Beyond Diversity: The Role of State Capacity in Fostering Social Cohesion in Brazil

Authors: Alexander Kustov and Giuliana Pardelli

Journal: World Development (2024)

DOI: 10.1016/j.worlddev.2024.106625

Resources

Figures & Tables

Abstract

A long-standing scholarship argues that greater ethnic diversity harms social cohesion. However, recent research also suggests that these outcomes are primarily influenced by the strength of state institutions. We evaluate these arguments using new geocoded historical data from Brazilian municipalities. Our initial analysis confirms that local racial diversity is negatively associated with social cohesion indicators such as trust, civic participation, belonging, turnout, and crime. Nonetheless, further analysis indicates that this relationship cannot be directly attributed to the effects of diversity, but rather hinges on the concentration of historically (dis)advantaged racial groups within particular areas. Finally, we demonstrate that both the spatial distribution of these groups and current levels of social cohesion are linked to past state capacity across municipalities. These results suggest that local social cohesion is more strongly associated with the historical development of state institutions across the national territory than with their contemporary levels of racial diversity. \ Keywords: Racial Demography, Diversity, Trust, Social Cohesion, State Capacity

Introduction

Do more ethnically and racially homogeneous communities have higher levels of trust and social cohesion? If so, can we attribute the observed variation in these outcomes across communities to the ethnoracial composition of their populations? The answers to these questions are crucial because they can help us understand whether the pressing social challenges experienced by contemporary societies facing high levels of immigration can be traced to the rising levels of diversity that accompany such demographic changes.

Previous research has revealed a negative association between ethnic diversity and social cohesion, encompassing a range of desirable outcomes such as interpersonal trust, civic engagement, and prosocial behavior [@Alesina2002, Costa2003, Putnam2007]. However, recent investigations have emphasized a more nuanced relationship between demographic heterogeneity and social cohesion, one that is significantly influenced by contextual factors [@VanderMeer2014, Dinesen2020]. Additionally, some scholars have raised questions about the very existence of a causal link or any discernible relationship between these two variables [@Portes2011a, Abascal2015a, Dragolov2016]. Amidst these advancements in research, a fundamental question remains unresolved: why is ethnic homogeneity associated with increased social cohesion for some groups but not for others?

This paper reevaluates the association between ethnoracial homogeneity and social cohesion by investigating a common antecedent factor that influences both outcomes. Rather than assuming a direct causal relationship between these variables, we propose that the link between demographic composition and social cohesion is influenced by the level of state capacity—understood as the material resources and organizational competencies that enable states to implement governing projects—accessible to different local communities. We argue that in contexts where the state has historically favored one identity over others—often due to the enduring legacies of slavery or colonialism—different groups may have encountered distinct obstacles and incentives to settle in various parts of the country. Consequently, a particular racial or ethnic group may have consistently enjoyed superior access to capable state institutions than others. When this occurs, homogeneity will be associated with better social cohesion for one group but not for others. By emphasizing the fundamental role of state institutions and their historical development, this perspective offers a potential explanation for the contradictory findings within the existing literature.

Our analysis relies on a new geocoded dataset covering 5,565 municipalities in Brazil, a country known for its significant racial diversity and low levels of trust. (Note: See, for example, “How does trust impact your quality of life” (Inter-American Development Bank, accessed on June 23, 2022); “30% of adults say most people can be trusted” (IPSOS, 24 March, 2022).) Drawing on historical and contemporary administrative sources, along with nationally representative surveys, our dataset combines information on social cohesion, racial demography, local fiscal capacity, and various socioeconomic covariates. We also use multilevel regression and poststratification to create original local estimates of various subjective social cohesion indicators, including generalized interpersonal trust, institutional trust, civic participation, and national belonging across Brazilian municipalities.

Our initial findings confirm the existence of a positive correlation between local racial homogeneity and social cohesion, consistent with previous research. However, further analysis reveals that racial homogeneity is positively associated with social cohesion only in white-majority municipalities. In contrast, no significant correlation is observed in municipalities with a (historically disadvantaged) Afro-descendant majority. These findings highlight the unequal “benefits” of homogeneity across different racial groups and the need to further understand the factors behind their uneven geographic distribution.

Next, we investigate the role of past state capacity in influencing the differences in demographic composition across municipalities. In the case of Brazil, Afro-Brazilians historically had compelling reasons to avoid regions with a strong state presence. While this strategy shielded these groups from excessive exploitation and coercion, it also deprived them of the trust-building benefits that robust state institutions can offer [@Rothstein2008]. Consistent with these expectations, our findings show that the historical presence of a more substantial state apparatus influenced both the spatial distribution of racial groups and the capacity of local governments to foster cooperation and engender trust within communities. In sum, these results suggest that social cohesion and racial homogeneity are both functions of past levels of local state capacity.

Our findings contribute to an extensive and diverse literature that underscores the influence of state institutions on the construction of ethnic difference [@Marx1998, Lieberman2012], settlement choices of different groups [@Pardelli2022, Trounstine2018b], and the emergence of trust and civic participation across communities [@Bustikova2017, Dell2018, Jensen2020, Abramson2022]. We expand upon these insights by suggesting that in contexts where the state sustains asymmetric relations with different groups, the correlation between homogeneity and social cohesion may be spurious for at least two reasons. First, homogeneity often coincides with a higher proportion of the dominant group, leading to the mistaken understanding that homogeneity itself generates favorable outcomes. Second, the spatial distribution of various groups across a country’s territory is not random but rather shaped by various historical forces, of which local state capacity is just one example. These forces influence the settlement choices of different groups over time, resulting in the demographic distribution observed across the territory in the present.

In summary, our findings resonate with the scholarship that cautions against making broad claims about the causal effects of demographic homogeneity and diversity [@Portes2011a, Abascal2015a, Pardelli2022]. Specifically, we emphasize the importance of disentangling the effects of demographic characteristics from those produced by unequal relations with the state when attempting to identify the benefits or challenges associated with specific racial compositions through observational data. This requires understanding the asymmetries that permeate the state’s interactions with various groups in society and the historical factors that explain why certain groups have come to settle in specific areas.

