Measures of Ethnic Politics in the Electoral Arena

Ethnic identity conventionally refers to any descent-based identity, including tribe and language (e.g., Horowitz Reference Horowitz1985), with measures of ethnic politics focusing on the extent to which ethnic identity explains, and predicts, vote choice. Ethnic politics in the electoral arena can be disaggregated into two intertwined phenomena: ethnic voting, namely voting for a candidate or party of the same ethnic background; and the existence of ethnic parties, or ethnic candidates, namely those that garner a disproportionate share of their electoral support from their coethnics. In this sense, ethnic candidates and parties are the opposite of national political actors with broadly representative, multiethnic electorates. The existence of ethnic parties is not determined by whether an entire ethnic group votes for the same party but by whether a party relies disproportionately on the support of a single group.

Conventionally, a party dominated by a single ethnic group is viewed as an ethnic party whether it explicitly seeks to represent only a specific ethnic group or not. This approach is important in the African setting,Footnote ² given that the formation of parties based on ethnic identities is outlawed in many African countries. Parties and candidates across Africa typically make discreet verbal appeals to ethnicity during campaigns without leaving a paper trail of their intention to favor a certain group. In this context, looking for overt appeals to ethnicity, instead of studying parties’ distribution of support, would most certainly undercount the number of ethnic parties in Africa.

Scholars have in many ways improved the operationalization of Horowitz’s concept of an ethnic party, namely one with an overwhelming share of its support coming from a specific ethnic group, by classifying ethnic parties and ethnic candidates in terms of “degree” and not just “either/or” coding, which avoids establishing arbitrary thresholds.Footnote ³ Such continuous measure of ethnic politics is at the basis of several indices, such as Cheeseman and Ford’s (Reference Cheeseman and Ford2007) ethnic polarization and ethnic diversity of political parties, Elischer’s (Reference Elischer2013) Party Nationalization Scores (PNS), and Dowd and Driessen’s (Reference Dowd and Driessen2008) Cramer’s V Ethno-linguistic Voting Index (CVELI). Despite subtle differences, these measures generally lead to similar conclusions because they all use similar data to assess the relationship between voters’ ethnic characteristics and vote choice. One limitation of these off-the-shelf indices is that they typically consider one dimension of ethnic identity, such as language, and evaluate the association between that particular identity and vote choice even though we know that ethnic identity is multidimensional in most African societies (Laitin Reference Laitin1986; Posner Reference Posner2005). Studying ethnic politics in African countries requires considering the politicization of different dimensions of ethnicity, such as language or tribe (Posner Reference Posner2005).

The field has produced an empirically rich catalogue of manifestations of ethnic voting, harnessing both large quantities of existing electoral data and producing new experimental data. The co-existence of these approaches is productive because it allows to compensate for shortcomings of either type of data taken separately. Aggregated electoral data mapped onto ethnic census data often present ecological fallacy problems, making it difficult to attribute electoral patterns to individual voting behavior.Footnote ⁴ Individual-level data, including public opinion polls, can provide a more accurate link between ethnic identity and vote choice, but they are unfortunately comparatively limited across Africa.Footnote ⁵

These data limitations are probably one of the reasons why experimental work has become so common among scholars studying ethnic politics in Africa. Experiments generate data that can reveal ethnic political preferences at the individual level – for example, when respondents are required to assess fictitious candidates and the researcher manipulates the ethnic characteristics of the candidates (e.g., Dunning and Harrison Reference Dunning and Harrison2010). Such experiments, however, have limitations as well, especially when it comes to external validity. For example, many studies have been located only in urban areas,Footnote ⁶ even though the majority of Africans live in rural areas, raising questions whether the effects of ethnicity on electoral preferences would be the same in different localities. The frequent use of fictitious candidates to assess electoral preferences also comes with trade-offs, including concerns whether such treatment is realistic and whether it reflects real life behavior. In sum, experimental work has produces additional evidence of ethnic politics, which is useful when it complements observed patterns of ethnic politics based on electoral and census data.

