“Traditional” Resolution of Land Conflicts: The Survival of Precolonial Dispute Settlement in Burundi

Where and how have precolonial institutions of conflict resolution remained intact? Although it is often argued that “traditional” institutions can play a key role in managing communal conflicts, little is known about the conditions of their “survival.” This article argues that historical, political, and cultural topographies are essential to understanding patterns of the persistence and demise of precolonial institutions. Traditional modes of conflict resolution remain strong where they have been internalized over centuries: in the cultural and political centers of precolonial states. I use original geocoded survey data and historical spatial information on precolonial Burundi to analyze this hypothesis. The estimations yield robust correlations between the geographic patterns of the precolonial kingdom and current modes of resource-related conflict resolution.


Introduction
Violence escalated in Burundi when incumbent president Pierre Nkurunziza announced in April 2015 that he would run for a third term in the next presidential elections. Beneath ethno-political cleavages, local land conflicts have played an essential role in fomenting violence in the country. Because of Burundi's very high population density and its economic dependence on agriculture, land is a key resource: "Around Burundi, brothers are killing brothers. Sons are killing fathers. And it's all for land" (Keenan, 2015). For centuries, the precolonial institution of the bashingantahe has been the central mechanism for the peaceful management of resource-related disputes; although the institution has lost its influence across large parts of the country, in the opinion of many observers, it remains the primary communal instrument for conflict settlement and reconciliation (Dexter & Ntahombaye, 2005; International Crisis Group [ICG], 2003;Kamungi, Oketch, & Huggins, 2004). Understanding the determinants of its persistence and demise is therefore essential for understanding and effectively capitalizing on local-level capacities for conflict resolution in Burundi.
Disputes over access to resources are important drivers of local-level tensions in countries as culturally, politically, and economically diverse as Sudan, Brazil, Ghana, Nicaragua, Cambodia, Kenya, or Yemen (Adano, Dietz, Witsenburg, & Zaal, 2012;Alston, Libecap, & Mueller, 1999;Bigagaza, Abong, & Mukarubuga, 2002;Broegaard, 2005;Daudelin, 2003; for example, Flint & de Waal, 2005;Fred-Mensah, 1999;Thompson, 2010). Across these cases, local institutional configurations have been found to be essential in determining where scarcity and distributional conflicts result in actual insecurity and violence. Notably, so-called "traditional institutions" 1 have received much attention in this regard. Precolonial modes of conflict resolution are viewed by many as particularly effective in settling land disputes because they draw strength from high levels of traditional legitimacy and due to their specific restorative, consensus-based character (Boege, 2006;Yamano & Deininger, 2005;Zartman, 2000). Consequently, where they have "survived," they are deemed to be particularly well placed to prevent the escalation of resource-related communal strife (Kibreab, 2002;Tubiana, Tanner, & Abdul-Jalil, 2012). 2 Despite the academic and policy interest in traditional conflict resolution, and precolonial institutions more generally, little is known about the factors that have allowed them to withstand colonial and postcolonial reform and retain their key role in local-level governance. Focusing on the case of Burundi, this article contributes to filling this gap: Where and how do precolonial institutions of conflict resolution remain socially salient?
Academic and policy studies on communal conflict resolution tend to explain the persistence of precolonial institutions with reference to formal states' absence, weakness, or noninterference (Beall, Mkhize, & Vawda, 2005;Deininger & Castagnini, 2006;Mercy Corps, 2011;Tubiana et al., 2012). From such a perspective, traditional conflict resolution is seen to have remained intact in rural peripheries where colonial and postcolonial interventions were marginal and where state presence has always been fragile. In short, the population continues to resort to precolonial modes of conflict resolution to solve their resource-related conflicts because alternative formal institutions are not available or not effective.
Such a state-based perspective disregards the fact that many precolonial institutions are deeply entrenched in local communities, making them resilient to change even in areas of strong state presence and effective enforcement of formal state institutions. Consequently, colonial and present-day political topographies may be less relevant in explaining the patterns of persistence and demise of traditional institutions. Rather, precolonial institutions are likely to remain salient in the historical strongholds of the precolonial political and cultural systems where traditional modes of conflict resolution have been particularly deeply internalized and enculturated over several centuries.
I investigate this argument in the case of Burundi. For centuries, the socalled bashingantahe was the key conflict-resolution institution with a strong role in mediation of disputes over access to land. Colonial and postcolonial reforms substantially weakened the institution. Finally, in 2010, the bashingantahe was formally stripped of its marginal remaining functions in locallevel jurisdictions. Still, whereas in many parts of the country, people resort to formal state agents for the resolution of land-related conflicts, they continue to turn first to the bashingantahe in many others. This article investigates the determinants of the spatial patterns of these institutional choices.
I draw on three main sources for the empirical analysis: I measure variation in current modes of local-level resolution of resource-related conflicts with an original opinion survey implemented in Burundi in late 2014. I identify geographical variation in the historical strength of precolonial institutions based on published and unpublished historical works on the precolonial Burundian kingdom. The empirical analysis lends support to the argument that traditional conflict resolution remains particularly strong in the historical, political, and cultural centers of the kingdom. Current patterns of state presence and capacity, however, are very weak predictors of the persistence of the bashingantahe.
These findings add to the extant literature in three main ways: They contribute to research on communal violence by clearly showing that historical political conditions can continue to shape present-day capacities for, and patterns of, settlement of resource-related conflicts. Although spatio-temporal long-term path dependencies are widely neglected in peace research, they play an essential role in explaining how communities deal with local-level conflict. Second, by shedding light on the determinants and spatial patterns of institutional "survival," the findings can inform recent research that focuses on the effects of traditional governance on various economic and political outcomes (Acemoglu, Reed, & Robinson, 2014;Baldwin, 2013;Díaz-Cayeros, Magaloni, & Ruiz-Euler, 2014). Finally, the results contribute to analyses of the patterns of effective state penetration. People's institutional choice in favor of precolonial institutions exposes the limits of formal institutions' claims to authority. Consequently, the findings shed light on how historical factors can affect regional variations in the state's ability to impose rules that are accepted by the population (Tamanaha, 2008).

