Ancestral institutions and the salience of African ethnicity: Theory and Evidence

Abstract

This paper advances a pre-colonial institutional thesis to explain the variation in the salience of ethnicity in African societies. It posits that pre-colonial political centralization facilitated the accumulation of economic and institutional advantages, positioning descendants of centralized ethnic groups to benefit from these advantages within postcolonial states. Social identity choices are rational; therefore, descendants of centralized ethnic groups, who enjoy greater advantages within the nation, find less incentive to choose their ethnicity over their national identity. Examples from Ethiopia and Ghana as well as the evidence from combining individual-level survey data from the Afrobarometer with historical data on pre-colonial political centralization support the theoretical claim. In particular, the paper presents both theory and evidence indicating that individuals with ancestors from politically centralized pre-colonial societies are less likely to favour their ethnic identity over their national identity . These findings underscore the importance of considering pre-colonial legacies when promoting national unity.

Introduction

In this paper, I propose an institutional theory attributing the salience of ethnicity in present-day Africa to precolonial political centralization. Precolonial political centralization facilitated the accumulation of economic and institutional advantages (Michalopoulos and Papaioannou, 2020). These ancestral advantages position descendants of precolonially centralized ethnic groups to benefit more from the advantages generated by those ancestral pathways in contemporary African states (Bandyopadhyay and Green, 2016; Michalopoulos et al., 2019). Since identities are rationally constructed (Laitin, 1998), descendants of precolonially centralized ethnic groups have fewer incentives to prioritize their ethnic group identity over national identity, given that their pre-colonial institutions have positioned them to benefit the most in the present nation. I thus hypothesize that precolonial political centralization leads to a weaker relative strength of ethnic versus national identification.

To substantiate the theoretical proposition, I initially provide illustrative cases from Ethiopia and Ghana. Next, I combine data from the Afrobarometer surveys with historical data sourced from the Ethnographic Atlas (Murdock, 1967). Results from both exercises reveal that individuals from precolonially centralized ethnic groups are less inclined to identify with their specific ethnic group than with national identity. The mechanism is prosperity (Ahlerup et al., 2017); descendants of pre-colonially centralized ethnic groups enjoy better conditions (Michalopoulos and Papaioannou, 2020), making it less rational to prefer ethnic identity over national identity (Laitin, 1998).

There is one important point worth noting in passing. The results demonstrate that precolonial political centralization is a robust predictor of ethnicity salience. However, additional analysis is warranted to ascertain whether precolonial centralization has the potential to counteract the positive influence of colonialism on the salience of ethnic identity. It is essential to investigate the specific contexts, locations, and time periods where it is likely that precolonial centralization can weaken ethnic identification by overturning the legacies of colonial policies.

The paper does not aim to overturn conventional wisdom but, instead, provides an explanation that firmly roots the salience of ethnicity in the history of Africa. If anything, the paper suggests that the salience of African ethnic identity is a construct that emerges from the interplay between past and present institutions and economic conditions, both of which are partly influenced by precolonial political centralization (Michalopoulos and Papaioannou, 2020). One factor for the salience of ethnicity in Africa is colonialism (Ali et al., 2019; McNamee, 2019). However, colonialism introduced institutions that overlapped with pre-colonial institutions. Thus, over time, it is possible that the positive effects of pre-colonial institutions counteract the adverse institutional effects of colonial policies (Gennaioli and Rainer, 2006), thereby reducing the salience of ethnic identity (Larcom, 2019). In line with this assertion, McNamee (2019) suggests that the influence of indirect colonial rule on contemporary ethnic identification in Africa diminishes over time due to the varied developmental and political trajectories of post-colonial African states. However, it's essential to recognize that these post-colonial paths have pre-colonial roots (Ali and Fjeldstad, 2023; Amodio et al., 2022; Gennaioli and Rainer, 2007; Michalopoulos and Papaioannou, 2013). I thus argue that the gradual reduction in the impact of colonial policies on ethnic identification over time can be attributed to the positive developmental legacies of precolonial centralization in the post-colonial era. Another factor for ethnic identification in Africa includes contextual factors, such as political competition (Eifert et al., 2010). However, recent research by Amodio et al. (2022) indicates that higher levels of precolonial centralization lead to a decrease in contemporary political competition. Therefore, precolonial centralization may shape the effects of both colonial and contemporary factors on the salience of ethnic identity in Africa. The argument presented echoes Larcom's (2019) assertion that nations with greater political centralization during precolonial periods are likely to exhibit lower levels of ethnic fractionalization in contemporary times.

The paper builds upon and adds to various threads of existing scholarship. First and foremost, the findings align with the conclusions of Robinson (2014) in challenging the notion that the colonial legacy creates insurmountable obstacles for the development of widespread territorial nationalism in Africa. Likewise, these findings are in line with the perspective presented by Michalopoulos and Papaioannou (2020), who argue that the contemporary effect of precolonial centralization is significant and extends beyond the influence of colonialism. Moreover, the findings are consistent with the evidence provided by Maseland (2018), suggesting that colonialism has generated a substantial yet temporary institutional shock. The paper also adds to the literature on the historical and institutional origins of ethnic salience (Ali et al., 2019; Cervellati et al., 2019; McNamee, 2019) and/or ethnic diversity (Ahlerup and Olsson, 2012; Cervellati et al., 2019; Larcom, 2019; Leeson, 2005; Posner, 2005). Furthermore, it contributes to the burgeoning literature on the long-lasting impact of precolonial institutions (Amodio et al., 2022; Chlouba et al., 2022; Michalopoulos and Papaioannou, 2020). At a broader level, the paper is related to the literature documenting the long-term impacts of historic events such as the slave trade (Fenske and Kala, 2017; Nunn and Wantchekon, 2011), colonial and pre-colonial institutions (Michalopoulos and Papaioannou, 2020), the Habsburg Empire (Becker et al., 2016), and the Middle Age Italian free cities (Guiso et al., 2016). It also adds to the identity economics literature (Atkin et al., 2021; Shayo, 2020), and the literature on nation-building (Blouin and Mukand, 2019; Depetris-Chauvin et al., 2020). It adds to this strand of the literature by demonstrating that social identity is an endogenous construct influenced by both historical and contemporary socio-economic advantages associated with membership in statehood societies. Finally, it offers evidence for the predatory theory of the state, which relates the predatory view to the construction of identity and cultural assimilation (Caskey and Murtazashvili, 2022; Murtazashvili and Murtazashvili, 2020). It contributes to this area of economic theories by presenting evidence that more powerful states opt to implement more expensive universal regulations, whereas less powerful states implement less expensive identity-based rules (rules that are determined and enforced based on the social identity of the individuals involved, such as ethnicity, or language).

