Cultural Foundations of Discrimination

The resilience of caste identities, as we claim, lies in their embedding within the cultural fabric of stratified societies.Footnote ⁵ Cultural mechanisms, such as endogamy, not only reinforce the hereditary nature of caste membership but also limit mobility across social strata, ensuring that caste distinctions remain an enduring feature of societal organization. Restricted mobility across caste boundaries is reinforced by the fact that, as a consequence of endogamy, family names become associated with particular caste groups. Surnames thus serve as easy and public markers of social status, facilitating the everyday enforcement of exclusionary practices.

Exclusionary cultural practices are perpetuated through informal institutions, such as family and community networks, which socialize individuals into accepting caste hierarchies from an early age. By framing caste distinctions as intrinsic to social harmony and moral order, these frameworks discourage challenges to the status quo. Members of stratified societies find themselves entangled in webs of interrelated ties that perpetuate notions of status ranking, making it difficult to challenge instances of caste discrimination. Even in modernizing societies, where legal and institutional reforms promote equality, cultural practices often remain resilient, operating beneath the surface to sustain stigmatizing norms (Srinivas [Reference Srinivas1966] 1995).

The fusion of stigmatizing norms with structural inequalities—such as disparities in educational opportunities, economic roles, and political representation—further entrenches caste hierarchies, creating a self-reinforcing social equilibrium (Betancourt and Gleason Reference Betancourt and Gleason2000; Hanna and Linden Reference Hanna and Linden2009). Under such conditions, even when caste-based distinctions are publicly acknowledged as socially archaic or unjust, members of these societies can find it difficult to defect from cultural practices that reinforce stigma. Higher-status individuals may feel compelled to comply with discriminatory norms to preserve their social standing and group identity even when they privately oppose such practices. Similarly, lower-status individuals may avoid openly challenging these norms due to the risk of social ostracism or economic penalties, further reinforcing the perception that caste-based discrimination is universally accepted within their society.

This mutual reinforcement of stigma and inequality creates significant barriers to dismantling caste discrimination. While cultural norms shape the collective meanings attached to a caste identity, they are reproduced through individual decisions made in contexts where social conformity is expected. The resilience of caste discrimination thus lies in the interlocking practices of exclusion that operate across multiple levels of social life. At the micro level, families regulate marriage and daily interactions through adherence to endogamy. At the meso level, community networks pressure individuals to comply through reputational sanctions. And at the macro level, symbolic codes and rituals normalize status hierarchies that are seen as reflecting the natural state of the world. Because such mechanisms operate simultaneously, caste discrimination can persist even when other egalitarian ideals begin to gain acceptance in society.

Compliance in Caste Maintenance

Caste discrimination persists because individuals often comply with its cultural norms despite their own personal beliefs or preferences. Individuals may choose to discriminate as a way of securing their own position within the hierarchy, especially when failing to do so could threaten their personal status. Moreover, if individuals derive part of their self-perception from membership in social groups, those who belong to higher-status groups should be especially likely to hold views that rationalize their privileged positions vis-à-vis lower-status groups. For those who benefit from these distinctions, maintaining the status quo ensures that their social standing, and their access to its associated privileges, remains uncontested.

Compliance with caste discrimination is particularly strong when individual choices are linked to collective outcomes. In the realm of marriage, for example, the stakes of upholding caste distinctions extend beyond the individual to the entire family unit. The choice to comply with norms of endogamy not only affects the couple but also their families, who may stand to gain or lose social prestige depending on whether the union upholds group boundaries. Individuals understand that noncompliance risks their own standing as well as the social and economic well-being of their broader family, including their future offspring. Individuals are therefore under significant social pressure to ensure that marriages align with caste expectations, since a transgressive marriage could jeopardize the entire family’s reputation, access to community networks, and even material resources. The everyday practice of such norms makes it challenging to dismantle caste-based discrimination even in modernizing societies (Dirks Reference Dirks2002).

Compliance with caste norms may thus reflect strategic adaptation to social incentives rather than deep personal commitment. Individuals may act in accordance with caste expectations in public or high-stakes settings because failing to do so risks social sanction. Some may uphold discriminatory norms despite privately opposing them, while others may conceal caste-based views they anticipate would be disapproved of in a more egalitarian public environment. These forms of strategic conformity—whether motivated by fear of exclusion, reputational concern, or conflict avoidance—can sustain the appearance of social consensus around the legitimacy of caste distinctions even when private beliefs diverge.

Providing socially desirable responses to meet situational expectations can create a veneer of equality that masks the persistence of caste hierarchies. In interpersonal interactions, individuals may tailor their behavior to present themselves favorably when engaging with social superiors or inferiors (Goffman Reference Goffman1959; Paulhus Reference Paulhus1984). Lower-status individuals may, for example, exhibit upward deference when interacting with higher-status counterparts, offering responses that reinforce the social order out of respect or fear of sanction. Conversely, higher-status individuals may downplay their caste-based views to project tolerance or avoid appearing prejudiced, particularly in contexts where overt discrimination would generate conflict. These socially desirable responses, shaped by context-specific expectations, can create a misleading impression of declining discrimination or mutual respect even as unjust social orders endure (Kuran Reference Kuran1997).

