Institutions connect “the past with the present and the future” (North, Reference North1991: 97). The study of institutional development has therefore been central to the social sciences. Yet, the institutional trajectory of traditional political institutions—governing subnational communities based on customary legitimacy—is mostly excluded from the systematic analysis of institutional persistence and change. This is despite wide-ranging evidence of their contemporary power (Logan, Reference Logan2013; Baldwin and Ricart-Huguet, Reference Baldwin and Ricart-Huguet2023) and importance for economic development, public goods provision, elections, and conflict (e.g., Logan, Reference Logan2013; Baldwin, Reference Baldwin2016; de Kadt and Larreguy, Reference de Kadt and Larreguy2018; Wig and Kromrey, Reference Wig and Kromrey2018; Baldwin and Holzinger, Reference Baldwin and Holzinger2019; Brierley and Ofosu, Reference Brierley and Ofosu2022; Henn, Reference Henn2022).
The lack of research on traditional institutions’ change is worrying since a large and growing literature reports robust long-term effects of their precolonial predecessor institutions on development (Gennaioli and Rainer, Reference Gennaioli and Rainer2007; Michalopoulos and Papaioannou, Reference Michalopoulos and Papaioannou2013), public goods (Wilfahrt, Reference Wilfahrt2018, Reference Wilfahrt2021; Archibong, Reference Archibong2019), political attitudes (Chlouba et al., Reference Chlouba, Smith and Wagner2022), political violence (Wig, Reference Wig2016; Paine, Reference Paine2019), as well as national-level democracy (Giuliano and Nunn, Reference Giuliano and Nunn2013; Sinding Bentzen et al., Reference Sinding Bentzen, Hariri and Robinson2019) in Africa. These analyses mostly rely on Murdock's (Reference Murdock1967) well-known measure of jurisdictional hierarchy of group-level precolonial institutions. The measure counts the number of hierarchical levels of political organization and is a widely used conceptualization of political complexity in anthropology and archaeology (e.g., Diamond, Reference Diamond1997; Currie et al., Reference Currie, Greenhill, Gray, Hasegawa and Mace2010).
The mechanism typically invoked to explain the long-term effects of past institutions on contemporary outcomes is institutional persistence: the view that today's traditional institutions closely mirror their precolonial predecessors. Yet, so far, we lack an empirical test of such persistence, which is also due to the scarcity of research on the causes of present-day traditional political institutions and their power (for exceptions see. e.g., Logan, Reference Logan2013; Baldwin and Ricart-Huguet, Reference Baldwin and Ricart-Huguet2023). Without such evidence, institutional persistence is observationally equivalent to persistent effects of precolonial institutions—historical effects of institutions (e.g., leading to past development) that have persisted independent of the institutions themselves. We provide this “missing middle” on the path from past to present (Cirone and Pepinsky, Reference Cirone and Pepinsky2021: 23). We use data on contemporary traditional institutions in Africa to examine whether they have persisted until today or whether they have undergone systematic institutional change, triggered, in particular, by different modes of colonial rule.Footnote 1
The extent of such institutional persistence is contested. Consistent with the idea of ubiquitous persistence, Herbst (Reference Herbst2000) argues that there has been no significant change in traditional political institutions since the precolonial era. Correspondingly, De Juan (Reference De Juan2017) finds cultural centers of the precolonial Burundi kingdom to persist as customary courts today. Concerned with cultural persistence, Bahrami-Rad et al. (Reference Bahrami-Rad, Becker and Henrich2021) find that Murdock's (Reference Murdock1967) precolonial measures of kinship organization, social norms, and customs significantly correlate with survey-based counterparts from today. A similar finding—that Murdock's polygyny measure corresponds to high levels of polygyny today—has been documented by Dalton and Leung (Reference Dalton and Leung2014).
Yet, political institutions also reflect and determine the distribution of political power (Acemoglu et al., Reference Acemoglu, Egorov, Sonin, Alberto and Giovanni2021) and may thus change when power shifts. From this perspective, many argue that political engineering, attempts to abolish traditional institutions, and the invention of institutions resulted in an institutional present of traditional authority that differs substantively from its precolonial past (Ranger, Reference Ranger1983; Young, Reference Young1994; Englebert, Reference Englebert2002). In particular, European colonial rule on the African continent is argued to have systematically changed, suppressed, or invented “traditional” institutions. In this vein, French direct rule is found to have destroyed traditional institutions while British indirect rule integrated traditional institutions into the colonial state, thereby fostering their persistence (Crowder, Reference Crowder1968; Ali et al., Reference Ali, Fjeldstad, Jiang and Shifa2019; Müller-Crepon, Reference Müller-Crepon2020).
Theoretically, we differentiate between two previously conflated dimensions of political complexity. The vertical dimension—political centralization—captures the existence of institutions that allow for supra-local governance through multiple layers of political hierarchy. The horizontal dimension—functional differentiation—focuses on the degree of institutional specialization in the provision of various types of governance. While these two dimensions are correlated, they do not always co-occur (McIntosh, Reference McIntosh1999). We argue that these two dimensions are both susceptible to the impact of direct colonial rule. Yet, since functional differentiation is closely related to actual governing practices, we expect French direct rule to be more detrimental to this dimension of political complexity. Political hierarchies were also often dismantled, yet the nominal existence of centralized political structures—such as kings—is less likely to disappear entirely and more likely to resurface after independence when compared to specialized governing practices.