The structure of the article is as follows. We begin by providing a clear definition of social cohesion, outlining competing perspectives, and explaining how it is measured in the paper. Next, we provide a concise overview of the existing literature on diversity and social cohesion. Subsequently, we explore the literature examining state capacity, its measurement, and how it shapes both diversity and social cohesion. In the next two sections, we present the specific context of Brazil and describe the data sources and methods employed in the paper. We then present our findings regarding the link between diversity and social cohesion, investigating whether state capacity can emerge as a potential predictor for both variables. We conclude by discussing the implications of our findings for future research.

Although social cohesion is widely recognized as a desirable outcome, often described as “the glue that holds societies together,” scholars and practitioners commonly differ in their definitions and approaches to operationalizing it. (Note: We use “social cohesion” instead of the related, narrower term “social capital” to emphasize the collective nature of these outcomes [@Delhey2023]. Unlike social capital, social cohesion cannot be defined as an attribute of individuals.) Recent research has largely embraced a conceptualization founded on “organic” solidarity, which prevails in societies characterized by complex interdependencies, labor differentiation, and the extension of prosociality beyond close-knit networks [@Portes2011a, Abascal2015a]. This perspective focuses on the extent to which individuals adhere to universal norms of reciprocity and cooperation, rather than relying on ethnicity or other ascribed group membership as the primary basis for prosocial behavior. (Note: For a comprehensive review on different definitions of social cohesion, see @Dragolov2016.)

In line with this view, @Chan2006 define social cohesion as “a state of affairs concerning both the vertical and the horizontal interactions among members of society as characterized by a set of attitudes and norms that includes trust, a sense of belonging and the willingness to participate and help, as well as their behavioural manifestations.” This definition has several noteworthy features. First, rather than characterizing social cohesion as an attribute of specific groups, it focuses on communities or society as a whole. Second, it encompasses both vertical and horizontal interactions among members of society. In other words, interactions among different individuals and groups matter as much as the relationship between society and the state. Finally, this conception incorporates a wide range of both objective and subjective components.

Building upon recent advances in this tradition, we conceive of social cohesion as a multifaceted concept encompassing three fundamental aspects that define the quality of social cooperation within a collective? resilient social relations, positive emotional connectedness, and orientation toward the common good [@Dragolov2016, Schiefer2017]. Social relations encompass the horizontal networks that connect individuals and groups within a society. Connectedness, in turn, relates to the bonds shared among individuals and their affiliation with the social entity they belong to. An orientation toward the common good is exemplified through actions and attitudes that manifest a sense of solidarity and active involvement in the broader community. (Note: Importantly, our conceptualization of social cohesion does not encompass certain elements found in communitarian definitions, such as shared values or population homogeneity. These traditional conceptions, founded on “mechanical” solidarity derived from similarity, inherently imply a positive relationship between homogeneity and social cohesion by their very definition.)

Empirically, these three aspects of social cohesion are typically assessed through subjective and objective indicators. Social relations are often measured using variables such as interpersonal trust and the strength of social networks. Connectedness is evaluated by considering the strength of people’s identification with their social entity, feelings of belonging, and levels of institutional trust. An emphasis on the common good is evidenced by individuals’ willingness to abide by social rules and levels of civic participation [@Delhey2023]. While these dimensions often overlap, empirical research has demonstrated that they are not identical [@Janmaat2011], and their relationship to ethnoracial demographics is not self-evident.

A substantial literature in the social sciences has shown that an increase in the size and number of ethnic minority groups is associated with reduced social cohesion. Scholars have proposed various mechanisms to explain this negative association. Among these, the “homophily” principle and the related concept of “parochial altruism” suggest that individuals inherently prefer interacting with those who are similar to themselves [@Alesina2005, VanderMeer2014, Kustov2020a]. Given this assumption, mere exposure to individuals from different ethnic groups is anticipated to undermine social cohesion [@Dinensen2015], resulting in higher levels of ethnic diversity being associated with reduced social interactions, decreased civic engagement, and diminished trust in generalized others. Additionally, it has also been hypothesized that the linguistic or cultural differences of diverse populations can encumber intergroup communication and impede coordination [@Alesina2002, Habyarimana2009, Desmet2009], obstructing the emergence of generalized prosociality. These observations imply that greater homogeneity should result in increased levels of trust and cohesion, regardless of which group constitutes the local majority. However, previous studies have contradicted this expectation, with majority groups often being the only ones responding negatively to increased diversity [@Stolle2008, Fieldhouse2010, Abascal2015a]. This begs the question of what accounts for these disparities.

One potential explanation for these discrepancies hinges on the distinction between contextual and compositional effects. Prior studies have traditionally assumed that the effects of diversity are primarily contextual. Contextual effects imply that individuals’ generalized trust decreases as the proportion of outgroups in their environment increases. Consequently, it is expected that trust will diminish in more diverse settings, regardless of which group constitutes the local majority. However, this perspective does not adequately account for variations in the relationship between homogeneity and social cohesion across different groups. An alternative view proposes that the negative relationship between ethnic diversity and social cohesion may stem from the tendency of disadvantaged minority groups to be less trusting [@Smith2010]. The underlying idea here is that minority status itself can lead to lower levels of trust and reduced engagement. Since minorities often reside in more ethnically diverse areas, compositional differences in the makeup of local populations could explain the negative correlation between diversity and social cohesion. However, they do not elucidate why particular groups consistently display lower levels of trust, even when they constitute the majority and reside in homogeneous areas. This points to the possibility of other, unidentified factors influencing these trust dynamics.

The varying relationships between ethnic homogeneity and social cohesion among different groups may be partly explained by the intrinsic link between ethnic demography and levels of social advantage or disadvantage. As many scholars have highlighted, the observed effect of ethnic composition on social trust might be confounded by the prevailing socioeconomic conditions within the local area. If ethnic diversity typically coincides with socioeconomic deprivation, unconditioned diversity effects could be spurious due to unobserved contextual heterogeneity. (Note: For a recent meta-analysis of these and other similar findings, see @Dinesen2020. For critical reviews of this literature, see @Portes2011a; @VanderMeer2014; @Baldassarri2020.) However, other work has shown that, even after accounting for socioeconomic deprivation and other structural factors, the negative effects of diversity persist [@Twigg2010, Dinesen2020], and continue to vary across groups [@Fieldhouse2010, Stolle2008]. Thus, the lingering question is: what factors could elucidate the diverse effects of homogeneity among distinct groups? We propose a potential explanation, one that is rooted in the role of historical state capacity in influencing both local social cohesion levels and groups’ settlement patterns.