Precolonial Institutions for Conflict Resolution
Empirical studies indicate substantial cross-country variation in the salience of precolonial modes of conflict resolution. For example, in Uganda, the vast majority of land conflicts (76%) are resolved by formal rather than informal means (Deininger & Castagnini, 2006). In neighboring Kenya, however, more than 90% of the land conflicts are brought to informal institutions first (Yamano & Deininger, 2005). Finally, Liberia represents a middle ground with approximately half of litigants in land conflicts turning to precolonial conflict-resolution institutions before they access other forms of dispute settlement (Hartman, 2010).
Such cross-national differences may easily be attributed to countries' social, political, and historical specificities. Similar variations, however, also persist at the subnational level. Thus, for example, a survey of litigants in Ghana has found that depending on the sampling area, respondents where substantially more or less likely to turn to traditional institutions of conflict resolution (Crook, 2005). Similar observations were made in survey-based studies in Liberia and Uganda: The modes of conflict resolution display substantive regional variations in both countries, with precolonial mechanisms being substantially more relevant in some areas than in others (Hartman, 2010;Rugadya, 2008). What factors, then, explain these subnational differences in the persistence and social salience of precolonial conflict-resolution institutions?

State-Centered Perspectives
A straightforward answer to this question focuses on variations in the present-day capacity of formal state institutions (e.g., Ntsebeza, 2004Ntsebeza, , 2005Ribot, 2002). This perspective is based on the assumption that cost-benefit calculations determine people's behavior in contexts of institutional pluralism. Adherence to precolonial modes of conflict resolution is more costly and less attractive in areas where "modern" institutions are present and enforced (Helmke & Levitsky, 2004). People abandon traditional dispute settlement because legal provisions curtail the authority and effectiveness of informal rulings, making formal systems more advantageous: The latter are legally sanctioned, are based on written, "objective" rules, provide binding adjudication, and are backed by effective means of law enforcement. Thus, variation in the salience of precolonial norms may be a consequence of the variation in the availability and capacity of formal systems of conflict resolution.
Although formal modes of conflict management are theoretically uniformly accessible, their presence and enforcement varies starkly within many countries, depending heavily on the state's infrastructural penetration and reach (González & King, 2004;Mann, 1984). Legal reforms and institutions need to be implemented and monitored on the ground (Levitsky & Murillo, 2009). The greater the distance to administrative centers, the lower the state's ability to effectively ensure availability and enforcement (Herbst, 2000;Soifer, 2006). Thus, the persistence of precolonial conflict-resolution mechanisms is most likely to occur in pockets of state weakness-peripheral areas where formal state institutions are not present, not capable, or not effectively enforced (Beall et al., 2005;Helmke & Levitsky, 2004;North, 1990;Tajima, 2013).
The main shortcoming of this state-centered perspective is that it focuses one-sidedly on factors that counteract and displace precolonial institutions of conflict resolution. Conversely, it neglects conditions that may stabilize and preserve precolonial practices. In the remainder of this section, I will present an alternative theoretical perspective highlighting path dependencies of precolonial political and cultural topographies. Drawing heavily on previous theoretical and empirical works in historical and sociological institutionalism (Hall & Taylor, 1996), I first suggest a general norm-based argument on the long-term persistence of precolonial institutions of conflict resolution before I discuss its potential scope conditions. I apply the argument to the case of Burundi and address and investigate potential alternative mechanisms of institutional stickiness in the Exploring the Mechanism section below.

Norm-Based Arguments on Institutional Persistence and Change
In many countries, colonial and postcolonial reforms may have abolished precolonial organizations, but they did not affect underlying precolonial norms in a similar way (Mahoney & Thelen, 2009;Roland, 2004;Streeck & Thelen, 2005). Norms are sticky and can hardly be changed in an authoritative and intentional manner. If they change at all, they change via longer term processes (Roland, 2004). Thus, in many cases, precolonial norms may continue to determine people's perception of what is the "right" form of conflict resolution and thereby affect institutional choices. From such a perspective, precolonial modes of conflict resolution remain strong not where alternative formal institutions are absent, but where precolonial dispute settlement is particularly deeply entrenched in local traditions and culture.
Traditional norms derive their strength from precolonial social and political systems that are still imprinted in local customs, myths, and rituals (Lund, 2006;Miller, 1968;Morapedi, 2006). However, similarly to "modern" institutions, although precolonial rules of conflict resolution may, in principle, have applied uniformly within precolonial political entities, their enforcement and resulting internalization was often not uniform across all of the territory. Consequently, I argue that the salience of precolonial norms will tend to vary geographically, mirroring the spatial unevenness of the respective precolonial states.
Precolonial state capacity was strongest in the political centers of the precolonial states and often marginal in peripheries (Herbst, 2000). In the centers, power-holding elites were most effective in ensuring compliance with precolonial institutions, thereby "regularizing" and stabilizing them (Falk- Moore, 2000). Long-term enforcement rooted precolonial conflict-management institutions in these local communities (see also Douglas, 1986;Giddens, 1984), and over centuries, these institutions became regular social practices perceived by the local populations as the "appropriate" modes of dispute settlement (March & Olsen, 2010;Powell & DiMaggio, 1991). Other institutional designs came to be seen as less suitable, limiting the menu of acceptable institutional alternatives (Bell, 2011;for example, Knill & Lenschow, 2001;Peters, Pierre, & King, 2005). Norms for conflict resolution were passed on from generation to generation: "The young are enculturated by the previous generation, while they in turn enculturate the next generation" (Zucker, 1977, p. 728). Strong ideational foundations supported the process of intergenerational transmission. Sacred sites were the locations of major burial, coronation, or marriage rituals. These cultural centers acted as physical manifestations and persisting reminders of related myths, traditions, and symbols, ensuring the recurrent collective reproduction of the spiritual foundations of the precolonial norms (Lund, 2006). Thus, internalization and enculturation were likely strongest in the vicinities of the precolonial political and cultural centers and lowest in the peripheries of the precolonial states.
These general norm-based arguments lead to this article's main hypothesis: Hypothesis 1: Subnational variations in precolonial institutions' persistence stem from the geographical setup of the precolonial state, with precolonial institutions being more likely to persist in the former strongholds of the precolonial state, where associated conflict-resolution norms are most strongly entrenched in local communities.