Why Ethnic Politics Are Prevalent

Probably the richest component of the literature has been its treatment of conditions and mechanisms responsible for ethnic politics. Because ethnic politics are very prevalent in Africa, the question of why we get ethnic politics has been widely studied. Just like scholars focusing on other parts of the world, researchers working on Africa consider two broad categories of motives behind ethnic politics: symbolic (also called expressive or psychic) and material (also called instrumental). By electing coethnic politicians, voters can gain these two broad types of benefits. For example, a coethnic’s electoral victory can make voters feel validated (hinting at a psychic benefit), but it can also improve their access to resources (indicating material benefits). While people can be motivated by many factors and benefits are not mutually exclusive, the majority of studies of ethnic politics in Africa have concluded that instrumental motivations are paramount. This differs from some leading scholars of ethnic politics in other parts of the world, such as Horowitz (Reference Horowitz1985), who stress expressive drivers of ethnic politics, asserting that people vote for coethnics because it helps their self-worth, derived from a sense of their group’s prestige. In contrast, some of the classic works on ethnic politics in Africa, including those by Bates (Reference Bates1974), Kasfir (Reference Kasfir1979), Skinner (Reference Skinner1975), and Joseph (Reference Joseph1987), suggest that groups use ethnicity to advance their material goals. From their perspective, ethnic groups offer efficient vehicles to advance group demands and to help voters access resources. Voters are thus amenable to ethnic mobilization because they expect to benefit materially. They also see ethnicity as instrumental in negotiating access to power (Ajulu Reference Ajulu2002).

Work by scholars such as Posner (Reference Posner2005) and van de Walle (Reference van de Walle2007) further supports the view that ethnic electoral mobilization is about defending material interests of different groups. Posner (Reference Posner2005) provides a very clear rationale as to why voters and politicians rely on ethnic identity. He argues that in situations of information scarcity, ethnic affiliation gives voters credible information about which groups will benefit if a given party or candidate wins election. He suggests that voters widely believe in ethnic favoritism, namely they expect their coethnics to help them materially more than a non-coethnic would, and stresses that it is the perception of ethnic favoritism that suffices rather than an established empirical pattern of preferential treatment. Posner adds that ethnic affiliation can help to enforce politicians’ promises because, while an individual voter cannot successfully punish a politician who reneges on his promise, an entire ethnic group can do so by withholding future support. Ethnic affiliation can thus serve as an enforcement mechanism, making coethnic politicians’ promises more credible. Politicians, in turn, take advantage of voters’ perceptions of ethnic favoritism. They rely on ethnicity for political mobilization because they can employ it as a cheap information shortcut when appealing to voters.

Ferree (Reference Ferree2010) presents a similar understanding of ethnic politics, or what she calls census elections, wherein parties and voters rely on party images as cognitive shortcuts to guide their behavior. She also highlights how politicians can sow doubts in voters’ minds about whether a non-coethnic party would have their interest at heart. Ferree convincingly shows that part of the ANC’s strategy in South Africa is to paint their opponents in racial terms to question their commitment to non-coethnic voters. As she argues, presenting rival parties as “white” helps to delegitimize and discredit them in the eyes of black voters. Ethnic mobilization can thus rely both on leading voters to believe that their coethnics would favor them and on raising doubts as to whether non-coethnic politicians would care about them. This dynamic is also consistent with Lynch’s (Reference Lynch2014) account of how William Ruto in Kenya successfully convinced his coethnics not to vote for Raila Odinga by stoking fear of the consequences of his victory. Lynch (Reference Lynch2008; Reference Lynch2011) further documents the persistent use of language of Kalenjin persecution and state bias to mobilize Kalenjin voters. Both Ferree’s and Lynch’s work discussed here also show clearly the role of political elites in accentuating ethnic differences and keeping them politically salient. In addition to fear of non-coethnics, Padró i Miquel (Reference Padró i Miquel2007) suggests that voters’ support for coethnics is further strengthened by the belief that other groups will vote for their coethnics.