State Capacity as a Common Antecedent Factor

State and the Spatial Distribution of Ethnic Groups. A rich tradition in political science and sociology has studied the sources of ethnicity and its relationship with state formation and other historical developments [@Anderson1983, Horowitz2000]. More recently, scholars have highlighted the critical role of state capacity in determining the emergence and persistence of ethnic diversity, both within and between countries. The fundamental idea here is that robust states can effectively encourage cultural, linguistic, and identity homogeneity through a range of measures, including education and coercion. Consequently, contemporary ethnic heterogeneity can be viewed as an outcome of historically weak state capacity.

Scholars have employed the term state capacity in various ways across different disciplines. (Note: For insightful overviews of this extensive and burgeoning field, see, e.g., @Soifer2008, Hendrix2010, Cingolani2013, CKY2017, JohnsonKoyama2017, Berwick2018, Hanson2021, DinceccoWang2022, Suryanarayan2023.) Despite the diversity of definitions, these works share the central idea that state capacity relates to the state’s ability to implement its goals or policies [@Cingolani2013]. Most uses are in the sense of @Mann1984 calls `infrastructural power:’ the capacity of the state to penetrate society and implement political decisions throughout the realm. The unifying theme among these definitions is the focus on state capacity as a gauge of “potential” [@Williams2021], described as the “organizational and bureaucratic ability to implement governing projects” [@CKY2017], or as the resources that “enable states to get things done” and more efficiently implement policies [@BesleyDray2022]. As @Hanson2021 aptly summarize, “state capabilities include material resources and organizational competencies internal to the state that exist independently of political decisions about how to deploy these capabilities.”

Recent research has emphasized the ways in which the use of these state capabilities can shape the composition of national populations. @Wimmer2018, for example, shows that states with the ability to exert influence uniformly across their territory, offered incentives for minorities to adopt the language and culture of the dominant group, ultimately leading to a decrease in diversity over time. This perspective suggests that ethnolinguistic diversity is not an exogenously given phenomenon but rather a product of slow and gradual processes of assimilation. @Abramson2022 provide further support for this idea. Their findings indicate that regions characterized by frequent border changes, which disrupted historical state-building processes and constrained the state’s capacity to establish local administrative structures, exhibited higher levels of ethnic diversity.

Another strand of the literature has brought attention to the role of geography in both limiting the reach of the state and fostering diversity. @Michalopoulos2012, for instance, demonstrates that areas of the world characterized by greater disparities in land quality and more rugged terrain tend to exhibit higher levels of ethnolinguistic diversity. The geographical remoteness that results from the natural barriers of these areas allows for the emergence and persistence of greater ethnic and linguistic diversity by both limiting interaction among different communities and creating obstacles for state repression [@Carter2019]. Scott’s (Scott2014) seminal work on the “Zomia” region in Southeast Asia further reinforces this idea, highlighting that populations residing in hills and mountains display significantly greater cultural diversity when compared to those in densely populated valleys. The geographical inaccessibility of these areas made them natural regions of refuge for those who had a reason to flee the state or the group that controlled it. A crucial implication of these studies is that remote and inaccessible areas, characterized by either weak or absent state institutions, tend not only to exhibit greater ethnic diversity but also to house a larger proportion of groups that have contentious relations with the state.

However, it is important to note that the propensity of state capacity to provoke the widespread flight of specific populations extends beyond extreme scenarios where the state is actively engaged in suppressing or eliminating certain ethnic or racial groups. This effect can arise in response to various forms of state intervention, such as the implementation of land use regulations, changes in immigration policies, and the enforcement of vagrancy laws [@Huggins1985, Andrews1991, Trounstine2018b]. It suffices for state policies to have heterogeneous effects on different ethnic groups for regions with stronger state capacity? and therefore a greater ability to enforce these policies? to produce distinct incentives (or obstacles) for specific groups to relocate there. In such cases, the strategic decisions made by dominant and non-dominant groups can result in enduring disparities in the quality of state institutions accessible to each of them [@Pardelli2022].

State and Social Cohesion. A longstanding debate in the social sciences revolves around the role of the state in either fostering or undermining trust, and by extension, social cohesion. One perspective within the literature suggests that a strong state can function as a substitute for an active and engaged civil society [@Putnam1993, Fukuyama1995]. This view posits that the state disrupts traditional social relationships and informal institutions, increasing the reliance of local residents on the state and consequently diminishing the necessity for trust. It challenges the idea that effective state institutions are the essential underpinnings of trust and, instead, associates distrust with the state [@Gellner1988, Taylor1987].

An alternative perspective regards the state as a catalyst for social cohesion rather than an obstacle to it [@Tarrow1996]. According to this view, when the state is not efficacious, trust may indeed develop and persist within small, close-knit communities. However, without robust state institutions, it is less likely to extend to external groups and strangers [@RothsteinStolle2008]. As @Levi1998 emphasizes, the state is the sole entity capable of fostering generalized interpersonal trust—a form of trust that transcends in-group boundaries and encompasses society at large. Strong state institutions are expected to create an environment where trust can extend beyond the confines of close-knit friend or family groups and encompass a broader, more diverse society, ultimately enhancing social cohesion. (Note: Generalized trust connects individuals with those who are different from themselves, in contrast to particularized trust, which is confined to one’s in-group. As @Rothstein2005 clarify, “generalized trust reflects a bond that people share across a society and across economic and ethnic groups, religions, and races.” This concept is integral to various measures of social cohesion.)

