Grievances, identity, and political opportunity: The effects of corporate and liberal power-sharing on ethnic conflict – Dissertation summary

Andreas Juon

Supervisors: Dr. Kristin M. Bakke, Dr. Nils Metternich (University College London)

1. Introduction

The costs of inter-ethnic violence remain devastating around the world. The bloody dissolution of Yugoslavia, genocide in Rwanda, and sectarianism in Iraq are only its most drastic examples (cf. Stewart 2008). Prominent conflict scholars recommend pacifying such places through power-sharing practices (e.g., Bormann et al. 2019; Cederman et al. 2013; Cohen 1997; Hartzell & Hoddie 2008; Norris 2008). By including diverse ethnic groups in government and providing them with political autonomy, these hold the promise of alleviating grievances and averting conflict. Following such considerations, the United Nations increasingly promotes inclusive institutions designed to bring about power-sharing practices (Bogaards 2000; McCulloch &McEvoy 2018; Sisk 1996). As a consequence, there has been a global trend towards institutionalized power-sharing, both in its horizontal (providing for government inclusion) and vertical dimensions (providing for autonomy) (see figure 1).

Yet, the increasing global reliance on power-sharing institutions coincides with highly uneven experiences of their ”users”. Nepal’s power-sharing constitution brought its decade-long civil war to an end, but immediately led to violent contestations with Madhesi groups who remained sidelined. Power-sharing enabled a peaceful democratic transition in post-Apartheid South Africa. Yet, in Rwanda, it was quickly dismantled by Hutu majority extremists and proved unable to prevent the 1994 genocide. Power-sharing constitutions ended the destructive civil wars in Lebanon and Bosnia. Yet, they left behind polarized societies facing the specter of renewed conflict.

These contrary experiences have spawned similarly diverging scholarly assessments on the merits of institutionalizing power-sharing (Horowitz 1985; Lijphart 2008; McGarry & O’Leary 2008). Even among its proponents, there is fundamental disagreement on how to design power-sharing institutions. A first option is corporate power-sharing, which relies on clear, ethnically-based criteria. For example, it enshrines ethnic government quotas, veto rights, and ethnically-defined autonomous areas. In spite of directly addressing the demands of aggrieved groups (Lijphart 1995), some scholars conclude that its side-effects are too costly (McCulloch 2014; McGarry & O’Leary 2009). A second option is liberal power-sharing. This avoids explicit group recognition and encourages power-sharing practices indirectly (McCulloch 2014). It does so through low vote thresholds for government inclusion, proportional electoral systems, and federalism with ethnically-mixed territorial units. While prominently recommended by some (Lijphart 1995), others argue that it only insufficiently reduces conflict risks (Bogaards 2019b).

In this dissertation, I address this debate with three key contributions. First, I formulate a novel theoretical argument on how corporate and liberal power-sharing institutions affect conflict risks. Second, I provide new global data on both types since World War 2. Finally, I quantitatively test my propositions. Going beyond previous research, I differentiate between the direct effects of power-sharing on included groups and its side-effects on others and between its diverging effects in the short- and long-term. Empirically, I not only consider how power-sharing affects conflict, but also the intermediate implications of my argument for mass attitudes of ethnic groups. In what follows, I first define the key terms I use and outline the gaps in previous research. I then discuss each of these contributions in turn. Finally, I conclude.

2. Key terms

Ethnic groups differ in their access to political power. Often, the state is controlled by a specific group, while others have only limited access to it (Bormann et al. 2017). To denote this, I borrow the term core group from Mylonas (2012). I thereby refer to the ethnic group whose elites have the strongest ”capacity to enforce its decisions within the borders of a state” (ibid: 24). Non-core groups are all other groups lacking this capacity. Often, this distinction corresponds to the one between the demographic majority and the rest of the population. For instance, the core-group in Switzerland are the majority Swiss Germans. However, they are not always equivalent. For example, in Rwanda, the minority Tutsi dominate state organs and are hence the core-group (Kaufmann & Haklai 2008). Where power-sharing is in place, I further differentiate between two different types of non-core groups. Included non-core groups are those that enjoy power-sharing, for instance by being given cabinet seats or autonomy. Conversely, excluded non-core groups are those left out of institutionalized power-sharing.

Finally, I assume that each group is split into elites and ”masses”. Both seek to protect their rights and interests and mutually constrain one another. Elites are constrained in their efforts to attain power by varying levels of mass support. Conversely, masses face coordination problems and are unable to mobilize absent elite leadership. Figure 2 gives an overview.

3. Gaps addressed

As I explain in chapter 2 of my dissertation, five critical gaps remain in the expansive literature on power-sharing and ethnic conflict.

1. Gap 1 – Side-effects on ”other”groups: Existing work focuses on the ”direct” effects of power-sharing practices on included non-core groups, which are assumed to be uniform across time (e.g., Bormann et al. 2019; Cederman et al. 2013, 2015; Cohen 1997; Hartzell & Hoddie 2008; Norris 2008). As I argue, these approaches must be theoretically and empirically complemented in five ways. First, we lack knowledge on the side-effects of power-sharing on ”other” groups. Most importantly, we need to know how it affects non-core groups that remain outside their inclusive umbrella. For example, this refers to Bosnia’s Jewish and Roma groups, which remain excluded from its extensive power-sharing provisions (Agarin et al. 2018; Stojanović 2018). We also need to know how power-sharing institutions affect the formerly dominant core group. As the examples of Nepal and Rwanda underline (Straus 2006), backlashes by excluded non-core and core groups risk violently overturning volatile power-sharing arrangements.

2. Gap 2 – Institutional types: Second, we need to distinguish between different institutionaltypes of power-sharing on which power-sharing practices are based. Importantly, case studies suggest that the ethnically-based protections of corporate power-sharing more strongly alleviate the grievances of included non-core groups than their liberal alternative (Bogaards 2019a; Lijphart 1995; McCulloch 2014). Yet, they inevitably include some groups at the expense of others, and may be especially prone to violent backlashes (Juon 2020).

3. Gap 3 – Time-differentiated effects: Third, we need to systematically consider the diverging effects of power-sharing across different time horizons. Notably, by constitutionally enshrining ethnicity, corporate power-sharing might increase ethnic salience and boost the risk of renewed conflict in the long-term. However, the literature offers diametrically opposed arguments, has mostly refrained from investigating time-differentiated effects, and offers only limited direct evidence for them (Rothchild & Roeder 2005b).