Scope Conditions of the Norm-Based Argument
This main hypothesis rests on the assumption that effective precolonial norm enforcement can foster processes of norm internalization. This, however, may only be plausible under certain conditions. Most important, effective norm enforcement requires a certain degree of political centralization and state capacity. Many precolonial states, however, were rather fragmented with little state control over territories outside of individual political centers-arguably due to high costs and limited benefits of extending power into peripheries (Herbst, 2000;Scott, 2009). Moreover, processes of internalization may have been impeded by specific precolonial settlement patterns. First, as agued by Herbst (2000), people's mobility and distribution across vast territories constituted challenges to precolonial statehood, reducing the ability of power holders to effectively enforce their rule. Second, enculturation of shared norms is less likely to develop among widely dispersed and nomadic groups with little regular interaction among each other.
Thus, the suggested long-term effects of precolonial political topographies are likely to be strongest and most relevant in regions that belonged to comparatively centralized and compact precolonial political systems. I draw on data provided by Murdock (1967), Gennaioli and Rainer (2007), and Nunn (2008) to show that these characteristics apply to a significant population of cases in sub-Saharan Africa. Murdock provides information on various characteristics of African ethnic groups before European colonization; Nunn (2008) has geocoded the respective ethnic settlement areas. Using these data, I create three binary explanatory variables: one for centralized precolonial entities using Gennaioli's and Rainer's (2006) measure ("large paramount chiefdoms/small states" and "large states"), another one for compact and relatively permanent settlement patterns ("compact and relatively permanent" and "complex settlements"), and a combined variable identifying ethnic groups displaying a high level of precolonial ethnic centralization and compact, permanent settlements.
Around 25% of all ethnic groups for which the relevant information is available in the data (386) can be considered "centralized" in precolonial times according to the measure introduced above (incl. Burundi); 54% display "compact and relatively permanent" or "complex" settlement patterns. Around 20% of the groups fulfill both these conditions. They comprise, for example, the Ashanti in Ghana, the Fur in Sudan, or the Songhai in Mali. Thus, although I focus my main empirical analysis on the case of Burundi as an example of a particularly highly centralized precolonial state, I believe that the findings can inform analyses of conflict resolution in a substantially wider population of countries.

The Case of Burundi
Burundi is one of the most densely populated countries in sub-Saharan Africa. Its economy is highly agriculture dependent (van Leeuwen & Haartsen, 2005). The high value and scarcity of land has been accompanied by persisting neopatrimonial mechanisms of land allocation, insufficient land tenure rights, and waves of conflict-induced migration and repatriation, all of which have combined to generate conflicts that have often become violent (ICG, 2014;Matignon, 2015). There is no precise estimate of the total number of land disputes in Burundi. The fact, however, that more than 70% of all cases filed with local communal courts pertain to land issues (Kohlhagen, 2009a), testifies to the high social and security relevance of land conflicts in Burundi.
Land conflicts are officially within the jurisdiction of local courts (see, for example, Kohlhagen, 2009a), but the formal judicial system is weak, suffering from deficits in terms of material and human resources. On average, communal courts are staffed with only eight judges per 100,000 inhabitants (Kohlhagen, 2009a). Magistrates tend to be secondary school graduates with only 6 months of legal training (Dexter & Ntahombaye, 2005). More important, courts are widely viewed as being corrupt and clientelist, which undermines legitimacy in the eyes of the population (Dexter & Ntahombaye, 2005;Kohlhagen, 2008;van Leeuwen & Haartsen, 2005).
In this context, the traditional bashingantahe institution is seen by many as an important local resource for conflict resolution. As a report by the ICG highlights, "the bashingantahe institution is the only one whose involvement in the sustainable restoration and maintenance of a fair and equitable order (intahe) can be useful for the peace process and the political neutralization of the land time bomb" (ICG, 2003, p. 12). Similarly, other policy-oriented studies stress that the bashingantahe represent a key local capacity for peacefully resolving land disputes (Dexter & Ntahombaye, 2005;Kamungi et al., 2004). The following subsection provides additional information on the background and evolution of the bashingantahe.

The History of the Bashingantahe
From the establishment of the Burundian kingdom in the 17th century, the bashingantahe were the primary moral and judiciary authorities. They were local notables charged with preserving unity and harmony through counseling, mediation, and reconciliation pertaining to all kinds of local-level conflicts. Etymologically, the term mushingantahe (singular of bashingantahe) derives from an expression that translates as "knocking the bar on the ground" and from "gushing intahe," which means to decide something. The name can be traced back to the fact the bashingantahe used to carry a stick made from holy wood that symbolized their administrative and moral authority (Ingelaere & Kohlhagen, 2012;Laely, 1995). Early on, the institution of the bashingantahe was introduced across all of the territory of the precolonial kingdom as a consequence of bottom-up institution building and top-down attempts at establishing uniform and decentralized institutions of conflict resolution (Dexter & Ntahombaye, 2005;Laely, 1995).
From the beginning of colonialization, the bashingantahe underwent a process of transformation and marginalization. The Belgian administration initially relied heavily on the bashingantahe. With time, however, the regime established indigenous tribunals with state-appointed judges who gradually took over functions traditionally carried out by the bashingantahe (Deslaurier, 2003). Eventually, the institution was formally integrated into the colonial judiciary system and reduced to a mere witness and juror in tribunals held under the auspices of local chiefs (Gahama, 1983;Laely, 1995;Weilenmann, 1996).
In 1966, only 4 years after independence, the Burundian monarchy was formally abolished, together with its local governance system. This stripped precolonial conflict resolution of the remnants of its institutional foundations. The monarchy was replaced by a military regime that politicized and instrumentalized the bashingantahe for its own ends. The institution briefly regained some acceptance in the 1990s, but the postwar government-which aimed to establish full control over the country-was wary of this parallel source of authority (Deslaurier, 2003;Shaw, Waldorf, & Hazan, 2010). In 2005, the Burundian constitution was reformed. Contrary to previous versions, it no longer made any reference to the bashingantahe. Finally in 2010, with the reorganization of the communal administration, the bashingantahe was stripped of its remaining, and already marginal, legal basis (Ingelaere & Kohlhagen, 2012).
Even though the bashingantahe was progressively deprived of its organizational and legal foundation, it never ceased to play an important role in the eyes of the local population. A survey conducted in 1956, toward the end of the colonial period, revealed continued support for the bashingantahe as the "natural counselor" of the population (Gahama, 1983). Similarly, in the postcolonial period, the bashingantahe has remained a respected institution of substantial moral authority (Deslaurier, 2003;Ingelaere & Kohlhagen, 2012). This enduring social salience, however, is not uniform across the country. Ethnographic works indicate that precolonial modes of conflict resolution remain particularly relevant in the central regions of Burundi, where the political and cultural centers of the former kingdom were located (Kohlhagen, 2009a;Weilenmann, 1996). This specificity can be traced back to the bashingantahe's strong link to the king. The bashingantahe spoke their rulings in his name and acted as his intermediaries (Laely, 1995). This strong association is best expressed by the verse that was exclaimed when the bashingantahe were inducted: "Receive this stick of justice that has been given by Ntare Rushati [the first king of Burundi in the seventeenth century] to your father and your grandfather" ("Recois ce bâton de justice qui a été donné par Ntare Rushati à ton père et à ton grandpère"; Mworoha, 1987, p. 209). Thus, it is reasonable to assume that the institution of the bashingantahe was particularly strongly enforced and embedded in the political and cultural center of the kingdom, and that this is the region where the collective memory and normative foundation of the precolonial order remains particularly vivid until today (Kohlhagen, 2009a;Weilenmann, 1996). The following subsection briefly introduces the specific political topography of the precolonial state.