The persistence of caste distinctions ultimately depends on public conformity, regardless of individuals’ private beliefs. When individuals outwardly comply with caste norms they reinforce the social order by sustaining the appearance of consensus, regardless of whether they endorse those norms. This public performance of norm adherence can make caste boundaries appear to be less contested than they might be, thus reducing the visibility of inequality and limiting the potential for collective resistance. This dynamic enables caste hierarchies to adapt to evolving cultural norms without losing their core structures of exclusion. By understanding how flexible cultural frameworks shape outward compliance with caste norms, we can gain greater insight into the mechanisms through which hierarchical systems can persist even in societies formally committed to equality.

The Structure of West African Caste Systems

Today, at least 15 ethnic groups across West Africa have caste structures, encompassing an estimated 90 million people. Scholars have long described these caste structures as a hierarchically organized system of social status determined by birth, rooted in historical occupational roles and reinforced by endogamy norms (Gueye, Audibert, and Delaunay Reference Gueye, Audibert and Delaunay2018; Irvine Reference Irvine1973; Markovitz Reference Markovitz1970; Mbow Reference Mbow2000; Todd Reference Todd1977). While variations exist in the specific categories and terminologies used among different ethnic groups, these hierarchies share significant structural similarities (Tamari Reference Tamari1991). For example, the caste system of the Wolof, the largest ethnic group in Senegal, is composed in descending status of nobles (garmi), peasants (geer), occupational castes (niégno), and the descendants of slaves (diaam). This final category traces its origins to domestic slavery. Enslaved individuals were typically acquired through warfare, and slave status was inherited matrilineally. Precolonial conflicts produced a substantial enslaved population in central Senegal and along the Senegal River, particularly in the homelands of the Wolof and the Peul (Moitt Reference Moitt1989).Footnote ⁷

Although the French abolished slavery in West Africa in 1905, the category did not cease to exist in the public imagination because of its inherited status and transmission through family. The occupational caste category as a whole is considered socially inferior to nobles and peasants, but there are also perceived status distinctions among various occupational subgroups such as blacksmiths and jewelers (teugue), leatherworkers (oude), and weavers (laabé). The griots (gewel) are typically considered to have the lowest status among the occupational castes (Irvine Reference Irvine1973). In everyday usage in Senegal, members of various occupational groups are typically grouped together into one category casté and contrasted with “freemen” (non-casté), which includes both nobles and peasants (Villalón Reference Villalón1995). Such caste systems exist among all major Senegalese ethnic groups with the exception of the minority Diola (around 5% of the population). There are thus very few Senegalese who do not have a place in the caste system.

West African caste systems are fundamentally descent-based, meaning that an individual’s status identity is determined by birth rather than occupation. Even if a person no longer practices the trade historically associated with their caste, their status remains fixed. A person descended from blacksmiths, for example, is still considered part of the blacksmith caste regardless of whether they possess the skills of the trade and regardless of their actual occupation.Footnote ⁸ Even a highly educated government minister or university professor can be commonly referred to as a blacksmith or jeweler. In this way, West African caste systems meet the anthropological definition of “caste”: hereditary membership obtained at birth through endogamous practices, ranked status rooted in historical occupations, and norms of purity (Tamari Reference Tamari1988; Todd Reference Todd1977).Footnote ⁹ Senegalese and West Africans themselves regularly use the language of caste in French to denote status distinctions among themselves (Izsák Reference Izsák2016).

West African caste hierarchies are reinforced by a suite of sociocultural practices, such as ritual complementarity and residential segregation. But strong endogamy norms have perhaps done more to allow status-based differentiation to persist, since patronymic associations easily facilitate the enforcement of status discrimination at the individual and communal levels (Richter Reference Richter1980). In Senegal, Wolof family names like Diop or Fall would be widely recognized as being of noble origin, while Mbaye or Mboup would be understood to be names linked to occupational castes (Wright Reference Wright, Arens and Karp1989). Such family names can signal caste status at first glance, enabling the enforcement of social boundaries in everyday interactions.

Even when family names are of ambiguous ranking, additional details such as place of origin or family history can clarify an individual’s social status. For example, among the contemporary descendants of slaves, some of their ancestors may have attempted to conceal their inherited status by adopting the family names of former masters and migrating to urban centers (Villalón Reference Villalón1995). Yet their lineage will typically remain known and remembered within their communities of origin, perpetuating the stigma across generations (Moitt Reference Moitt1989). As noted by Irvine (Reference Irvine1973, 74), “even if a slave denies his station and moves away to escape it, nobles still remember who their slaves were; and the bond remains residually because the slave or his descendants can come back and reinstate it.” While out-migration provided descendants of slaves some opportunities to hide their status, it did not eliminate it entirely and not everyone chose to leave. The possibility of concealing one’s status was even more limited for members of occupational castes, whose names often provide clear cues about their status.

Caste in Contemporary Senegalese Society

While caste structures persist as purely informal institutions without legal recognition in contemporary West Africa, in Senegal they affect about 90% of the population who belong to ethnic groups with such hierarchies. Inherited status continues to regulate social interactions that remind individuals of their relative standing in society and instruct them to behave accordingly (Diop [Reference Diop1981] 2012; RADDHO 2012). Descendants of nobles are incentivized to maintain the system as it grants them access to public esteem, if not power and resources. Members of occupational castes, though subordinate, may also comply with the strictures of the caste system because it defines their place within the community and provides some cultural roles that shield them from complete exclusion. Griots, for example, still enjoy a socially recognized role as oral historians in many public and government ceremonies.

