruguay who identify as White, or ‘blanco’ (W), earn on average 30% more than those in the intermediate skin tone category. On the other end of the spectrum, individuals identifying as mulatto (‘mulato’, MU) earn roughly 30% less compared to the median skin tone benchmark.

Figure 9 distinctly reveals a striking correlation between income disparities and both perceived skin tone as assessed by interviewers and self-identified ethnoracial categories throughout the Americas. A consistent, albeit varying, linear relationship is evident in numerous countries: lighter skin tones often correlate with higher per capita household incomes. This pattern is particularly noticeable in countries, such as Mexico, Paraguay, Uruguay, Ecuador and Nicaragua, although the magnitude of income disparities associated with skin color differs across regions. For instance, Guatemala exhibits the most pronounced income difference for individuals with the lightest skin tone when contrasted with countries like Nicaragua and Chile, where income differences based on skin tone are less pronounced, as reflected in the survey samples from these countries. The relationship between skin color and income sometimes shows a strong, sometimes a weak correlation, and in rare instances, such as in Panama and Honduras, the pattern is somewhat reversed. These variations notwithstanding, the overarching trend underscores a systematic structuring of social inequality along the lines of skin color hierarchy in Latin America.

Regarding the relationship between income and ethnoracial category divides, the LAPOP national surveys demonstrate that individuals who identify as White occupy mostly the upper echelons of the ethnoracial hierarchy. This predominance is consistent in 14 out of the 18 surveyed countries in Latin America. Figure 9 further elucidates this point by illustrating the persistent socioeconomic disadvantages faced by individuals who self-identify as Indigenous, particularly in countries, such as Panama, Chile and Colombia, where the disparities are most pronounced.

In our analysis of racial category hierarchies across Latin America, only 10 countries had enough individuals self-identifying as Afro-descendants in our samples to be included in the study. Within these nations, Afro-descendants occupy the lowest socioeconomic tier in four specific countries: Nicaragua, Brazil, El Salvador and Ecuador. The SES of Afro-descendants in the remaining countries is variable; however, they are typically at a disadvantage when compared to individuals identifying as White. Notably, in some instances, those who identify as Afro-descendant are situated higher in the income hierarchy than other non-White groups—e.g. mestizos in Uruguay and morenos in Venezuela. Interestingly, the data from Panama and Honduras are outliers, showing that individuals identifying as Black are positioned above all other ethnoracial groups. Moreover, the analysis presented in Fig. 9 indicates that in most Latin American countries that recognize mestizo or other intermediate categories, these groups consistently find themselves in a middle-income bracket within the national ethnoracial hierarchies.

Race versus class in income inequality

Expanding on this debate, we examine the mediating role of social origin—proxied by maternal education—in the relationship between skin color and income (see Bailey et al., 2016).¹⁰ We begin by replicating the analysis presented in Fig. 9 above, which used data from 2012, with a later 2014 LAPOP survey round. We then introduce social origin in the model. We graph the results in Fig. 10. The column for each country contains two sets of results. First, on the left, we present the predicted mean per capita household income value of color points (relative to color point 5) from models using color alone; on the right, we present results from models controlling for maternal education.

Figure 10

Predicted mean per capita household income by skin color and maternal education, LAPOP 2014. Source: LAPOP 2014. Adapted from Bailey et al. (2016). Notes: values are mean per capita household income for skin color categories. The mean per capita household income of skin color category five is the reference (0%) for each country. ‘M.Ed’ is ‘Mother’s education’

The data presented in Fig. 10 offers another compelling visualization of the relationship between skin color and household income across various countries. Most countries surveyed demonstrate a trend where lighter skin color is associated with higher per capita household income. Argentina and Mexico stand out for their linear correlation between skin color and income—suggesting a more consistent gradient where each incremental change in skin tone correlates with a stepwise change in income levels. In contrast, consistent with the analysis of the earlier data presented in Fig. 9, our results suggest reversals of the near-universal disadvantage of the darkest skin colors in a few countries, namely Costa Rica, Honduras and Panama.