The Topography of the Burundian Kingdom
In the 16th century, what was to become the kingdom of Burundi consisted of two main centers of political power. In the 17th and early 18th centuries, these centers merged. This moment has been understood as the actual founding moment of the Burundian kingdom. In the following years, the first king, Ntare Rushatsi, began to establish a number of key political, ritual, and economic domains delineating what was to remain the center of the precolonial kingdom for centuries to come (Chrétien, 1993;Laely, 1995;Mworoha, 1987;Vansina, 1972).
Two longer term trends were essential in determining the location of the political and cultural center. First, recurrent droughts in the 17th century had led to population movements from eastern regions into areas of higher precipitation to the west. Second, raids by the northern Banyoro tribe triggered population movements to the south (Chrétien, 1993;Laely, 1995;Mworoha, 1987). These movements led to a concentration of population and economic activity in a region bordered by natural landmarks and obstacles-namely, the hills of Ruvubu to the north, the mountains of Kibira to the west, and the river Ruvjironza to the southeast (Laely, 1995;Mworoha, 1977). The upper left map in Figure 1 illustrates the original two centers of power and these three essential landmarks that demarcated the political and cultural center of the kingdom.
Within this region, subsequent Burundian kings established so-called "royal domains" that constituted the sociopolitical capitals of the kingdom (see Mworoha, 1987, for discussions of the concept of "royal domains"). The most important domains were the political ones, where the king spent most of his time. Other domains were needed for the kingdom's economic production (e.g., specific pastures). Finally, the cultural domains were the locus of all essential rituals, especially burials and coronations. Two largely idiosyncratic factors played an essential role in determining the exact location of these domains: The first was the location of trees/forests deemed sacred in the royal cult. These could have been single, very large, and isolated trees or the areas where particular plants needed for the production of artifacts used in sacred rituals grew (Mworoha, 1987). The second was the residences of the kings' spouses (Laely, 1995).
The resulting geographical structure of the early kingdom created strong path dependencies. Despite major social, political, and economic changes as well as the substantial geographical expansion of the kingdom, the location of the core center of power and of individual royal domains remained largely unchanged. Although every new king founded new domains, the majority of sites were "inherited" from predecessors (up to 75% according to Mworoha, 1987; see also Laely, 1995;Vansina, 1972). In the early 20th century, two thirds of the royal domains were still located in the original center of power. We can see in the maps in Figure 1 that the main natural landmarks mentioned above continued to define the outer boundaries of the kingdom's center; the major sites of political power remained within these core areas (Vansina, 1972). Consequently, it was largely exogenous factors that led to the specific spatial pattern of the precolonial kingdom that was to last from the 17th to the early 20th century.
Within few decades of colonial and postcolonial rule, the political topography of Burundi was systematically reshaped. The colonial state reorganized the territorial setup of the country, systematically stripping former royal domains of their specific and privileged political position. The German regime established its regional political capital in Usumbura, to the west of the former royal center. Usumbura remained the regional capital of the Belgian colony of Ruanda-Urundi and later became today's capital city Bujumbura. The capital of Burundi and the residence of the colonial administration was established in Kitega, southeast of the royal center.
The maps (a) and (b) of Figure 2 display the fundamentally distinct logic of political organization of the precolonial and the colonial state: Darker shades indicate longer distances to the nearest royal domain (late 19th century) and the nearest administrative headquarter (1930; "Chef-lieu de territoire"), respectively. Contrary to the concentration of political power in a naturally grown center in the precolonial area, we see that colonial agents established administrative hubs in a way that partitioned the colonial territory into roughly equal-sized administrative units ("Territoires"). Map (c) shows that this logic was carried over into the postcolonial period where the number of regional provincial capitals increased in an attempt to further strengthen penetration. Thus, already few years after the beginning of the Belgian colonial period, little was left of the specific political topography of the precolonial area.
Before I move on to investigate, if the spatial patterns of precolonial statehood nonetheless continue to shape institutional choices in conflict resolution, I draw on unpublished colonial sources to assess the underlying claim of my main hypothesis that the center of the precolonial kingdom did indeed originally differ from its periphery in terms of power relations and the salience of the precolonial norms.

Specificities of the Political and Cultural Center of the Kingdom
In 1929, the Belgian colonial government undertook an administrative survey. The exercise aimed to collate systematic information on the colony's territory. Among other things, it included assessments of individual chiefs' attitudes toward the colonial government and missionaries, as well as of their leadership skills and their attachment to the "traditional beliefs" that constituted the spiritual foundations of the kingdom. The assessments were carried out by Belgian administrators who had limited knowledge of the local context and who were deeply resented by many of the chiefs. Nonetheless, considering the key social, economic, and political role of the chiefs, their alleged attitudes and relations to the population provide rare insights into the social and political conditions of the kingdom at the end of its sovereign existence.
Copies of the original colonial assessments are included in the J.M. Derscheid collection hosted by the University of Florida Digital Collection. 3 The collection contains a total of 66 assessments. I have been able to attribute 56 of them to specific geographical locations using maps of chieftainships produced by Gahama (Gahama, 1983). This allows me to compare the assessments of chiefs from within the core areas of the precolonial kingdom with those of chiefs from the periphery. To allow for more systematic comparisons, I have created a small data set based on information coded from the qualitative chief assessments. Table 1 shows the results of a number of simple t tests. Although the results have to be treated cautiously due to the very limited number of observations and the purely bivariate nature of the correlations, they do allow for some assessment of the plausibility of the argument that specific norms and power relations were particularly strong in the center of the precolonial kingdom even before the first massive colonial interventions in 1929.
There does not seem to have been any systematic difference across the two groups with regard to the personal characteristics of the chiefs, such as age, literacy, or ethnic identity. Similarly, on average, chiefs within and outside the royal area barely differed in terms of the number of years they had been in office when the survey was carried out or the period of time they remained in office after the survey. 4 However, according to the Belgian assessments, chiefs within the royal areas were more strongly attached to the "traditional" non-Christian belief system associated with the precolonial political order (80% within the royal areas as compared with 40% in the periphery). The survey also tried to gauge the quality of local-level power relations. The assessments feature information on the acceptance of chiefs by their respective constituencies. Whereas all chiefs within the core royal areas were credited a certain degree of local authority, nearly 20% of the chiefs outside of the royal areas lacked local legitimacy according to Belgian administrators (as measured, for example, by the population's complaints on their chiefs filed with the colonial administration). Thus, on average, the spiritual foundations of the precolonial system of governance seem to have been more strongly embedded in the center. Moreover, local-level authority sanctioned by the king was more easily accepted within the core areas of the kingdom.
Although the survey does not include any information on the bashingantahe itself, the patterns presented above lend some additional support to my underlying assumption that specific norms and power relations associated with the precolonial kingdom were more strongly entrenched in the core areas of the regime, thus contributing to the internationalization and enculturation of norms associated with the kingdom-such as the resolution of conflict by the bashingantahe.
The following section provides an empirical analysis of the argument that these specificities continue to affect present-day spatial patterns of conflict resolution.