The impact of caste discrimination is most evident in marriage. Even in a rapidly changing Senegalese society, marriage across caste lines remains controversial (Babou Reference Babou2009; Irvine Reference Irvine1973). Endogamous marriage is justified by many as being essential for ensuring the quality of offspring, echoing long-standing beliefs about lineage and behavior. People of noble origin are thought to inherit traits such as honor (kerse in Wolof), resulting in a perception that those who belong to this status group are more likely to possess and pass on qualities such as self-restraint, courage, and dignity.Footnote ¹⁰ By contrast, many Senegalese believe that members of occupational castes “lack personal dignity, manners, etiquette and decorum” (Dilley Reference Dilley2000, 154). Consequently, families will go so far as to hire genealogists to investigate the lineage of potential spouses and thus ensure their children are not marrying lower-status partners, especially from occupational castes (Arté and Baozi Production, France24, and RFI 2016). As a result, intercaste marriage in Senegal is far less frequent than marriage across ethnic groups.

Focus groups conducted with university students in Dakar, the capital, indicate that even educated youths expect their families to discourage intercaste relationships, possibly disowning them for breaking social norms.Footnote ¹¹ A higher-status woman stated that she could never marry a gewel because of the social stigma that she would encounter in her village. A woman from an occupational caste confessed that she felt that her marriage prospects were constrained by her identity. Another participant had a personal experience to prove this point: his girlfriend’s family was against their relationship because he was a “blacksmith.” The biggest obstacle in such unions is the fate of the offspring. Our focus-group participants explained that a child from an intercaste marriage might not be socially accepted by their families or communities. As one participant noted, many people simply believe that endogamous unions will produce better children.Footnote ¹²

Prejudices rooted in beliefs about caste identity produce a broader set of social constraints, including access to land and burial, and the ability to participate in public life (Auriol and Demonsant Reference Auriol and Demonsant2012). Silla (Reference Silla1966, 746–48) describes a case in Senegal’s Thiès Region in the 1960s in which griots were buried in the community cemetery at the insistence of a local Catholic priest. When a terrible harvest occurred the following year, the villagers not only attributed the spoiled harvest to the burial of the griots, but then also chased many of the remaining griots away from the village. Such beliefs are not relegated to the past. Fifty years later, residents of another village in the Thiès Region similarly refused to bury a griot in the community cemetery, insisting instead that she be interred in a baobab tree, the customary burial location for griots to avoid bad luck for the community (RFI 2021).Footnote ¹³

Caste discrimination also manifests itself in everyday civic life by limiting acceptable public behaviors. Such prejudice includes notions that it is inappropriate for individuals from occupational castes to speak up in public. The famed Senegalese director Ousmane Sembène weaves expectations about caste behavior into some of his films. In Moolaadé, when a female griot speaks up at a village gathering, she is quickly interrupted by the village chief and reminded that she, as a griot, should know her place and should not speak (Sembène Reference Sembène2004). Similar forms of discrimination continue to occur in many other locations across Senegal. An employee of Trust Africa, a nongovernmental organization fighting social discrimination, recounted instances when people “would not express themselves freely when they are from lower castes.”Footnote ¹⁴ In some communities, villagers from occupational castes are not allowed to pray in the front rows of the mosque (RADDHO 2012, 4). And some persist in such discrimination even when it is costly to them. When a USAID program sought to promote health awareness in the Kaffrine Region, many villagers refused to attend because the facilitators hired to run the sessions belonged to occupational castes.Footnote ¹⁵ Caste status similarly affects educational opportunities. Gueye, Audibert, and Delaunay (Reference Gueye, Audibert and Delaunay2018) found that increased school attendance by children from occupational castes decreased attendance by nobles’ children, suggesting nobles’ reluctance to have their children mingle with those from lower-status families. Given such treatment, many individuals from occupational castes consider out-migration as a way to escape the stigma attached to their background (Auriol and Demonsant Reference Auriol and Demonsant2012).

Despite recognition by anthropologists and historians that caste-based identities continue to shape social interactions across West Africa, political scientists lack systematic knowledge of how caste shapes people’s lived citizenship. This is due, in part, to the unspoken nature of this hierarchy. Systematic observation and analysis of caste in West African societies is challenging because it is generally considered inappropriate to directly discuss a person’s caste status in public. In Senegal, speaking about caste is close to a social taboo, and it is almost always done behind closed doors. As the country’s first president, Léopold Sédar Senghor, noted in this article’s epigraph, he considered caste but never spoke of it. This reflects the approach that most Senegalese take to caste: they are acutely aware of its existence but rarely mention it. The Senegalese state similarly acts as if caste does not exist, collecting no data on this form of identity and keeping little to no written record of it.Footnote ¹⁶ Caste is so thoroughly invisible to casual observers that it would be easy to visit Senegal without realizing that caste hierarchies undergird even the most basic social interactions.

Sampling Strategy

We followed a multistage strategy to reach our 2,160 respondents. Because we focus on identifying whether caste bias exists and, if so, what mechanisms drive it, we targeted the two largest ethnic groups of Senegal, Wolof and Peul, for our sampling frame, rather than seeking a nationally representative sample.Footnote ¹⁹ As a consequence, we cannot speak to the generalizability of our findings to the country as a whole. We sampled one hundred enumeration areas (with approximately 20 respondents per enumeration area) among the most populous departments that are dominated by the Wolof and Peul: Saint-Louis, Dagana, Podor, Louga, Kébémer, and two populous suburbs of Dakar, Pikine and Guédiawaye. Because these departments have both rural and urban locations, we are able to construct a sample that is half urban and half rural. The samples from Pikine and Guédiawaye are exclusively urban, while the remaining departments have a mix of rural and urban samples.