A third pattern is demonstrated in some countries, where income disparities are not uniformly spread across the color spectrum. Instead, they display a form of clustering where certain ranges of skin tones do not show significant differences in household income. This is observed in Brazil, where the data points suggest that income levels are relatively similar among the darker skin tones, irrespective of the variations in skin tone. Such clustering implies that while skin color does impact income, the effect is only sometimes consistent or gradual. There may be thresholds or specific points along the color scale where income levels plateau, indicating other factors might be at play that either amplify or mitigate the economic impact of skin color.

The second subcolumn in Fig. 10 presents predicted household income by color point controlling for maternal educational achievement. Suppose social origins play an important role in mediating the effect of skin color in Latin America. In that case, we expect these color points to be substantially closer to the zero line than those in the first subcolumn. Contrary to that expectation, Fig. 10 suggests that controlling for maternal education does not substantially reduce skin color differences; the color differences with and without this control are relatively similar.

Figure 11 more succinctly summarizes how controlling for social origin affects color inequality in household income by graphing the relative change in the coefficients for color points when introducing maternal education as a control. Argentina shows the most negligible impact of social origin on the relationship of color to income inequality, with an average of roughly 2% difference between the gross effect of each color point and the effect of each color point controlling for maternal education. Reductions in the association of color with income inequality are similar in Brazil, the Dominican Republic, Uruguay and Nicaragua. At the other end of the spectrum, Fig. 11 shows that social origin has a greater impact on color inequality in El Salvador, Peru and Panama. Countries such as Chile and Costa Rica have relatively little color-based inequality, so the changes observed in these countries when maternal education is introduced should be interpreted cautiously.

Figure 11

Coefficient change for skin color after controlling for Mother’s education, LAPOP 2014. Source: LAPOP 2014. Adapted from Bailey et al. (2016). Notes: Values are relative change in coefficients for color categories after controlling for mother’s education based on OLS regression

The overall implication is that skin color has an independent effect on income inequality in Latin America, i.e. an impact that is not fully explained by social class as proxied by maternal education. This indicates that color-based discrimination and advantages likely play a role in economic disparities separate from the class-based differences that are also present and deep.¹¹ Although we believe that conceptualizing an ‘independent effect’ of skin color on income inequality in relation to social origin is useful and heuristic, we do not mean to reduce these input variables to independent impacts. Indeed, we would argue that skin color’s relationship to racial income gaps, as well as to social stratification across all domains, is more complex than the ‘independence’ assumption suggests. Durable and obstinate racial stratification is a type of ‘complex inequality’ that operates intersectionally with social origin and class, gender and other institutionalized factors to impact social structure (Wilde 2018). Nonetheless, the dominant tendency has been class reductionism, and our research narrative necessarily acts to contradict and project corrections by positing instead the racialized complexity of social stratification in Latin America.

Occupational inequality, census data

Our final SES measure of ethnoracial inequality in Latin America is occupational prestige. We rely on data from all national censuses that we could access through IPUMS and with the requisite variables. In particular, our dependent variable is the percent representation in administrative and professional occupations by ethnoracial category. The findings, summarized in Fig. 12, reveal notable disparities in occupational prestige among ethnoracial groups. Similar to the patterns observed in university completion rates (Fig. 3), individuals of Asian descent in Costa Rica and, notably, in Puerto Rico emerge as outliers, occupying positions at the high end of the spectrum. This suggests a positive correlation between educational attainment and occupational prestige.