Several foundational and recent studies have associated the roots of interpersonal and institutional trust with historical levels of state capacity. @Levi1998, for example, argues that the trustworthiness of a competent and honest bureaucracy serves as a fundamental element in fostering citizens’ cooperation and support for governmental policies. @Herreros2008 further specifies that it is the absence or inefficacy of the state that erodes trust within society. These arguments have found empirical support in a variety of settings. For instance, @Dell2018 show that citizens have been more successful in mobilizing through civil society to secure public goods provision in Vietnamese villages with a longstanding tradition of local bureaucratic institutions dating back to pre-colonial times. Focusing on the United States, @Jensen2020 provide further evidence that historical investments in formal capacity are strongly associated with contemporary indicators of social cohesion. @Abramson2022 demonstrate that in European regions marked by frequent border changes, which disrupted the development of interpersonal social networks and impeded state-building, individuals tend to exhibit lower levels of trust in others and in their governments. (Note: The connection between the quality of state institutions and trust has been substantiated through extensive empirical investigations. Nonetheless, the potential for a reciprocal relationship between these variables still exists [@Robbins2012]. It is crucial to emphasize that this possibility does not undermine our argument concerning the endogenous association between racial demography and social cohesion.)

Variation in state capacity can influence trust in state institutions through a variety of mechanisms. A robust state apparatus can foster trust by clearly defining and consistently enforcing rules and regulations [@Levi1998, North1990], thereby bolstering the perception of impartiality within state institutions. This trust is further deepened through the effective delivery of public services by a capable and dependable bureaucracy [@Bustikova2017]. The efficient use of public resources and their allocation to highly visible public goods contributes to increasing citizens’ confidence in the government’s judicious use of tax revenues. Similarly, a skilled and trustworthy bureaucracy underpins collaborative interactions between citizens and the state, thereby laying the groundwork for institutional trust.

Moreover, effective resource extraction often goes hand in hand with some degree of accountability, ensuring that public resources serve the collective interest. For this, as scholars in the fiscal exchange literature have highlighted, “credible commitment” mechanisms are essential [@Bates1985, Levi1988, Dincecco2011, Garfias2019, Jensen2023]. These mechanisms ensure the state’s appropriate use of tax revenues and foster a sense of quasi-voluntary compliance among citizens. Notable strategies for establishing such commitments include legislative assemblies [@North1989], elections [@Ferejohn1986], political parties [@Stasavage2007, Timmons2010], and an impartial public bureaucracy [@Miller2000]. Essentially, when the state is sufficiently constrained and the preferences of specific groups are reflected in governance, trust is likely to grow within those groups that feel represented. Consequently, as governance becomes more attuned to the needs of the broader population, as is characteristic of democratic settings, trust in state institutions is expected to expand across larger segments of society. As @Levi1998 highlights, “the major sources of distrust in government are promise breaking, incompetence, and the antagonism of government actors toward those they are supposed to serve.” In other words, citizens’ trust in the state is conditional on their belief that it will act in their best interest and implement fair procedures. Collectively, these elements contribute to building institutional trust, establishing a positive relationship between the efficacy of state operations and the level of trust it commands from the population.

The strength and effectiveness of state institutions also influence patterns of interpersonal trust. A robust state infrastructure fosters an environment that reduces uncertainty in human interactions and exchanges, cultivating expectations of reliability and commitment [@Levi1998, North1990, Farrell2003, Herreros2008]. @Rothstein2000 advances a theory in which trust is cultivated from the top down: the trustworthiness of efficient institutions engenders interpersonal trust, which subsequently promotes social capital. By contrast, in settings where the state is unable to effectively perform its core functions, such as collecting revenue, providing basic public services, maintaining law and order, and promoting social inclusiveness, (Note: Beyond the mechanisms outlined in existing literature, the capacity of the state to enact inclusive policies, ensure equality of opportunity, and provide impartial treatment constitutes another pathway through which state capacity can foster both institutional and interpersonal trust across the wider population. The tolerance for inequality tends to diminish when citizens perceive their chances for socioeconomic advancement as limited [@Hirschman1973]. Such perceptions of constrained mobility, in turn, can significantly undermine social cohesion, especially in the face of observable inequalities. In line with these observations, scholars have underscored the role of social exclusion in eroding the cohesiveness of communities and societies at large [@Letki2008, VanStaveren2017]. Thus, states that demonstrate greater effectiveness in confronting these challenges are likely to significantly enhance societal cohesion.) individuals may become more reliant on close-knit friend or family networks and relationships to satisfy their needs. This reliance can limit interactions with individuals outside of one’s immediate circle, reducing opportunities to build trust with a broader range of people. Essentially, in the absence of a functional state to uphold contracts and safeguard rights, citizens are more likely to hesitate before engaging in trust-based interactions. Overall, low state capacity can erode interpersonal trust by fostering an environment where skepticism of state institutions prevails and reliance on close family and friends become the norm, limiting opportunities for individuals to develop trust in a broader social context.

One important question concerns the choice of an appropriate metric for assessing variation in state capabilities and their potential effects on both institutional and interpersonal trust. State capacity has been described as encompassing three main dimensions: extractive, coercive, and administrative [@Soifer2008, Berwick2018, Hanson2021]. Extractive capacity consists of the state’s ability to secure resources. This dimension is pivotal, not only because the ability to impose and collect taxes is a fundamental requirement for the effective execution of virtually all other state activities [@Hendrix2010]; but also because efficient tax collection depends on the fulfillment of other crucial state duties, including the acquisition of accurate information, the recruitment of skilled personnel, the supervision of transactions, and the effective enforcement of regulations. Thus, the ability of local governments to generate revenue serves as a reasonable measure of their overall effectiveness in executing a broad range of essential activities. In light of this, our empirical investigation employs historical data on local per capita taxation as the primary indicator of municipal-level state capacity. However, recognizing the multifaceted nature of state capacity and aiming to ensure the robustness of our findings across different operational definitions, we also incorporate measures of local coercive and administrative capacities, such as the presence of law enforcement personnel and the size of municipal bureaucracies [@Garfias2018, Lee2019, Fergusson2022, Suryanarayan2021, Pardelli2024].