Despite a quasi-consensus that ethnic mobilization is about competition for material goods, this scholarship is largely agnostic about the specifics of groups’ wants. There has been a tendency to believe that all groups want essentially the same thing, namely access to scarce state resources, development, and security (e.g., Melson and Wolpe Reference Melson and Wolpe1970; Joseph Reference Joseph1987). In contrast, Lieberman and McClendon (Reference Lieberman and McClendon2013) suggest that different ethnic groups have distinct priorities and preferences over how resources should be spent. The question of what groups want deserves more attention because it has implications for the durability and perpetuation of ethnic politics. If groups essentially want the same goods, ethnic politics might wither if politicians could credibly ensure equal access to resources for all groups, for example, through universal redistributive programs. In contrast, if groups have different wants, ethnic politics are more likely to persist, reflecting underlying policy preferences of different groups.

Overall, one of the main strengths of the existing research on the factors driving ethnic politics is the persistent finding of the importance of material benefits in generating ethnic voting. This view emerges consistently across many different cases, over time and across methodologies. In contrast, one of the weaknesses of the existing scholarship is that certain key assumptions are not questioned because of widespread belief in them. For example, as discussed, Lieberman and McClendon’s (Reference Lieberman and McClendon2013) work suggests that we should scrutinize rather than assume that all ethnic groups have similar wants. Another assumption that is rarely questioned is the belief that ethnic ties provide the most logical basis for clientelist networks. One of the consequences of this widespread belief in the utility of ethnic networks for accessing material benefits is that the literature is much better at explaining why politicians rely on ethnic mobilization than why political entrepreneurs do not mobilize along ethnic lines.

The logic of ethnic politics analyzed in Africa raises doubts about whether ethnic politics are driven by the same logics everywhere. The empirical cases coming from the continent show similarities to other developing countries such as India where instrumental motivations seem to be crucial (e.g., Chandra Reference Chandra2004). In contrast, ethnic appeals in Africa relatively rarely refer to issues such as language instruction, culture, or desire for sovereignty or federalism, which appear to be more common in Europe (e.g., De Winter Reference De Winter, De Winter and Tursan1998). As the review highlighted, expressive motivations for ethnic voting, such as those highlighted by Horowitz’s seminal work, have found less empirical support in Africa.

There are at least two key contextual differences between Africa and Europe that shape scholars’ understanding of ethnic politics. First is the legacy of colonialism and its impact on ethnic politics. It is a well-established view that European colonialism affected ethnic groups and regions differently, creating unequal access to resources, be it through differential development, uneven missionary activity and access to education, or employment patterns (Ajulu Reference Ajulu2002; Berman Reference Berman1998; Lonsdale Reference Lonsdale2014; Ndegwa Reference Ndegwa1997). By highlighting and accentuating group differences, colonialism played an important role in politicizing ethnic identities. The colonial experience mattered in several ways: it made some identities salient and led to the creation of others (Ndegwa Reference Ndegwa1997); it unleashed an acute contestation over resources and bore witness to the uneven impacts of capitalist penetration (Ajulu Reference Ajulu2002); and it contributed to the centrality of ethnic patron-client networks of support and protection over other networks (Berman Reference Berman1998). Some scholars, such as Mafeje (Reference Mafeje1971), also argue that the whole concept and vocabulary of ethnicity and tribalism is a European colonial import. The contribution of these works is that they highlight the historical origin of politicization of ethnicity and the centrality of ethnicity in determining access to resources. The importance of the colonial experience to the origin of ethnic politics presents a stark contrast with cases of ethnic politics in the absence of colonial history.

Second, ethnic politics in Africa are often understood not just through the prism of benefits of voting for a coethnic but also through the limitations of alternative politician-voter linkages, such as programmatic appeals. Many scholars believe that one of the key reasons why ethnic politics are prevalent on the continent is the absence of ideological divides, which could serve as alternative electoral cleavages and help distinguish parties and candidates (e.g., van de Walle Reference van de Walle2003; Reference van de Walle2007).Footnote ⁷ Ethnicity is thus, in some ways, filling the void in electoral politics left by low salience of programmatic and ideological issues (Ajulu Reference Ajulu2002). Empirically, it is uncommon in Africa to see ethnic politics coexisting with ideological polarization. Indeed, the value of ethnic markers and shared ethnicity is often predicated on an information-poor environment (Posner Reference Posner2005). This creates comparative contrasts with other parts of the world, where ethnic politics exist in environments in which voters have much better access to information.