There are no official statistics on the percentage of the Senegalese population belonging to lower-ranked groups in the caste hierarchy. Informal estimates place the number around 10–15% (Tamari Reference Tamari1991). To ensure adequate statistical power for analyzing their experiences, we oversampled this population. In our final sample, approximately 31% identify as members of occupational castes or as descendants of slaves and 57% as members of higher-status peasant and noble groups.

Our sampling strategy differed across rural and urban areas to account for context-specific patterns of residential segregation. Rural areas are more likely to have residential segregation, with members of occupational castes often living in their own villages, or, if not, in specific neighborhoods. For this sample, we compiled a list of villages where the last name of the village chief indicated the likelihood of most residents being from occupational castes, since only such villages traditionally appointed chiefs from these backgrounds. We confirmed and revised this list through consultations with local officials in each department. We then created parallel lists of predominantly occupational and nonoccupational caste villages and randomly selected from each to ensure representation at the departmental level. For the urban sample, where residential patterns are more fluid, we contacted local officials in each city to identify neighborhoods with known populations from occupational castes. Since urban areas rarely have neighborhoods that are exclusively populated by occupational castes, our goal was to increase the likelihood of reaching them. We then randomly selected neighborhoods to sample from both occupational and nonoccupational backgrounds. Again, in both rural and urban areas, we randomly selected villages and neighborhoods to produce samples representative at the departmental level.

The sampled population is limited to adults of voting age in Senegal with equal shares of women and men. The casted population in rural areas tends to live in small, geographically clustered communities, and their households are also often clustered into larger compounds. We therefore sampled those households in proportion to the total number found across compounds within a village. The nonoccupational caste population in rural areas is much larger, and we sampled respondents by dividing each village into four sections. Sampled households were selected by choosing a household from the fifth compound in a random walk procedure. In urban areas, sampling followed the same random walk procedure as in the noncasted rural villages. For all locations, respondents were randomly selected at the household level from among voting-age men and women after seeking approval from the head of the household.

Our sampling strategy allows us to gain insight into the views of Wolof and Peul Senegalese who belong to occupational and nonoccupational caste groups as well as a small number of slave descendants. Although we cannot make generalizable claims about the country’s population as a whole, the results offer the first systematic analysis of Senegalese attitudes toward caste that we are aware of. Summary statistics on the demographic profile of our respondents can be found in table 1. Descriptive statistics for all variables utilized in this paper can be found in the supplementary materials. Data and replication files can be found in Arriola, Koter, and Wilfahrt (Reference Arriola, Koter and Wilfahrt2025).

Table 1 Summary Statistics of Respondents’ Demographic Profiles

Note: Education measured from one to 11, ranging from no formal schooling to university completed. A three thus represents some primary school.

Quasi-Random Assignment of Cross-Caste Interactions

We introduce one element in our survey design that approximates randomization: whether a respondent was interviewed by someone from their own caste or a different one. Because the enumerators introduced themselves at the beginning of the interview using their full names in accordance with local norms, we expect that the respondents could make a reasonable inference about the enumerator’s caste status.Footnote ²⁰

When respondents are surveyed by an enumerator from a different caste than their own, we expect such interviews to subtly prime awareness of social hierarchy. In most of Senegal, where caste structures are embedded within ethnic groups, social interactions involve a process of greeting that allows participants to effectively rank each other within the status hierarchy (Irvine Reference Irvine1973, 293). In this context, cross-caste pairings for survey interviews offer an opportunity to assess whether hierarchical awareness affects individual responses.

Since enumerators from occupational and nonoccupational caste backgrounds worked together in mixed teams—and were not assigned to respondents based on caste—the likelihood that a respondent was interviewed by an enumerator from an occupational caste was effectively exogenous to the respondent’s own caste status. Although this setup does not meet the criteria for true randomization, it approximates quasi-random assignment. When a survey team arrived in a neighborhood or village, all enumerators followed the same random walk procedure to identify households, regardless of their own caste identity.

Evidence of quasi-random assignment can be found in difference-of-means tests reported in the supplementary materials. We find no statistically significant differences across caste status in the likelihood of being interviewed by an enumerator from an occupational caste. We find that 9.5% of respondents from nonoccupational caste backgrounds and 7.5% from occupational backgrounds were interviewed by such enumerators. This is a statistically insignificant difference.Footnote ²¹

We build on existing work on enumerator effects in survey research. Survey respondents have been shown to systematically alter their responses when their own ethnic (Adida et al. Reference Adida, Ferree, Posner and Robinson2016) or racial (Campbell Reference Campbell1981) identities overlap with those of enumerators or, more diffusely, when enumerators’ self-presentation signals things like religious piety (Blaydes and Gillum Reference Blaydes and Gillum2013). Scholars generally assume that respondents alter their answers out of a desire to appear agreeable or due to social desirability bias, whereby respondents seek to present themselves in line with norms of social appropriateness even when their true beliefs may deviate. These effects are expected to be particularly large in questions sensitive to identity characteristics. For example, coethnicity response bias should be particularly prevalent in questions pertaining to ethnicity (Adida et al. Reference Adida, Ferree, Posner and Robinson2016).