Empirically, we ask to which degree contemporary traditional institutions reflect their precolonial predecessors and whether systematic changes are due to colonial styles of direct and indirect rule. While Baldwin and Holzinger (Reference Baldwin and Holzinger2019: 1748) acknowledge that today's traditional institutions are not “accurate reflections of historic governance practices,” we so far lack a systematic and continent-wide examination of the link between historical and contemporary traditional institutions.
To address the need for complete comparative data,Footnote 2 we use new expert-coded data on 566 African ethnic groups’ contemporary traditional institutions that are similar to yet more detailed than Murdock's (Reference Murdock1967) measure of jurisdictional hierarchy.Footnote 3 We rely on a battery of characteristics of contemporary institutions that together measure their political complexity and its vertical and horizontal sub-dimensions. The vertical dimension includes the hierarchical level of political organization and leaders. The horizontal dimension includes indicators of the existence and number of types of traditional leaders and institutions—such as houses of elders and customary courts—, as well as their functions and ties to the state.
Our analysis shows a robust relationship between past and present levels of institutional centralization. However, this general persistence is primarily driven by ethnic groups in former British colonies where indirect rule was applied in particular to centralized groups. More direct rule by the French led to a substantively and statistically insignificant relationship between past and present levels of traditional authorities’ institutionalization. Furthermore, this French–British difference is particularly pronounced for the functional differentiation dimension of political complexity. This pattern is also reflected in the contemporary importance of traditional authorities for the daily lives of group members. Suggesting a continuation of patterns of colonial indirect rule through the post-colonial period, we finally find that countries that formerly experienced indirect rule are more likely to have integrated traditional institutions into their constitutions today.
1. Traditional institutions in Africa: persistence and change
The literature on the long-term effects of precolonial institutions typically attributes the assumed institutional persistence to path-dependence, a “historical causality” rooted in the institution itself (Page, Reference Page2006: 87). The initial institution sets out the path, which can result in persistence with no institutional changes over time or gradual path-dependent changes, endogenous to the institution (e.g., Mahoney and Thelen, Reference Mahoney and Thelen2009; Acemoglu et al., Reference Acemoglu, Egorov, Sonin, Alberto and Giovanni2021; Gerschewski, Reference Gerschewski2021). No change in the institution—“stasis”—is the predominant form of persistence assumed in the social sciences (Acemoglu et al., Reference Acemoglu, Egorov, Sonin, Alberto and Giovanni2021: 367).
While stasis and gradual institutional build-up can occur conditionally on the path, institutional destruction is often non-gradual and induced by external factors. Analyzing the political evolution of Austronesian-speaking societies, Currie et al. (Reference Currie, Greenhill, Gray, Hasegawa and Mace2010) show that increases in political complexity—e.g., from an acephalous society to a chiefdom with one administrative level—occur incrementally. However, the political breakdown of centralized states into acephalous societies is often abrupt.
Beyond such non-persistence, the assumption of stasis may mask variation in the type of institutional features that are persistent. Political complexity, in our case, can, for example, be disaggregated into two dimensions: a vertical dimension that captures institutions’ political centralization and a horizontal institutional dimension that comprises their functional differentiation (Figure 1). Political centralization is the degree to which political power is hierarchically ordered in a territory and its people. Centralized institutions are often associated with top-level executive leadership that governs from a political capital and delegates power to lower hierarchical levels.
Political complexity: political centralization and functional differentiation.
Functional differentiation, conversely, pertains to what different parts of the political institutions do and the degree to which they fulfill a broad array of governance tasks. Differentiation highlights the existences and roles of various actors in the system, such as the existence of checks and balances through judiciary bodies or the inclusion of councils in decision-making processes. These two concepts are non-exclusive: a very centralized political system may be highly differentiated. However, a centralized system can lack meaningful differentiation and vice versa. These concepts provide a framework for better-assessing variation in institutional persistence.
1.1. The effect of colonial rule on institutional persistence
Institutional change, and in particular, destruction can be induced by causes external to the institution, such as through environmental changes or external intervention (e.g., Gerschewski, Reference Gerschewski2021). Powerful external actors that want to subject and govern over existing institutions decrease the likelihood of persistence by altering, alienating, or destroying them. As institutions both reflect and determine political power, we have to consider systematic alterations of the distribution of political power as most likely moments of rapid institutional change.
European colonial rule constituted the most rampant and continent-wide external shock to indigenous institutions. The Scramble for Africa—reaching its violent climax after the Berlin conference in 1884/1885—established European rule across the African continent hitherto governed by indigenous institutions. While all colonizers relied on traditional institutions at the very local level (Mamdani, Reference Mamdani1996; Herbst, Reference Herbst2000), the directness of rule at higher administrative levels varied between colonizers, in particular between the French and British empires which ruled over most of the African continent and population (Crowder, Reference Crowder1968; Asiwaju, Reference Asiwaju1970; Miles, Reference Miles1994).