Effects of Politicized Ethnicity

The way that scholars and journalists view ethnic politics seems to vary around the world. Whereas identity politics and symbolic representation of ethnic groups are often referred to as progressive in common parlance in the US and are largely lauded in Latin America (see Van Cott Reference Van Cott2010 for review), ethnic politics are overwhelmingly viewed as negative phenomena in Africa. The different frames used to talk about ethnic politics on different continents are problematic and invite further debate. There is no simple explanation as to why identity politics in Africa are viewed more negatively than in many other places, and it is not solely driven by Western negative stereotypes of the continent. While some of the terminology such as “tribalism” frequently employed in the media contribute to the persistently negative connotations, what is striking is that African citizens are also harsh critics of ethnic politics. For example, Posner (Reference Posner2005) describes Zambians’ widespread criticism of ethnic politics and the considerable social stigma attached to it. A recent Kenyan documentary, Softie (2020), chronicles the pernicious undertow of ethnic mobilization in Kenya and many Kenyans’ distaste of it. The protagonist, Boniface Mwangi, a candidate for parliament, firmly believes that ethnic politics have held the country back and he (unsuccessfully) tries to run a non-ethnic campaign. In his telling, ethnic politics are regressive, and non-ethnic representation would constitute progress. Politicians throughout Africa accuse each other of playing the ethnic card, and these pronouncements are invariably meant as criticism, so much so that many political candidates avoid ethnic mobilization in public and resort to it behind the scenes.

This is in stark contrast with supportive language surrounding indigenous political mobilization in Latin America where, as Van Cott argues, researchers avoid criticism even when it is merited (Reference Van Cott2010, 401). Part of the explanation why ethnic (indigenous) mobilization in Latin America is perceived positively is that it represents “empowering oppressed people” (Van Cott Reference Van Cott2010, 401). Yet many ethnic groups in Africa are marginalizedFootnote ⁸; but assessment of ethnic mobilization among them in no more positive than among larger and more powerful groups.

Setting these different standards and expectations aside, scholarship on ethnic politics in Africa can contribute to our understanding of the effects of ethnic politics. Researchers have identified many (potentially) problematic consequences of ethnic politics, many of which could apply to other contexts. This differs markedly from more benign views of ethnic politics found in cases outside of Africa (e.g., Ishiyama Reference Ishiyama2001; Chandra Reference Chandra2005). Scholars focusing on other parts of the world can thus learn about potential pitfalls of ethnic politics from African cases, as described below.

First, in contrast to economists who often focus on sheer ethnic diversity and its effects (e.g., Easterly and Levine Reference Easterly and Levine1997; Kimenyi Reference Kimenyi2006), political scientists tend to believe that it is the politicization of ethnicity, and not diversity per se, that is harmful to societies. Posner (Reference Posner2004) finds that his measure of politically relevant ethnic groups (PREG), namely politicized ethnic divisions, has a much more negative impact on growth than mere ethnic diversity, as measured by the commonly used ELF index (Posner Reference Posner2004, 860). Lake and Rothchild (Reference Lake and Rothchild1998) further claim that ethnic competition for resources has negative economic consequences because when each group calls for group-specific benefits, they typically distort the economy and can reduce national wealth in the long run (Reference Lake and Rothchild1998, 10). Lieberman (Reference Lieberman2007) shows that ethnic politics lead to diminished expenditures on HIV/AIDS. Wimmer, Cederman and Min argue that it is ethnic politics, as measured by their Ethnic Power Relations index, and not mere diversity that is likely to lead to violent conflict (Reference Wimmer, Cederman and Min2009). Lieberman and Singh (Reference Lieberman and Singh2012) demonstrate with cases from southern Africa that institutionalized identities pose problems as they erect boundaries between ethnic friends and foes.

Within the realm of electoral politics, ethnic voting and the existence of ethnic parties limit voters’ choices, as politicians focus only on their own group. Political competition along ethnic lines encourages parties to abandon national platforms and to cater to specific groups (Wantchekon Reference Wantchekon2003). In other words, it reduces the inclusiveness of parties and candidates. Ethnicized political competition also creates a zero-sum game where an ethnic group as a whole is either in power or not. Ethnic groups who lose might be marginalized and left without resources.