Concretely, quasi-random assignment allows us to examine whether an enumerator’s caste identity affects how respondents answer questions pertaining to caste relations in Senegal. Although we are unable to directly estimate the true preferences of respondents, systematic differences in survey answers depending on the enumerator’s caste would be indicative of caste’s continued salience in social interactions in Senegal. In this context, we interpret variation in survey responses as indirect evidence of the social desirability pressures generated during interpersonal dynamics between individuals with different caste rankings.

Estimation Strategy

We present results from a series of ordinary least squares (OLS) and logit models for the empirical analysis. We report robust standard errors clustered at the enumeration area for each. Peasants, the largest number of respondents in our sample, serve as the baseline category in all models.Footnote ²²

All models control for several preregistered control variables, including characteristics associated with less traditional values such as a respondent’s education level, their socioeconomic status,Footnote ²³ and urban residency. In interviews and focus groups, three other factors were commonly noted as being associated with more conservative values on questions of caste: older generations, women, and ethnic Peul are expected to hold more traditional views about caste, with women invested in marriage markets and the Peul seen as generally more conservative overall. In contrast, it is often argued that religiosity mitigates attitudes toward caste. In particular, it has been anecdotally claimed that devout Muslims are less likely to embrace and maintain caste-based traditions.Footnote ²⁴ We further control for a respondent’s social hierarchy orientation. This is an additive index of how strongly respondents agree with the statements that some individuals or groups are more worthy, more deserving, or inferior to others.

We note at the outset that there is no relationship between caste status and socioeconomic outcomes. As presented in the supplementary materials, members of occupational castes do not have statistically different outcomes in socioeconomic status or a measure of the quality of their water access. This suggests that caste is not simply proxying for class, in contrast to South Asia, where these are highly correlated. In the supplementary materials, we explore how socioeconomic status may condition our results, finding that where there are differences, wealthier respondents often report more conservative attitudes. We further show that members of occupational castes score lower on education outcomes.Footnote ²⁵ Respondents from peasant and noble backgrounds, on average, have completed primary school, while respondents from occupational castes have an average outcome of incomplete primary education. This may be a product of occupational caste families opting to invest in occupational apprenticeships instead of traditional schooling.

Caste Discrimination Is a Social Reality

We begin by demonstrating the persistent reality of caste-based discrimination in Senegal today. Respondents in our sample overwhelmingly confirm that caste remains a relevant social identity in their lives. The most immediate evidence comes from the fact that 99% of our respondents could identify the caste status of their family name, a polite means of asking respondents about their own caste status.Footnote ²⁶ Furthermore, over 96% of respondents reported personally identifying with their caste status, reinforcing the salience of caste as a meaningful identity.Footnote ²⁷ Simply put, Senegalese recognize their caste-based identities with the same frequency as their ethnic ones.Footnote ²⁸

Table 2 reports the results of models estimating the relationship between caste status and three measures of perceived discrimination. Across all models, we find that members of occupational castes are significantly more likely to report experiencing discrimination due to their inherited social rank. Models 1 and 2 focus on perceived unfair treatment by asking respondents, “[D]uring the last year, how often, if at all, were you personally treated unfairly by other Senegalese on the basis of [respondent’s caste]?” Results show that respondents who descend from occupational castes are more likely to report unfair treatment compared with peasants, while nobles are significantly less so.Footnote ²⁹

Table 2 Caste Identity Affects the Likelihood of Discrimination

Note: Dependent variables listed in column titles. The reference group for all comparisons is “peasant.” Robust standard errors, clustered by enumeration area (EA), are in parentheses. Models 1 and 2 present coefficients estimated using OLS regression, while models 3 through 6 report odds ratios from logistic regression. * p < 0.10; ** p < 0.05; *** p < 0.01.

In models 3 and 4, we examine insults tied to caste status. Respondents were asked if they had ever been subjected to insults or slurs based on their family surname.Footnote ³⁰ The odds that respondents from occupational castes report such insults are over 2.5 times higher compared with that for peasants.Footnote ³¹ Interestingly, nobles are the most likely to report receiving insults based on their last name. This likely reflects cousinage or joking kinship (Dunning and Harrison Reference Dunning and Harrison2010), which is built around dyadic insults between patronyms. This finding reflects the nuanced role of hierarchical identities in shaping Senegalese social dynamics. While cousinage is often argued to reduce ethnic tensions, it is built, in part, on reciprocal insults across the caste hierarchy. These joking relationships, while superficially egalitarian, reflect deeper power imbalances that perpetuate, or at least mask, caste-based inequalities.

Finally, models 5 and 6 investigate whether respondents had ever been denied a service or right (e.g., a job, housing, transportation, education, government services, access to social events, or community participation) due to their caste status.Footnote ³² Respondents from occupational castes are nearly 1.8 times more likely to report such denial compared with those from peasant backgrounds.Footnote ³³ Individuals of slave descent have particularly acute experiences with discrimination. Models 5 and 6 reveal that descendants of slaves are 210% more likely to experience a denial of services or rights compared with peasants. These findings underscore both the material and nonmaterial consequences of caste-based discrimination. The fact that this same group does not significantly differ from peasants when experiencing name-based insults likely reflects the fact that many Senegalese of slave status adopted the family surname of their former masters after the abolition of slavery, obscuring their status in more anonymous public interactions.