Empirical Analysis
My empirical analysis relies primarily on two sources. The first is an opinion survey we undertook in 2014 in Burundi. 5 A stratified two-stage design was used to select a total of 100 rural communities (based on the 2008 Population and Housing Census). Across these communities, a total of 1,500 households were drawn randomly from household lists maintained by village chiefs. Survey instruments were translated from English into French and Kirundi and back-translated for validation. In addition to the individual-level survey, we undertook an accompanying village-level survey, collecting various information on the sampled communities such as distances to specific infrastructure, development projects, and local conflicts. The data were collected through group interviews with local elites (chiefs, teachers, members of the local administration). Data collection took place with hand-held devices that were also used to geolocate interview sites using precise longitude and latitude measurements. The second main source is historical information drawn from ethnographic works on the precolonial kingdom. The Burundian historian Émile Mworoha prepared detailed accounts of the early genesis of the Burundian state (Mworoha, 1987). His work includes a number of widely cited geographical representations of the state-building process featuring information on the primary royal domains and the approximate outer boundaries of the kingdom's core areas.
The following three subsections provide information on the concrete operationalization of my main outcome, explanatory, and control variables.

Outcome Variables
To measure the main outcome variable, I have relied mainly on one item from the abovementioned survey. The respondents were asked, "If you were involved in a conflict about natural resources (water, land, pasture, forest), what are the first institutions you would turn to resolve the conflict?" The first three responses were recorded by the interviewer and ranked according to their sequence. Figure 3 illustrates the response patterns across actors and in geographical terms. We see that the majority of respondents would first turn to the chef de colline, who is the lowest level state actor (locally elected, paid by the state). However, approximately one fifth of the population would turn first to a bashingantahe to resolve a conflict related to natural resources. My main outcome variable is coded "1" for all respondents who would first turn to the bashingantahe and "0" otherwise.
Looking at the right-hand map in Figure 3, we can see that the prioritization of the bashingantahe seems to be clustered. In line with my theoretical expectations, the village-level average of people ranking the bashingantahe as the most relevant actor in local-level conflict resolution appears to be higher in areas closer to the original center of the precolonial kingdom than in other, more peripheral regions of the country.

Explanatory Variables
I want to investigate whether the response patterns of people within the center of the precolonial kingdom really differ in a significant way from those of people in other areas of the country. To do so, I rely on the aforementioned ethnographic works on the precolonial kingdom to operationalize my main explanatory variable.
I first create a binary variable that spatially assigns each village in the sample to either the precolonial kingdom's center or its periphery (see the dark-red zone in the middle of the map on the left side of Figure 4). The second indicator considers additional variation outside the kingdom's center. On behalf of the king, the periphery was administered mainly by descendants of one of the two main aristocratic lineages, the Bezi and the Batare (Gahama, 1983). The Bezi controlled a north-south axis in the middle of the kingdom, whereas the Batare where dominant in the eastern periphery. In some mostly western and northern regions, power was transferred to local elites in families with past ties to the aristocracy or to those in families without any direct ties to the main royal lineages. I have created binary variables for the main zones of "sociopolitical proximity" to the kingdom's center of power, namely, areas under control of families without ties to the aristocracy, zones under control of the Batare and those regions controlled by the Bezi (Weilenmann, 1996).
The map on the right-hand side of Figure 4 displays the location of the kingdom's main royal domains in the 19th century. I use each sampled village's distance to the nearest royal domain as a third explanatory variable. The main models presented below focus on the distinction between the kingdom's center and its periphery, while the results of the two alternative measurements are presented in the "Robustness Checks" section.

Control Variables
I include a number of control variables in the model to consider potential alternative determinants of people's assessment of traditional conflict-resolution institutions. First, I include two "historical controls" intended to capture pretreatment conditions before the establishment of the Burundian kingdom that may have affected the location of the royal domains and various current outcomes. As emphasized above, climatic conditions played a role in the formation of the early kingdom. As climate data are not available for the 17th century, I proxy historical conditions with precipitation data for the period 1950 to 2000 from the Worldclim data set. I also include dummy variables for respondents' location within the former territory of the scattered political entities that preexisted the foundation of the kingdom of Burundi and were later partly integrated into the kingdom. Next, I include two "colonial controls." Precolonial institutions may have been able to remain stronger in regions that were less affected by colonial interventions. Moreover, such interventions may have been shaped by precolonial institutional characteristics. I therefore control for administrative and missionary interventions. From 1929 on, the Belgian regime intervened massively into local political and social organization, replacing the majority of the established chiefs with others deemed more suitable and loyal to the colonial administration. Chief replacements most often lead to the simultaneous ouster of associated local elites such as the so-called sous-chefs (secondary chiefs) and bashingantahe (Gahama, 1983). Gahama (1983) provides maps of the approximate locations of all individual chieftainships as well as the names of chiefs in power in 1929, 1933, 1937, 1945, and 1954. I have georeferenced and digitized these maps. With this information, I have been able to attribute the total number of colonial interventions in terms of chief replacements that took place in the area of each surveyed village across these five distinct periods of time.
The administrative reforms were accompanied by increasing missionary presence and activity. Missions had the explicit objectives of replacing traditional beliefs with Christian ones and of alienating people from precolonial power holders (Mvuyekure, 2003, p. 117). Given the comparatively low number of Protestant missionaries in Burundi, this function was mainly carried out by the Catholic mission. To control for variation in exposure to missionary activity, I draw on information on the locations of Catholic missionary stations. The information stems from a map in the Atlas du Burundi (Association pour l'Atlas du Burundi, 1979) that displays missionary stations existent up to 1970, a few years after independence. I identify the nearest missionary station for all villages included in the survey. I use the distance to the nearest station to gauge exposure to missionary activity.
Finally, I include a number of "current controls" from the 2014 opinion survey. These are the usual demographic variables such as age (and age squared), gender, and two dummy variables for levels of education (no formal education and completed primary education). I also include a dummy variable for self-reported civil-war exposure, an index of material asset possession (additive: generator, car, motorbike, bicycle, television, radio) and a Likert-type scaled variable on self-reported material well-being to control for the possibility that people's preferences in terms of institutional modes of conflict resolution are shaped by previous victimization or by their economic situation. I draw on the village-level survey to proxy state presence and accessibility of formal institutions of conflict resolution, in line with the alternative explanation for the persistence of traditional institutions presented above. I control for information on the travel distance to the nearest police station, the nearest court of justice, the respective provincial capital, and the national capital Bujumbura.