The rest of this paper is organized as follows: Section ‘Related literature’ reviews related literature. Section ‘Theoretical background’ presents the theoretical argument of why precolonial political centralization shapes identity today and corroborates it with illustrative examples. Section ‘Data’ describes the data. Section ‘Identification strategy’ sketches the identification strategy. Section ‘Results’ presents the empirical results. Section ‘Conclusion’ summarizes the paper.

Related literature

Several theories attempt to provide an explanation for Africa's underdevelopment. One of these is the ethnic theory of African stagnation. This theory, which has its origin in the seminar works of Easterly and Levine (1997), attributes Africa's underdevelopment to its ethnic heterogeneity. In that line of the research, ethnic diversity, commonly measured by an ethnolinguistic fractionalization index (ELF), is often treated as an exogenous covariate in African economic growth regressions. However, recent research casts doubt on the exogeneity of ELF (Alesina and Ferrara, 2005). Regarding the origin of ELF, there are two main hypotheses. The first is the evolutionary approach, which contends that ethnic differences have deep roots in history and ecology and should be studied in an evolutionary context (Ahlerup and Olsson, 2012). The second is a constructivist view, which claims that ethnic diversity is essentially a product of modern states (Bates, 1974; Chandra, 2012; Fearon and Laitin, 2000) and/or economic costs (Atkin et al., 2021).

A growing body of literature attempts to endogenize ethnic diversity. For a sample of societies around the world, Michalopoulos (2012) argues and provides evidence that geographical variation in each area reduced inter-regional migration and led to more ethnic groups. According to the theory and empirical evidence of Ahlerup and Olsson (2012), ELF emerged among peripheral populations in response to an insufficient supply of collective goods. For Africa, Cervellati et al. (2019) theorizes and provides evidence that premodern populations relied on sexual endogamy to limit malaria prevalence, and thus, ethnic diversity in Africa today is the result of ancestral malaria in Africa. Whatley and Gillezeau (2011) claim and provide evidence that Africa's ELF is an endogenous outcome of the social conflict associated with the slave trade. Furthermore, according to Leeson (2005), destructive fractionalization is a result of the poor institutions that generate it. Leeson (2005) illustrates this by examining how heterogeneous agents in pre-colonial Africa used social distance-reducing cues to enhance trade. His ‘endogenizing fractionalization’ theory suggests that colonial institutions disrupted these signals, hindering agents from realizing the advantages of broader exchanges.

It is thus fair to claim that ethnic diversity is not an exogenous construct. The question now is not whether ethnic diversity is exogenous or not. It is rather whether it is correct to assume that ethnic diversity makes ethnicity salient. Not only does the ethnic theory assume that ELF is exogenous but also makes an implicit assumption that ethnic diversity makes ethnicity salient. Nevertheless, both assumptions have received little empirical backing in the existing research. For a cross-section of individuals covered by the third round of the Afrobarometer, Robinson (2014) reports a null association between ELF and the salience of ethnic identity. Masella (2013) provides a similar finding outside Africa. Both Chad and Zambia are in the top decile on the ELF measure, yet it is the former that is a conflict thorn country while there has been no major conflict in the latter (Desmet et al., 2012). Likewise, both Zambia and Malawi have similar ELF, yet the Chewa and Tumbuka ethnic groups are friendly in Zambia while they are enemies in Malawi (Posner, 2004, 2005). This does not mean that ELF is not a correlate of development at all. It only means that the effect of ELF is likely to be limited to places where ethnic identity is salient.

So, the inquiry becomes: What causes ethnicity to be more salient in certain locations compared to others? One strand of research shows that the salience of ethnicity depends on contemporary political factors. For example, Eifert et al. (2010) found that ethnicity becomes more salient as elections approach, while Green (2020) demonstrated that the degree to which respondents identify with their nation or ethnic group depends on whether there is a co-ethnic president in power. However, another related line of research argues that historical factors matter above and beyond these factors. For instance, McNamee (2019) showed that differences in the colonial rule are related to the relative strength of ethnic identification, while Cervellati et al. (2019) showed that ethnic diversity and salience are the results of exposure to ancestral malaria. This paper adds to this research by rooting the salience of ethnicity in the precolonial statehood of Africa.

According to a review by Michalopoulos and Papaioannou (2020), if pre-colonial institutions have survived anywhere, it is in Africa because (a) the influence of colonial powers was limited to the centre; (b) indirect rule reinforced them, and (c) areas far from the centre still largely rely on ethnic institutions. Recent research illuminates that pre-colonial political centralization across and within African countries translates into democratic functioning (Amodio et al., 2022), institutional quality (Gennaioli and Rainer, 2006), and development (see Michalopoulos and Papaioannou, 2020, for a recent review). The current work builds on this strand of literature and contributes to it by examining the influence of pre-colonial political development in Africa on the relative strength of national and ethnic identification in the contemporary context. Through this investigation, the research provides valuable insights into the potential mechanisms through which precolonial centralization may contribute to economic development in Africa (Michalopoulos and Papaioannou, 2013).

The aim is not to paint precolonial centralization in an overly positive light. Indeed, my argument emphasizes the positive effects of precolonial statehood on the relative strength of national versus ethnic identification. However, it is important to acknowledge that centralization can have negative consequences as well. Hariri (2012) argues that early statehood is associated with autocratic political systems and can hinder the development of democratic institutions. This may be because centralized power structures tend to concentrate power in the hands of a few individuals or groups, making it difficult for citizens to have a meaningful say in their governance. Similarly, Chlouba et al. (2022) found that precolonial centralization is associated with positive attitudes towards autocracy. They suggest that this is because centralized political systems are more efficient at providing public goods and services, and citizens may be willing to accept a more authoritarian style of government in exchange for these benefits. However, the concentration of power in the hands of a few individuals or groups can also lead to corruption, nepotism, and other forms of abuse of power. Therefore, while I argue that precolonial statehood can have positive effects on the relative strength of national versus ethnic identification, it is important to recognize the side effects of centralization.