4. Gap 4 – Accounting for the origins of power-sharing: Fourth, investigating power-sharing and conflict poses problems of selection bias and reverse causation. Often, it is precisely the most unstable states that institutionalize power-sharing, especially its corporate type (Anderson & Costa 2016; Grigoryan 2012; McCulloch 2014). Some studies address this problem with sophisticated methodological approaches (Cederman et al. 2015; Wucherpfennig et al. 2016). Yet, these do not differentiate between corporate and liberal power-sharing on which inclusive practices are predominantly based and which might strongly differ in their effects on conflict.

5. Gap 5 – Evidence for attitudinal mechanisms: Fifth, to test the posited causal mechanisms by which power-sharing affects conflict risks, evidence on their intermediateimplications is required. Most studies of power-sharing investigate its effect on conflict directly and assume that these are mediated by changes in attitudes (Cederman et al.2013). Yet, direct tests of these attitudinal effects are needed to bolster evidence that power-sharing actually reduces included non-core groups’ grievances. Additionally, they are needed to attain evidence on the controversial question of how it shapes ethnic salience.

4. Theoretical argument

In chapter 3 of my dissertation, I address these gaps with a novel theoretical framework. I go beyond previous research by not only discussing the effects of power-sharing on included non-core groups. Specifically, I also consider its side-effects on excluded non-core groups, who remain outside its inclusive umbrella, and on the core group, whose dominance is circumscribed by it (gap 1). Second, I go beyond the previous focus on power-sharing practices by considering whether these are based on corporate or liberal power-sharing institutions (gap 2). Third, I differentiate between their variable impact in the short- and long-term (gap 3). Thereby, I reconcile previously contradictory arguments on how they shape ethnic salience.

Figure 3 gives an overview on my main expectations. These explain how corporate and liberal power-sharing affect the conflict risks of included non-core, core, and excluded non-core groups over time. These expectations are the result of two trade-offs posed by the institutionalization of power-sharing: one along group lines, the other across time horizons. I now summarize my arguments on both trade-offs in turn.

4.1. The trade-off along group lines

The first trade-off is along group lines. I expect the explicitly ethnic guarantees of corporate power-sharing to strongly reduce included non-core groups’ grievances and, in turn, their conflict risks. Conversely, I expect liberal power-sharing to do so only in a weaker manner. However, by excluding specific non-core groups, such as Bosnia’s Roma and Jews, and visibly circumscribing majority rule, such as of Rwanda’s Hutu majority, corporate power-sharing should increase the grievances, and conflict risks, of excluded non-core groups and the core group.

Direct effects of power-sharing institutions on included non-core groups

My expectations are based on the following considerations. As I explain, power-sharing institutions reduce the grievances of included non-core groups in three main ways. First, by including their representatives in political office, they increase their masses’ perception of being part of the political process (Aarts & Thomassen 2008; Anderson & Guillory 1997). This translates into fewer grievances (Hänni 2017; Mansbridge 1999; Sanchez & Morin 2011). Second, if their members are represented in the political arena, the decisions taken there are more likely to reflect their preferences as well (Anderson & Guillory 1997). This enhances their support for the political system and further reduces grievances (Christensen 2015; Lijphart 1999; Kittilson & Schwindt-Bayer 2010). Third, power-sharing signals accommodative intent on behalf of core group elites (Cederman et al. 2013; Stewart 2008). Through this, it reduces the power of injustice frames, whereby continuing inequality would be blamed on the state. Together, these grievance-alleviating effects reduce the conflict risks emanating from included non-core groups.

By relying on clearly ethnically-differentiated criteria, corporate power-sharing strengthens all three pathways. First, its clear guarantees ensure that de-jure rights actually translate into de-facto practices. For example, ethnic government quotas offer comparably enforceable guarantees for enduring group representation. Second, it visibly enshrines inclusiveness and most strongly reduces the power of injustice frames. In this way, corporate power-sharing clearly alleviates included non-core groups’ grievances and, in turn, reduces their conflict risks. This explains why their representatives frequently demand precisely such guarantees (Lijphart1995; McCulloch 2014).

Conversely, liberal power-sharing is more difficult to enforce. Often requiring inter-group cooperation, it is less effective in encouraging the representation of specific non-core groups (Lijphart 1995). For example, to exploit (liberal) parliamentary super-majority requirements, cooperation between multiple groups is vital. This means that liberal power-sharing might not result in de-facto inclusive practices (Bogaards 2019b). It also lacks the clear signal of accommodative intent inherent in corporate power-sharing. Thereby, it may fail to alleviate perceptions of discrimination, even where non-core group representatives achieve political influence. In this way, liberal power-sharing only weakly alleviates included non-core groups’ grievances and in turn, only weakly reduces their conflict risks.

Side-effects of power-sharing institutions on excluded non-core groups and the core group

Hence, corporate power-sharing more effectively reduces the grievances of included non-core groups than its liberal alternative. However, the inverse situation applies to non-core groups who remain excluded from its protections and to the formerly-dominant core group whose power is visibly diminished by them.

Corporate power-sharing should increase the grievances of excluded non-core groups, such as the Roma and Jews in Bosnia, in three ways. First, they are now not only de-facto outside of political power, but de-jure formally barred from it. This accentuates their previously marginalized political position and increases their grievances (Stojanović 2018; Juon 2020). Second, corporate power-sharing increases the likelihood that their marginalized status becomes politicized through injustice frames. Most importantly, it bolsters norms of group entitlements. This makes the status of excluded groups all the more jarring and provides a ”higher” reference point to which they it to. Finally, corporate power-sharing also provides a comparably visible target to mobilize against. In sum, excluded non-core groups should therefore possess higher grievances and face higher conflict risks under corporate power-sharing, compared to other institutional contexts.

A similar situation applies to core groups under corporate power-sharing, for example the Hutu under Rwanda’s short-lived 1994 power-sharing arrangement. Where they were able to freely dominate the political scene before, corporate power-sharing imposes severe restrictions on this. Their loss of de-facto dominance increases disaffection with the political process among core group members (Bühlmann & Hänni 2012; Hänni 2017). Corporate power-sharing also more ostensibly deviates from norms of majority rule than its liberal alternative. Similar to excluded non-core groups, this should make the activation of injustice frames among core groups more likely. Together, this means that corporate power-sharing is most likely to engender core group backlashes and boost their conflict risks.