Models and Results
To deal with a binary outcome variable, I use simple logistic regressions with standard errors clustered by province to account for heteroskedasticity. Table 2 summarizes the results (detailed tables including coefficients of all control variables are provided in the online appendix).
The first column presents the results of a simple bivariate model, comparing interviewees' responses within and outside the center of what was at one time the precolonial kingdom. The second model approximates a simple matching approach. Rather than comparing respondents across the country, I create a subsample including only respondents within and near (less than 10 km from the outer limits of) the kingdom's center. This helps me to implicitly control for factors that cluster spatially. The third model includes two "historical" (pretreatment) controls. The fourth model adds controls for colonial interventions. The fifth model additionally includes control variables drawn from the individual and village-level surveys. In the final model, I also include province-level fixed effects to account for any unobserved variation across provinces. Most notably, this should help me account for any variation in the effectiveness of postcolonial administrative reforms that undermined the bashingantahe and strengthened formal state institutions of conflict resolution. Models 3 to 5 should be interpreted with some caution. They include controls that have been measured posttreatment meaning that they are endogenous to precolonial institutions. The results presented in Table 2 confirm the expected positive correlation between respondents' location within the center of what was formerly the precolonial kingdom and their appraisal of the bashingantahe's role in locallevel conflict resolution. Across all the models, I find a statistically significant correlation in the expected direction. Proxies for colonial and postcolonial state presence and penetration fail to account for the variation in local-level conflict resolution. None of the variables considered turns out to be correlated with modes of conflict resolution at conventional levels of statistical significance. Detailed results of the models are presented in the online appendix (Table A1).
Due to the nonlinear nature of the specifications, we cannot judge the substantive significance of the associations. I therefore estimate the average marginal effects of the respondents' location within the center of the precolonial kingdom based on the likelihood that interviewees would first turn to the bashingantahe in cases of conflict over natural resources. Depending on the model, respondents within the center of what was the precolonial kingdom are 11% to 18% more likely to make use of precolonial conflict-resolution institutions than respondents in other regions of the country. The effect size of my main explanatory variable is barely affected by the inclusion of colonial and current controls. It is only 1% to 2% lower than in models including historical controls only, making it appear implausible that precolonial topographies affected current modes of conflict resolution only by shaping colonial and postcolonial state building.

Robustness Checks
The correlations presented above lend support to the article's main argument. To further scrutinize these results and address some of the shortcomings associated with the specifications of the main models, I have implemented four robustness checks.
To further scrutinize whether my findings actually mirror the long-term effects of the precolonial state rather than other (random) spatial patterns in the data, I have randomly relocated the polygon representing the center of the precolonial kingdom. This has resulted in four new treatment and control groups equaling the sample sizes of the original ones. I have estimated all the main models using these alternative explanatory variables. I do not find similar patterns of systematic variation of the outcome variable across treatment and control groups in any of them (see online appendix , Tables A2-A5).
Next, I have estimated models with alternative proxies for my main explanatory and outcome variables. I have reestimated models with the two alternative specifications of the explanatory variable outlined above: the role of precolonial sociopolitical hierarchies (online appendix, Table A6) and the distance to the nearest precolonial royal domain (online appendix, Table A7). I have also probed alternative outcome measures. I have first created an analogous binary variable that is positively coded if the bashingantahe is among the first three actors interviewees would turn to in cases of land conflict (online appendix, Table A8). Second, I have used people's subjective assessments of the bashingantahe's contribution to the local security situation as an alternative outcome (online appendix, Table A9). All these specifications yield results that mirror the main findings presented above.
All the previous measurements of my outcome variable rely on individual assessments drawn from the household survey. Moreover, the previous models include only rather crude proxies for the features of the formal legal system-namely, distances to courts and administrative centers-that may not do justice to the main state-based counterargument presented above. I rely on an additional data set to assess an alternative indicator of the social salience of the bashingantahe and to control for more nuanced characteristics of the formal legal system. A descriptive statistical overview of the Burundian judicial system provided by Kohlhagen (2009b) contains rich information on the resources and practices of a sample of 51 of 129 communal courts (focusing on the period 2006-2008; the number of communal courts has changed since 2009). I have geographically attributed the respective communes to either the center or the periphery of the precolonial kingdom. I rely on information on a specific legal practice as proxy for the social salience of the bashingantahe per commune: A legal reform implemented in 1987 stipulated that prior to submitting a case to the communal courts, litigants needed to submit their case to the bashingantahe. Only if the claimant did not agree with the solution proposed by the bashingantahe was he or she entitled to present the case to the court. Judges had to keep the minutes of the proposition made by the bashingantahe as part of the documentation of the case (Kohlhagen, 2009b;van Leeuwen & Haartsen, 2005). Although these specific requirements were formally abolished in 2005, many communal courts continue to voluntarily collect and document the minutes, indicating that solutions proposed by the bashingantahe are deemed particularly relevant in the respective regions (Kohlhagen, 2009b). The share of cases with documentation containing these minutes serves as my alternative proxy. Table 3 provides the results of linear regressions controlling for a number of court characteristics that more persuasively represent the capacity of the formal legal system than distances alone-namely, the number of judges per population, delays in court rulings, and delays in the enforcement of judgments. The results lend further support to the argument that the precolonial topography plays an essential role in explaining the persistence of precolonial conflict-resolution institutions: Courts within the former royal center have been collecting and archiving minutes of meetings with the bashingantahe much more systematically than courts in the former periphery of the precolonial kingdom. This correlation is independent from the capacities and performance of the formal courts. Standard errors clustered at the province level; standard errors in parentheses. OLS = ordinary least squares. † p < .10. *p < .05.**p < .01. ***p < .001.