Figure 12

Percent in administrative and professional occupations, ages 25–60. Source: Harmonized National Censuses for each ‘nation (year)’ accessed through IPUMS. Variable(s): ‘race’ and ‘occisco’ in harmonized IPUMS data for each ‘nation (year)’; race categories across nation (year): White, Black (including Black African, Black Caribbean, Afro-Ecuadorian, other Black), Indigenous (including American Indian and Latin American Indian), Asian (including Chinese, Japanese, Korean, Vietnamese, Filipino, Indian, Pakistani, Bangladeshi, other Asian), Mixed race (including Brown (Brazil), Mestizo (Indigenous and White), Mulatto (Black and White), colored (South Africa), two or more races). Occupational prestige categories across nation (year): (1) Legislators, senior officials and managers, (2) Professionals, (3) Technicians and associate professionals, (4) Clerks, (5) Service workers and shop and market sales, (6) Skilled agricultural and fishery worker, (7) Crafts and related trades workers, (8) Plant and machine operators and assemblers, (9) Elementary occupations and (10) Unknown

Our analysis also highlights Cuba as a unique case of minimal ethnoracial stratification in administrative and professional occupations, significantly diverging from the broader Latin American context. This striking anomaly can be attributed to Cuba’s state-managed labor market, constraints on income distributions, and the absence of significant color stratification across educational levels, among other factors. This example of the low salience of racial stratification in occupational prestige as measured through Cuban Census data suggests the striking influence of a complex web of systemic factors on social stratification in Latin America (de la Fuente and Bailey 2021).

In sharp contrast, Ecuador and Panama exhibit among the highest disparities in occupational prestige based on ethnoracial background within the region, yet they present distinct patterns of hierarchical organization. In Ecuador, the relatively small White population enjoys a significant advantage over the country’s indigenous groups, a gap mirrored in El Salvador’s socio-economic landscape. Conversely, Panama presents an unusual hierarchy where the Black population appears to lead in occupational prestige, as depicted in Fig. 12, markedly separated from the indigenous communities’ lower standings. This divergence emphasizes the unique socio-economic dynamics in Panama, where, notably, the census lacks a specific category for the White population, urging a cautious approach to interpretation.

The Panama anomaly

The case of Panama, or the ‘Panama Anomaly,’ presents a notable departure from the typical ethnoracial ordering observed in many Latin American countries, where Afro-descendant populations face significant relative socioeconomic disadvantages in contrast to non-Black and non-Indigenous populations. In contrast, Afro-Panamanians hold a relatively elevated position within Panama’s hierarchy of occupational prestige (Fig. 12), a distinction underscored by their higher average years of education and university completion rates compared to counterparts in the region (Figs 4–6). This divergence is further supported by the mean per capita household income analysis, where Afro-Panamanians and individuals with darker skin tones report higher income levels than other demographic groups within Panama (Figs 9–11).

This section delves deeper into the ‘Panama Anomaly’ by examining additional socioeconomic indicators derived from Panama’s census data, alongside specific traits unique to Panama’s Afro-descendant population and of Panama City itself. A component of this analysis involves the Panamanian Census’s detailed ethnoracial classification system, which allows for a nuanced understanding of the diversity within Panama’s Black population. This refined approach to classifying ethnoracial identities enables a more granular examination of the SES of Afro-Panamanians, shedding light on the factors that contribute to their distinctive position in Panama’s social and economic fabric.

The data presented from the 2010 Census in Table 1 first highlights the ethnoracial composition of Panama. The collective Black population, referred to as ‘All Negros’ in the first column, constitutes ~9% of Panama’s total population. This demographic is smaller than the Indigenous population, which accounts for ~12%. The largest demographic segment by far is categorized under a ‘residual race’ classification, comprising nearly 80% of the population. This statistical category encompasses non-Black and non-Indigenous Panamanians, likely representing a blend of individuals typically classified as White and mestizo in other Latin American contexts.