Specifically, historical evidence suggests that the French ruled more directly than the British. Following a “Republican spirit” (Cohen, Reference Cohen1971), French colonizers met precolonial political institutions with hostility (Huillery, Reference Huillery2010) and stripped old elites of most of their power and transferred it to “commandants de cercle”—administrators who rotated too often to acculturate themselves (e.g., Crowder, Reference Crowder1968; Cohen, Reference Cohen1971; Conklin, Reference Conklin1997). British colonial rulers, conversely, are oftentimes described as co-opting precolonial institutions and elites to indirectly rule through them (Crowder, Reference Crowder1968). Such differences were also driven by different legal traditions of the colonizer, which were transplanted to the conquered territories: the more state-centered French civil law tradition has been argued to be “inquisitorial,” whereas the British common law tradition has been based on more bottom-up procedures of legal precedents (La Porta et al., Reference La Porta, Silanes and Shleifer2008: 288–289).
If direct rule permanently destroyed some of the precolonial institutions they had conquered and indirect rule integrated them, we would expect less institutional persistence in French than in British colonies. We would, furthermore, expect the difference between indirect and direct rule to affect both the persistence of political centralization and functional differentiation. However, functional differentiation may be more susceptible to the impact of direct rule, as it is closely linked to those tangible governance processes replaced by direct rule and incorporated into indirect rule. Conversely, political centralization may be more prone to nominal persistence and institutional resurfacing or re-invention after independence (e.g., Englebert, Reference Englebert2002).
The contemporaneous impact of colonialism on political centralization is documented by Müller-Crepon (Reference Müller-Crepon2020), who shows that among 124 African precolonial centralized states, $70$
percent of the ruling lines of succession persisted until independence under British rule, while only $30$
percent did so under French rule. Yet, given that hereditary practices are widespread, nominal centralization may resurface if descendants in former French colonies re-adopt old titles without the full extent of their original powers. Not dissimilar to former European monarchies, residual royal titles then exist nominally, yet lacking governing responsibilities. This is the case in the Rwandan kingdom, which was abolished at independence. After the 1994 genocide, exiled King Kigeli V sought to reinstate the monarchy, including a 2001 visit to Democratic Republic of the Congo's President Laurent Kabila. The Rwandan government pushed back against these efforts and they ultimately failed (Englebert, Reference Englebert2002: 54). Today, the royal line persists in exile, with a new king succeeding in 2017.Footnote 4
On the other hand, the functional differentiation of precolonial institutions may be more critically affected by direct rule. Functional differentiation facilitates effective collective action: differentiated systems are specialized and distribute tasks among different institutions and their leaders—prerequisites for effective political coordination and political responsiveness. Functional differentiation is, therefore, much more closely linked to actual governance. The mechanisms based on functional differentiation are often invoked to explain long-term effects of precolonial centralization.
We can thus suspect French direct rule to have a much bigger impact on the destruction of functional differentiation. Where colonizers used indirect rule, local governing capacities were supported and used under the supervision of colonizers. Including traditional authorities in the day-to-day governance ensures the persistent demand and supply of governing functions of traditional authorities. Letsa and Wilfahrt (Reference Letsa and Wilfahrt2020) report that anglophone Cameroonians are generally more imbued in local politics, have more trust in traditional authorities, and approve of their jobs more frequently than their francophone counterparts. Indirect rule may indeed have contributed to differentiation due to the active integration, but also the active addition of new administrative roles within the colonial regime (Apthorpe, Reference Apthorpe1960: 218).
The Lozi kingdom exemplifies a case where British colonizers left a comparatively light mark, granting some autonomy to the Lozi's Barotseland (Mainga, Reference Mainga2010 [1973]: 190; Zeller and Melber, Reference Zeller and Melber2019: 299).Footnote 5 Precolonially, the institutions of the Lozi kingdom in today's Zambia has been both centralized and functionally differentiated. This included “a hierarchy of ‘officers of state’ and ‘a general Council’ comprising ‘state officials,’ chiefs and subordinate governors” (Mainga, Reference Mainga2010 [1973]: 38). These were required for an array of economic functions and to organize labor such as “[m]ound building and canal construction for drainage and transportations” (Mainga, Reference Mainga2010 [1973]: 32). By 1924, Barotseland became a British “protectorate within a protectorate” (Zeller and Melber, Reference Zeller and Melber2019: 302) and in 1936, the Barotse Native Authority was established, which governed the territory with “far-reaching responsibilities in the fields of land and natural resource management, jurisdiction, and law enforcement” (Zeller and Melber, Reference Zeller and Melber2019: 302). While their political system certainly did not remain untouched by colonial rule, the Lozi preserved their institutions and governance functions more than many other traditional political systems (Ranger, Reference Ranger1968: 228), and direct British administration was limited only to “regions which had never been more than loosely under Lozi control” (Mainga, Reference Mainga2010 [1973]: 190).
Post-independence, the Lozi initially gained autonomy within Zambia through the 1964 Barotseland Agreement. Yet, the government soon tried to dismantle these rights, seeking “to relocate and isolate Lozi authorities from the realm of everyday administration into a sphere of depoliticized ‘folklore”’ (Zeller and Melber, Reference Zeller and Melber2019: 306). However, efforts to overrule and replace the Kingdom and its institutions were unsuccessful—partly due to Zambia's weak central state (Herbst, Reference Herbst2000; Zeller and Melber, Reference Zeller and Melber2019)—resulting in continued demands for independent Lozi governance.