Exploring the Mechanism
I have highlighted a norm-based explanation of the long-term transmission of precolonial political and cultural topographies in the Norm-based Arguments on Institutional Persistence and Change section. However, at least two other mechanisms may equally well explain the general correlations presented above: The bashingantahe may have survived as a consequence of active state promotion or because the institution represented an equilibrium of interests that prevented substantive institutional change. In this section, I will introduce both these alternative explanations and discuss their plausibility in the light of available qualitative and quantitative evidence.

Institutional Persistence as a Consequence of State Promotion
From the first alternative perspective, it may in fact have been the colonial and postcolonial states that contributed to preserving the precolonial institution. As argued above, both regimes as well as individual ruling parties have pursued a dual strategy of weakening and instrumentalizing the bashingantahe. Attempts at integrating the bashingantahe into the state judiciary and using it as a political tool on the local level may have contributed to sustaining it (for similar arguments in other contexts, see, for example, Boege, Brown, & Clements, 2009;Muriaas, 2011;Ntsebeza, 2004Ntsebeza, , 2005Ribot, 2002;Ubink, 2008). Potential spatial variation in the intensity of active promotion may then explain geographical patterns of the present-day salience of the bashingantahe. If this alternative argument would hold true in the case of Burundi, we would expect to see independent or conditional effects of state strength and/ or political party influence on the salience of the bashingantahe in the quantitative analysis: A positive correlation or interaction effect would reflect the assumption that the bashingantahe has been sustained were the state or ruling parties have been capable of effectively promoting it. A negative correlation or interaction effect would result from the state's or parties' promotion of the bashingantahe where own institutions and political influence were weak. I find little evidence for any direct or indirect effects of postcolonial state presence (see main effects in Table A1 and interaction effects in Tables A10 and A11) or party influence (using commune-level election results of Union for National Progress [Union pour le Progrès national, UPRONA] in 1993 and CNDD-FDD [Conseil National pour la Défense de la Démocratie-Forces pour la Défense de la Démocratie] in 2010; see Tables A12-A15). Colonial missionary and administrative interventions, however, seem to mitigate the association between precolonial topographies and present-day conflict resolution (see Tables A16 and A17). I turn to qualitative information to assess whether this finding does indeed reflect the consequences of deliberate attempts at strengthening the institution in areas of relative state weakness.
Colonial and postcolonial state policies toward the bashingantahe were highly inconsistent, oscillating between politicization, negligence, sidelining, and active suppression. Apart from short-lived phases of meaningful state promotion in the 1980s and 1990s, colonial and postcolonial attempts at integrating the bashingantahe into state and party structures were accompanied by even more pronounced measures, stripping the institution of its main functions (Deslaurier, 2003). The colonial Belgian administration was aware of the bashingantahe's influence and systematically tried to undermine it (Laely, 1995). It therefore turned the institution into a subordinate state agent transferring its original competencies to newly created institutions. The regime also broke with the tradition of community-based investitures and turned toward top-down political appointment to detach the bashingantahe from local communities (Laely, 1995). The ruling UPRONA continued this approach transforming the bashingantahe from a locally rooted institution into a tool of the ruling party elite (Deslaurier, 2003;Dexter & Ntahombaye, 2005). Thus, overall qualitative studies strongly suggest that colonial and postcolonial policies have contributed to the bashingantahe's loss of responsibilities, influence, and acceptance rather than to its persistence (e.g., Dexter & Ntahombaye, 2005;Kohlhagen, 2009b;Laely, 1995). The observable interaction effects of colonial policies are, therefore, more likely to indicate that administrative interventions have affected interests and capabilities of influential elites (in line with the equilibrium-based perspective; see below) and that missionary interventions have undermined norm transmission through education and proselytization (in line with the norm-based perspective).

Institutional Persistence as a Consequence of an Equilibrium of Interests
Norm-based arguments on institutional persistence are often contrasted with an equilibrium-based perspective (Hall & Taylor, 1996;Knill & Lenschow, 2001;Mahoney & Thelen, 2009;Streeck & Thelen, 2005;Thelen, 1999). The latter has its roots in a rational-choice approach: Institutions are created to serve the interests of actors and represent an equilibrium of their cumulative costs and benefits. Consequently, institutions persist as long as either a majority of individuals and groups or particularly influential elites perceive of the costs of leaving the equilibrium as being higher as the potential benefits (Hall & Taylor, 1996;Mahoney, 2000). Applied to the main research question of this article, one may argue that precolonial institutions were able to persist where they continued to be perceived by the local population and/or influential elites as the most beneficial arrangement.
Both, the equilibrium-based and the norm-based arguments are in line with the findings of the quantitative analysis presented above. I therefore need to identity additional empirical tests that generate evidence that is more compatible with the observable implications of one argument than with those of the other. I exploit variation in the duration of exposure to the precolonial state to provide for such a test. More specifically, I compare the effects of strong long-term exposure with the precolonial state with the effects of strong short-term exposure. The duration of exposure is essential from a norm-based perspective that highlights long-term processes of norm internalization. There is, however, little reason to believe that duration of exposure should matter from the equilibrium-based perspective that highlights the interests and capabilities of influential actors.
Relying on the ethnographic and historical sources introduced above, I have geolocated precolonial royal domains in the 16th, 17th, and 18th centuries. I assigned all surveyed villages into three groups: (a) communities that have never been strongly exposed to the precolonial state (i.e., they have always been more than 10 km away from the nearest royal domain); (b) communities that have been "newly" exposed in the 19th century but were apart from royal domains in the 16th, 17th, and 18th centuries (i.e., new royal domains were established within a radius of 10 km from these villages in the 19th century); and (c) communities that have continuously been exposed to the colonial state throughout the entire precolonial period of the kingdom from the 16th to the 19th centuries (i.e., throughout this time period, royal domains were present within a 10-km radius from respective villages). Limiting the sample to these three groups, I have estimated effects of "new" exposure and continuous long-term exposure on the salience of the bashingantahe. I find only the latter's coefficients to be statistically significant (see Table A18), lending support to the norm-based explanation of institutional stickiness: Norms of conflict resolution have only taken root where the precolonial state had been in existence for an extended period of time. I now turn to qualitative information to further scrutinize both arguments.
Careful qualitative analyses consistently stress that the institution of the bashingantahe has profoundly changed as a consequence of colonial and postcolonial interventions. Most notably processes of training, selection, and investiture were undermined and adapted over time reshaping the institutional manifestation of the bashingantahe (for similar arguments in other contexts, see Davidson & Henley, 2007;Moore, 1986). Most influential elites who originally enforced and sustained the institution were forcefully replaced-not only in the center of the kingdom but eventually across all of its territory. Regime transitions also substantially influenced local elites' allegiances, incentive structures, and capabilities (Laely, 1995). As a consequence, costs and benefits associated with the institution of the bashingantahe changed relative to newly created colonial and postcolonial institutions. From a purely equilibrium-focused perspective, macropolitical changes and their effects on actors' cost-benefit calculations led to the erosion and eventual abandonment of the bashingantahe in favor of a new equilibrium deemed more advantageous for influential actors and/or the population at large.
Although the formal and material foundations of the precolonial kingdom have been abolished, the locations, principles, and beliefs of the precolonial cultural and political system remained deeply rooted in people's collective memory (Ntabona, 2002). It has been highlighted that geographic characteristics of the country have been particularly conducive to this preservation: Socialization has been strongly circumscribed to individual collines leading to a particularly homogenized and effective internalization (Laely, 1995). Fieldwork and survey-based research shows that the bashingantahe survived as a specific principle that continues to determine to a great extent the expectations people have of one another and of the "right" ways of settling conflicts (Ingelaere & Kohlhagen, 2012;Uvin, 2008). Thus, the bashingantahe's influence is not based on people's rational assessment of the potential consequences of defiance but on the internalization of the underlying values and principles. It is the "the fear of being marginalized or, simply, being considered ridiculous" that makes people turn to the bashingantahe rather than to other institutions (Ingelaere & Kohlhagen, 2012).