Cultural Norms Sustain Caste Discrimination

Having established that caste-based discrimination persists in Senegal, we now turn to examining why individuals engage in caste discrimination and who is most likely to enforce status-based distinctions. To understand the underlying justifications for such distinctions, we asked respondents why people from different castes were historically kept apart in Senegal. The four most common responses were “Biological differences (blood) between groups,” “Economic competition,” “Religious instruction,” and “Cultural practice of our community.”Footnote ³⁴

As shown in figure 1, respondents overwhelmingly cited cultural practices as the primary driver of caste distinctions, with 80% attributing caste-based separation to cultural norms.Footnote ³⁵ This finding is important for two reasons. On the one hand, cultural practices are inherently mutable, suggesting potential levers for changing such norms.Footnote ³⁶ On the other hand, our field-based qualitative data suggests that cultural beliefs remain deeply embedded. For example, our focus-group participants frequently justified endogamy by expressing concerns that children from cross-caste unions would face familial or social rejection or limited life prospects. These sentiments highlight the durability of cultural practices that shape everyday decisions even among the urban educated.

Figure 1 Perceived Drivers of Caste Differences

Enforcement of Cultural Norms

If most Senegalese agree that caste distinctions are maintained through cultural practice, who enforces these norms? Table 3 examines respondents’ self-reported comfort with cross-caste social interactions. We find that nearly a quarter of respondents (24%) explicitly endorsed “the continued respect of practices that keep people from different castes apart. There are good reasons for doing so.” However, as models 1 and 2 of table 3 reveal, nobles are significantly more likely to support maintaining such practices compared with peasants. In contrast, in models with demographic controls, descendants of slaves are less likely to endorse exclusionary practices. For their part, respondents who ascribe differences to a biological basis—the most immutable of reasons—are the least likely to support challenges to the caste hierarchy.Footnote ³⁷ These findings suggest that endorsement of caste separation is strongest among the traditionally privileged.

Table 3 Comfort with Cross-Caste Relations

Note: Dependent variables listed in column titles. The reference group for all comparisons is “peasant.” Robust standard errors, clustered by enumeration area, are in parentheses. Models 1 and 2 report odds ratios from logistic regression, while models 3 through 6 present coefficients estimated using OLS regression. * p < 0.10; ** p < 0.05; *** p < 0.01.

To measure cross-caste interactions, we asked respondents how comfortable they would be in various scenarios involving individuals from other groups, such as allowing them into their home, sitting on their beds, or exchanging gifts or money. This set of interactions was workshopped with Senegalese respondents to identify areas of potential stigma. Analyzing an index of these interactions, models 3 and 4 of table 3 show that, compared with peasants, all other groups are systematically more comfortable with cross-caste interactions. These patterns suggest that norms of separation are not solely imposed by high-status groups but may also be reinforced by middle-ranking groups like peasants.

Models 5 and 6 in table 3 extend the analysis to broader questions about the acceptability of cross-caste interactions. We created an index of respondent answers to a battery of questions asking about the acceptability of members from different castes working together, living in the same neighborhood, being friends, and marrying. Consistent with earlier findings, nobles are significantly less accepting of cross-caste relations when compared with the baseline category of peasants, while slave descendants are significantly more accepting. Notably, we do not find similar results when the same questions were asked about cross-religious or cross-ethnic relationships. For example, nobles have a negative coefficient on every component of our caste acceptance index, but a significantly positive coefficient when asked about acceptance of cross-ethnic relations with the exception of marriage (see the supplementary materials). This suggests that caste hierarchies may be enforced through a distinctive set of social norms that operate independently from, and sometimes in tension with, broader norms of ethnic inclusion.

Policing Status Boundaries in Marriage Markets

Marriage markets are a focal point for the enforcement of caste-based norms. As shown in table 4, respondents from higher-status groups are more likely to discourage caste-based marriages, highlighting the centrality of endogamy in preserving the advantages of status in a caste hierarchy. When asked how they would advise their child if they revealed plans to marry someone from another status group, occupational caste and slave-descended respondents are significantly less likely to advise their child to end a cross-caste relationship compared with nobles and peasants.Footnote ³⁸ We illustrate the underlying descriptive differences in figure 2.

Table 4 Higher-Status Groups Enforce Marriage Restrictions

Note: Dependent variables listed in column titles. The reference group for all comparisons is “peasant.” Robust standard errors, clustered by enumeration area, are in parentheses. Models 1 and 2 report odds ratios from logistic regression while models 3 and 4 present coefficients estimated using OLS regression. * p < 0.10; ** p < 0.05; *** p < 0.01.

Figure 2 Opposition to Cross-Caste Relationships between Children

Note: Figure depicts the percentage of respondents from each caste who would advise a child to end a cross-caste relationship.

Models 3 and 4 of table 4 extend the analysis by measuring respondents’ actual exposure to cross-caste marriages within their social networks, ranging from zero (no known marriages) to four (marriages known in each of the following categories: immediate family, extended family, friends, and neighbors). The results indicate that occupational caste respondents report, on average, almost one additional type of cross-caste marriage within their social networks when compared with peasants. Nobles, in contrast, report fewer such relationships, indicating that the highest-status groups enforce stricter social boundaries in practice as well as in attitude. These results reflect a broader social reality revealed in the survey: while nearly 7% of respondents in our sample reported that their parents were from different ethnic groups, only 3% reported that their parents had a cross-caste marriage. This suggests that, in Senegal, caste may entail a stronger social boundary than ethnicity.Footnote ³⁹

The key takeaway across these analyses is that caste norms are more rigidly policed by higher-status groups, particularly nobles and peasants.Footnote ⁴⁰ Occupational caste respondents, by contrast, generally display greater tolerance for cross-caste interactions, as measured both in their subjective attitudes and in the composition of their social networks. This divergence in attitudes and experiences underscores a fundamental tension in the persistence of caste hierarchies. If norms surrounding marriage remain stratified—with higher-status Senegalese being less likely to engage in cross-caste interactions or marriages—then we can begin to understand why caste continues to shape Senegalese social life and constrain intergroup relationships.