Table 1

Panamanian socioeconomic indicators by race, 2010

	Distribution	University degree	Mean income	Mean years of schooling	% Urban	% In Capital city	% Self-employed	% Professional and administrative occupations
All Negros	8.92	22.00	4582.91	11.65	88.13	36.79	19.95	21.04
Negros coloniales	2.32	25.87	4248.25	12.21	89.27	39.20	17.64	23.07
Negros antillanos	1.93	26.77	4881.94	12.47	92.70	41.60	18.25	23.41
Negros	4.22	16.80	4647.66	10.88	85.15	33.19	22.62	15.07
Negros otros	0.45	27.45	4408.47	12.14	85.42	35.09	16.45	22.64
Indigenous	12.2	2.85	5262.83	5.59	43.43	22.71	40.65	5.89
Residual Race	78.88	19.03	4048.23	10.46	70.67	29.15	24.11	17.11

Source: 2010 census of Panama.

The SES of Black Panamanians, as detailed in Table 1, sets them apart within Panama’s social structure. On average, Black Panamanians possess higher levels of university attainment and mean years of schooling, higher mean income and greater occupational prestige, compared to both the Indigenous population and other non-Black Panamanians. Results in Table 1 using Panama’s 2010 Census and the three SES measures—education, income and occupation—are supported by the findings from our other datasets and alternative angles presented above.

In addition, the data in Table 1 highlight two other pivotal factors often associated with higher SES in Latin America: urban residency and capital city habitation. Urban areas often provide a concentration of resources, services and opportunities less readily available in rural areas, facilitating greater social and economic mobility for those living there. Place of residency is particularly relevant in Latin America, where there can be stark contrasts between urban and rural quality of life (Torche and Lopez-Calva 2013). Data in Table 1 show that Black Panamanians are more likely to reside in urban areas than those in the residual category, with 88.1% living in cities versus 70.7% of non-Black and non-Indigenous individuals.

Urban compared to rural residency might further amplify advantages for residents of Latin America’s capital cities, where access to resources, networks and opportunities notably surpasses that of non-capital cities. Indeed, Table 1 shows that Black Panamanians exhibit a higher propensity to live in Panama City, at ~38 compared to 29% of Panamanians in the residual majority category. However, the impact of this capital city-centric divide in Panama on SES might be even more significant than in other Latin American countries due to some of Panama City’s unique characteristics.

According to Sigler (2013), Panama City is a unique transit zone in the global capitalist economy. It occupies a place even beyond ‘top tier’ urban hierarchies around the globe, partly by exogenous (US) design and partly due to its geography (the Isthmus of Panama). Sigler conceptualizes Panama’s capital as a ‘relational city’ alongside, e.g. Dubai: Panama City is an intermediary space with a niche intermediary role between the North and the South, a platform for international services, capital flows and offshoring, and an expatriate business elite. Panama City could even be viewed as the middleman of the Western hemisphere, shaped by a long history of exogenous influence and designed with an exogenous orientation.

The Panama Railroad (completed in 1855) was the first permanent interoceanic link, and American businessmen and engineers of the Panama Railroad Company spearheaded it. The mammoth project included the recruitment of a significant number of Afro-Caribbean laborers, to whom we return below. However, even more impactful toward the eventual making of Panama City into a relational city today was the territorial establishment of the US-sovereign Canal Zone in 1903. The Zone was established shortly after Panama became independent from Colombia (through significant US support), and the US President appointed its American governor. In addition to the interoceanic canal itself (completed in 1914), the zone also had military bases, towns and schools and was a de facto American enclave within the country of Panama. These projects and settlements necessitated significant migrant labor and employed many Afro-Caribbean descendants of the earlier migrant stream.

The building of Panama City into one of the world’s few ‘relational’ cities for global capitalism provides context for understanding Black Panamanians’ urban-centric lifestyle today, particularly their concentration in Panama City. We make the connection by leveraging the nuanced ethnoracial categorization scheme provided by the Panamanian Census. It differentiates between Antillean Blacks (descendants of earlier railroad and canal migrant laborers), colonial Blacks (negros coloniales—presumably descendants of formerly enslaved peoples), and others (including those uncertain of their Afro-descendant origin). We observe the pronounced levels of urban residence and capital city habitation of Antillean Blacks. Table 1 shows that 92.7% live in urban areas compared to the majority/residual population’s 70.2%, and 41.6% of Antillean Blacks reside in Panama City compared to 29.1% of the population of the residual category. Their spatial location as a product of historical forces suggests an explanation for their relative higher SES across outcomes.