Generalizability
In how far can the general findings of this study be transferred to other cases? Similar analyses in other country contexts are needed to assess generalizability. Such an extended analysis is beyond the scope of this article. I instead assess the plausibility of a key observable implication of my main argument in a larger set of countries. As argued above, my hypothesis rests on the assumption that strong and concentrated precolonial political systems foster the enculturation of norms that, in turn, allow precolonial institutions to persist. If this would hold true beyond the case of Burundi, I should be able to observe-across country contexts-that the present-day salience of precolonial institutions is stronger in areas that were formally part of compact and centralized precolonial states.
For this analysis, I draw on replication data by Nunn and Wantchekon (2011) that have attributed respondents of the 2005 round of the Afrobarometer survey to settlement areas of ethnic groups included in the Murdock data set introduced above. I add an additional variable from the Afrobarometer survey to gauge my outcome variable: "During the past year, how often have you contacted any of the following persons about some important problem or to give them your views: A traditional ruler?" I create a binary variable differentiating respondents who have not contacted a traditional ruler (77%) and those who have contacted a ruler at least once (23%). As my main explanatory variable, I use the binary measure for centralized precolonial entities with compact and relatively permanent settlements introduced above. Similar to my main analysis, I consider only rural areas, leaving me with around 12,000 respondents in 17 sub-Saharan countries. Results of simple bivariate logistic regression with and without country-level fixed effects lend some initial support to the assumption, that traditional institutions are more likely to remain socially salient within the territories of centralized and compact precolonial states. These findings do not, of course, allow me to make strong claims on the applicability of my argument to other cases. They show, however, that the underlying assumption of my hypothesis is plausible in a wider context beyond the single case of Burundi: Precolonial social and political conditions can in fact continue to shape people's institutional choices.

Conclusion
There is still little knowledge about the conditions behind the persistence of precolonial institutions. This article has suggested and empirically investigated an argument highlighting the internalization and enculturation of precolonial modes of dispute settlement.
In many parts of Burundi, precolonial institutions remain the primary framework for local-level conflict resolution. In other regions, traditional dispute settlement has largely lost its social significance. This article's empirical analyses demonstrate that precolonial political topographies are a much better predictor of these geographical patterns than the spatial features of present-day Burundian state presence and capacity. Precolonial institutions have been able to persist in the former political and cultural centers of the kingdom, where the associated norms have been internalized and enculturated over a period of more than three centuries. Although the case of Burundi is specific in a number of respects, additional analyses indicate that similar dynamics may be at play in other country contexts.
These findings provide several important insights: First, they highlight and specify the historical sources of local conflict-resolution capacity in the highly relevant context of land conflict, informing programs intended to revive or strengthen traditional forms of dispute settlement. Second, they may contribute to understanding the background of uneven state building in many developing countries: The success of states' attempts to enforce formal judicial institutions uniformly across their territories is not just a matter of state capacity and state presence but also depends on the salience of competing institutions that draw their strength and resilience from precolonial times. Third, adding to a growing body of work on the long-term economic effects of precolonial institutions, the analysis demonstrates how precolonial political entities can also continue to shape present-day political and social outcomes by functioning as the source of salient social norms. This may be relevant beyond conflict resolution in a narrower sense. Subsequent studies might possibly analyze whether the topographies of precolonial political entities affect current variations in other social and political conditions such as social capital, religious beliefs, or state-society relations more generally.

Acknowledgments
I thank Marie Asmus for excellent research assistance. I am grateful to Max Montgomery and Johannes Vüllers for comments on a previous version of the article.

Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article has been prepared in the framework of the research project "From Civil War to Social Contract" funded by the Federal Ministry for Economic Cooperation and Development and KfW Development Bank.

Notes
1. Institutions can be understood as the "rules of the game"-they constrain human behavior, and thereby structure human interactions (North, 1990). I use the terms traditional institutions and precolonial institutions interchangeably for those rules that have developed independently, without importation from the outside, and that have been practiced for an extended period of time (Boege, 2006;Zartman, 2000). 2. More skeptical voices make the critique that exclusionary traditional norms contradict universal human rights and stress that their persistence prevents the entrenchment of formal state institutions (Barron, Kaiser, & Pradhan, 2009;Beall, Mkhize, & Vawda, 2005; van Donge & Pherani, 1999). 3. http://ufdc.ufl.edu/derscheid 4. The latter finding contradicts the assumption that areas within the precolonial kingdom were treated differently by the Belgian administration in terms of interventions into local governance. Additional information from the data set indicates that the strongest predictor of chief replacement was previous disciplinary punishment. 5. The survey took place within the framework of a project funded by the German Federal Ministry of Economic Cooperation and Development (BMZ) as well as the German Development Bank (KfW). Data collection was undertaken by a local commercial firm, Development Through Expert Consultancy (DevEC).