This chapter addresses the challenge of endogeneity, testing Master Hypothesis 4. Some critics of the grievance-based literature argue that endogeneity may undermine the link between exclusion and conflict, and by implication the opposite one linking inclusion to peace. Endogeneity derives from the fact that the governments' decisions to include or exclude could be motivated by the anticipation of conflict. We counter this threat to inference by articulating a causal pathway that explains ethnic groups' access to power independently of conflict. Focusing on postcolonial states, we exploit differences in colonial empires' strategies of rule to model which ethnic groups were represented in government at the time of independence. This identification strategy allows for estimating the exogenous effect of inclusiveness on conflict. We find strong and systematic evidence that - at least for the post-colonial world - inclusion in governmental power sharing systematically reduces the likelihood that ethnic groups become involved in ethnic civil war. Our instrumental-variable analysis confirms Master Hypothesis 4, because we have found that governments tend to co-opt potential rebels rather than excluding them.

Many scholars have detected a decrease in political violence in recent decades, but the causes of this decline remain unclear. As a contribution to this debate, this chapter revisits the controversy over trends in conflict after the end of the Cold War. While several scholars made ominous predictions of surging ethnic warfare, Ted Robert Gurr presented evidence of a pacifying trend since the mid?1990s and predicted a further decline in ethnic conflict in an article on"ethnic warfare on the wane." Leveraging more recent data on ethnic groups and their participation in ethnic civil wars, this chapter evaluates if Gurr was right about the decline of ethnic conflict, and if he was right for the right reasons. We assess whether an increase in governments' accommodative policies toward ethnic groups can plausibly account for a decline in ethnic civil war. Our findings are largely compatible with Gurr's observations and stand in stark contrast to various pessimistic projections that were made in the early post-Cold War period. Among a number of empirical dimensions, we have found that this relatively optimistic perspective holds up well despite a surge in civil conflict in recent years. Ethnic, as opposed to non-ethnic, civil wars appear to have subsided after the mid-1990s, and this decline is at least partially attributable to an increase in governments' accommodative policies toward ethnic groups, such as the granting of group rights, regional autonomy, and inclusion through governmental power sharing, as well as democratization and peacekeeping.

In this chapter, we explore an actor-centric approach that emphasizes strategic interaction and offers an alternative to IV-estimation to pursue the evaluation of Master Hypothesis 4. We proceed in two steps. First, we introduce a simple formal model that captures the most important strategic dimension and highlights what is crucial for reverse causation: The conditions under which governments are likely to share power depending on the level of threat posed by a domestic challenger. As we have shown in Chapters 3 and 6, previous research on the origins of power sharing highlights either government incentives for the co-optation of threatening challengers through power sharing arrangement or risk-diversion by means of exclusion so as to avoid perilous infighting. Rather than favoring either consideration at the expense of the other, the model presented here unifies these mechanisms and shows that both have their place - depending on the challenger's level of threat as well as the government's ability to provide credible guarantees. In a second step, the chapter provides a novel statistical estimator that closely mirrors the strategic logic of the theoretical model, but also accounts for the selection on unobservables. Applied to data at the level of ethnic groups around the globe since World War II, the model offers strong evidence that governments do indeed strategically use power sharing as a way of managing the risk of conflict. In short, power sharing is systematically endogenous to conflict as suggested by Master Hypothesis 4. Moreover, the results reaffirm that power sharing systematically causes peace, and that naïve analyses that do not account for endogeneity tend to underestimate this effect.

This chapter sets the stage for the empirical analyses that are presented in the following chapters. We first introduce data on power-sharing practices which are drawn from the Ethnic Power Relations (EPR) dataset, which documents ethnic groups' access to power since 1946. This part of the chapter introduces the data structures and describes trends in power sharing globally and by world region. It is shown that both territorial and governmental power sharing have seen increased use for several decades, with a possible tapering off in more recent years. We present a series of simple cross tabulations and regression analyses to set the baseline for the chapters to come. In the second part of the chapter, we turn to data on formal institutions. This part of the chapter relies on the new Inclusion Dispersion and Constraints (IDC) dataset that measures formal power-sharing provisions along several institutional dimensions. The chapter closes with an overview of how these data are used to address our Master Hypotheses in the subsequent empirical chapters.