While enhancing extractive capabilities is crucial, such improvements may not immediately influence societal trust. Doubts about the integrity and efficiency of local state institutions, especially concerning their dedication to unbiased enforcement and prudent financial management, can temper the influence of state capacity development on trust levels in the short term. In line with this intuition, scholars have observed that changes in state institutions do not instantly alter citizens’ perceptions and behavior [@Rothstein2000]. Building credibility and trustworthiness is a gradual process, deeply anchored in the historical track record of state institutions. Therefore, in our investigation of the link between state capacity and social cohesion, we rely on historical data on fiscal capacity instead of contemporary figures. This approach acknowledges the incremental process through which state effectiveness cultivates societal trust.

Brazil is a highly unequal and racially diverse society that has been commonly found to have one of the lowest levels of interpersonal trust in the world [@Rothstein2008]. The conventional wisdom implies that the country’s considerable degree of racial diversity plays a significant, if not a predominant, role in shaping this outcome. However, recent research suggests this understanding may be mistaken given that the distribution of racial groups across the country is itself not random, but rather closely tied to the spatial distribution of historical state capacity [@Pardelli2022].

Brazil was the largest receiver of enslaved migrants during the Atlantic slave trade era and the last country in the West to abolish slavery in 1888. Between the 16th and 19th centuries, 4.8 million Africans were forcibly displaced to the country. During this entire period, enslaved populations actively resisted domination by Portuguese colonial authorities and, later, by Brazilian officials and slaveholders. One of the basic forms of resistance was the formation of independent, self-sustaining communities of fugitive slaves in the hinterlands. These settlements, called quilombos or mocambos were numerous, long-lasting, and widespread across the country—something that scholars associate with the openness of the Brazilian frontier [@Klein2009].

The establishment and persistence of quilombos provide an important example of the spatial segregation patterns that can emerge from the uneven distribution of state strength at the local level. In the case of Brazil, the existence of slavery and the associated brutal repression and exploitation of enslaved persons by dominant groups, bolstered by state support, prompted Afro-descendant communities to adopt self-isolation strategies and deliberately select regions with minimal state presence for their settlements.

More generally, any official policy of persecution, discrimination, or oppression can constrain the settlement choices of targeted groups, giving rise to a systematic association between racial demography and state capacity that can persist long after these institutions cease to exist. This is exemplified by fugitive slaves who had compelling reasons to avoid regions where the local government possessed a strong and effective apparatus. Consequently, homogeneous Afro-descendant communities were more prone to establish themselves in remote areas of the country where state institutions were practically nonexistent [@Pardelli2022].

1 The fact that the primary responsibility for financing local public services in Brazil was in the hands of municipal governments meant that without a robust fiscal apparatus capable of generating revenue effectively, local authorities could not fulfill their monitoring, sanctioning, or guarantor functions. If, on the one hand, this made hard-to-reach areas more attractive to groups seeking to avoid the grip of the state; on the other hand, it may also have deprived these communities of the trust-producing benefits engendered by strong state institutions.

In the following sections, we investigate this possibility and evaluate whether, in the case of Brazil, the alleged causal association between racial demography and social cohesion can be traced to common antecedent factors.

Data and Methods

We start our investigation by examining the relationship between local racial demography and social cohesion across municipalities. Specifically, we evaluate whether homogeneity is related to positive societal outcomes per se or only insofar as it reflects an increase in the share of the dominant group in the local population. We then explore whether the association between social cohesion and racial homogeneity (of the dominant group) can be traced to historical levels of local state capacity. Given its significance as a key outcome in this literature, we include the main results for interpersonal trust in our figures and tables alongside those for the general social cohesion index. (Note: Generalized interpersonal trust, also referred to as social trust, is perhaps the most commonly studied dimension of social cohesion. Scholars contend that this type of trust is particularly relevant in contemporary societies characterized by extensive interactions among strangers [@Dinensen2020], owing to its positive effects on cooperation [@Gachter2004].) We use an original dataset of 5,565 Brazilian municipalities as defined by the 2010 census—the lowest and socially meaningful administrative level at which relevant data are available—including contemporary and historical racial demography, social cohesion, state capacity, and geography variables. (Note: Note that, following previous research in this context, we exclude the country’s capital, Bras'ilia, from our analyses.)

Municipalities constitute a suitable unit of analysis for investigating the implications of local state capacity on trust and social cohesion in Brazil for several reasons. First, municipalities represent the most proximate level of government for citizens and are responsible for administering a variety of public services and local policies. As a result, the efficacy of municipal bureaucracy has a direct bearing on people’s everyday experiences, influencing their assessment of the trustworthiness of the average community member and their trust in institutions. Second, Brazilian municipalities possess significant autonomy in decision-making and resource allocation. Therefore, by examining municipalities, we can delve into the historical role of local bureaucratic effectiveness in fostering trust and specific demographic configurations. Lastly, a wealth of pertinent social, economic, and political data are collected and made available at the municipal level in Brazil. (Note: We acknowledge that, despite their smaller scale compared to states or regions, municipalities may still display substantial heterogeneity in terms of state institution availability and quality, as well as the demographic composition within their confines. Nevertheless, this internal variation should not undermine our analysis; if anything, it may lead to the underestimation of the association between state effectiveness and trust existing at even more granular geographic levels, such as neighborhoods or communities.)

Social Cohesion and Trust. To capture the three core aspects of social cohesion outlined earlier (relationships, connectedness, common good), we utilize a variety of objective and subjective indicators, employing state-of-the-art measurement techniques designed for both national and regional evaluations [@Dragolov2016]. (Note: Specifically, we employ seven distinct measures that align with the three core aspects of our adopted definition of social cohesion. First, social relations are gauged using a measure of interpersonal trust. Second, connectedness is evaluated based on the intensity of people’s feelings of national belonging and their trust in major institutions. Lastly, a focus on the common good is reflected in various indicators, including civic and political participation (e.g., voter turnout, the number of civil society organizations, self-reported civic engagement), as well as the willingness to adhere to social norms (crime rates).) We adapt these measures to the Brazilian context based on data availability at the local level. Most importantly, we create original municipal-level estimates of generalized interpersonal trust, institutional trust, (self-reported) civic participation, and feelings of national belonging by aggregating all available relevant survey data from Latinobarometro, LAPOP, and ESEB. These estimates are further refined using the multilevel regression with poststratification (MRP) technique. To obtain an objective measure of civic participation, we include IPEA data on the number of civil society organizations per capita across municipalities. We also incorporate two outcomes that are commonly viewed as direct manifestations of social cohesion (or its absence), namely, homicide rates and electoral turnout. To minimize measurement error, all of these measures are aggregated over the 2000-2020 period. Collectively, these variables aim to capture the essence of the “social glue” that binds communities together. To create a comprehensive measure, we thus aggregate these variables into a single index ranging from 0 to 1. This index is obtained by taking the average of all the components, which we then utilize in our main specifications (see Figure fig:cohesionmap). For details on the index construction and the MRP estimation, see Appendix.