As implied by the Lozi case, we note that indirect rule was not applied uniformly across the British colonies, which further underpins the need for our comparative analysis. The British integrated centralized and hierarchical precolonial institutions, e.g., the Lozi Kingdom (Mainga, Reference Mainga2010 [1973]), the Fulani Emirates (Miles, Reference Miles1994), or the Buganda Kingdom (Reid, Reference Reid2002), into the colonial state by co-opting their leaders who retained much of their accustomed powers. However, where societies lacked centralized institutions, the creation of new institutions was imperative for the roll-out of colonial rule (e.g., Fortes and Evans-Pritchard, Reference Fortes and Evans-Pritchard1940; Hicks, Reference Hicks1961). This led to a more direct style of colonial rule (Gerring et al., Reference Gerring, Ziblatt, Gorp and Arévalo2011; Müller-Crepon, Reference Müller-Crepon2020). Some newly created local institutions were headed by (invented) local elites such as the “Warrant Chiefs” in previously acephalous southeastern Nigeria (Afigbo, Reference Afigbo1972). To the degree that these embedded themselves locally, we would expect some limited centralization of previously decentralized ethnic groups, leading to institutional change. Given that all colonizers relied on local indigenous elites (Mamdani, Reference Mamdani1996; Herbst, Reference Herbst2000), such institutional “upgrading” in decentralized areas was not a phenomenon limited to British colonies.
For the difference between the French and British styles of colonial rule to consistently affect traditional institutions until today, postcolonial governance arrangements between the state and traditional authorities must roughly correlate with colonial ones. Otherwise, postcolonial change in traditional institutions could have slowly washed out the effects of colonial rule. Our results will shed light on this mechanism.
2. Data and research design
To analyze whether precolonial ethnic institutions persist, we combine two datasets that provide information on ethnic institutions in the precolonial past and the present. The first is Murdock's (Murdock, Reference Murdock1959, Reference Murdock1967) Ethnographic Atlas (EA) which contains information on ethnic groups’ political institutions around the time of colonization. Second, the Traditional Governance Groups DatasetFootnote 6 (TGG, see also Baldwin and Holzinger, Reference Baldwin and Holzinger2019) is derived from an online survey of 1122 experts who provided detailed information on traditional institutions across the African continent today.
Murdock's Ethnographic Atlas
The EA is a dataset that compiles information on social, political, and cultural traits of around 1200 ethnic groups worldwide around or before European colonization. It was coded by anthropologist Murdock, who relied on in-depth ethnographic work from secondary sources. Murdock claims to have surveyed “[p]ractically the entire ethnographic literature” (Reference Murdock1967, 1) at the time and used (translated) material in all languages to avoid selection biases. Murdock argues that—at the time—the African ethnographic literature was the most complete (Reference Murdock1967, 7).
Throughout the 1960s, Murdock published the EA in 21 installments in the journal Ethnology and the corresponding book was published in 1967. Despite skepticism toward the comparative exercise from within the anthropological discipline (see Boas, Reference Boas1896; Tobin, Reference Tobin1990; Bahrami-Rad et al., Reference Bahrami-Rad, Becker and Henrich2021: 1), the EA has lately become an important source for quantitative analyses in the social sciences. Bahrami-Rad et al. (Reference Bahrami-Rad, Becker and Henrich2021, Appendix pp. 1–2) list 64 recent peer-reviewed articles that use the EA, of which 17 use Murdock's measure of jurisdictional hierarchy, our focus in this article.
EA's variable no. 33—“Jurisdictional Hierarchy Beyond Local Community (v33)”—is an ordinal measure that counts an ethnic group's levels of institutional hierarchy. It is one of the two variables of the EA that captures political institutions.Footnote 7 It is zero where there exists no political authority beyond the local community. Groups with one level beyond the local community are called “petty chiefdoms,” followed by “large chiefdoms,” “states,” and finally “large states” on level four.
Murdock describes the variable as “a measure of political complexity” (Murdock, Reference Murdock1967: 52). This equation of centralization and complexity likely stems from Fortes and Evans-Pritchard's (Reference Fortes and Evans-Pritchard1940) influential work, which distinguishes societies with centralized authorities from those without. In this view, centralized states have the highest level of political complexity and are “characterized by a centralized political bureaucracy that contains a number of more specialized administrative offices” (Currie et al., Reference Currie, Greenhill, Gray, Hasegawa and Mace2010: 801). The econometric literature that uses Murdock's variable often interprets it as “centralization,” but sees it as a measure of complexity too (see, e.g., Michalopoulos and Papaioannou, Reference Michalopoulos and Papaioannou2013; Wig, Reference Wig2016; Archibong, Reference Archibong2019; Chlouba et al., Reference Chlouba, Smith and Wagner2022).
However—and as argued above—political centralization and complexity are not necessarily equivalent. The anthropological literature has challenged the notion that complexity is exclusive to centralized societies. McIntosh (Reference McIntosh1999: 22) argued that the dominant focus on hierarchy as a proxy of complexity needs to be counterbalanced “with an understanding of flexible hierarchies, multiple overlapping hierarchies, and horizontal differentiation as alternative modes of complex organization.” While we cannot distinguish between the above-discussed vertical and horizontal dimensions of political complexity in the EA, we introduce two respective outcome variables below.