4.2. The trade-off across time horizons

A second trade-off is across time horizons. I expect time-differentiated effects to be most pronounced for included non-core groups under corporate power-sharing. Specifically, its short-term positive effects, as described above, should reverse in the long-term. I base this expectation on two arguments: First, over time, corporate power-sharing enhances the salience of ethnic divisions, which incentivizes elites to mobilize on an ethnic basis. Second, it provides these elites with gradually-accumulating institutional resources. These reduce the costs of mobilizing against the state. As a consequence, I expect included non-core groups under corporate power-sharing to have higher conflict risks in the long-term. I now summarize both arguments in turn.

Corporate power-sharing and ethnic salience

Instead of revisiting the prominent debate on ethnic identities are ”primordial” or socially ”constructed” (Hale 2004), I introduce a cognitive concept of ethnicity based on the social identity approach (Hornsey 2008). This is not only able to encompass both primordialist and constructivist aspects of ethnicity. Instead, it also allows us to more firmly theorize on their malleability through power-sharing. Thereby, it helps me reconcile previously diverging expectations on how it affects ethnic salience.

I theorize on three links between power-sharing and ethnic salience, which are countervailing but time-differentiated. First, power-sharing diminishes horizontal inequalities. Thereby, it reduces the correlation between ethnic identity and group status. As a result of this decreased cognitive fit (cf. Abrams & Hogg 1990; Mullin & a. Hogg 1999; Turner 1987), the role that ethnicity can play as a ”rule of thumb” for cognition is diminished (cf. Hale 2008). Hence, this should uniformly, but weakly, decrease ethnic salience. Second, power-sharing reassures included non-core groups of their continued existence. Thereby, it decreases the need to defend these elements of self-esteem in a reactive way, and reduces ethnic salience further (cf. Brewer& Gaertner 2008). These reassurances should be strongest in the immediate wake of conflict when power-sharing is first institutionalized. However, as time progresses, this effect loses traction, as immediate identity threats fade. Hence, through these first two links, both corporate and liberal power-sharing decrease the salience of ethnicity, especially in the short-term. However, over the long-term, these effects reverse for included non-core groups under corporate power-sharing through a third link. By explicitly relying on ethnic categories as a determinant of political inclusion, corporate power-sharing increases the chronic cognitive accessibility of should have a countervailing, positive impact on the salience of ethnic identifications, which increases over the long-term.

In sum, while corporate power-sharing initially decreases ethnic salience, over time this effect reverses. Thereby, it incentivizes elites of included non-core groups to mobilize on an ethnic basis in the long-term.

Corporate power-sharing and institutional resources

Exacerbating these destabilizing long-term incentives, corporate power-sharing also provides elites of included non-core groups with institutional resources, which they can exploit to initiate violent conflict. First, they increase their capabilities vis-a-vis the state. For instance, they provide them with institutional levers to create political deadlock (Roeder 2005; Rothchild & Roeder 2005a; Spears 2000) or they redistribute coercive resources to them. This is especially the case where corporate power-sharing includes territorial autonomy in its vertical dimension, which creates subnational government institutions that included non-core group elites can control (Grigoryan 2015; Roeder 2007; Treisman 1997).

These effects are time-differentiated. In the short-term, included non-core group elites’ control over the institutional resources provided by corporate power-sharing is still unconsolidated. Thus, it does not yet result in significantly increased capabilities. However, in the long-term, this changes, as elites consolidate control over corporate power-sharing institutions, and in turn face lower costs to revolt. Thereby, they increase their capabilities to wage violent conflict in the long-term.

In sum, I expect corporate power-sharing to exert the opposite impact on included non-core groups in the long-term than in the short-term. Not only does it gradually deepen divisive ethnic identities, but it also provides increasing capabilities to their elites, which they can exploit to mobilize for conflict.

4.3. Expectations

Based on these group-, institution-, and time-differentiated considerations, I formulate six hypotheses on how power-sharing institutions affect group-wise conflict risks (figure 3):

H1: For included non-core groups, corporate power-sharing strongly decreases conflict risks in the short-term.

H2: For included non-core groups, corporate power-sharing increases conflict risks in the long-term.

H3: For non-core groups, liberal power-sharing weakly decreases conflict risks.

H4: For core groups, corporate power-sharing strongly increases conflict risks, especially in the short-term.

H5: For core groups, liberal power-sharing weakly increases conflict risks, especially in the short-term.

H6: For excluded non-core groups, corporate power-sharing increases conflict risks, especially in the short-term.

5. Data

The disaggregated theoretical framework presented above requires similarly disaggregated data to investigate its hypotheses comparatively. Addressing this need, in chapter 4 of my dissertation, I present three new data sources. These include a new dataset on power-sharing, which differentiates between their corporate and liberal type, newly ethnically-attributed mass survey data on grievances and ethnic salience, to test the intermediate implications of my argument for grievances and ethnic salience, and a geo-coded dataset on administrative boundaries, which played an auxiliary role in both. I now present all three data contributions in turn.

5.1. Constitutional Power-Sharing Dataset (CPSD)

The most important data contribution of my dissertation is the Constitutional Power-Sharing Dataset (CPSD). Drawing on hand-coded constitutional documents and autonomy statutes, I provide detailed information on corporate and liberal power-sharing on the ethnic group-level, with the group selection given by the widely-used Ethnic Power Relations dataset (Vogt et al. 2015). Preventing selection bias, the CPSD covers 180 countries and 1441 ethnic groups within them for the period between 1945 and 2018. This permits me to go beyond the investigation of power-sharing practices and to distinguish between their underlying institutional types on the group-level and diverging impact over time (gaps 1-3).

First, I systematically screened legal source documents to collect raw, country-level data for numerous indicators for corporate and liberal power-sharing, covering both its horizontal (government inclusion) and vertical dimensions (political autonomy). I relied on three principal sources: Constitutions, electoral laws, and formal autonomy statutes. Using a combination of these sources, I noted down the specifics of each power-sharing institution within the CPSD on a country-year level. Most importantly, as one of the biggest time investments of my dissertation, I manually coded over 700 unique constitutional texts, focusing on power-sharing institutions fitting my criteria.