Theoretical background

Why does pre-colonial statehood matter for contemporary identity? The argument is as follows: suppose that at a certain point in history, societies split into two types based on statehood. The first group formed a state, while the second remained stateless.¹ Following Fortes and Evans-Pritchard (2015), I refer to the state-based groups as type A and the stateless groups as type B. Control of the state is a wealth-creating asset and a source of group economic power and status. Therefore, type A societies are better equipped to accumulate economic resources. Access to power and economic benefits mutually reinforce each other, with early statehood amplifying the capacity of ethnic groups to secure the economic and political advantages within contemporary states. These mechanisms are not mutually exclusive, and a path dependence emerges where economic resources and state control mutually bolster each other. Put simply, precolonial centralization facilitated the accumulation of economic and institutional advantages, enabling the descendants of centralized ethnic groups to capitalize on the resulting economic and political benefits in modern state arrangements.

Which identities are chosen and why? Laitin (1998) proposes that identity is a constructive process that involves the selection and manipulation of cultural markers to achieve certain goals. He suggests that individuals have multiple identities to choose from, which are constructed in a context-specific manner based on their needs and goals. The next question is: Which society, type A or type B, prioritizes ethnic identity over national identity? Laitin (1998) highlights the role of instrumental rationality and social comparison in identity construction, emphasizing the agency of individuals in constructing their identities. Instrumental rationality suggests that individuals choose an identity that benefits them economically (Bates, 1974; Chandra, 2007). Meanwhile, social identity theory posits that individuals strive for positive self-worth or positive social identity, which is achieved through inter-group comparison (Tajfel, 1982; Turner, 1975). In other words, individuals tend to align with groups that provide them with a positive sense of self and a sense of belonging.

I mainly draw from the ‘instrumental rationality’ explanation to argue that rational agents tend to prefer a social identity that serves their material well-being most effectively. Since it is the rich that benefit from the economic security provided by the nation, then the advantaged should prefer its national identity over ethnic identity. In a similar manner, descendants of pre-colonially centralized ethnic groups are more likely to choose national over ethnic identity for two reasons. Firstly, individuals from statehood societies are more likely positioned to receive economic benefits than those in stateless societies. Research lends support to this argument. For instance, Michalopoulos and Papaioannou (2013) provide strong evidence of a positive association between pre-colonial statehood and contemporary development in Africa.² Similarly, Gennaioli and Rainer (2007) offer evidence that pre-colonial political institutions have a lasting impact on the provision of public goods in African countries. In addition, Bandyopadhyay and Green (2016) report similar findings for Uganda. Furthermore, Gennaioli and Rainer (2006) demonstrate that pre-colonial statehood is positively correlated with contemporary institutional quality in Africa. Secondly, access to power and economic benefits reinforces each other, and early statehood enhances ethnic groups' ability to capture state power in the present. In this regard, Green (2020) argues and provides evidence that one of the key determinants of national identification in Sub-Saharan Africa is membership in a ‘core’ ethnic group or Staatsvolk, and whether that group is in power. However, the origins of such ethnic groups or Staatsvolk are not explicitly addressed by Green (2020). My argument builds on Green (2020) and posits that politically centralized groups are more likely to control state power and become the Staatsvolk. This claim can be substantiated with the help of Green (2020)'s ethnic cores in 22 African countries, such as Fon in Benin, Kikuyu in Kenya, Cheba in Malawi, Bambara in Mali, Yoruba in Nigeria, Bremba in Zambia, and Shona in Zimbabwe, among others. These ethnic groups held power and were recognized as the true rulers of their respective societies during the pre-colonial era (Mamdani, 1996), and this system was reinforced by the colonial powers and continued even after independence (Müller-Crepon, 2020).

Alternative explanations exist regarding the origins of the salience of ethnic identity in Africa, one of which relates to the logic of indirect rule during colonialism. In directly ruled colonies, like those under French control, colonial administrators were less inclined to utilize existing traditional institutions. Conversely, under the indirect rule, the opposite was true. Recent research has shed light on how differences in colonial rule shape the salience of ethnic identity in Africa (Ali et al., 2019; McNamee, 2019). Alternatively, other research argues that the salience of ethnicity may also be influenced by factors beyond the extent of colonial policies. For instance, Green (2020) posits that belonging to a dominant ethnic group or ‘Staatsvolk’ and whether that group holds political power are significant factors in determining national identification in Africa. Similarly, Eifert et al. (2010) find strong evidence that political competition can strengthen ethnic identities in Africa. While my approach does not entirely overturn this conventional wisdom, it goes beyond it by delving deeper into African history. I argue that, in addition to colonial policies and contemporary factors, the significance of pre-colonial statehood must be considered when explaining contemporary identity in Africa. Such an approach helps gain a clearer understanding of the intricate dynamics that shape identity in the region today.

Why does pre-colonial centralization explain contemporary identity despite colonial policies and contemporary factors? First, the influence of colonial powers was limited to the centre or indirect rule reinforced pre-colonial institutions (Michalopoulos and Papaioannou, 2020). Countries with stronger pre-colonial statehood were less likely to be colonized, and if colonized, were more likely to experience indirect colonial rule, which was contingent upon pre-colonial centralization (Hariri, 2012). Moreover, British colonizers were more likely to delegate power to native authorities and employ less administrative effort in territories with centralized institutions, as argued by Müller-Crepon (2020). According to Wucherpfennig et al. (2016), indirect colonial rule in Africa resulted in lower levels of exclusion from power in the postcolonial era. The authors suggest that indirect rule allowed for the emergence of local elites who were incorporated into the colonial administration and were able to gain political power and experience that they later leveraged in the postcolonial period. This, in turn, reduced the exclusion of these elites and their ethnic groups from political power in the postcolonial era. If anything, excluded groups with centralized pre-colonial institutions can rely on these institutions to bargain more credibly with the state and thus improve their chances of achieving economic and political benefits (Wig, 2016). In fact, colonialism introduced new institutions that overlapped with existing pre-colonial institutions in Africa. Thus, the influence of colonial institutions played a significant role in shaping the prominence of ethnicity (McNamee, 2019). Nonetheless, McNamee (2019) argues that the legacy of indirect colonial rule on contemporary ethnic identification in Africa appears to diminish over time due to the diverse developmental and political trajectories of post-colonial African states. There is ample evidence that the varied developmental and political paths taken in post-colonial African states bear precolonial footprints (Amodio et al., 2022; Gennaioli and Rainer, 2007; Michalopoulos and Papaioannou, 2013). I thus argue that, over time, the enduring and positive institutional and developmental legacies of pre-colonial centralization in Africa have the potential to counteract the negative institutional and developmental effects of colonial influences. Consequently, these pre-colonial legacies increasingly shape the salience of identity in contemporary African society.