Without affecting the results substantively, we reduce the influence of the outlying four “large states” by recoding them to “states” (3). Our measure of Jurisdictional Hierarchy—(v33) therefore ranges from 0 to 3. The EA was geocoded by Nunn and Wantchekon (Reference Nunn and Wantchekon2011) using Murdock's (Reference Murdock1959) map of African ethnic “homelands,” our unit of analysis. Because many of Murdock's 841 groups’ settlement areas span across several countries, we split them into 1321 groups nested within today's country borders.
There is an emerging debate on the EA's validity and its geocoded ethnic “homelands” (see, e.g., Wilfahrt, Reference Wilfahrt2018, Reference Wilfahrt2021; Paine et al., Reference Paine, Qiu and Ricart-Huguet2021). While there is a need for more research, we are most interested in the ethnic group category (rather than its exact geolocation). Furthermore, newer measures of precolonial states (e.g., Paine et al., Reference Paine, Qiu and Ricart-Huguet2021) exclude non-centralized groups, which are crucial to our analysis.
Traditional Governance Groups Data
We measure contemporary traditional institutions with data collected via a global online expert survey on ethnic groups’ traditional institutions, their leaders, and functions. Similar to the EA, the survey was conducted among experts who were mostly anthropologists.Footnote 8 The survey was fielded between May 2016 and June 2017 in English, French, Spanish, Portuguese, and Russian to prevent selection bias. The resulting subsample for Africa combines information from 1122 experts on the contemporary Footnote 9 traditional institutions of 746 African groups (1.5 experts per group). Data on groups with multiple expert answers generally show high levels of agreement and were aggregated manually to incorporate additional comments provided by the surveyed experts. Appendix A.1 further describes the data collection and aggregation and reports inter-coder agreement.
The universe of groups for the expert survey is mainly based on the All Minorities at Risk list of socially relevant ethnic groups (Birnir et al., Reference Birnir, Wilkenfeld, Fearon, Laitin, Gurr, Brancati, Saideman, Pate and Hultquist2014). Social relevance denotes that “people notice and condition their actions on ethnic distinctions in everyday life” (Fearon, Reference Fearon2006: 852). Thereby, the population of the contemporary data on traditional institutions is not affected by any form of institutional or political organization, which could result in post-treatment bias by selecting specific groups on the outcome variable. Furthermore, the population is sufficiently fine-grained to be matched with Murdock's groups.
Ethnic matching
To link the TradGov Groups data (TGG) to Murdock's EA in a coherent and replicable manner, we draw on the Linking Ethnic Groups in Africa project by Müller-Crepon et al. (Reference Müller-Crepon, Pengl and Bormann2022) and leverage the universe of known languages and dialects to link the two datasets. Doing so has the advantage of accommodating the varying ethnic labels and levels of aggregation at which the two datasets enlist ethnic groups, and ensures the replicability of our procedure. We link groups from the two datasets if they share a language or dialect as a defining ethnic marker—with very few exceptions, ethnic groups in both datasets are linguistically defined. We match 579 (84.3 percent) groups from the TGG data to a total of 731 (55.3 percent) groups enumerated by Murdock. As we demonstrate in an analysis of the determinants of finding a match in Appendix A.3, many unmatched groups in the EA are small group segments that spill across international borders.
An important concern with this intertemporal matching is the introduction of selection bias. If groups whose traditional institutions were destroyed at some point since colonization are less likely to be enumerated in TGG, we would underestimate the amount of change. We, therefore, conduct a descriptive analysis of the attributes of EA groups that lack a link to TGG (Appendix A.3). In general, we find no systematic effect of precolonial centralization but larger and more populous groups are more often present in the TGG data. However, we do find that centralized groups in former French colonies are less likely to appear in the TGG data than centralized groups in former British colonies. Appendix C.4 therefore analyzes the resulting potential for selection bias and finds our main results to be robust.
Main outcomes
To assess the political complexity of traditional political institutions (TPI) today, we combine variables from the TGG data that capture the structure of groups’ institutions, their leadership, functions, and ties to the state. To disaggregate compound “complexity” into theoretically meaningful components, we focus on groups’ vertical level of political centralization and their horizontal, functional differentiation. Our measure of the former resembles most closely the EA's coding of political hierarchy by capturing the hierarchical level at which political institutions and leadership exist. The latter—functional differentiation—is in turn concerned with the horizontal differentiation of institutions and the functions they fulfill, including their ties to the formal state.Footnote 10
Political centralization:
TPI Level: Coding of experts of the highest level of traditional organization, indicating whether a group is acephalous, organized on the district or regional level, or a higher level. This is the variable that coincides most with Murdock's ordinal coding.
Max Leader: Again oriented along Murdock, we encode the maximum hierarchical level on which a leader exists with kings being on level 3, chiefs on level 2, and headmen on level 1. Groups that have none of these political leaders are coded as 0.Footnote 11
Functional differentiation:
The Institution Index is the mean of a series of dummy variables that encode whether an ethnic group features a council of elders and/or the king, assemblies, dispute resolution mechanisms such as courts, and native customary rules.