Additional elements may help explain the relatively higher SES of Black Panamanians beyond urban residency and capital city habitation. Antillean Blacks, or ‘negros antillanos’, as descendants of early migration waves, may benefit from the originating migrant selection process. The characteristics of migrating West Indians probably positioned them favorably in terms of labor market demands and social mobility channels. English-speaking skills, e.g. may have been a part of that favoring mix given the significant US presence and the city’s exogenous orientation.

This analysis suggests that the historical migration from the West Indies has been instrumental in shaping Panama’s socioeconomic and ethnoracial fabric. The demographic composition of this migration provided Black Panamanians of Antillean descent with distinct relative socio-economic advantages. Such findings highlight the enduring legacy of migration on Panama’s ethnoracial landscape and underscore the broader implications of migration patterns on socio-economic stratification within the region.

Public policy to combat ethnoracial inequality

This final section focuses on public policy for combating ethnoracial inequality. We narrow in on the case of disparities in higher education attainment. Despite the rise in higher education enrollment in Latin America over the past two decades (Rivas 2015; Mainero 2020), significant ethnoracial disparities remain and necessitate targeted approaches. In the United States, affirmative action, particularly in universities, increased the size of the Black middle class. Still, racial quotas, a particular version of affirmative action, have been ruled unconstitutional by the US Supreme Court. Nonetheless, other affirmative strategies have persisted and continue to build on past gains. Affirmative action is a recent development in Latin America, and Brazil is an exemplary case and includes racial quotas. In Brazil, racial quotas exist alongside class-based quotas and are particularly robust, even in comparison to the United States. Moreover, despite some resistance to race quotas, research on public opinion reveals strong and consistent support across all ethnoracial categories for race-targeted affirmative action (Bailey 2004; Bailey et al., 2018).

In 2001, the first university racial admissions quotas were legislated in Rio de Janeiro, Brazil. State deputies passed a law establishing a 40% quota in state universities—the State University of Rio de Janeiro (UERJ) and the State University of North Fluminense (UENF)—for ‘self-declared Blacks (Negros) and Browns (Pardos)’. The year before, the state legislature of Rio de Janeiro passed a law allocating 50% of the slots in the same universities to students from public schools (a proxy for low income). By 2005, 24 of the 95 public universities in Brazil had adopted various affirmative action policies in admissions, some based on class, some on race and some on both. By 2007, that number had grown to 37, and as of 2012, 73 had some type of affirmative action in admissions. Then, in 2012, Congress passed the Quotas law, requiring that all federal universities adopt class- and race-based university quotas. The same year, in a historic decision, the Brazilian Supreme Court unanimously ruled that quotas for Black and indigenous students were constitutional (Telles and Paixão 2013; Peria and Bailey 2014). This decision is consistent with the Brazilian Constitution; it addresses the need to reduce social inequalities and specifically suggests affirmative action as one approach (Telles and Paixão 2013).

Changing ethnoracial inequality in higher education

Although ever-increasing data on Brazil’s novel and expansive race-targeted affirmative action policies in higher education reveals their success on several levels, scholars continue to study their impact in challenging historical inequality. In the hope of contributing to that conversation, we explore university attendance rates from 1992 to 2021 using data from the ‘Pesquisa Annual de Amostra de Domicilios’ (PNAD) for 18–25 year olds.

The results of our analysis in Fig. 13 show that in 1992, our baseline year, <10% of individuals self-classified as preto and pardo attended college, compared to 30% of individuals self-classified as White and 45% as Asian.¹² The percentage of Whites, pretos and pardos in the university remained relatively flat throughout the 1990s. In the early 2000s, which marked the beginning of racial quotas in Brazilian universities, the percentage of Brazilians attending college began to increase.