Second, contemporary factors, such as political competition (Eifert et al., 2010), are themselves endogenous to pre-colonial institutions. In recent work, Amodio et al. (2022) show that pre-colonial centralization is associated with decreased levels of political competition. If anything, my theory helps explain and reconcile these findings. As implied by my theory, pre-colonial political centralization may have set the stage for contemporary political dynamics by creating enduring political and economic hierarchies. For instance, pre-colonial centralization may have enabled certain groups to accumulate economic resources and gain political power, which in turn may have reinforced their position of dominance in the post-colonial era. This may have contributed to the exclusion of other groups from political power and resources, leading to increased ethnic identification among marginalized groups. Conversely, in societies with weaker pre-colonial centralization, there may have been more fluid and decentralized political and economic systems that allowed for greater competition and a more diverse array of power brokers.

Illustrative examples

Ethiopia

Ethiopia's multi-ethnic nature and its history of never being colonized make it an ideal laboratory for investigating the relationship between historical political centralization and the salience of ethnic identity. Ethiopia is home to more than 80 ethnic groups, with the Oromo and Amhara groups being the largest. According to Levine (2000), the historical characteristics of these two groups are vastly different, with the Amhara group being centralized and dominating the Ethiopian state for centuries under the Solomonic dynasty (1270–1974). This historical legacy made the Amhara a ‘core' ethnic group where the Amhara identity is synonymous with the Ethiopian identity (Levine, 2000).

In 1974, a military group known as ‘the Derg’ overthrew the last king of the Solomonic dynasty and controlled state power until 1991. In 1991, the Derg regime was overthrown. Following this, Melese Zenawi, a Tigrayan ethnic group member, rose to power, and nine ethnic federations were created. This process equipped the nine ethnically designed regions with their own regional legislatures, media, police forces, and political parties which many Ethiopian pundits claim may encourage Ethiopians alike to transfer their allegiance toward their ethnic groups from national identity. One speculation is that members of the Amhara ethnic group may identify more as Amhara than as an Ethiopian. First, the Tigrian-dominated political party controlled the Ethiopian state for about 30 years, and this makes the Amharas incur a high psychic loss from identifying with a nation supposedly run by a minority ethnic group (Kahneman et al., 1990). Second, the institutionalization of ethnicity in post-1991 politics may disincentivize identifying as an Ethiopian among larger ethnic groups such as the Oromo. Third, members of the Amhara ethnic group who live outside the Amhara region may also choose to pass to other ethic groups since identifying as Amhara and/or an Ethiopian has lower incentives in an ethnically governed society. Perhaps for this reason, though debatable the Amhara ethnic group has experienced a substantial decline in population size in comparison to other ethnic groups, notably the Oromo (Green, 2020). In the 1984 Ethiopian census, the population size of the Amhara was 0.7% smaller than the Oromo. However, in the 1994 census, the Amhara had become 2.0% smaller than the Oromo, and this difference further increased to 7.6% in the 2007 Ethiopian census. Altogther, these factors may lead one to assume that the Amhara have reduced sense of national attachment.

Nonetheless, data from the eighth round of the Afrobarometer data for Ethiopia contradict the above expectation. The data show that respondents from the Amhara ethnic group are more likely to identify as Ethiopian than as an Amhara, while the opposite is true for the Oromo and Tigrian ethnic groups. The size of the Oromo ethnic group is the largest in the country and the Tigre ethnic group has had strong presence in state power. Thus, neither ethnic group size nor access to contemporary political power or the politicization of ethnicity alone does explain this data. Alternatively, the institutional history of the Amhara is a persistent factor in shaping their national identity, providing further support for our hypothesis.

Ghana

The prominent ethnic groups in Ghana comprise the Akan, accounting for 47.5% of the population, followed by the Mole-Dagbon (Dagaare and Dagbanli) at 16.6%, the Ewe at 13.9%, and the Ga-Dangme at 7.4%. The Akan (also known as the Ashanti) and the Ewe ethnic groups are known to have had a more centralized system of governance in pre-colonial times, while the Mole-Dagbon group was relatively less centralized. During the period of post-colonial coups in Ghana, the Ashanti and Ewe ethnic groups emerged as significant players vying for increased political power. For instance, when Acheampong, who was of Ashanti ethnicity, staged a coup in 1972, the Ashanti people exerted considerable influence in the political landscape, while the Ewe community revived their secessionist aspirations. Conversely, when Rawlings, who had an Ewe mother and a Scottish father, assumed power in 1979, the Ashanti made attempts to overthrow Rawlings in order to counterbalance the growing dominance of the Ewe community within the state. Anecdotal evidence shows that both the Akan and Ewe are the politically advantaged groups in the country.

Evidence from the third round of the Afrobarometer survey suggests that the Akan and the Ewe, compared to Mole-Dagbon, have a stronger sense of identification with the nation than the Mole-Dagbon (Dagaare and Dagbanli) group, likely due to their historical economic and political dominance. This observation supports my instrumental rationality theory, which suggests that early statehood enhances a group's ability to capture state power and access economic resources, leading to a greater sense of identification with the nation.

Data

The Afrobarometer surveys

The Afrobarometer constitutes the first major source of data. The Afrobarometer comprises a series of nationally representative surveys designed to assess the attitudes of African citizens on a wide range of issues. Afrobarometer interviews are conducted in the local languages, and questions are standardized so that responses can be compared across countries. This paper relies on 5 (3–7) rounds of the Afrobarometer surveys. These rounds have respondent-level survey data for countries that include Benin, Botswana, Burkina Faso, Cape Verde, Ghana, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mozambique, Namibia, Nigeria, Senegal, South Africa, Tanzania, Uganda, Zambia, and Zimbabwe.