In a second step, I transferred the recorded institutional specifics to the ethnic group-level. In a nutshell, indicator data for ethnic groups was directly coded on the group-level by linking institutions to ethnic groups by name; institutions targeting organizations were linked to ethnic groups via organizational claims; data for country-level electoral variables were taken over for each ethnic group active in a country equally; and data for territorial units were aggregated additively onto the ethnic group-level by spatially intersecting group-wise settlement areas (provided by the Geo-EPR dataset, Vogt et al. 2015) with geo-coded administrative units provided by the Significant Administrative Unit Dataset (section 5.3).

The CPSD covers a wide range of institutional power-sharing measures (see table 1). First, it covers institutions in the horizontal dimension of power-sharing, broadly corresponding to Arend Lijphart’s (1977) first three “pillars” of consociationalism. These provide for the representation of minorities through executive and legislative power-sharing and give their representatives veto rights over policy-making. Second, it includes institutions in its vertical dimension, broadly corresponding to the fourth ”pillar”, segmental autonomy (Lijphart 1977). These include territorial forms of autonomy, such as federalism, and ethnic ones, such as group-based autonomy. Within both dimensions, I differentiate between corporate power-sharing institutions and liberal ones and aggregate these into separate indices. The corporate indices encompass all pre-determined, ethnically-based, institutions. The liberal indices include only electorally-based, group-blind, or territorial institutions whose boundaries cross-cut ethnicity. Each index ranges from 0 (no power-sharing of the respective type and dimension in place) to 1 (full power-sharing) (see figure 1 for global averages of these indices). Together, they allow me to investigate my arguments in a quantitative manner.

	Horizontal	Vertical
Corporate	Executive: Ethnic quotas, appointment vetoes, and rotation rules. Legislature: Ethnic quotas, appointment vetoes, and rotation rules. Ethnic veto rights over constitutional amendments and legislation.	Territorial autonomy along ethnic settlement patterns, group-based autonomy.
Liberal	Executive: Supermajority for appointment, low vote thresholds for party inclusion. Proportional electoral systems. Supermajority for constitutional amendments and legislation.	Territorial autonomy that cross-cuts ethnic settlement patterns.

5.2. Ethnically-attributed mass surveys

The second new data source I provide in my dissertation is a combination of extensive, newly ethnically-attributed mass surveys. These allow me to consider the intermediate attitudinal steps of my arguments, most notably for mass grievances and the salience of ethnic identifications (gap 5). To my knowledge, this collection of attitudinal data represents the most extensive survey sample employed to analyze the attitudinal effects of inclusive institutions so far. It includes 148 countries with, on average, 9 unique survey waves for each. The coverage is based on large cross-regional mass surveys measuring individual attitudes for a global or regional sample, such as the World Values Surveys (Inglehart et al. 2014), the Afrobarometer (Afrobarometer Data 1999), and the Latinobarometro. In all surveys, I attribute respondents to their ethnic groups, based on question items that ask for a respondent’s ethnic identity, language, religion, and place of residence. I then identify survey items that capture grievances of group members against the state and their degree of ethnic identification. These serve as dependent variables in my attitudinal analyses (section 6).

5.3. Significant Administrative Unit Dataset (SAU)

A third data contribution of my dissertation is the Significant Administrative Units (SAU) dataset. This provides geo-coded internal administrative boundaries for all country years in my sample. In this dissertation, it played an auxiliary role in the collection of the first two data contributions. First, for the CPSD (section 5.1), it enables the aggregation of territorial power-sharing provisions onto the ethnic group level. Second, for the ethnic attribution of survey respondents (section 5.2), it enables me to conduct a more precise attribution procedure, by combining information on a respondent’s place of residence with said place’s ethnic demography.

In a first step, I compiled a list of all geographic units for which geo-coded polygons are required, based on Law (2010). Second, I geo-coded their corresponding polygons in the GIS software QGIS. I relied heavily on data from the Database of Global Administrative Areas (GADM, http://gadm.org/data.html) from which I obtained the most recent administrative boundaries. I then created a time-varying version of this dataset by geo-referencing historical maps and by backwards aggregating the GADM polygons. The result is the time-varying Significant Administrative Units dataset. It contains more than 8900 unique polygon-periods (figure 4). Beyond this dissertation, SAU also enables more detailed analyses for other research applications, for example enabling scholars to study the spatial distribution of violent conflict events across administrative units.

6. Empirical analyses

6.1. Overview on analyses

In chapters 5, 6, and 7 of my dissertation, I test my hypotheses by relying on these new datasources. I conduct five main types of analyses. In what follows, I first give an overview on these procedures before summarizing my main results for each hypothesis.

1. Quantitative analyses: from power-sharing institutions (PSI) to groupconflict: First, I quantitatively investigate the impact of corporate and liberal power-sharing institutions on the main outcome of interest: the outbreak of violent conflict. Using a series of group-level logistic regressions, and relying on my group-wise indices of corporate and liberal power-sharing from the CPSD (see section 5.1), I study three dependent variables: the outbreak of (governmental and territorial) civil wars for non-core groups (Gleditsch et al. 2002; Vogt et al. 2015) and the instigation of coup attempts by politically dominant core groups, whose violence is by definition unlikely to take the form of civil wars (Roessler 2011). This allows me to address gaps 1-3, which call for a systematic analysis of how power-sharing affects conflict risks, disaggregated by group, type of underlying institution, and time.

2. Instrumental variables analyses: from colonial legacies to PSI to groupconflict: Second, I conduct an instrumental variable analysis. Specifically, I exploit the fact that British and French colonies systematically differed in whether they awarded corporate and liberal power-sharing institutions to their non-core groups, depending on their size and settlement areas. These colonial institutions were then inherited by their successor states. Instrumenting for power-sharing with variables capturing these colonial legacies allows me to focus on power-sharing institutions whose creation was plausibly exogenous to a group’s baseline conflict risk. I thereby go beyond similar approaches that instrument for power-sharing practices (Cederman et al. 2015; Wucherpfennig et al. 2016) by additionally differentiating between their corporate and liberal types. Using this approach, I am able to address gap 4, which highlights the importance of accounting for self-selection and reverse causation of the two types of power-sharing.