Leader Index: The average existence of a series of leadership roles, ranging from a king or paramount chief, over chiefs, headmen, judges, healers, to spiritual leaders.
The Functions Index is the average existence of official or unofficial responsibility of a group's traditional institutions for the governance of land, culture, family matters, dispute resolution, health, security, religion, and infrastructure.
The State-ties Index is the average response to the question of whether traditional authorities are (1) formally acknowledged by the state, (2) interact regularly through formal institutions, and on (3) the strength of traditional authorities’ informal ties to state politicians.Footnote 12
To derive a single measure of the political complexity of traditionally governed groups that minimizes information loss, we first linearly transform all variables to a range between 0 (no traditional institutions) and 1 (maximum traditional institutions) and extract the first principal component. Our institutional dimensions are all strongly and positively correlated with this first principal component, which explains 50.9 percent of their variance (see Appendix A.2). This component, named TPI Index hereafter, is a measure of the political complexity of the group and constitutes the main outcome of the empirical analyses.
We derive the two additional sub-indices Functional Differentiation and Political Centralization in the same manner. These are strongly but far from perfectly correlated with each other ($r = 0.54$
). We also present results for the constitutive parts of the index to investigate differential levels of persistence across the dimensions of traditional institutions.
Model specification
We assess the relationship between ethnic groups’ precolonial centralization and the index of today's traditional institutions (1) among all observations and (2) contrasting groups only from former British and French coloniesFootnote 13 using linear models:
where country-fixed effects α i net the data of all variation among ethnic groups i that is constant within (contemporary) countries.Footnote 14 In the baseline specification (1), the level of historical persistence is captured by the coefficient β 1. In Equation (2) which we run on a sample restricted to groups in former British and French colonies, β 2 and β 3 capture the level of institutional persistence in the British and French empires, respectively. The French–British difference in the effect of jurisdictional hierarchies is computed as the difference between β 2 and β 3 and is discussed in the results. We cluster standard errors on the level of ethnic groups (based on Murdock's coding), many straddling international borders.
We add a vector of control variables Xi to account for “pre-treatment” factors that may have simultaneously caused past and current institutions.Footnote 15 We sequentially add three vectors of controls to our model: baseline controls include groups’ population, area, distance to coast, and navigable river since size and the connection to larger economic markets may affect centralization. Building on research on the natural fundamentals of centralization (e.g., Herbst, Reference Herbst2000), nature controls include median altitude and slope, mean annual temperature, precipitation, and evapotranspiration, the ratio of the two, agricultural suitability, and soils’ suitability for cash crop production. Ethnic controls are the reliance on agriculture and pastoralism, and agricultural intensity—activities capturing the potential economic roots of centralization. See Appendix A.4 for details.
3. Results: traditional institutions: past and present
We start our analysis by visualizing the bivariate relationship between past and present African traditional political institutions in Figure 2.Footnote 16 Focusing first on the full sample to the left, we see a consistent and positive relation between Murdock's measure of jurisdictional hierarchy (v33) and our TPI Index, suggesting institutional persistence. While consistently linear when we disaggregate Murdock's measure into its ordinal levels, the correlation is far from perfect and disturbed by many “off-diagonal” cases resulting from institutional change or measurement error.
Correlation of precolonial centralization with the TPI Index across all observations and groups in former British and French colonies.
Note: Points show jittered observed values. Point estimates by level of precolonial centralization and linear correlation, both with 95 percent confidence intervals (CIs).
Splitting the sample between ethnic groups in former British (N British = 282) and French (N French = 161) colonies highlights the type of colonial rule as an important source of change. Traditional institutions are persistent in former British colonies, which were often ruled through rather than against precolonial institutions. In contrast, institutional destruction was more frequent under French rule, where precolonially complex groups are not, on average, more institutionalized today than precolonially less complex ones.
The Lozi in Zambia fit this pattern. As described above, the Lozi were able to preserve both their institutions and administrative functions under the British protectorate. This is reflected in their high scores on the TPI Index, which matches their high Jurisdictional Hierarchy. A case of institutional breakdown under French colonial administration is the Merina kingdom in Madagascar, which fell after the second Franco-Merina war (Kent et al., Reference Kent, Southall, Covell, Deschamps and Dresch2023) and scores low on the TPI Index today.
Estimating variations of Equation 1 in which we sequentially add the various control variables, the first block of coefficients of jurisdictional hierarchy in Figure 3 shows a robust positive relationship with the TPI Index across colonies (see also Table A5 in Appendix B). As we add our control variables in specifications 2–4, the size of the coefficient of precolonial centralization decreases only slightly and its precision remains high. An increase of precolonial centralization by one level leads to an increase of a fifth of a standard deviation of the TPI Index, hence moving from 0 (min) to 3 (max) on v33—Jurisdictional Hierarchy leads to a 60 percent increase of a standard deviation on the TPI Index.
Effect of precolonial jurisdictional hierarchy (Murdock's v33) on TPI Index and its vertical and horizontal sub-dimensions across specifications with 95 and 90 percent CIs. See Appendix B for details.