Furthermore, close electoral contests between ethnic parties also raise the specter of possible violence. Prominent examples when this scenario materialized with deadly consequences include the presidential elections in 2007 in KenyaFootnote ⁹ and in 2010 in Côte d’Ivoire. To be clear, ethnic politics do not automatically result in violence, but when disputes over elections arise, ethnic markers can be used to target members of ethnic groups associated with political opponents, as was the case in Kenya where the Kikuyu were attacked following Uhuru Kenyatta’s controversial win.

Scholar such as Bates (Reference Bates1974), Young (Reference Young1982), and Joseph (Reference Joseph1987) also believe that the use of ethnicity for political competition and material advancement further increases the salience of ethnicity in society. This would explain how ethnic mobilization, once started, would create a vicious cycle that would be long-lasting and hard to break. This cycle can be perpetuated for several reasons: once some political actors mobilize ethnicity successfully, this creates an attractive template for other actors to replicate, creating an ethnic outbidding effect. Ethnic mobilization by some political entrepreneurs also encourages defensive mobilization by other groups as they develop a sense of shared marginalization (Lynch Reference Lynch2011).

Ethnic favoritism in the distribution of material goods is another expected negative consequence of ethnic politics. As outlined, materialist approaches understand ethnic voting as driven largely by perceptions that having coethnics in power improves one’s access to resources. Hence ethnic favoritism seems a logical outcome. Problematically, for a long time, this expectation did not receive sufficient empirical scrutiny. Perhaps it seemed too obvious and was buttressed by anecdotal evidence or voters’ perceptions. Yet there is empirical evidence that brings into question whether voters benefit when their coethnics are in office. For example, Kasara (Reference Kasara2007) showed that in Kenya peasants from the president’s ethnic group are actually taxed at a higher rate than other ethnic groups. Van de Walle saw little observable ethnic favoritism in Uganda (Reference van de Walle2007). In what is a very productive development for the field, more recent studies began testing the evidence for ethnic favoritism more systematically and in a broader set of contexts. They highlight that effects are variable across countries, and they often depend on specific goods analyzed, applying to some benefits but not others (Franck and Rainer Reference Franck and Rainer2012; Kramon and Posner Reference Kramon and Posner2013). Ejdemyr, Kramon, and Robinson (Reference Ejdemyr, Kramon and Robinson2018) further produced the insight that ethnic segregation has consequences for ethnic favoritism. Their evidence from Malawi showed that politicians targeted local public goods to coethnics only when ethnic groups were sufficiently segregated. Michelitch (Reference Michelitch2015) also illuminated a temporal dimension to ethnic favoritism between ordinary people. She found that price discrimination in Ghanaian taxis depended not just on riders’ and drivers’ ethnicity, and their assumed political loyalties, but also on the proximity to elections. In sum, whereas perceptions of ethnic favoritism are very widespread and sufficient alone to motivate voting behavior, recent scholarship highlights that understanding when and how politicians actually employ ethnic favoritism is contingent on a number of factors, and it deserves further scrutiny.

Variation in Ethnic Politics

One of the biggest developments of the last decade has been a growing focus on the variation in ethnicity’s political effects both between and within countries (Hoffman and Long Reference Hoffman and Long2013; Ichino and Nathan Reference Ichino and Nathan2013; Koter Reference Koter2016). Ethnic identity is a powerful predictor of voting behavior in some contexts but not others, and these studies address the reasons for this variation. For example, at the country level, the degree to which ethnicity is a good predictor of vote choice varies dramatically, explaining between 35–40% of vote choice in countries such as Zambia, Kenya, and Benin, but only about 10% in Senegal or Botswana (Koter Reference Koter2016, 7).