Ethnic versus national identification (ENI)

The dependent variable measures the relative strength of ethnic identification in relation to national identification. It is based on a Moreno-type question (see Moreno, 2006), which the Afrobarometer surveys ask respondents as

Let us suppose that you had to choose between being a national ID and being a [Respondent's Ethnic Group]. Which of the following best expresses your feelings?

The relevant answers on this question take the values 1 for ‘I feel only (R's ethnic group)’, 2 for ‘I feel more (R's ethnic group) than national ID’, 3 for ‘I feel equally national ID, and (R's ethnic group)’, 4 for ‘I feel more national ID than (R's ethnic group)’, and 5 for ‘I feel only national ID’.

The above question and its answers constitute a five-point self-identification variable ranging from 1 to 5. In the main analysis, I recode this variable in such a way that higher values indicate that the respondent identifies more closely with his/her ethnic group relative to his/her nation. For brevity, I denote it by ENI, and this is an indicator of the salience of ethnic identity over national identity in the empirical analysis.

The Ethnographic Atlas

The major source of historical data is the EA. The EA describes a group of variables for a number of ethnic groups around the world. The EA includes variables that reflect precolonial prosperity, political systems, and economic arrangements (Murdock, 1967).³

Political centralization

In the analysis, the key explanatory variable is a measure of pre-colonial ethnic political centralization from the EA. Using QGIS, I generated the map in Figure 1. The map shows the degree of precolonial political centralization in Africa over the African ethnic groups' map digitized by Nunn (2008).

Figure 1. The degree of precolonial political centralization in Africa.

In the EA, the political centralization variable is coded as v33. This variable is referred to as the ‘jurisdictional hierarchy beyond the local community level’. The EA codes this variable as 0 for stateless societies that lacked any form of political centralization; 1 for petty chiefdoms; 2 for paramount chiefdoms; 3 and 4 for larger groups. Following the existing literature (e.g. Alsan, 2015), I create a dummy of political centralization which equals 1 if v33 ≥2 or 0 if v33 ≤1. The empirical analysis mainly utilizes this dummy.⁴ In the sample, about 58.7% of the respondents belong to precolonially centralized ethnic groups.

The study combines individual-level survey data from the Afrobarometer with historical data on pre-colonial political centralization from the EA. Matching data from EA to data from Afrobarometer is challenging since the names of ethnic groups in each of these data sets are differently recorded. To overcome this challenge, I turn to the matching concordance developed by Müller-Crepon et al. (2022). It is an R package known as Linking Ethnic Data from Africa (LEDA). Using this method, I matched 248 ethnic groups from the EA to the Afrobarometer.

Identification strategy

The baseline model is an OLS model of the form:

(1)$$ENI_{iek} = \alpha _w + \mu _k + \beta \;Centralized_e + \varepsilon _{iek}$$

where ENI_iek is the measure of ethnic versus national identification for individual i from ethnic group e in country k, Centralized_e is an index of a precolonial political centralization of ethnic group e to which a respondent i belongs to; α_w is Afrobarometer survey wave/round fixed effect; μ_k is country fixed effect and ɛ_iek is a normally distributed error term. As implied by the specification, the treatment occurs at the ethnic group level since this is the level at which precolonial political centralization is coded.

The parameter of interest in specification (1) is β. The identification of β is a challenging task. The first challenge comes from attempting to identify the impact from survey data on respondents sampled from multiple countries. That is, the results may be capturing cross-country differences in contemporary institutions, or the estimates may reflect indirect effects of state formation on national institutions (Gennaioli and Rainer, 2006). The result can also be due to cross-country differences in a country's stock of ethnic diversity (Bates, 2000; Collier, 2001) or the legacy of post-colonial nation-building policies (e.g. see Miguel, 2003). To deal with this confounding problem, I first control for country fixed effects. By comparing individuals with different ethnic backgrounds within the same country, this approach minimizes the concern of capturing cross-country differences in national or other cultural institutions.

Controls

I have also controlled for several other factors. I have two types of controls. For brevity, these controls are labelled as pre-treatment and post-treatment controls. The pre-treatment controls refer to those sets of variables that are either ecological or likely existed prior to the establishment of precolonial centralization. The post-treatment controls are sets of variables that are plausibly correlated with the current attitudes of individuals. In their work, Cervellati et al. (2019) show that malaria affected the origin of ethnic groups. Likewise, Acemoglu et al. (2001) argue that the disease environment early European settlers faced shaped the type of colonial institutions. I thus control for malaria ecology index (borrowed from Kiszewski et al., 2004). Moreover, ecological diversity affects ethnolinguistic diversity (Michalopoulos, 2012) and pre-colonial state formation (Fenske, 2014). For that reason, I further control for ecological diversity from Fenske (2014). I also added geographic controls that include soil fertility and temperature in the ethnic homelands of the respondents I am considering.

Michalopoulos et al. (2019) document that precolonial agriculture affects contemporary wealth and literacy. I thus control for the ancestral practice of intensive agriculture as well as animal husbandry (from Murdock, 1967). I also added ethnic homeland population density (from Alsan, 2015), latitude, and a dummy for the presence of a city in 1850 (all of which come from Murdock, 1967). The African slave trade has been shown to be a source of ethnic heterogeneity (Whatley and Gillezeau, 2011) and mistrust Nunn and Wantchekon (2011). By extension, it may also be affecting identity salience. I thus control the log of total slave exports from Nunn and Wantchekon (2011).

Ethnic homelands which historically had more developed institutions correspond to areas that are more developed today (e.g. see Gennaioli and Rainer, 2007; Michalopoulos and Papaioannou, 2013). In the presence of this evidence, one could argue that present-day institutions or current development, as opposed to historic state institutions, are shaping present-day ethnic identification. Education and religion influence participation in community (e.g. Alesina and Ferrara, 2000; Bjørnskov, 2007). To account for these and similar other explanations, I control for present development indicators such as night lights intensity, lived poverty index,⁵ and an index of public goods access. I also control for a respondent's education, a dummy for being male and a dummy for being an urban resident. I refer to this set of controls as post-treatment covariates.