Racial Demography. We measure local racial demography in two ways. First, we calculate an indicator of racial homogeneity using the inverse of the standard fractionalization index adopted in the literature—that is, we measure the probability that two randomly selected individuals in the community belong to the same group. Second, we use the share of white and Afro-descendant (Black and Brown) populations in each municipality. All measures are constructed using the 2010 census microdata. (Note: Racial and ethnic boundaries in Brazil are often described as fluid and seen as more nuanced and flexible compared to other contexts [@Davenport2020]. However, existing studies have indicated that this fluidity is more evident between Black and mixed-race populations, rather than between white and nonwhite [@Telles2002]. Our analyses are based on the white-nonwhite cleavage, therefore variations in self-identification should not influence our results. Additionally, @Pardelli2022 have demonstrated that there are no systematic differences in objective and subjective racial self-classification based on the level of local state capacity.)

Historical State Capacity Data. Our primary measure of historical state capacity at the local level focuses on the ability of municipal governments to extract fiscal resources. Specifically, our data reflects municipal tax revenues per capita in 1923. (Note: To map past levels of state capacity into modern-day administrative boundaries, we adopt the approach previously used to address changes in US county boundaries over time [@Acharya2016]. Following this method, the total amount of taxes per capita collected in 1923 is divided among the relevant municipalities in 2010 such that the proportion of taxes from 1923 municipality $i$ that is assigned to 2010 municipality $j$ is based on the size of their overlapping areas.) While high tax revenues can serve as a testament to local state power, they may not solely reflect the governments’ tax collection efficiency but could also be influenced by specific policy environments. Utilizing local tax revenue data for analysis mitigates these concerns, as such measures are generally less susceptible to the confounding influences of policy choices when compared to analyses based on national-level data [@Harbers2015]. To the extent that there is minimal variation in tax policies at the local level, this measure is more likely to reflect the capabilities of local governments in navigating the challenges of tax collection rather than differences in policy choices. Nonetheless, in an effort to ensure the robustness of our findings, we incorporate three alternative ways of operationalizing our primary independent variable: the number of law enforcement officials, representing the state’s coercive capacity; the extent of local railroad networks, reflecting the state’s infrastructural ability to reach different areas; and the count of local public officials, a vital input measure indicative of the state’s administrative resources. Notably, the latter serves as a particularly useful metric, as it embodies a fungible resource whose utilization does not depend on the policy orientation of current officeholders.

Control Variables. Our analysis integrates contemporary covariates that capture municipal characteristics known to impact both the racial composition of localities and social cohesion [@VanderMeer2014, Dinesen2020]. These characteristics include the size of the municipal area, demographic attributes such as the age distribution and education levels of the population, the rate of urbanization, local economic output, the degree of income inequality, and the poverty rate. Municipalities with larger areas and higher urbanization rates tend to attract more diverse populations and foster increased opportunities for social interactions. The age makeup of a population can correlate with racial identities and distinct social engagement patterns. Economic prosperity and higher education levels are often linked with enhanced social cohesion and may align with the demographic profiles of specific racial groups. Conversely, higher levels of inequality and poverty are known to erode social trust and tend to be more prevalent in areas with greater population diversity.

In our historical specifications, we incorporate a set of pre-treatment, time-invariant geographic covariates, such as altitude, rainfall, sunshine, proximity to the coast, and distance from the capital, which have been previously shown to influence our outcomes [@Naritomi2012]. These geographic features play a crucial role in shaping historical settlement patterns, economic activities, and infrastructure development, which in turn, can affect the extent of state reach, racial diversity, and social cohesion [@Pardelli2022]. However, it is important to note that the literature does not universally agree on the inclusion of control variables in studies exploring diversity and social cohesion [@Schaeffer2013]. The necessity and selection of controls are debated, with concerns that some may introduce bias if they are influenced by both ethnic diversity and social cohesion [@Cinelli2022]. To mitigate these concerns and enhance the transparency and robustness of our findings, we also present our main results without the inclusion of control variables (see Figure 2 and Table A2). Summary statistics for all variables are shown in Table A1.

Figure. Interpersonal Trust and Social Cohesion Index across Brazilian Municipalities

Analysis and Results

As Figure fig:cohesionmap illustrates, despite Brazil’s relatively low levels of interpersonal trust and social cohesion, there is a significant degree of variation in these outcomes across municipalities. On average, only 25% of respondents across the country say that most other people can be trusted. However, this average masks substantial regional differences, with certain municipalities in the South showing a majority of residents who trust others, while in some municipalities in the North, less than 10% of residents trust others. The question arises: can this significant variation in trust levels be attributed to differences in the racial composition of local populations?

Figure fig:cohesion depicts graphically our correlations of interest. The two panels in Column 1 show the association between local racial diversity and our two main outcomes—the original survey-based estimates of interpersonal trust and social cohesion. In line with the “diversity debit” literature, these figures show a strong positive association between racial homogeneity and social cohesion across municipalities. In Table A2, we regress each of our outcomes of interest on the homogeneity indicator. As the results show, the positive association persists even after we account for standard confounding factors, state fixed effects, and time-invariant geographic covariates.

The panels in Column 2 of Figure fig:cohesion depict the relationship between the white population shares across municipalities and our two main outcomes of interest. The positive correlations, in this case, are even stronger than those documented in Column 1. Table A3 confirms the robustness of this association to the inclusion of relevant controls reflecting local socioeconomic and geographic characteristics.