3. Mediation analyses: from PSI to power-sharing practices to group-conflict: I then turn to the intermediate implications of my argument. A key assumption is that power-sharing institutions are actually implemented. In a third step, I test this by conducting a set of formal mediation analyses, based on the framework by Imai and colleagues (2019). To do so, I rely on existing group-level measures for de-facto government inclusion and autonomy and see whether they are influenced by the underlying de-jure corporate and liberal power-sharing institutions (Vogt et al. 2015).

4. Attitudinal analyses: from PSI to mass grievances and ethnic salience: I then turn to the intermediate, attitudinal implications of my argument. Using the ethnically-attributed mass surveys presented in section 5.2, I use dependent variables for how aggrieved group members are, given by perceptions of state discrimination, and for ethnic salience, given by their prioritization of their ethnic vis-a-vis their national identity. By investigating the effects of power-sharing institutions on these attitudes in a set of hierarchical multi-level regressions, this allows me to address gap 5.

5. Process tracing with qualitative evidence from specific cases: I further probe the plausibility of my quantitative findings by tracing my hypothesized causal mechanisms in six typical cases: Bosnia, Burundi, Georgia, Iraq, Rwanda, and South Africa. I selected these cases based on the smallest possible residual in my main models, which is a common technique used in confirmatory process tracing (Seawright & Gerring 2008).

6.2. Main results

Corporate power-sharing: Included non-core groups – short-term (H1)

For corporate power-sharing, I expected a strong conflict-alleviating effect for included non-core groups in the short-term. The evidence I attain is mostly in accordance with these expectations. In my main quantitative analyses, I attain only a weak negative effect of corporate power-sharing on the onset of conflict (see figure 5 for relative frequencies). Yet, my instrumental variable analyses reveal that this weak association is due to strong endogeneity, whereby corporate power-sharing is predominantly given to the most conflict-prone groups. This is in line with observations from several of the most deeply-entrenched corporate power-sharing systems, many of which were adopted as a response to previous conflict (McCulloch 2014). For example, while continuing instability plagues Bosnia’s (Bieber & Keil 2009) corporate power-sharing system, this cannot be attributed to a causal effect of these institutions, which were adopted to end its civil war in the first place. Hence, it appears likely that the ”true” effect of corporate power-sharing is to reduce included non-core groups’ propensities of civil war participation.

My analyses on the intermediate implications of my argument support this interpretation that corporate power-sharing reduces the conflict risks of included non-core groups: First, my mediation analyses show that corporate power-sharing institutions are the strongest predictors of de-facto power-sharing practices, which in turn reduce conflict risks. My attitudinal analyses further confirm that this pacifying effect stems from reductions in included non-core groups’ grievances. Non-core group members are significantly less likely to perceive themselves as horizontal dimension, which provides for their government inclusion (see figure 6).

These findings are reflected in my process tracing of influential cases. For example, Bosnia already had limited, partially liberal power-sharing provisions in its 1992 constitution. Yet, in the context of the disintegration of Yugoslavia, these proved insufficient to stem the tide of conflict. In 1995, to end the civil war, these ”soft” provisions were consequently expanded and became more corporate in nature (Gordy 2015). While tensions persist, the combination of these policies indeed reduced the most severe mass grievances among the ethnic Serb population and alleviated at least the immediate risks of violent conflict (Basta 2016).

Hence, the overall picture offered by this dissertation’s empirical analyses supports the basic logic of hypothesis 1, illustrated by Bosnia: Corporate power-sharing is often adopted in difficult contexts, offers enforceable and visible guarantees of de-facto government inclusion and autonomy, thereby alleviates included non-core groups’ grievances, and in turn appears to decrease their conflict risks in the short-term.

Corporate power-sharing: Included non-core groups – long-term (H2)

Under the trade-off across time horizons, I expected these initially beneficial impacts to reverse in the long-term. My main quantitative analyses fail to offer statistically significant support for this expectation, however. Building on this non-finding, I then probed the intermediate implication of my argument, which posits that corporate power-sharing should increase ethnic salience among included non-core group members in the long-term. Indeed, I find such a relationship for its horizontal dimension, which is associated with higher levels of ethnic salience the more consolidated it becomes. However, I attain no similar finding for its vertical dimension. Hence, overall these findings seem to reflect the ”stable identifications” perspective that doubts whether power-sharing significantly affects ethnic identifications (Cederman et al. 2013), rather than arguments whereby it should gradually increase them through a reification process (Brubaker 2002; Hale 2008). In the cases I qualitatively analyzed, I also only find evidence in line with my hypothesis that is limited to specific cases, notably the successor states of the Soviet Union. Hence, it appears, that, contrary to my hypothesis 2, the conflict-alleviating effects of corporate power-sharing also persist in the long-term.

Liberal power-sharing: included non-core groups (H3)

For liberal power-sharing I expected a similar, but weaker conflict-alleviating effect as I did for its corporate alternative in the short-term. My main quantitative analyses offer support for this expectation: They suggest that its horizontal dimension, providing for broad-based government coalitions, makes the participation of non-core groups in governmental civil wars less likely, while its vertical dimension, providing for ethnically-mixed autonomous areas, does the same for their participation in territorial civil wars (see figure 5). This echoes hopes based on South Africa’s liberal, post-apartheid power-sharing constitution that its more flexible provisions might prove exportable to other contexts, effectively include diverse ethnic groups in government, alleviate mass grievances, and reduce conflict risks (Lijphart 1995; McCulloch 2014; Sisk & Stefes 2005).

However, my instrumental variables and mediation analyses counsel some caution in this interpretation: Liberal power-sharing appears to be a partial substitute for its corporate alternative, often offered to groups that are already less conflict-prone in the first place. Additionally, it only weakly influences de-facto inclusive practices. This hardens the suspicion that liberal power-sharing might, at least in some cases, reflect token measures offered to non-core groups with low bargaining power, rather than a strong institutional reassurance.

I additionally probed the intermediate attitudinal steps of my hypothesized causal mechanism, whereby liberal power-sharing should reduce non-core groups’ grievances. The findings for this were mixed: While I find that its vertical dimension (autonomy) strongly reduces grievances, its horizontal dimension (government inclusion) appears associated with higher, rather than lower grievances (see figure 6). This underlines criticism that liberal power-sharing might fail to appease aggrieved groups, who demand the more explicit reassurances provided by its corporate alternative (Bogaards 2019b; Lijphart 1995; McCulloch 2014). One example in line with this interpretation is post-Saddam Iraq, which installed a transitional liberal power-sharing regime in 2004. While this indeed initially resulted in the political inclusion and autonomy of the Kurd and Sunni minorities, this positive effect gradually faltered and proved unable to prevent further conflict (Bogaards 2019b; McCulloch 2014).