Results

Baseline estimates

Table 1 provides the baseline OLS estimates to specification (1).⁶ The unit of observation is the i ^th individual in the c ^th country belonging to ethnic group e. In columns 1–4, the dependent variable is the original 5-point scale measure of ethnic versus national identification (ENI). In column 5, the dependent variable is a dummy of ENI that equals 1 if 1≤ENI≤2 or 0 if 4≤ENI≤5. The ‘centralized dummy’ is a precolonial political centralization dummy that takes a value of 1 for centralized groups or 0 for stateless ethnic groups. I report robust standard errors in parenthesis. The estimates in all columns are from OLS. Country and survey wave fixed effects are considered in all columns except column 1. In column 2, only pre-treatment controls are included. These include slave exports, malaria index, latitude, temperature, soil quality, a dummy of city presence in 1800, precolonial dependence on agriculture, and animal husbandry. In columns 3, 4, and 5, the post-treatment controls are added.

Table 1. OLS estimates of precolonial centralization on identification

Notes: In columns 1–4, the dependent variable is a 5-point scale measure of the strength of ethnic versus national identification. In column 5, the dependent variable is a dummy of the strength of ethnic versus national identification. Robust standard errors in parenthesis. Wave refers to Afrobarometer survey rounds 3–7. OLS is ordinary least squares. FE is fixed effects. ***P < 0.01,**P < 0.05.

I now turn to the analysis of the estimates in Table 1. In column 1, the point estimate for β equals −0.218. It is statistically significant at 5%. Column 4 reports the point estimate when all controls are included. It is statistically significant. The point estimate in column 5 is from a linear probability model, where the dependent variable is a dummy of ethnic identification. Overall, the result shows that decedents of politically centralized groups are less likely to prefer their ethnic identity over national identity. For interpretation, I may rely on the result in column 5 of Table 1 since both the dependent and the explanatory variables are on the same scale. The estimate in column 5 shows that decedents of precolonially centralized societies are about 5.5% less likely to prefer their ethnicity over their national identity.

Omitted variable bias

There may be other factors besides long-run exposures to state history that could influence contemporary ethnic identification, and these factors may be unobservable. Therefore, the point estimates reported in Table 1 could be biased. To assess the danger of omitted variable bias, I use a heuristic proposed by Altonji et al. (2005) and formalized by Oster (2019). The approach is to gain insight into the magnitude of unobservable factors necessary to fully attribute an estimated relationship to omitted variables.

Recently, Oster (2019) provides a statistic known as δ for which one is to obtain β = 0. This statistic shows the degree of selection on unobservables relative to observables needed for the true effect of the treatment variable to be a statistical null. A value of δ > 1 indicates limited scope for unobservables to pose a threat to the results. Following this, I report estimates for Oster (2019)'s δ statistic. The estimates for Oster (2019)'s δ in Table 1 are higher than 3, indicating that the selection on unobservables would have to be more than three times the selection on observables to explain away the entire statistical relationship between state history and contemporary ethnic identification. This suggests that omitted variable bias does not entirely drive the results. However, it is important to note that this does not mean there are no omitted variables at all, just that they are less likely to have a significant impact on the results.

Instrumental variable strategy

In the preceding section, I presented suggestive evidence that selection on unobservables does not represent a significant concern. However, I must still address two challenges related to the measure of political centralization. First, the intensity of exposure to statehood is likely to matter, but the available EA data do not provide information on the length of time that an ethnic group was exposed to political centralization. Second, the results I report so far rely on the assumption that precolonial centralization is exogenous. However, precolonial African states are endogenous constructs resulting from long-distance trade (Bates, 1987; Fenske, 2014) and tsetse fly ecology (Alsan, 2015). For instance, Bates (1987) develops a ‘Ricardian’ theory of state formation. He argues and shows that the need to control trade items and routes made long-distance trade to be a primary cause of state formation in Africa. Fenske (2014) extends Bates (1987)’ work by showing, with a sample of 440 ethnic groups, that long-distance trade predicts precolonial political centralization. To address this and other similar concerns, I adopt an instrumental variable (IV) approach and use the Tsetse fly Suitability Index (TSI) from African disease ecology. Alsan (2015) demonstrates that the tsetse fly, an African bloodsucking fly that transmits sleeping sickness and nagana, impeded precolonial political centralization. Borrowing from Alsan (2015) and related literature (e.g. Chlouba et al., 2022), I instrument precolonial political centralization using TSI. In this application, the average tsetse fly suitability on the ancestral homeland is computed, and that average value is then assigned to all respondents belonging to that ethnicity.

The estimates obtained using the IV method are presented in Table 2. Columns 1–4 of Table 2 present the results of the two-stage least squares (2SLS) regression, where the TSI index is used as the instrument for precolonial political centralization. Columns 1–4 of Table 2 replicate the results in columns 1–4 of Table 1, with the exception that precolonial political centralization is instrumented for in the IV method. Across all specifications, the IV estimates are larger than the OLS estimates presented in Table 1. The preferred estimates are shown in column 4, which includes all control variables. In column 4 of Table 2, the 2SLS estimate of β is −0.344 and differs from the OLS estimate in column 4 of Table 1. These results provide suggestive evidence of endogeneity in the measure of political centralization.

Table 2. IV estimates of precolonial centralization on identification

Notes: In columns 1–4, the dependent variable is the 5-point scale measure of ethnic versus national identification in column 1, and it is a dummy of state centralization in column 5. Robust standard errors in parenthesis. ***P < 0.01, **P < 0.05.

Furthermore, the IV estimates in Table 2 show a significant negative (positive) association between ethnic (national) identification and precolonial political centralization, consistent with the OLS estimates in Table 1. Overall, the findings indicate a causal effect of precolonial political centralization on contemporary identity in Africa. Specifically, individuals whose ancestors lived in centralized precolonial states tend to identify more strongly with the nation rather than their ethnicity.

Instrument (IV) validity

The 2SLS estimate obtained here is valid only if the IV satisfies the instrument relevance and exclusion restriction conditions. The third row of Table 2 reports the first-stage F statistic. Irrespective of whether I limit our model to include no controls, only fixed effects, or expand it to incorporate varying sets of controls, the F statistic pertaining to the excluded instrument exceeds 10, signifying the absence of a weak instrument problem. The first-stage results show that tsetse fly is a negative predictor of state centralization in Africa. This is basically a replication of the results of Alsan (2015).