Figure 13

Percent in university by race in Brazil from 1992–2021, ages 18–25. Source: Pesquisa Nacional Para Amostra de Domicilios (PNADs), 1992–2021

The most significant gains, however, were for self-classified Whites. In 2003, 47% of young White Brazilians attended or graduated from college, compared to 16% of pardos and 14% of pretos. Thus, in absolute terms, college attendance had increased by 17 percentage points for Whites compared to ~6 to 7 percentage points for individuals self-classifying as preto and pardo. Even though racial quotas got most of the attention (and controversy), class-based quotas and an overall expansion of universities and university slots better explained significant gains in Brazilian higher education in the early 2000s.

From 2004 to 2015, Fig. 14 shows that college education continued to rise across all populations between 2004 and 2015, and the White-nonwhite gap appeared to close slightly. In 2015, 48–49% of pardos and pretos were college-educated compared to 70% of Whites. Thus, despite the 20% White-nonwhite gap, overall college attendance rates for pardo and preto youth increased significantly across time (from college attendance at a rate of <30% of Whites to roughly 70%). Note that in 2012, the Brazilian government mandated racial quotas for all federal universities on an annually growing scale to maximize the allotments in 2015. Overall, results in Fig. 14 suggest that university expansion and affirmative action (by race and class) in the first two decades of the 21st century produced significant increases in higher education attendance for all Brazilians, though ethnoracial disparities remain. Half of Black and Brown college-age Brazilians now attend college, a considerable gain from 10% two decades ago.

Conclusion

The study aimed to examine social stratification in Latin America by dissecting racial and ethnic dimensions. It utilized a range of measures and data sources for ~20 countries within the region, illustrating the vast diversity in racial composition—from the predominantly White population of Argentina to the mestizo and Indigenous majorities in Guatemala and Ecuador and the mainly Afro-descendant demographic in Brazil. The core of the analysis focused on disparities in education, income and occupational status among different ethnoracial groups. Census data, accessed through the Integrated Public Use Microdata Series (IPUMS), typically showcased entrenched racial hierarchies, with White or ‘mixed-race’ individuals (non-Indigenous and non-Black) often occupying advantaged socioeconomic positions. At the same time, Black and Indigenous peoples frequently occupied the lower end of the spectrum.

Recognizing that census formats and content were heavily influenced by national political climates, the study also incorporated data from the LAPOP. This data, with its more comprehensive ethnoracial categorization—including White, mestizo, Indigenous and Black—served to standardize comparisons across countries. The findings from LAPOP aligned with the census data, indicating consistent racial stratification across Latin American countries.

The study acknowledged the considerable racial fluidity in Latin America, particularly in survey scenarios that rely on self-identification. This phenomenon introduces measurement complications, such as endogeneity, that exceed the scope of this research (see Saperstein and Penner 2012; Bailey et al., 2013; Muniz and Bailey 2022; Villarreal and Bailey 2020). In response to these real challenges, this paper’s research and analytic approach leveraged comparison and multiple SES metrics and ethnoracial measures, particularly in its inclusion of skin color, to more comprehensively assess ethnoracial identification and stratification and thereby help to mitigate some of the noted measurement and estimation limitations.

The investigation into skin tone yielded the most uniform evidence for a racial hierarchy across nearly all countries in the region. There was a clear positive correlation between lighter skin color and higher SES throughout Latin America, with few exceptions (Telles and Steele 2012; Bailey et al., 2014; Bailey et al., 2016). This pattern led to the characterization of Latin America as a ‘pigmentocracy’, where individuals with lighter skin tones are predominantly at the top of the social hierarchy, and those with darker skin are at the bottom (Telles 2014). Most notably, we sometimes find differences in inequality using skin color compared to standard, census-type measures of race. In particular countries, inequality based on different ethnoracial measures can and does vary; moreover, the results produced by contrasting measures of race can even appear counterintuitive, such as our findings in Panama, suggesting instances in which non-Whites have higher SES than Whites. We believe these empirical variations by ethnoracial measure and across countries flow from the fact that ethnoracial self-identification reflects not only an individual’s perceived phenotype or skin color, which likely predicts social treatment or discrimination, but also responds to factors beyond phenotype, such as political and cultural attachments, social desirability, ethnic assimilation and exposure to racial ideologies (Telles et al., 2015).