The standardized effect of full centralization as compared to acephalous political structures ranges from 0.45 to 0.69,Footnote 17 an effect that is considered of medium size (Sawilowsky, Reference Sawilowsky2009). They compare favorably to other persistence studies that use Murdock (Reference Murdock1967). For instance, Sinding Bentzen et al. (Reference Sinding Bentzen, Hariri and Robinson2019: 688 and Appendix p. 4) report that precolonial democratic institutions have a standardized effect of between 0.26 and 0.75 on present-day democracy (see also Giuliano and Nunn, Reference Giuliano and Nunn2013). This similarity is despite the likely comparatively high level of noise in our TPI data as compared to well-established country-level data on democratic institutions. At the group level, Bahrami-Rad et al. (Reference Bahrami-Rad, Becker and Henrich2021) find standardized effect sizes ranging from 0.0022 to 0.12 for precolonial measures of kinship organization, social norms, and customs, effects which are well below our estimates. Compared to previous persistence studies, our effect sizes thus suggest a non-negligible association between past and present institutions. Yet, of course, the relationship is not perfect, due to noisy measurement of past and present institutions as well as unmodeled institutional changes (e.g., Englebert, Reference Englebert2002).
The second and third blocks of Figure 3 show the effect of precolonial jurisdictional hierarchy on the sub-indices of the TPI Index. Both indices are positively and significantly affected by Murdock's (Reference Murdock1967) v33. The coefficient for Political Centralization starts higher but becomes gradually smaller when adding control variables. The coefficient for Functional Differentiation is virtually unaffected by controls. In sum, we find both vertical and horizontal institutional persistence.
Figure 4 shows the relation between precolonial jurisdictional hierarchy and the single TPI Index components. Across outcomes, the effect of v33 is positive and significant. Worth noting are the outcomes TPI Level and the Institutions Index, where precolonial centralization has the biggest effect. TPI Level records the highest level of traditional organization and is therefore similar to Murdock's coding. The Institutions Index measures institutional differentiation, with higher numbers indicating more institutions, including councils, assemblies, and courts. The state-ties index is the only outcome that is not significantly related to v33 across all specifications, indicating that precolonial jurisdictional hierarchy did not affect these contemporary relationships on average.
Effect of precolonial jurisdictional hierarchy (Murdock's v33) on TPI Index and its components across specifications with 95 and 90 percent CIs. See Appendix B for details.
3.1. Colonial rule and institutional persistence
Figure 5 formally tests whether British rule led to the persistence of precolonial institutions as compared to more direct French rule (Equation 2). The plot shows the respective results for the British and French sub-samples, respectively as well as the French–British difference between the estimated effect of precolonial centralization on the indices.Footnote 18
Effect of precolonial jurisdictional hierarchy (Murdock's v33) on TPI Index and its vertical and horizontal sub-dimensions in former British and French colonies.
Note: The four coefficients in each cluster correspond (from left to right) to specifications 1–4 reported in Table A6.
The overall correlation between precolonial jurisdictional hierarchy and our TPI Index is almost exclusively driven by ethnic groups in former British, rather than French colonies. The respective coefficient (${\small \sf{v33}} \times {\small \sf {British}}$
) in the second block is slightly larger than estimated on the full sample and statistically highly significant. In turn, the estimated relation between past jurisdictional hierarchy and the TPI Index in former French colonies is close to zero. The difference between the two estimates in the fourth block turns statistically insignificant once we add the full vector of control variables but remains stable in size.Footnote 19
The second and third blocks of Figure 5 split the TPI Index into its vertical and horizontal components. Traditional institutions in former British colonies have persisted both in terms of their centralization and functional differentiation, with the former effect again more sensitive to the inclusion of controls. The effect size of v33 in former British colonies is larger for Functional Differentiation than for the TPI Index and Political Centralization. When considering former French colonies, we observe a positive, yet insignificant, effect for Political Centralization and a null effect for Functional Differentiation. Correspondingly, the difference between the two estimates is only statistically significant for the functional differentiation of present-day institutions. This suggests that the type of colonial rule was critical for the differentiation of traditional institutions. Precolonially more hierarchical groups in former British colonies feature more differentiated institutions and functions today than comparable groups in former French colonies.
We further disaggregate the analysis of French–British differences in institutional persistence using the constitutive parts of the TPI Index in Figure 6. Supporting the previous results, we find the most pronounced differences in the Institutions Index, traditional authorities’ ties to the state, and the extent of their functions and responsibilities. These are much stronger and positively correlated with precolonial centralization in former British than in former French colonies where a relationship is absent. In turn, differences in the TPI Level and the leadership characteristics are substantively less pronounced and statistically insignificant.
Effect of precolonial jurisdictional hierarchy (Murdock's v33) on all current outcomes in former British and French colonies.
Note: The four coefficients in each cluster correspond (from left to right) to specifications 1–4 reported in Table A6.
3.2. Contemporary relevance and mechanisms of post-colonial persistence
Does institutional persistence also come with greater contemporary relevance and power of traditional authorities? We address this question through two additional items from the TGG survey that asked experts about the importance of traditional authorities for (a) the everyday lives of group members and (b) national politics.Footnote 20 We report the full results in Appendix Table A12. Among former British colonies, precolonial jurisdictional hierarchy predicts the importance of traditional authorities for the daily lives of group members positively and with high precision. There is no such effect in former French colonies and the difference between the colonial regimes is large and significant. This supports the intuition that functional differentiation, which is most affected by colonial governance style, is related to actual governing functions that make traditional authorities relevant to people's lives. In contrast, we find that precolonial jurisdictional hierarchy does not have a stable or precisely estimated effect on traditional authorities’ importance for national politics, neither in former British nor French colonies.