First, several scholars have pointed to the limits of ethnic mobilization given the fact that in most African countries no single group makes up the majority of voters. As a result, relying exclusively on coethnics’ votes would make victory impossible in many settings, and there is increasing interest in studying when and how politicians reach out to non-coethnics. Arriola (Reference Arriola2013) argues that, where business is autonomous from state-controlled capital opposition, candidates can access the necessary resources to buy support from their potential competitors representing different ethnic groups, thus stitching together multiethnic coalitions in presidential elections. In his study of presidential elections in Kenya, where outreach to non-coethnics is mathematically necessary for victory, Jeremy Horowitz (Reference Horowitz2016) shows that the main presidential candidates concentrate their efforts on campaigning among unaffiliated (swing) non-coethnic voters, holding the most rallies among them, and delegating mobilization of coethnics to local level politicians. Devasher and Gadjanova (Reference Devasher and Gadjanova2019) study when politicians make cross-ethnic appeals, concluding that they are most likely to do so when such outreach carries low risk of alienating their own coethnics and when their appeals are credible. Gadjanova (Reference Gadjanova2017) also documented how politicians can expand their coalitions to include unaligned voters by using ethnic wedge issues, namely ethnic issue appeals meant to sow division between unaligned voters and rival coalitions. Adida et al. (Reference Adida, Combes, Lo and Verink2016) provide evidence that African politicians can also take advantage of cross-ethnic marriages to expand their electoral base. Together, these approaches highlight politicians’ strategies (and need) to construct alliances with non-coethnics to supplement their bank of coethnic voters. However, they do not explain why politicians eschew ethnic mobilization altogether.

In contrast, there is also an emerging body of work that explores the question of when ethnic politics are less likely to emerge. Different works concentrate on different levels of variation, ranging from the individual or regional to the national. Keeping in mind which level of analysis they address is important because it is often difficult to apply insights from arguments developed at one level of analysis to another. For example, individual-level explanations rarely shed light on cross-country differences, suggesting that differences between countries are not just an aggregate result of differences between individuals, as I discuss in the following paragraphs.

At the individual level, some scholars focus on the role of information since instrumental theories of ethnic voting contend that ethnicity plays an important role in elections due to information scarcity, such as lack of information about candidates’ stances on issues, their qualifications, or politicians’ performance in office, which would allow voters to assess candidates using criteria other than ethnicity. For example, Conroy-Krutz’s (Reference Conroy-Krutz2013) experiment in Uganda reveals that as voters gain more information, especially negative, about their coethnic politicians, they are less likely to support them. Conroy-Krutz’s findings would imply that more knowledgeable individuals should be less swayed by candidates’ ethnic profiles, since they are more likely to weigh them against other pieces of information.

Yet there are also reasons to doubt that increased information would transform ethnic politics. The scarcity of information among African voters is often asserted rather than measured. Barkan (Reference Barkan1976) argued already in the 1970s that African peasants are not as uninformed as we might think. While some sources of information such as newspapers or television are much more limited in Africa than elsewhere, most African voters have access to news on the radio.Footnote ¹⁰ It also does not appear that voters in Mali, where there is little ethnic voting, have any more information than in Benin,Footnote ¹¹ where ethnic politics are rampant. We thus cannot assume that if citizens with better information are less likely to vote along ethnic lines, then countries with more robust media are less likely to have ethnic politics. Finally, other studies show that individuals use information selectively in judging coethnic and non-coethnic politicians (Carlson Reference Carlson2015; Adida et al. Reference Adida, Gottlieb, Kramon and McClendon2017); thus, greater access to information is not bound to reduce pro-coethnic bias.

Recent research by Dulani et al. (Reference Dulani, Harris, Horowitz and Kayuni2021) also raised interesting questions about the electoral behavior of Africa’s growing number of multi-ethnic voters. The study found that relative to mono-ethnics, voters with parents of two different ethnic backgrounds are less likely to support the party associated with their stated ethnic group. This is an intriguing finding, but additional cross-country research is needed to establish whether there is a systematic link between country-level intermarriage rates and ethnic politics.

One of the most compelling explanations of sub-national variation comes from Ichino and Nathan (Reference Ichino and Nathan2013) who focus on the effect of ethnic geography of voters’ localities on voters’ electoral decisions. They provide evidence from Ghana to show that when voters are an ethnic minority in a district, they are less likely to vote for their coethnic politician. The authors argue that geographic contexts shape voting behavior because they modify the information conveyed by ethnicity. Voting for coethnic politicians becomes less attractive to voters when district geography favors other ethnic groups. They point out that many local public goods that voters desire are not excludable, therefore supporting a non-coethnic politician who brings public goods to the district will not deprive the voter of access to these benefits. An important implication of Ichino and Nathan’s argument is that expected benefits from coethnic and non-coethnic politicians will depend on the nature of goods in question and whether they are excludable. While the sub-national evidence in the case of Ghana is very persuasive, it cannot be easily adapted to understanding different outcomes in presidential elections across Africa.