The prevailing model for interpreting inequality in Latin America has traditionally emphasized social class origins, often downplaying or dismissing race as an inconsequential aspect or simply a byproduct of class. However, our research, aligning with numerous recent studies, contradicts this perspective by demonstrating that racial and color disparities cannot be solely attributed to social class. Racial inequality emerges as a distinct and pervasive source of disparity across various societal indicators in Latin America, including education levels, income, and occupation. Our findings add to the mounting evidence that race and ethnicity are fundamental axes of the social stratification observed in Latin American societies. In sum, the stratifying effects of historical and contemporary ethnoracial exclusion across the Americas continue to overdetermine socioeconomic outcomes and well-being, especially for its Afro-descendant and Indigenous populations.

STUDY FUNDING

This article was written for the Latin American and Caribbean Inequality Review, funded by the International Inequalities Institute at the London School of Economics and Political Science, the Inter-American Development Bank, Yale University, and the Institute for Fiscal Studies. The views expressed are those of the authors and not necessarily of the funders.

CONFLICT OF INTEREST

The authors have no conflict of interest to disclose.

DATA AVAILABILITY

The data used in this paper are all publicly accessible. The census data can be downloaded via https://www.ipums.org, the Latin American Public Opinion Project (LAPOP) data can be downloaded via https://www.vanderbilt.edu/lapop/, the Pesquisa Nacional para Amostra de Domicilios (National Household Survey) (PNAD) data can be downloaded via https://www.ibge.gov.br/en/ and the Sistema de Avaliação da Educação Básica (National Basic Education Assessment System) (SAEB) data can be downloaded via https://www.gov.br/pt-br.

Footnotes

1

For the first, we use census data samples for IPUMS, and for the latter two measures, we draw on data from the Latin American Public Opinion Project (LAPOP). A skin color measure has been included in the LAPOP surveys since 2010 and is increasingly used to understand ethnoracial identification dynamics and social stratification across the Americas (Bailey et al., 2014; Telles and PERLA 2014).

2

Henceforward, we generally use the descriptive label ‘ethnoracial’ for ‘race and ethnicity.’

3

We exclude Argentina, Peru, and Venezuela from all other analyses based on census data. Nonetheless, we include these cases in analyses using LAPOP and PERLA data.

4

The inclusion of the term ‘Afro-Brazilian’ alongside the term negro may have contributed to the doubling of this population segment in LAPOP.

5

Cuba and Puerto Rico were not included in LAPOP.

6

The LAPOP data on the Dominican Republic uses the Mulatto category. Similar to mestizo in other countries, it tops the educational hierarchy. See Telles et al. (2015) on how selectivity in self-identification produces this anomaly.

7

The overall results presented in Figures 6 and 7, though bivariate, are supported by multivariable statistical analysis using LAPOP data and skin color on mean years of schooling (Telles et al., 2015, Telles and Steele, 2012).

8

The regression model controls for each of the other variables that are not being predicted (parental occupation, ethnoracial classification and skin color) and various sociodemographic variables (sex, age, community size and region).

9

We note that our models are not exhaustive due to data limitations and the complexity that our broad comparative approach introduces; the possibility of unobserved factors impacting our estimates is important to note here (and throughout our analyses).

10

See Torche (2014) and Hout (2015) regarding the use of parental educational achievement as a proxy for respondents’ social origin in modeling SES.

11

Given that our analysis used only one measure of social origin, mothers’ education, due to data limitations, their results should be considered conservative estimates of social origin effects.

12

The numbers for Asian and Indigenous populations may be too small for analysis, given the erratic dips and peaks over time. The vast majority (99%) of Brazilians self-classify as preto, pardo, and branco.