Hence, the overall support for this third hypothesis is mixed. My analyses provide partial support for my overall expectation that liberal power-sharing reduces conflict and only weak support for the intermediate implications of my argument, whereby it should result in de-facto inclusive practices and lower grievances.

Corporate and liberal power-sharing: Core groups (H4-5)

In hypothesis 4, I expected corporate power-sharing to increase the conflict risks of core groups. Indeed, my main quantitative analyses support the view that such side-effects exist: I found corporate power-sharing, especially in its horizontal dimension, to be a strong predictor of core group coup attempts (figure 7).

My attitudinal analyses further boost this conclusion. I find that corporate power-sharing in the horizontal dimension is associated with a higher probability that core group members perceive themselves as discriminated against and that they are unsatisfied with the political system. Interestingly, this finding is most pronounced for less educated core group members, who hold authoritarian values and are consequently more likely to feel threatened by corporate power-sharing. Together, these findings reflect the experiences of several prominent cases of power-sharing, where corporate power-sharing sparked backlashes by core group masses, for example in 1994 Rwanda, post-Ohrid Macedonia (Brunnbauer 2002) and post-Dayton Bosnia (Basta 2016). In sum, my findings are in line with hypothesis 4, whereby I expected corporate power-sharing to boost core groups’ grievances and make conflict initiation on their part more likely.

Under the trade-off along group lines, I expected a similar, but weaker relationship for liberal power-sharing (H5). However, my analyses provide no evidence for such a relationship. Liberal power-sharing is neither a predictor of core group coups nor of core group grievances. This reflects the encouraging experience of South Africa, where liberal power-sharing proved acceptable to the core group and enabled a sustainable transition to peace without major backlashes (Sisk & Stefes 2005).

Corporate power-sharing: Excluded non-core groups (H6)

For excluded non-core groups, I expected corporate power-sharing to engender ”secondary grievances” among their members, and in turn increase their conflict risks. My main quantitative analyses do not lend support to these expectations, however: I found no evidence that exclusion from corporate power-sharing translates into higher risks of violent conflict.

My analyses of my arguments’ implications are more in line with my expectations. I indeed find that the exclusion from corporate power-sharing provisions further marginalizes non-core groups, as given by their de-facto government inclusion and compared to situations without any power-sharing in place. My attitudinal analyses further indicate that corporate power-sharing indeed increases the grievances of excluded non-core groups. Hence, my findings suggest that excluded non-core groups are indeed more likely to be de-facto marginalized and aggrieved, but that they lack the capabilities to translate these grievances into civil war participation.

In my process tracing, a typical case for these mixed findings is Bosnia. Its 1995 corporate power-sharing constitution excluded its Roma non-core group from the comprehensive rights exclusively reserved for its ”constituent peoples”. While this sparked widespread discontent, culminating in the condemnation of this discriminatory arrangement by the European Court of Human Rights (Juon 2020; Stojanović 2018), these grievances are unlikely to result in an armed insurrection due to the small size, dispersed settlement pattern, and general lack of resources of Bosnia’s Roma minority (Agarin & Brosig 2009).

Taken together, the evidence for my six hypotheses mostly supports the expected trade-off along group lines: Corporate power-sharing indeed reduces the conflict risks of included non-core groups. It does so by effectively providing for their government inclusion and autonomy and by alleviating the grievances among their masses. However, this comes at the price of boosting the grievances of excluded non-core groups, who are now even more marginalized. It similarly increases grievances among the core group, whose former dominance is now effectively circumscribed. While the former are unable to mobilize upon their increased grievances in overt violence due to a lack of capabilities, the latter become more likely to instigate coups to reassert their former dominance in central government. Conversely, liberal power-sharing is not found to have strong effects of grievances or conflict risks. Although it slightly increases the chance that non-core groups attain government inclusion, it does not substantially shape either the grievances among their masses and only inconsistently affects their resulting conflict risks. In contrast, I find only very limited evidence for the trade-off across time horizons, whereby I expected the initially pacifying effects of corporate power-sharing to reverse over the long-term.

7. Conclusion

Given the apparent benefits and normative desirability of power-sharing practices, it is no wonder that many states try to enforce them through their formal institutionalization. My dissertation discusses and empirically investigates the merits of the two most popular options available in this endeavor: First, corporate power-sharing, which mandates ethnic government inclusion and autonomy by relying on explicitly ethnic criteria or ethnically-defined territorial units. At the surface, these comparably enforceable provisions offer the most direct way to rectify the unequal status of non-core groups and, in turn, reduce their risk of participating in violent conflict. Second, liberal power-sharing, which seeks to indirectly encourage inclusive outcomes through electoral criteria and extensive individual rights. While constituting weaker reassurances, these promise to engender more widely-encompassing power-sharing practices and to avoid several problematic side-effects of the corporate type.

In contrast, my findings indicate that neither of these two institutional alternatives constitutes an easy, transferable ”fix” that can be used to engineer peace. Neither can comprehensively or sustainably remedy the risks of inter-ethnic violence in divided places. My findings reinforce some of the warnings about institutionalizing power-sharing, while attenuating others. On the one hand, they are in mostly in line with the trade-off along group lines. They indicate that corporate power-sharing indeed alleviates the conflict risks of included non-core groups at the cost of increasing the grievances of excluded non-core groups and the conflict risks emanating from the formerly-dominant core group. They also support the expectation that liberal power-sharing represents a far weaker institutional measure that may only inconsistently shape conflict risks. Conversely, I found only limited evidence for the long-term destabilizing impacts of corporate power-sharing, expected under the trade-off across time horizons.

References

Aarts, K., & Thomassen, J. (2008). Satisfaction with democracy: Do institutions matter? Electoral Studies, 27 (1), 5–18.

Abrams, D., & Hogg, M. A. (1990). Social Identification, Self-Categorization and Social Influence. European Review of Social Psychology, 1(1), 195–228.

Afrobarometer Data (1999). Afrobarometer Rounds 1-6.