The difference between French and British effects on the institutional persistence of traditional institutions is further underlined by our final analysis (Appendix D) which sheds light on an important driver of post-colonial persistence. As many independent states built their constitutions upon the metropolitan blueprints they inherited from their colonizers, the degree to which constitutions accommodate and grant powers to traditional authorities or not may explain part of the persistence in former British colonies and its absence in former French ones. Using data on the constitutionalization of traditional authorities for former French and British colonies in 2014 from Holzinger et al. (Reference Holzinger, Haer, Bayer, Behr and Neupert-Wentz2019), we find that constitutions in former British colonies acknowledge, regulate, and integrate traditional authorities to a substantively greater degree than those in former French colonies. This suggests that the post-colonial continuation of indirect arrangements of rule play at least some role in the persistence of precolonial institutions in former British colonies.
4. Conclusion
In a contribution to “decompress history” (Austin, Reference Austin2008), this article has assessed the degree to which precolonial institutions in Africa have persisted over the past century. Our analysis is motivated by a large and growing literature on the enduring effects of precolonial institutions, many of which are implicitly or explicitly assumed to be due to institutions’ persistence over time. To assess the empirical merits of this assumption, we have combined data on the precolonial centralization of ethnic groups with expert-coded information on their contemporary traditional institutions.
Our empirical analysis shows a robust association between past and present degrees of the political complexity of traditional authorities. This suggests that traditional institutions have been, on average, persistent over the past century. However, and consistent with arguments about the effects of direct and indirect rule on precolonial institutions, this result is almost exclusively driven by ethnic groups in former British colonies where indirect rule and its postcolonial constitutional legacies preserved local institutions. We observe the most pronounced differences in persistence between former French and British colonies with respect to functional differentiation and the importance of traditional authorities for the everyday lives of group members. While certain hierarchies and leadership roles may persist as relics of precolonial times across colonial regimes, functional governing activity and relevance are more restricted to areas under indirect colonial rule. This finding provides a potentially important starting point for analyzing variation in the long-term effects of precolonial institutions.
Our findings suggest additionally that future studies on the long-term effects of precolonial institutions on various contemporary outcomes (i.e., interested in the effect of X 0 on Y 1 as shown in Figure 7) should differentiate between two broader mechanisms. First, persistence (X 0 → X 1 → Y 1) of traditional institutions and the governance functions they fulfill is largely limited to former British colonies. Second, persistent effects of past institutions—a historical effect of institutions that has persisted independent of the institutions themselves (X 0 → Y 0 → Y 1)—are possibly geographically unlimited. These two mechanisms have previously been largely undifferentiated, also due to the lack of data measuring traditional institutions today.Footnote 21
Graph of different paths of long-term effects.
Note: X 0 are precolonial institutions, X 1 are contemporary traditional institutions. Y 0 is a precolonial outcome, such as development, and Y 1 is its contemporary equivalent. W is a confounder. Effects of X 0 on Y 1 driven by institutional persistence run through contemporary institutions X 1. Persistent effects of past institutions run through mediator Y 0, e.g., past development. If only precolonial institutions X 0 and the outcome Y 1 are measured, it is impossible to distinguish the different paths via X 1 or Y 0.
To analyze whether the long-term effects of precolonial institutions are driven primarily by institutional persistence or persistent effects in future studies, we suggest mediation analysis. This allows for estimating the indirect effect through X 1—institutional persistence—in producing the long-term effect. Yet, the options to causally identify such an effect may be limited (Acharya et al., Reference Acharya, Blackwell and Sen2016). For instance, it may be hard to measure appropriate confounders (W in Figure 7). The mediation analysis would require controlling for confounders of X 0 and X 1 (X 0 ← W → X 1) as well as of X 1 and Y 1 (X 1 ← W → Y 1). Especially for the first effect, observable confounders are largely limited to geographical features such as land quality in Fenske (Reference Fenske2013).
Our findings furthermore point to future research questions on variation in the long-term effects of precolonial institutions on outcomes such as economic development, public goods delivery, violent conflict, and the perception of traditional leaders. For instance, in the spirit of Cornell and Kalt (Reference Cornell and Kalt2000) and the findings of this paper, it could be that a match of past and present institutions can facilitate the production of public goods while institutional disruption undermines it. Finally, other aspects—such as ethnic group size, location, and political status—could affect persistence. We have only considered colonial state-level drivers of institutional change and their postcolonial legacies. Theorizing and analyzing postcolonial change may constitute a promising avenue to foster our understanding of the path to the present of traditional institutions in Africa.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2023.50.
To obtain replication material for this article, please visit https://doi.org/10.7910/DVN/IJ2MII.
Acknowledgements
We thank three anonymous reviewers, Pierre Englebert, Ane Foged, Constantin Ruhe, and Julian Schuessler, as well as participants at the APSA Annual Meeting 2019, DPSA 2020, and the AFK Methods workshop 2019 for very helpful comments and feedback.
Competing interests
None.