Policy Solutions to Caste Discrimination

Does a respondent’s caste background shape their policy preferences? Table 5 presents the results for a series of questions that asked respondents whether they thought the Senegalese government should intervene in cases of caste-based discrimination. These scenarios include being denied a job, housing, the right to bury a relative, or being relegated to pray at the back of a mosque.

Table 5 Policy Preferences Regarding Discrimination

Note: Dependent variables listed in column titles. The reference group for all comparisons is “peasant.” Robust standard errors, clustered by enumeration area, are in parentheses. Models 1–8 present coefficients estimated using OLS regression. * p < 0.10; ** p < 0.05; *** p < 0.01.

The results reveal clear policy preferences across caste groups. Members of occupational castes are significantly more likely to support government intervention in cases involving jobs and housing, areas where state regulation might be expected. Respondents of slave descent are more likely to support government intervention on the question of mosque access. In contrast, nobles are more skeptical of the appropriateness of government intervention, particularly in more social spheres such as burial and, in the base models, mosques and housing. These identity-specific responses suggest a general reluctance among higher-status groups to endorse policies that could undermine the real-world implications of social stratification.

Additional analyses reported in the supplementary materials show that support for government intervention negatively correlates with stronger social hierarchy orientation. This suggests that individuals who endorse the legitimacy of caste hierarchies more strongly are also systematically less likely to support state interventions to address discrimination. These results point to a deeper tension in that caste-based policy preferences are not just reflections of material interest, but also of personal investments in the preservation or transformation of status hierarchies.

Cross-Caste Interactions and Norm Persistence

We extend the analysis by examining whether respondents adjust their answers when interviewed face-to-face by someone of a different or similar caste. Because traditional social interactions in Senegal often involve greeting rituals that signal and affirm relative status, we interpret cross-caste survey pairings as a subtle but meaningful priming of the caste hierarchy. The quasi-random assignment of enumerators allows us to assess whether these caste pairings—whether across or within caste categories—influence how respondents answer questions about caste relations. Although we cannot directly observe respondents’ true preferences, systematic variation in survey answers based on the caste of the enumerator would suggest that this identity remains a salient feature of social interaction. In this context, we interpret shifts in reported attitudes as indirect evidence of the social desirability pressures that arise during interpersonal encounters between individuals with different positions in the caste hierarchy.

Based on the coefficients associated with being interviewed by an enumerator from an occupational caste, as reported in table 2 and table 3, we find that respondents from noble and peasant backgrounds are significantly less likely to report discrimination and more likely to report comfort with cross-caste interactions. This is made evident in the results presented in figure 3, which suggest a tendency toward downward deference among higher-status respondents when engaging with lower-status enumerators in an effort to signal a degree of egalitarianism.Footnote ⁴¹ Conversely, respondents from occupational castes or of slave decent tend to exhibit upward deference by reporting lower comfort with cross-caste interactions when interviewed by a higher-status enumerator. However, we find that the effects reverse when respondents are interviewed by someone of their own caste. Respondents of noble and peasant backgrounds now report higher discomfort with cross-caste interactions and stronger support for social exclusion, while respondents of the occupational caste and of slave descent report more comfort with cross-caste relations.

Figure 3 Effect of Enumerator’s Caste Relative to Respondent

Note: The “lower-status respondents” category combines occupational caste and slave status; “higher-status respondents” combines nobles and peasants. Full model results can be found in tables 8.13–8.15 in the supplementary materials.

One notable divergence from the broader pattern of downward deference among higher-ranked respondents emerges in attitudes toward status norms. As shown in figure 3, respondents of noble and peasant backgrounds are more likely to endorse the statement that status differences should be respected when interviewed by an occupational caste enumerator, which is a reversal from their otherwise more egalitarian responses on measures like cross-caste comfort. One interpretation of this finding is that, when the status hierarchy itself is called into question, higher-ranked individuals may feel compelled to affirm its normative legitimacy. This suggests that respondents of noble and peasant backgrounds do not consistently signal egalitarianism in cross-caste pairings. Rather, their responses are calibrated to context-specific concerns about status. It is one thing to signal a willingness to engage with subordinates; it is another to endorse dismantling the very hierarchy that defines those relationships.

A similar sensitivity to status is evident with regard to marriage. As reported in table 4, respondents from noble and peasant backgrounds do not soften their answers to marriage questions when interviewed by occupational caste enumerators. They consistently express support for in-group marriage regardless of the enumerator’s caste, which is reflected in figure 3. In contrast, respondents from occupational castes and of slave descent express greater willingness to advise ending cross-caste relationships when interviewed by higher-status enumerators. These patterns underscore how marriage remains a cultural red line that is policed with particular intensity. Marriage appears to remain a domain where normative expectations are especially rigid, likely because the stakes of transgression are perceived to extend beyond the individual and carry collective implications for a family’s social standing.