Figure. Social Cohesion, Racial Demography, and Past State Capacity

1 We then investigate the idea that it is not racial homogeneity per se that is related to positive social outcomes, but rather that homogeneity merely proxies for the share of whites across communities. If this is the case, then splitting our sample between majority-white and majority-nonwhite municipalities should yield positive homogeneity coefficients only among the former. In line with this expectation, the results in Table tab:demMINI Panels A and B show that homogeneity, unlike white population shares, is only positively associated with trust and social cohesion among majority-white municipalities. (Note: Table A4 shows the estimates for each component of the social cohesion index. The results are identical if we include an interaction of the demographic variables and a 50% binary threshold indicator for white shares in the full sample of municipalities (not shown).) This pattern is in line with previous findings in the US where out-group contact is found to undermine trust only among whites [@Abascal2015a]. Furthermore, Table A5 indicates that, when one includes both homogeneity and white group shares in a regression, the former is not statistically significant across most specifications and the latter always has a significantly larger coefficient.

These results are relevant theoretically not only because they change our expectations about which mechanisms might be at play—given that the channels through which homogeneity affects cohesion differ from those behind the effect of relative group sizes—but also because they suggest that self-selection into specific geographic areas may play an important role in explaining the observed link between racial demography and social cohesion.

Table. Racial Demography and Social Cohesion

The Role of Historical State Capacity

The results above do not provide support for the idea that racial homogeneity increases social cohesion; they do, however, show a strong association between racial demography (white population shares) and our outcomes of interest. Homogeneous white and nonwhite communities differ in important ways. Group shares can thus capture differences along a whole set of dimensions that may themselves influence local levels of social cohesion. As discussed above, these patterns might emerge from groups’ differential sorting into environments that are more or less propitious for the development of trust and social engagement. In this section, we investigate the plausibility of this interpretation, while considering the compositional and structural characteristics of communities.

Figure fig:cohesion, Column 3 provides descriptive evidence of the positive association between past levels of local state capacity and our main outcomes of interest, social cohesion and interpersonal trust. (Note: Figure A1 in the Appendix illustrates the associations between homogeneity, white population proportions, and historical state capacity with each of the dependent variables examined in this study.) Table tab:pastpresent further explores this relationship and shows that local taxes per capita 100 years ago explain approximately 8-20% of the variation in these contemporary outcomes. These relationships persist even after we include state fixed effects and account for a range of geographic controls. In line with the results obtained in previous work [@Pardelli2022], columns 5 and 6 in Table tab:pastpresent show that tax revenues per capita in 1923 are also associated with significantly higher proportions of whites across Brazilian municipalities today.

Table. Social Cohesion, Racial Demography, Past State Capacity

In order to assess the robustness of our findings, we utilize three alternative measures of local state capacity in Table tab:capacityALT. These measures capture non-fiscal aspects of state strength, specifically its administrative and coercive dimensions, as well as an indicator of its geographic reach throughout the territory. The specifications in Panel A use the size of the state’s bureaucracy across municipalities in 1920, whereas Panel B adopts the number of law enforcement officials as an independent variable (both variables were obtained from the 1920 Census). In Panel C, we use the number of railroads within each municipality as an alternative (even if somewhat imperfect) proxy for the local reach of the state. As the estimates in Table tab:capacityALT indicate, our results remain substantively unchanged. Municipalities that had a larger bureaucracy, a stronger coercive apparatus, and more accessible territories a century ago currently exhibit improved social cohesion outcomes and higher proportions of white residents.

Table. Social Cohesion, Racial Demography, Past State Capacity (Alternative Measures)

Finally, although social cohesion is by definition a collective outcome, we recognize that some scholars may also be interested in trust as an individual-level outcome. Consequently, we replicate our results at the individual level using the Latinobarometro data on generalized interpersonal trust (see Table A6). Remarkably, respondents’ propensity to trust others is strongly related to historical state presence in their locality, even after we account for racial identity and other demographic characteristics.

Conclusion

Examining the impact of ethnic diversity and demographic changes on social cohesion and trust is of great interest to social scientists and policymakers. Our paper makes two contributions to our understanding of this relationship. First, using both historical data and original contemporary social indicators across Brazilian municipalities, we demonstrate that the previously observed negative correlation between these variables is spurious, manifesting only when racial groups are considered to be interchangeable and their historical interactions with the state are assumed to be uniform. Second, our research reveals that past state capacity is a strong predictor of both racial homogeneity among dominant groups and increased levels of contemporary social cohesion. These findings offer valuable insights that complement recent research on the nuanced, and often spurious, relationship between ethnic diversity and the provision of public goods [@Gisselquist2016, Wimmer2018, Kustov2018, Pardelli2022] as well as the endogeneity of diversity and trust in the U.S. context [@Abascal2015a].

Overall, our results highlight the significant differences between local communities with diverse ethnoracial compositions, not only in terms of their contemporary socioeconomic makeup, but also their historical relationship to state institutions. These differences pose a serious challenge to the research on ethnic diversity and social cohesion that overlooks the historical asymmetries that have permeated the relationship between the state and different ethnic or racial groups in society. This oversight is exacerbated by the reliance on cross-sectional analyses that obscure the historical sorting of disadvantaged groups into relatively more remote regions, and of dominant groups into areas with a stronger state apparatus. (Note: It is important to note that local state capacity may itself result from the influence of other factors. This possibility, however, does not detract from our core argument regarding the biased distribution of groups across space, and the resulting endogeneity of racial demography and social cohesion.)

Moreover, our findings suggest that the benefits commonly associated with demographic homogeneity are most effectively harnessed when there are at least minimally capable state institutions in place. In the absence of such institutions, homogeneity plays a limited role in enhancing social cohesion. Notably, these findings do not contradict established research indicating that homogeneity can enhance reciprocity norms and sanctioning opportunities within communities. Nonetheless, they underscore how historical disparities in access to the state may impede certain groups from enjoying these advantages. This insight emphasizes the need for empirically exploring the specific mechanisms through which enhanced state capacity, particularly at the subnational level, can facilitate the development of social trust and cohesion within communities. Investigating these channels offers a promising path for future research.