At the country level, several different factors have been suggested to explain the varied salience of ethnicity in politics. Elischer (Reference Elischer2013) argues that ethnic parties and party systems tend to occur in countries with high ethnic fragmentation and without a core ethnic group, whereas non-ethnic party systems prevail where ethnic fractionalization is low and there is a core ethnic group. He suggests that high ethnic fragmentation poses an imperative for groups to seek unity and vote as a bloc. While this is an intuitively appealing argument, it does not explain all existing variation. For example, some African countries, such as Senegal and Benin, have very similar ethnic fragmentation and numerically dominant groups of roughly the same size, and yet they see very different outcomes in the level of ethnic politics.

The design of electoral systems is another factor that some suspect might explain this variation (e.g., Brambor, Clark, and Golder Reference Brambor, Clark and Golder2007; Mozaffar, Scarritt, and Galaich Reference Mozaffar, Scarritt and Galaich2003). These approaches suggest that differences in electoral rules, such as district size, can affect politicians’ mobilization strategies and the resulting electoral outcomes. However, while electoral rules are certainly relevant to politicians’ calculations and strategic choices, we find ethnic voting across the whole range of electoral systems in Africa, as well as divergent outcomes in countries with similar electoral institutions (see, for example, Elischer Reference Elischer2013). This suggests that electoral systems might matter more for determining how, rather than if, ethnicity will feature in electoral politics. For example, Posner’s (Reference Posner2005) study of Zambia shows how changes in electoral institutions, namely a shift from one-party to multiparty politics, did not result in politicians resorting to more or less ethnic politics but in the activation of different ethnic cleavages. Furthermore, while electoral institutions, in conjunction with ethnic demographics such as concentration and fragmentation, affect the number of resulting parties, they do not necessarily determine their (non)ethnic nature (Mozaffar, Scarritt, and Galaich Reference Mozaffar, Scarritt and Galaich2003).

Other prominent arguments include the expectation that cross-cutting cleavages can prevent crystallization of electoral competition along a single identity cleavage. For example, Dunning and Harrison (Reference Dunning and Harrison2010) provide experimental evidence from Mali to show that cross-cutting ties based on an informal institution of joking kinship, or cousinage, can counterbalance ethnic ties. Because voters feel affinity toward both candidates who are their coethnics as well as non-coethnic “cousins,” the effect of ethnicity is dampened. For Dunning and Harrison, this countervailing force of cross-cutting cleavages explains why we do not see ethnic politics in Mali, despite the salience of ethnicity in social life. Yet, empirically, Dunning and Harrison’s finding does not easily travel to other settings in West Africa, such as Senegal, where cousinage ties exist but are not invoked in politics. This suggests that cousinage ties have the potential to counterbalance ethnic ties, but this outcome depends on how they are actually used.

In contrast, I have suggested in previous work (Koter Reference Koter2013; Reference Koter2016) that in order to understand the variation in ethnic politics one should focus on existing alternatives to ethnic mobilization. I demonstrate that where there are powerful local notables, such as traditional and religious leaders, they can serve as intermediaries between politicians and voters, helping politicians to reach out to non-coethnic voters. Intermediaries have much more bargaining power than individual voters, and they have incentives to work with politicians who can deliver the most resources, irrespective of ethnic identity. Consequently, because intermediaries can help to forge linkages between politicians and non-coethnic voters, mobilization through local leaders produces more ethnically diverse electorates. Greater strength of local leadership thus makes the emergence of ethnic politics less likely because it broadens politicians’ mobilization options. Yet, because the power of local leaders varies greatly across Africa, this strategy is viable only in some places.