We start this conceptual and theoretical chapter by recapitulating previous work on exclusion and civil war by outlining how it connects political exclusion and grievances with the onset of conflict. Consequently, if power sharing reduces inequality and exclusion, peace becomes more likely through a grievance-reducing effect. In addition, power sharing can also bring peace through confidence building. After defining the key notion of power sharing practices, we introduce our first Master Hypothesis about linking both governmental and territorial power-sharing practices to the reduction of civil conflict. The rest of the chapter advances four additional master hypotheses that correspond to the four initial challenges confronting research on power sharing and conflict that were introduced in Chapter 1. First, we derive Master Hypothesis 2 by arguing that practices channel the main conflict-reducing effect of formal power-sharing institutions and also reduce conflict even in the absence of formal institutions. According to Master Hypothesis 3, power-sharing practices have a pacifying effect both before and after the first conflict but the risk of conflict onset is generally higher in the latter case. With Master Hypothesis 4, we address the important issue of endogeneity by showing that governments introduce power-sharing arrangements mostly as a way to co-opt potentially violent challenges to their sovereign power. Finally, under the heading of Master Hypothesis 5, we show how territorial power sharing, especially in relationships already characterized by past violence, may be insufficient to build confidence in support of stable peace. In such cases, autonomy and other territorial approaches are more effective if combined with central power-sharing practices.

In this concluding chapter of the book, we summarize what we have found in the previous chapters and discuss the theoretical significance and limitations of our findings. Furthermore, based on these reflections, we provide some recommendations for policy making. In the first section we return to our master hypotheses that we introduced in Chapter 3. Most importantly, we conclude that there is robust evidence that both governmental and territorial power-sharing practices tend, on average, to reduce conflict compared to situations characterized by their absence (see Master Hypotheses 1a and 1b respectively). We summarize the findings pertaining to the other Master Hypotheses as well. The discussion on the limitations of our findings focuses primarily on implications of power sharing on other outcomes than peace and conflict, such as democracy and economic development, but also some of the simplifying assumptions that have supported our analysis, such as the unity of ethnic groups. Finally, the discussion of policy implications reminds the reader that our results are probabilistic rather than deterministic. Still our analysis sheds light on why critics of power sharing may have overstated their case. For instance, failure to consider implementation difficulties could render power sharing ineffective or even counter-productive. In particular, such practices may be particularly effective before the first outbreak of violence, which confirms the importance of conflict reduction through preventive measures, rather than merely through conflict resolution once conflict has already erupted.

This chapter explores how formal power-sharing institutions relate to power-sharing practices and demonstrates the importance of the latter. Applying causal mediation analysis, we use the Ethnic Power Relations dataset to measure power-sharing practices and the Inclusion Dispersion and Constraints dataset to capture formal power-sharing institutions. The first part of this chapter evaluates whether and why our argument might hold. First, formal power-sharing institutions are not always formally implemented as our analysis clearly demonstrated. Second, in addition to not being implemented, formal power-sharing institutions often fail to result in practices that accommodate ethnic groups. Third, practices that accommodate ethnic groups often emerge even in the absence of formal institutional provisions. These three points highlight that the exclusive institutional focus typically present in existing studies of the effect of power sharing is likely to be misleading. In the second part, we assess more systematically how formal institutions affect the likelihood of conflict onset through practices or other channels. Throughout our analyses, a common theme emerges: If formal institutions affect conflict onset at all, this effect is mainly mediated through power-sharing practices. We find the strongest, mediated effects for formal governmental power-sharing institutions. In contrast, the effects for territorial power sharing are less clear-cut. For governmental power sharing we have also been able to show that the effect of practices on conflict onset depends less on formal power-sharing institutions than on other factors. This result underlines even more forcefully our argument that practices are playing a pivotal role in the link between power sharing and conflict.

Given the huge number of writings covering power sharing in general, and its relationship to internal conflict in particular, it is essential to summarize what has been written in order to position the current work in this context. Thus, the first half of the chapter offers a straightforward overview of the relevant literatures. While this summary pays attention to conceptual issues and how the literature has analyzed the consequences of power sharing in broad terms, the primary focus is on conflict. Proceeding chronologically, the chapter starts with an account of the central divide in comparative politics between those perspectives that endorse power sharing, and those that oppose it. We retrace this classical divide back to the early writings of Lijphart, who introduced his consociational approach to stable democracy in divided societies in opposition to majoritarian democracy as practiced in Anglo-Saxon countries. The review of the literature covers its main stages together with work by other scholars working within this tradition as well as its main critics, including most prominently Horowitz, Rothchild and Roeder, and more recently research by Gates, Strøm and others. We then turn to the literature on post-conflict agreements. A section summarizing the main arguments for and against power sharing follows. The chapter ends with a discussion of the four challenges identified in Chapter 1.