Lastly, our analysis indicates that the often-touted advantages of homogeneity might actually stem from deeper structural conditions rather than demographic composition alone. Although our study does not rule out that ethnic and racial demography may have an independent effect on trust and cohesion, it highlights that the demographic composition of communities can itself be intertwined with distinct historical experiences and interactions with the state. Simultaneously, local state capabilities have a non-negligible effect on the perceived trustworthiness of the generalized other. Taken together, these insights suggest that government policies solely focused on preserving homogeneity or preventing demographic changes may not necessarily result in discernible benefits when it comes to promoting greater trust and social cohesion among their residents.

Online Appendix\

Beyond Diversity: The Role of State Capacity in Fostering Social Cohesion in Brazil

roman

Details of Variables and Data Sources

Social Cohesion: Subjective Indicators. We create original municipal-level estimates of generalized interpersonal trust, institutional trust, (self-reported) civic participation, and feelings of national belonging by aggregating all available relevant survey data (Latinobarometro, LAPOP, and ESEB) and improving the resultant estimates with the MRP technique. (Note: Our estimates are based on publicly available microdata. The replication code for our MRP estimates can be shared upon request.) While interpersonal trust and many other indicators of social cohesion are by design self-reported survey measures, none of the existing surveys provide representative population samples at the municipal level. The multilevel regression with poststratification (MRP) method allows for the construction of reliable trust estimates for each municipality within Brazil by adjusting estimates from (nationally representative but locally non-representative) samples with information about the population distributions using linked census data. MRP is commonly used for adjusting biases in estimating aggregate preferences at the level of regions, districts, and municipalities [@Tausanovitch2013]. While MRP is widely acknowledged as the “gold standard” for estimating aggregate preferences in small areas, it is rarely applied outside of the US due to the lack of detailed regional data and its computational intensity [@Leemann2017]. The rapid increase in computational resources and the increased availability of survey and census data over the last decade, however, allows us to provide reliable municipal estimates for trust and other self-reported aspects of social cohesion.

We apply the MRP method to estimate the degree of interpersonal trust and other self-reported social cohesion indicators across Brazilian municipalities using all available nationally representative survey data from the last two decades and the last 2010 census. First, we identify and compile all high-quality surveys that asked relevant questions (e.g., for interpersonal trust: “Generally speaking, would you say that most people can be trusted, or that you can’t be too careful in dealing with people?”). After aggregating all relevant survey waves from the Latin American Public Opinion Project (LAPOP), Brazilian Electoral Study (ESEB), and Latinobarometro available in the 2000-2020 period, we obtain more than 38,000 answers to this question from respondents across the country alongside their sociodemographic characteristics (e.g., age, gender, race, education, etc.). Importantly, even with this amount of data, many municipalities only had a few respondents, precluding any representative estimation of their preferences without the MRP. To address this concern, we initially merged these responses with comprehensive census microdata detailing the relative prevalence of various respondent characteristics across municipalities (such as the proportion of educated individuals in each municipality). Subsequently, we employed multilevel linear regression techniques to estimate how individuals’ trust levels may vary based on their other characteristics (e.g., age, education, etc.). We then used this information to refine our municipal-level trust estimates. For instance, municipalities with a higher proportion of educated residents were expected, on average, to exhibit higher levels of trust compared to municipalities with a lower proportion of educated residents. Finally, we used an objective indicator of municipal crime rates (a common component of interpersonal trust and social cohesion) to validate our novel estimates across Brazilian municipalities. The specific wording used for various social cohesion component measures was as follows:

Interpersonal trust: “Generally speaking, would you say that most people can be trusted, or that you can’t be too careful in dealing with people?”
Institutional trust: “To what extent do you trust? the municipal government / federal government / political parties / Congress / Supreme Court / Police / Armed Forced?” (an average of all questions)
Civic participation: “In order to solve your problems, did you ask for any help or cooperation? any federal, state representative or senator / any local authority / any ministry, public institution, or state body?” “Now I will read a list of groups and organizations. Please say whether you attend meetings at least once a week, once or twice a month, once or twice a year, or never? a religious organization / a parents’ association from school / a neighborhood’s association / a union / a party?” (an average of all questions)
National pride: “To what extent are you proud to be Brazilian?”

Social Cohesion: Objective Indicators. To obtain an objective measure of civic participation, we include IPEA data on the number of civil society organizations per capita (log) across municipalities. We also include two outcomes that are commonly viewed as direct manifestations of social cohesion (or its absence), namely, homicide rates (Note: Specifically, our variable reflects the number of homicides per 100,000 respondents, which has been previously shown to be a good indicator of crime in the Brazilian context [@Ishak2022].) and electoral turnout. (Note: We calculate this by dividing the number of votes cast in Brazil’s general elections by the municipality’s electorate size.) To minimize measurement error, we aggregated all of these measures over the period from 2000 to 2020. These variables collectively represent the “social glue” that holds communities together. We created a single 0-1 index by averaging all the components, which is used in our main specifications.

Racial Demography**. All demographic variables were constructed using microdata from the 2010 Census from the Instituto Brasileiro de Geografia e Estat'istica (IBGE). Our main indicator of Afro-descendant shares is based on the sum of two Census categories (pardos* and pretos), but all results are robust to considering individual categories as separate groups.

*State Capacity**. The data on fiscal capacity across municipalities in 1923 were entered from Brazil’s 1926 statistical yearbook of public finances. We obtained data on the count of public administrative and law enforcement officials in 1920, as well as information on the presence of public officials in 1872 from the respective census records.

Control variables**. The data come from the National Institute of Geology (INGEO) and include *Area size (logged), Altitude, Distance to Coast (logged), Distance from the Capital (logged), Rainfall, Sunshine, Latitude and Longitude. Some specification also include age and education structure, urbanization, local output, interpersonal inequality (GINI), and poverty rate. For details, see @Naritomi2012.

Tables and Figures

Atable

Table. Descriptive statistics (n = 5565)

Afigure

Figure. Social Cohesion, Racial Demography, and Past State Capacity

Table. Racial Homogeneity and Social Cohesion Indicators

Table. White Population Shares and Social Cohesion Indicators

Table. Racial Demography and Social Cohesion Indicators

Table. Racial Homogeneity, White Population Shares, and Social Cohesion

Table. Individual Interpersonal Trust and Past State Capacity