Agarin, T., & Brosig, M. (Eds.) (2009). Minority Integration in Central Eastern Europe: Between Ethnic Diversity and Equality. No. 18 in On the Boundary to Two Worlds. Amsterdam: Rodopi.

Agarin, T., McCulloch, A., & Murtagh, C. (2018). Others in Deeply Divided Societies: A Research Agenda. Nationalism and Ethnic Politics, 24(3), 299–310.

Anderson, C. J., & Guillory, C. A. (1997). Political Institutions and Satisfaction with Democracy: A Cross-National Analysis of Consensus and Majoritarian Systems. American Political Science Review, 91(01), 66–81.

Anderson, L., & Costa, C. (2016). Survival of the Fittest: Explaining the Success of Ethnic Autonomy Arrangements. Swiss Political Science Review, 22(4), 516–544.

Basta, K. (2016). Imagined Institutions: The Symbolic Power of Formal Rules in Bosnia and Herzegovina. Slavic Review, 75(4), 944.

Bieber, F., & Keil, S. (2009). Power-Sharing Revisited: Lessons Learned in the Balkans? Review of Central and East European Law, 34(4), 337–360.

Bogaards, M. (2000). The Uneasy Relationship Between Empirical and Normative Types in Consociational Theory. Journal of Theoretical politics, 12(4), 395–423.

Bogaards, M. (2019a). Formal and Informal Consociational Institutions: A Comparison of the National Pact and the Taif Agreement in Lebanon. Nationalism and Ethnic Politics, 25(1), 27–42.

Bogaards, M. (2019b). Iraq’s Constitution of 2005: The Case Against Consociationalism ‘Light’. Ethnopolitics, (pp. 1–17).

Bormann, N.-C., Cederman, L.-E., Gates, S., Graham, B. A. T., Hug, S., Strøm, K. W., & Wucherpfennig, J. (2019). Power Sharing: Institutions, Behavior, and Peace. AmericanJournal of Political Science, 63(1), 84–100.

Bormann, N.-C., Cederman, L.-E., & Vogt, M. (2017). Language, Religion, and Ethnic Civil War. Journal of Conflict Resolution, 61(4), 744–771.

Brewer, M. B., & Gaertner, S. L. (2008). Toward Reduction of Prejudice: Intergroup Contact and Social Categorization. In R. Brown, & S. Gaertner (Eds.) Blackwell Handbook of Social Psychology: Intergroup Processes, (pp. 451–474). John Wiley & Sons.

Brubaker, R. (2002). Ethnicity without groups. European Journal of Sociology, 43(2), 163–189.

Brunnbauer, U. (2002). The Implementation of the Ohrid Agreement: Ethnic MacedonianResentments. JEMIE – Journal on Ethnopolitics and Minority Issues in Europe, 1, 1–23.

Bühlmann, M., & Hänni, M. (2012). Creating a ‘We-feeling’ in Heterogeneous Nations: A Cross-country Investigation of the Impact of Inclusion and Autonomy on the National Pride of Ethnic Minorities. Comparative Sociology, 11(3), 325–353.

Cederman, L.-E., Gleditsch, K. S., & Buhaug, H. (2013). Inequality, Grievances, and Civil War. Cambridge: Cambridge University Press.

Cederman, L.-E., Hug, S., Schädel, A., & Wucherpfennig, J. (2015). Territorial Autonomy in the Shadow of Conflict: Too Little, Too Late? American Political Science Review, 109(02),354–370.

Christensen, H. S. (2015). Power Sharing and Political Dissatisfaction: A Multilevel Analysis of the Link Between Power-Sharing Institutions and Kinds of Political Dissatisfaction in 24 European Democracies. European Politics and Society, 16(2), 280–300.

Cohen, F. S. (1997). Proportional versus majoritarian ethnic conflict management in democracies. Comparative Political Studies, 30(5), 607–630.

Elkins, Z., Ginsburg, T., & Melton, J. (2014). The Comparative Constitutions Project: A Cross-national Historical Dataset of Written Constitutions. http://www.comparativeconstitutionsproject.org.

Gleditsch, N. P., Wallensteen, P., & Eriksson, M. (2002). Armed Conflict 1946-2001: A New Dataset. Journal of Peace Research, 39(5), 615–637.

Gordy, E. (2015). Dayton’s Annex 4 Constitution at 20: Political stalemate, public dissatisfaction and the rebirth of self-organisation. Southeast European and Black Sea Studies, 15(4),611–622.

Grigoryan, A. (2012). Ethnofederalism, separatism, and conflict: What have we learned from the Soviet and Yugoslav experiences? International Political Science Review, 33(5), 520–538.

Grigoryan, A. (2015). Concessions or Coercion? How Governments Respond to Restive Ethnic Minorities. International Security, 39(4), 170–207.

Hale, H. E. (2004). Explaining Ethnicity. Comparative Political Studies, 37 (4), 458–485.

Hale, H. E. (2008). The Foundations of Ethnic Politics: Separatism of States and Nations in Eurasia and the World. Cambridge University Press.

Hänni, M. (2017). Presence, Representation, and Impact: How Minority MPs Affect Policy Outcomes: Presence, Representation, and Impact. Legislative Studies Quarterly, 42(1), 97–130.

Hartzell, C., & Hoddie, M. (2008). Crafting Peace: Power-Sharing Institutions and the NegotiatedSettlement of Civil Wars. University Park: Penn State University Press.

Hornsey, M. J. (2008). Social Identity Theory and Self-categorization Theory: A Historical Review. Social and Personality Psychology Compass, 2(1), 204–222.

Horowitz, D. L. (1985). Ethnic Groups in Conflict. Berkeley, Los Angeles, London: University of California Press.

Imai, K., Kim, I. S., & Wang, E. (2019). Matching Methods for Causal Inference with Time-Series Cross-Sectional Data.

Inglehart, R., Haerpfer, C., Moreno, A., Welzel, C., Kizilova, K., Diez-Medrano, J., Lagos, M., Norris, P., Ponarin, E., & Puranen, B. (2014). World Values Survey: All Rounds – Country-Pooled Datafile Version.. Madrid: JD Systems Institute.

Juon, A. (2020). Minorities overlooked: Group-based power-sharing and the exclusion-amid-inclusion dilemma. International Political Science Review, 41(1), 89–107.

Kaufmann, E., & Haklai, O. (2008). Dominant ethnicity: From minority to majority: Dominant ethnicity. Nations and Nationalism, 14(4), 743–767.