Focusing on regional autonomy arrangements, this chapter investigates, to what extent, and in what form, territorial power sharing mitigates civil conflict (see Master Hypothesis 1b). Our point of departure is again our past research indicating that exclusion of ethnic groups increases the risk of internal conflict. As argued in Chapter 3, however, such results do not automatically imply that regional inclusiveness will guarantee peace, especially if the relationship between an excluded group and the incumbent government has already seen violence. Based on a global sample of ethnic groups as provided by the Ethnic Power Relations dataset, here we show that, in such situations and on its own, regional autonomy is likely to be"too little, too late." It is too little because only full inclusion through governmental power sharing reduces conflict propensity significantly (see Master Hypothesis 5); and it is too late since regional autonomy could be effective, but only if offered in a timely, preventive fashion before group-government relations turn violent (see Master Hypothesis 3). Accounting for endogeneity, we also instrument for autonomy in postcolonial states by exploiting that French, as opposed to British, colonial rule rarely relied on decentralized governance. This identification strategy suggests that naïve analysis tends to underestimate the pacifying influence of decentralization.

This chapter opens by stating the main research question, namely whether power sharing reduces civil conflict or not. After briefly illustrating successful and unsuccessful cases of power sharing, we introduce our approach, which builds on a stream of work in conflict research stressing how exclusion of ethnic groups increased the risk of civil conflict. If this relationship is correct, one would expect the reverse to be true as well. That inclusion of ethnic groups through both territorial and governmental power sharing brings peace is indeed our working hypothesis. Still, there are good reasons to expect that this relationship may be more complicated. There are four main challenges that need to be addressed. First, it is essential not to lose sight of how power sharing is channeled through practices rather than merely being expressions of formal institutions. Second, analysts need to consider full samples rather than focusing only on cases that experienced conflict. Third, rather than being exogenously imposed, power sharing is usually enacted with an eye to future outcomes and is therefore profoundly endogenous to conflict. Finally, it is insufficient to analyze territorial power sharing without considering how pacific outcomes may hinge on whether this type of power sharing is being combined with governmental power sharing at the center. In fact, a failure to come to grips with these difficulties go a long way toward explaining why some researchers find no conflict-reducing effect, and sometimes even a conflict-increasing impact.

Taking a step back from the question whether power sharing reduces conflict, this chapter investigates the drivers of power sharing itself. While there is a growing consensus that ethnic inclusion produces peace, less is known about what causes transitions to power sharing between ethnic groups in central governments in multi-ethnic states. The few studies that have addressed this question have proposed explanations stressing exclusively domestic factors. Yet, power sharing is spatially clustered, which suggests that diffusion may be at play. Inspired by studies of democratic diffusion, we study the spread of inclusive policies with an "open polity model" that explicitly traces diffusion from inclusion in other states. Our findings indicate that the relevant diffusion processes operate primarily at the level of world regions rather than globally or between territorial neighbors. Thus, the more inclusive the region, the more likely a shift to power sharing becomes. Shifts away from inclusion to dominance are less common since World War II, but they are more likely in regional settings characterized by ethnic exclusion.

Focusing on regional autonomy arrangements, this chapter investigates, to what extent, and in what form, territorial power sharing mitigates civil conflict (see Master Hypothesis 1b). Our point of departure is again our past research indicating that exclusion of ethnic groups increases the risk of internal conflict. As argued in Chapter 3, however, such results do not automatically imply that regional inclusiveness will guarantee peace, especially if the relationship between an excluded group and the incumbent government has already seen violence. Based on a global sample of ethnic groups as provided by the Ethnic Power Relations dataset, here we show that, in such situations and on its own, regional autonomy is likely to be ``too little, too late.'' It is too little because only full inclusion through governmental power sharing reduces conflict propensity significantly (see Master Hypothesis 5); and it is too late since regional autonomy could be effective, but only if offered in a timely, preventive fashion before group-government relations turn violent (see Master Hypothesis 3). Accounting for endogeneity, we also instrument for autonomy in postcolonial states by exploiting that French, as opposed to British, colonial rule rarely relied on decentralized governance. This identification strategy suggests that naïve analysis tends to underestimate the pacifying influence of decentralization.