A successful IV design relies on the assumption of exclusion restriction, which means that the instrument used (in this case, TSI) should only impact the dependent variable through the independent variable of interest (precolonial centralization). If this assumption is not met, the results may be biased. One potential source of bias is economic development (Alsan, 2015). To address this, I follow the literature (e.g. Chlouba et al., 2022) and control for several indicators of present development such as nightlight intensity and the present population size of precolonial ethnic homelands. I also include other variables that may affect the exclusion restriction if left unaddressed. However, I acknowledge that there is still a possibility that the exclusion restriction may not be fully satisfied. The alternative is to relax the assumption and perform inference with an ‘imperfect instrumental variable’ (IIV) (Conley et al., 2012; Nevo and Rosen, 2012).

I follow the ‘IIV’ approach from Nevo and Rosen (2012) and present results. The estimates from this approach are plausible under two main assumptions. The first assumption needed for obtaining IIV is ρ_zɛρ_xɛ ≥0.⁷ This assumption is known as Nevo and Rosen (2012)'s ‘assumption 3’ and states that the instrument has (weakly) the same direction of correlation with the omitted error term as the endogenous variable. The second assumption for employing IIV is that |ρ_zɛ| ≤ |ρ_xɛ|. This is Nevo and Rosen (2012)'s ‘assumption 4’ and states that the correlation between the instrument and the error term shall be less than the correlation between the original endogenous variable and the error term. This is to say that the IV is less endogenous than the endogenous variable of interest (x). These assumptions yield what Nevo and Rosen (2012) refer to as an IIV, which is defined as an IV that has the same direction of correlation with the unobserved error term as x, however, is less endogenous than x.

Under one more additional assumption, Nevo and Rosen (2012)'s approach can be used to obtain upper and lower bound IV estimates. This last assumption is that the instrument (z) is negatively correlated with the endogenous variable (x). In my case, this is indeed the case since TSI and political centralization are negatively correlated (see column 5 of Table 2).

Following this, I present results from Nevo and Rosen (2012)'s approach in Table 3.⁸ These results are computed conditional on the fact that Nevo and Rosen (2012)'s assumptions hold. That is, |ρ_zɛ| ≤ |ρ_xɛ|, ρ_zɛρ_xɛ ≥ 0, and ρ_xz < 0. As can be seen from Table 3, the estimated coefficient on the centralization dummy is between −0.354 and −0.2377 (with a confidence interval of −0.605 and −0.119). The OLS estimate of −0.199 (from Table 1) is inside the confidence interval. Likewise, the IV estimates lie inside the boundary of the confidence interval. Thus, the results I obtain remain robust even when the exclusion restriction assumption is relaxed.

Table 3. Nevo and Rosen (2012)'s imperfect IV bounds

Notes: The upper and lower bounds (along their 95% CI for centralized) are computed using Nevo and Rosen (2012)'s approach. Centralized is the dummy of precolonial political centralization. The dependent variable is the 5-point scale measure of ethnic identification. All controls are included.

Auxiliary results

In a supplementary material, I reported results from a supplementary analysis. Firstly, I replicated the results in Table 1 using an ordered logit model to account for the ordinal nature of the dependent variable. Secondly, I used an alternative measure of the dependent variable – a dummy for ethnic and national identification – to ensure that the results do not suffer from social desirability bias. Finally, I demonstrated the robustness of the findings after controlling for colonial policies and contemporary factors.

The main finding remains robust through these exercises. Among other things, the results hold even after considering the influence of colonial styles. One possible explanation for this is that colonialism introduced new institutions that overlapped with existing pre-colonial institutions in Africa. This explanation implies that the influence of colonial institutions played a significant role in shaping the salience of ethnicity. However, over time, the temporary effects of colonial influences diminished, allowing precolonial institutions to regain their significance and increasingly shape the salience of identity. This explanation is consistent with the conclusions of Robinson (2014), which challenges the notion that the colonial legacy creates insurmountable obstacles to the development of widespread territorial nationalism in Africa. Likewise, this explanation is consistent with the evidence provided by Maseland (2018), suggesting that colonialism has generated a substantial yet temporary institutional shock. Moreover, this perspective aligns with the argument presented by Michalopoulos and Papaioannou (2020), who advocate for the limited impact of colonialism.

Early statehood and ethnic conditions: some evidence

This section aims to provide some suggestive evidence on the theoretical mechanisms underlying my argument that descendants of pre-colonial states are less likely to prioritize their ethnicity over national identity. The proposed mechanism was that pre-colonial political centralization played a crucial role in enabling the accumulation of economic and institutional advantages, thereby granting descendants of centralized ethnic groups the ability to benefit from the economic and political advantages in post-colonial states. This claim has empirical support. Stronger precolonial political institutions allowed colonial and postcolonial African governments to better implement modernization programmes in rural areas, resulting in the provision of public goods such as education, health, and infrastructure in African countries (Gennaioli and Rainer, 2007). To further support this claim, I turn to data from the Afrobarometer.

The Afrobarometer asks how often respondents perceive their ethnic group as being treated unfairly by the government. In particular, the Afrobarometer asks ‘How often is [Respondent's ethnic group] treated unfairly by the government?’ The relevant answers are 0 for ‘Never’, 1 for ‘Sometimes’, 2 for ‘Often’, and 3 for ‘Always’. I use this as a measure of ethnic grievances. As reported in column 1 of Table 4, political centralization is negatively associated with perceived ethnic mistreatment by the government. This may suggest that statehood societies are less likely to be excluded from state power, providing further evidence for my argument.

Table 4. Early statehood and ethnic conditions today

Notes: Estimates are from OLS. Robust standard errors in parenthesis. ***P < 0.01.

Similarly, the Afrobarometer provides data on how respondents perceive the economic conditions of their ethnic group compared to others in the country. In particular, the Afrobarometer asks respondents as ‘Think about the condition of [Respondent's ethnic group]. Are their economic conditions worse, the same as, or better than other groups in this country?’ The relevant answers to this question are coded as 5 for ‘Much worse’, 4 for ‘Worse’, 3 for ‘Same’, 2 for ‘Better’, and 1 for ‘Much better’. I use this as a measure of the economic condition of ethnic groups. The results are reported in column 2 of Table 4. As can be seen, politically centralized groups are less likely to have worse economic conditions. This provides further evidence that descendants of pre-colonial states are better off, and their ethnic groups are politically dominant, making them less likely to prioritize their ethnicity over national identity.