The findings from table 5 show that cross-caste interactions systematically shift answers to policy questions. As shown in figure 4, the effect of being interviewed by an enumerator from the same or opposite caste is especially pronounced. Respondents of noble and peasant backgrounds consistently inflate their support for antidiscrimination government policies when interviewed by an occupational caste enumerator. This patterned response may, again, suggest a form of downward deference. Occupational caste and slave-descent respondents, by contrast, do the exact opposite when facing a higher-ranked enumerator by expressing significantly less support for government intervention. Strikingly, these effects reverse when the enumerator is of the same caste background as the respondent. We interpret these results as evidence that caste awareness is activated in face-to-face interactions, leading individuals to adjust their stated views in accordance with what they perceive to be socially desirable. This shift may reflect an effort by respondents to avoid offending or unsettling the enumerator by aligning their stated views with the desired attitudes they infer from the other person’s perceived social rank.

Figure 4 Policy Preferences: Effect of Enumerator’s Caste Relative to Respondent

Note: The “lower-status respondents” category combines occupational caste and slave status; “higher-status respondents” combines nobles and peasants. Full model results can be found in tables 9.4 and 9.5 of the supplementary materials.

We stop short of claiming that these patterns reflect the actual distribution of preferences, since our data cannot directly uncover respondents’ true preferences. However, the systematic variation in responses suggests a strong influence of social desirability bias around questions of caste. Occupational caste respondents tend to signal greater conformity with hierarchical expectations, while respondents from noble and peasant backgrounds emphasize more egalitarian attitudes. Regardless of their private beliefs, individuals often present a conciliatory front that sustains the appearance of consensus and inhibits disruption of the social order. This results in a stable social equilibrium that makes entrenched hierarchies more difficult to challenge.

Revisiting Political Mobilization

Prior research demonstrates that ethnicity becomes politically salient in the competition for resources and power (Arriola Reference Arriola2012; Bates Reference Bates1974; Lynch Reference Lynch, Cheeseman, Anderson and Scheibler2017; Posner Reference Posner2004). However, our findings suggest that the assumption of ethnic groups mobilizing as cohesive units must be reexamined in contexts where identity stratification exists. Social hierarchies within ethnic groups can generate additional layers of friction, impeding the manifestation of group solidarity or cohesion. For instance, when lower-status leaders dominate the emergence of an ethnic-based party, higher-status individuals may withdraw support to avoid subordinating themselves to social inferiors, undermining the party’s potential relevance. Analytical frameworks must therefore shift toward understanding how elites manage intraethnic divisions, navigating the competing priorities of status subgroups alongside ethnic-specific goals. This approach could illuminate previously unexamined dynamics, such as the strategic balancing of power between status groups to maintain broader group solidarity.

Revisiting Public Goods

Ethnic identities have been shown to shape public goods distribution, influencing both access to resources and the trust necessary for collective action (Casey Reference Casey2015; Lieberman Reference Lieberman2009; Miguel Reference Miguel2004). Yet our findings suggest that social hierarchies operate as an additional stratification mechanism, conditioning access to services such as education and healthcare even within ethnically homogeneous communities. For example, caste hierarchies can create disparities in preferences or deepen mistrust, challenging assumptions about policy or process interventions to foster inclusion. Mechanisms like information dissemination, often seen as a remedy for ethnic favoritism, may prove less effective in addressing exclusions rooted in hierarchy, as caste-based norms are more resistant to change. Future research might therefore explore the strategies that local actors might use to disrupt entrenched social hierarchies, such as targeted inclusion programs that address intragroup disparities within broader ethnic coalitions.

Revisiting Identity Salience

Studies of ethnic identification highlight the complexities of how individuals align with group identities amid shifting political dynamics (Eifert, Miguel, and Posner Reference Eifert, Miguel and Posner2010; Robinson Reference Robinson2024). Our findings suggest that hierarchical divisions, such as caste, further complicate these processes, reinforcing intraethnic exclusions even as broader societal changes occur. For instance, entrenched norms of exclusion may ensure that lower-status members remain disadvantaged by identity markers, such as their last names, even in more democratic or transparent contexts. This persistence suggests that modernization may intensify hierarchies within groups even if they serve to soften the boundaries between other identities. Mechanisms such as migration or urbanization could exacerbate stratification if higher-status groups are able to adapt more quickly to new environments while lower-status individuals face continued exclusion. Acknowledging these dynamics could refine hypotheses about identity formation, emphasizing the need to analyze stratified groups as internally heterogeneous entities.

Toward a Multidimensional Understanding

The analytical benefits of more explicitly incorporating social hierarchy into political science are substantial. Hierarchies like caste are often nested within ethnic groups, but they are not reducible to ethnicity. While nested identities can sometimes form the basis for new political coalitions (Ferree Reference Ferree and Chandra2012), caste typically functions as an internal fault line that can undermine ethnic solidarity rather than reinforce it. Theoretically, scholars can develop models that examine how elites within stratified groups navigate internal hierarchical divisions to maintain cohesion or assert dominance, and how marginalized subgroups mobilize in response to exclusion. Integrating insights from historical, anthropological, and sociological studies would enrich our understanding of how social stratification interacts with institutions. Empirically, incorporating measures of intragroup inequality would enable more precise analyses of how hierarchies intersect with ethnicity to shape political behavior. By combining these approaches, political scientists can advance a multidimensional understanding of identity that accounts for the impact of both horizontal and vertical cleavages in diverse societies.