These works collectively have produced many insights into factors reducing the likelihood of ethnic politics. Much more needs to be known about how different factors, especially individual- and country-level ones, interact with one another. Additional empirical tests in different contexts would help gauge their scope conditions and external validity. For example, it would be beneficial to know if theories tested in single cases (e.g., Dunning and Harrison Reference Dunning and Harrison2010 in Mali; Ichino and Nathan Reference Ichino and Nathan2013 in Ghana) would yield similar results in other countries.

Ethnicity vis-à-vis Other Social Identities

Studying ethnicity in Africa also invites questions as to how it interacts in the political sphere with other socially salient identities, such as national identity, class, or gender. When does it become more or less salient vis-à-vis other identities? Scholars of ethnic politics in Africa have provided important insights using public opinion data, such as the Afrobarometer surveys which record ethnic identification relative to other identities, along a host of other variables. For example, Eifert, Miguel, and Posner (Reference Eifert, Miguel and Posner2010) find that respondents’ ethnic identification is sensitive to political events, such as the timing of elections.

One such line of research questions the relationship between ethnicity and national identity. Historically, national identity in Africa has been viewed as much weaker than ethnic identity and is sometimes portrayed as merely superimposed over ethnicity; ethnic identity, in turn, is sometimes described as an obstacle to national integration (see Koter Reference Koter2019 for review). Ethnicity has received exponentially more attention in African politics than national identity. Only the last few years have seen an uptick in studies examining national identity, after years of neglect. Because of the perceived tension between ethnic and national identification, scholars have focused on factors that boost national versus ethnic identification, with mixed conclusions and some unanswered questions. The Afrobarometer, which fields the equivalent of the Moreno question, asking respondents to assign relative weight to their ethnic versus national identity, provides important data over time and across individuals and countries and has been at the center of this line of research. Robinson (Reference Robinson2014) found that factors associated with modernization theory, such as wealth, literacy, and urbanization, make respondents more likely to privilege their national identity, though more recent rounds of data do not support this conclusion. Green (Reference Green2020) shows that when the core ethnic group, namely the group with the same ethnic identity as the president, is in power, its members identify more with the nation, but when this group is out of power members identify more with their ethnic group. I (Koter Reference Koter2019) find that following an election of a coethnic president, national identity (versus ethnic identity) rises among the president’s coethnics. This suggests that ethnic power dynamics have implications for national identification. In contrast to many other parts of the world where the Moreno question is used inconsistently in survey research (more typically only in areas with history of separatist sentiment), the data gathered in Africa, along the emerging line of inquiry, provide a unique opportunity to draw insights between the interaction of ethnic and national identities.

Future Directions

Because of the salience of ethnicity, the field of ethnic politics in Africa is large and robust. There are still unanswered questions that need to be pursued. There is a consensus in the literature that ethnicity is an important, but not the only, factor shaping electoral behavior. Future work could improve our understanding of how ethnicity interacts with other factors, such as incumbent performance or clientelism (Lindberg and Morrison Reference Lindberg and Morrison2008; Weghorst and Lindberg Reference Weghorst and Lindberg2013). There is a growing sense that even if we do not see much programmatic politics in Africa, campaign issues do matter (Bleck and van de Walle Reference Bleck and van de Walle2013). There is still much that we could learn about how issues, evaluations of politicians’ performance, and voters’ characteristics interact. We certainly know that ethnicity does not trump all other considerations, but the relative weight of different factors is harder to ascertain.

Second, while some patterns discussed here are robust, there are areas were our collective wisdom would be enhanced if findings were replicated in a broader set of settings to know if observed dynamics are generalizable or idiosyncratic. This particularly pertains to experimental work, which is often limited to a single setting in a single country with the use of rather artificial treatment (e.g., Dunning and Harrison Reference Dunning and Harrison2010; Conroy-Krutz Reference Conroy-Krutz2013). If people show a preference for coethnics in an experiment using a vignette about a fictitious candidate, can we assume that they would do so in real life at the ballot box? Can we be confident in a mechanism uncovered in a lab setting in Uganda’s capital without replicating such work in other settings? Third, Africa is a rapidly changing continent, both demographically and economically. We do not know how this changing environment, with growing urbanization, increasing wealth, and a younger population, will shape ethnic politics. These changes offer unique opportunities to study the consequences of demographic changes for ethnic politics.