Overall, I find support for my claim that descendants of pre-colonial states have better conditions, making them less likely to prioritize their ethnicity over national identity.

Conclusion

Ethnicity remains a salient feature in African societies, with significant implications for politics, economics, and social development. However, there is substantial variation in the strength of ethnic identification across the continent. This has led scholars to question what explains this variation and how it can be understood from an institutional perspective.

Building on institutional literature, this paper proposes an institutional hypothesis of identity salience that posits precolonial statehood as a critical factor in shaping ethnic identification. Specifically, the paper argues that precolonially centralized ethnic groups have a greater likelihood of identifying with their nation rather than their ethnic group. This hypothesis is tested using individual-level survey data from the Afrobarometer and historical data on precolonial political centralization from the EA. The paper finds a strong association between precolonial political development in Africa and the contemporary relative strength of national versus ethnic identification. Individuals whose ancestors are from politically centralized pre-colonial societies are now characterized by lower ethnic identification. The evidence suggests that precolonial statehood matters above and beyond colonial and post-colonial factors.

The research presented in this paper has important implications for our understanding of the institutional origins of identity salience in Africa. The findings indicate that ethnicity is not an immutable feature of African societies, but rather a construct that is shaped by the interplay between past and present conditions. This underscores the need for policymakers to consider historical institutional legacies when seeking to promote national unity and social cohesion in ethnically diverse societies. It is worth noting that the paper does not aim to displace existing orthodoxy, but rather to bolster it with empirical evidence drawn from the African context.

The study's findings are consistent with the broader literature on the institutional underpinnings of identity salience. Specifically, the paper contributes to the scholarship on the institutional origins of ethnic fractionalization (Ahlerup and Olsson, 2012; Larcom, 2019; Leeson, 2005) and ethnic identification (Eifert et al., 2010; Green, 2020; McNamee, 2019). This body of work posits that institutions play a crucial role in shaping the salience of ethnicity, and the current study's results lend support to this proposition. As such, the research underscores the importance of paying greater attention to the role of institutions in driving the salience of ethnicity, and how this can inform policy interventions aimed at managing ethnic diversity and promoting social cohesion.

The analysis offers fruitful avenues for future research. First, considering heterogeneity among ethnic groups would be important. Some early states tried to build a common ethnic identity (notably the Zulu Kingdom or the Mandinka Empire), others were loose confederacies where no attempts at shaping a common identity were made (such as the Ashanti Empire). It is thus important to show how the legacies of early statehood on contemporary identity were shaped by such heterogeneity. The findings strongly suggest that there is a significant relationship between precolonial political centralization and the salience of ethnicity. However, more in-depth analysis is required to ascertain whether precolonial centralization can effectively mitigate the positive influence of colonialism on the prominence of ethnic identity. Specifically, it is essential to identify the specific circumstances, geographical areas, and historical periods in which this phenomenon is most pronounced.

It is also useful to extend the analysis by including the predatory state perspective. Caskey and Murtazashvili (2022), in their article ‘The Predatory State and Coercive Assimilation: The Case of the Uyghurs in Xinjiang’ is a relevant reference to consider, explore how the predatory state employs coercion to construct identity and assimilate cultural groups. This and other similar research make it apparent that there are similarities between the coercive assimilation of the Uyghurs in Xinjiang and the historical attempts of colonial and post-colonial states to impose their own identities and institutions on African societies. Given the significant role of pre-colonial institutional legacies in shaping identity salience in Africa, future research could benefit from exploring the interactions between predatory states and institutional legacies in shaping identity formation and cultural assimilation.

The paper presents arguments and empirical cases demonstrating that precolonial centralization influenced the salience of ethnic identity through economic prosperity. However, it is important to acknowledge that this is not the only possible mechanism at play. Historical political centralization may have also impacted the number of ethnic groups, with politically centralized groups more likely to be found in countries characterized by lower ethnic diversity and greater homogeneity. This, in turn, could contribute to a reduced ethnic identity salience. Additionally, while the paper suggests that precolonial centralized societies are better equipped to accumulate economic resources, the specific nature of these resources remains unclear. For example, it is worth considering whether precolonial centralization facilitated the construction of infrastructure such as roads, which could enhance exchanges between ethnicities and promote economic development. Similarly, the presence of public buildings and large markets could have shaped interethnic interactions and economic opportunities. Although the paper focuses on the channel of economic prosperity, it is important to recognize the potential significance of these alternative mechanisms. Future studies could further explore these aspects to shed light on their influence on the salience of ethnic identity.

Another promising area for future research stemming from my analysis pertains to the role of customary and cultural institutions in confronting predatory states. Research by Murtazashvili and Murtazashvili (2016) on Afghanistan suggests that informal private property rights, as established through customary organizations, may be more effective than formal private property rights, given their greater capacity, constraints, and legitimacy vis-à-vis the state. This perspective is especially relevant in states, such as Afghanistan, that were not colonized, and where diverse cultures organized along tribal lines faced varying degrees of state intervention, which may have shaped patterns of development. Accordingly, exploring the evolution of customary or cultural institutions, their interaction with state institutions, and their role in shaping institutional change and persistence in the face of state predation would be interesting.

Additionally, it is important to consider the literature on the evolution of institutions, which has not been extensively engaged in this paper. Evolutionary economists such as Geoffrey M. Hodgson have written extensively on this topic. Furthermore, Bednar and Page's research on the co-evolution of culture and political institutions is particularly relevant to this discussion (e.g. see Bednar and Page, 2018). By incorporating these perspectives, a more nuanced and comprehensive understanding of the role of cultural institutions in shaping political institutions can be developed.

Finally, American Indians faced a centralized state and were coerced into adopting new organizational systems that undermined their identities. Dippel (2014) argues that this forced integration of autonomous polities into a shared governance system had severe negative long-term consequences. Moreover, Caskey and Murtazashvili (2022) employed the predatory theory of the state to analyse China's violent assimilationist campaign targeting the Uyghurs, revealing how cultural genocide could occur in tandem with economic growth. This raises important questions about the preservation of cultural institutions (Crepelle et al., 2022; Murtazashvili and Murtazashvili, 2016), which merits further investigation.