Kittilson, M. C., & Schwindt-Bayer, L. (2010). Engaging Citizens: The Role of Power-SharingInstitutions. The Journal of Politics, 72(4), 990–1002.

Law, G. (2010). Administrative Subdivisions of Countries: A Comprehensive World Reference. Jefferson, NC: McFarland & Company.

Lijphart, A. (1977). Democracy in Plural Societies. A Comparative Exploration. New Haven and London: Yale University Press.

Lijphart, A. (1995). Self-Determination versus Pre-Determination of Ethnic Minorities in Power-Sharing Systems. In W. Kymlicka (Ed.) The Rights of Minority Cultures, (pp.275–287). Oxford: Oxford University Press.

Lijphart, A. (1999). Patterns of Democracy. Yale: Yale University Press.

Lijphart, A. (2008). Thinking about Democracy: Power Sharing and Majority Rule in Theory and Practice. London; New York: Routledge.

Mansbridge, J. (1999). Should Blacks Represent Blacks and Women Represent Women? A Contingent ”Yes”. The Journal of Politics, 61(3), 628–657.

McCulloch, A. (2014). Consociational settlements in deeply divided societies: The liberal-corporate distinction. Democratization, 21(3), 501–518.

McCulloch, A., & McEvoy, J. (2018). The international mediation of power-sharing settlements. Cooperation and Conflict, (p. 001083671876176).

McGarry, J., & O’Leary, B. (2008). Integration or accommodation? The enduring debate in conflict regulation. In S. Choudhry, & R. Simeon (Eds.) Constitutional Design for Divided Societies. Integration or Accommodation?, (pp. 41–90). Oxford: Oxford University Press.

McGarry, J., & O’Leary, B. (2009). Power Shared after the Death of Thousands. In R. Taylor (Ed.) Consociational Theory: McGarry and O’Leary and the Northern Ireland Conflict, (pp. 15–84). London: Routledge.

Mullin, B.-A., & a. Hogg, M. (1999). Motivations for Group Membership: The Role of Subjective Importance and Uncertainty Reduction. Basic and Applied Social Psychology, 21(2),91–102.

Mylonas, H. (2012). The Politics of Nation-Building. Making Co-Nationals, Refugees, and Minorities. Cambridge: Cambridge University Press.

Norris, P. (2008). Driving Democracy: Do Power Sharing Institutions Work? Leiden: CambridgeUniversity Press.

Regan, P. M., Frank, R. W., & Clark, D. H. (2009). New datasets on political institutions and elections, 1972—2005. Conflict Management and Peace Science, 26(3), 286–304.

Roeder, P. G. (2005). Power Dividing as an Alternative to Ethnic Power Sharing. In P. G. Roeder, & Rothchild (Eds.) Sustainable Peace: Power and Democracy after Civil Wars, (pp.51–82). Ithaca, London: Cornell University Press.

Roeder, P. G. (2007). Where Nation-States Come from: Institutional Change in the Age of Nationalism. Princeton University Press.

Roessler, P. (2011). The Enemy Within: Personal Rule, Coups, and Civil War in Africa. World Politics, 63 (02), 300–346.

Rothchild, D., & Roeder, P. G. (2005a). Dilemmas of State-Building in Divided Societies. In P. G. Roeder, & Rothchild (Eds.) Sustainable Peace: Power and Democracy after Civil Wars (pp. 1–26). Ithaca, London: Cornell University Press.

Rothchild, D., & Roeder, P. G. (2005b). Power Sharing as an Impediment to Peace and Democracy. In P. G. Roeder, & Rothchild (Eds.) Sustainable Peace: Power and Democracy after Civil Wars, (pp. 29–50). Ithaca, London: Cornell University Press.

Sanchez, G. R., & Morin, J. L. (2011). The Effect of Descriptive Representation on Latinos’ Views of Government and of Themselves: Descriptive Representation and Latino Views of Government and Selves. Social Science Quarterly, 92(2), 483–508.

Seawright, J., & Gerring, J. (2008). Case Selection Techniques in Case Study Research: A Menu of Qualitative and Quantitative Options. Political Research Quarterly, 61(2), 294–308.

Sisk, T. (1996). Power Sharing and International Mediation in Ethnic Conflict. Washington, DC: United States Institute of Peace Press.

Sisk, T. D., & Stefes, C. (2005). Power Sharing as an Interim Step in Peace Building: Lessons from Africa. In P. G. Roeder, & D. Rothchild (Eds.) Sustainable Peace: Power and Democracy after Civil Wars, (p. 32). Ithaca, New York: Cornell University Press.

Spears, I. S. (2000). Understanding Inclusive Peace Agreements in Africa: The Problems ofSharing Power. Third World Quarterly, 21(1), 105–118.

Stewart, F. (2008). Horizontal Inequalities and Conflict: Understanding Group Violence in Multiethnic Societies. London: Palgrave Macmillan UK Imprint: Palgrave Macmillan.

Stojanović, N. (2018). Political marginalization of “Others” in consociational regimes. Zeitschrift für Vergleichende Politikwissenschaft, 12, 341–364.

Straus, S. (2006). The Order of Genocide: Race, Power, and War in Rwanda. Ithaca, NewYork: Cornell University Press.

Tajfel, H., & Turner, J. (1979). An Integrative Theory of Intergroup Conflict. In W. Austin, & S. Worchel (Eds.) The Social Psychology of Intergroup Relations, (pp. 33–47). California: Brooks-Cole.

Treisman, D. S. (1997). Russia’s “Ethnic Revival”: The Separatist Activism of Regional Leaders in a Postcommunist Order. World Politics, 49(02), 212–249.

Turner, J. C. (1987). Rediscovering the Social Group: A Self-Categorization Theory. B. Blackwell.

Vogt, M., Bormann, N.-C., Ruegger, S., Cederman, L.-E., Hunziker, P., & Girardin, L. (2015). Integrating Data on Ethnicity, Geography, and Conflict: The Ethnic Power Relations Data Set Family. Journal of Conflict Resolution, 59(7), 1327–1342.

Wucherpfennig, J., Hunziker, P., & Cederman, L.-E. (2016). Who Inherits the State? Colonial Rule and Postcolonial Conflict. American Journal of Political Science, 60(4), 882–898.