Why Some Power-Sharing Governments Sustain Peace – and Others Don't

When societies fracture deeply along ethnic lines, the risk of civil war rises sharply. The remedy most commonly proposed is forming a multi-ethnic coalition government in which former enemies share power. By including representatives of all ethnic communities, these governments aim to address the grievances and fears that come with political exclusion. But power-sharing creates its own challenge.

Once in government, representatives of rival ethnic communities must actually govern together – a formidable task, especially when they were recently on opposite sides of a battlefield. For this reason, multi-ethnic coalitions are prone to paralysis and instability, and do not always prevent conflict from occurring or reoccurring.

My dissertation argues that understanding why power-sharing works better in some cases than in others requires looking beyond the formal rules of government. What matters is the broader dynamics of the ruling coalition: who shares power, how power-holders interact with one another, and how they relate to actors outside the coalition – opposition parties, civil society organisations, and ordinary voters.

Three logics of power-sharing

The dissertation distinguishes between three types of multi-ethnic coalitions, each with its own way of managing ethnic conflict – and its own distinctive risks.

1. Consociational coalitions – first described by Arendt Lijphart – are formed when each ethnic community independently selects its own representatives, who then come together in a "grand coalition" after elections. Ethnic conflict is managed by organising each community into its own separate pillar, leaving elites to negotiate solutions to disputed issues once in government. The risk is that such a heterogeneous coalition is prone to infighting, entrenches wartime divisions, and slides into immobilism.
2. Centripetal coalitions – associated with the work of Donald Horowitz – emerge when ethnic moderates ally at the centre before elections, with the explicit aim of excluding extremists from power. Because they are built on prior agreement, these coalitions tend to be more cohesive and thus less prone to internal disagreements. The danger, however, is that the excluded hardliners remain a constant threat from outside, and their pressure can ultimately lead to coups or a return to violence.
3. Associational coalitions – a new type introduced by the dissertation — emerge in cross-ethnic party systems, where most parties are themselves multi-ethnic even though ethnicity is politically salient. In such systems, ethnic communities are represented within each party rather than across separate parties.

Unlike consociational arrangements, ethnic divisions are not institutionalised but actively bridged by the parties themselves; disagreements between groups are worked out internally rather than played out in public for electoral gain. And unlike centripetal power-sharing, there are no excluded extremists waiting on the sidelines to destabilise the system; election results are therefore less likely to be disputed by the losing side.

Associational power-sharing does, however, face its own distinctive challenge. When political parties are all multi-ethnic, minority groups have no choice but to seek representation through parties in which they are also a numerical minority. The risk is that their representation becomes cosmetic – they hold seats but lack real influence. This tokenism can breed frustration among minorities and, over time, become a source of instability in its own right.

What the evidence shows

The dissertation maps the three types in a new dataset of multi-ethnic coalitions as they exist in practice. Starting from cases of de facto power-sharing identified in the Ethnic Power Relations dataset, it disaggregates these cases into the three types based on the structure of the party system and the timing of coalition formation.

The results are striking. In Africa, associational power-sharing is by far the most common form, accounting for 62% of cases (in country-years), compared to 30% for consociational and 25% for centripetal arrangements. This finding challenges a long-standing assumption in the academic literature that African party politics is inescapably organised along ethnic lines. Multi-ethnic coalitions on the continent are, in fact, frequently formed within cross-ethnic party systems.

While each type has been able to achieve long-lasting peace in some cases, statistical analysis reveals distinct probabilistic tendencies. Associational power-sharing is associated with longer periods of peace than the alternatives, and centripetal power-sharing proves most unstable. The threat of extremist outbidding may thus be the most destabilising force for multi-ethnic coalitions.

Associational power-sharing's tokenism problem appears less likely to tip a country into civil war. It nevertheless remains a serious democratic concern. A case study of Mauritius explores how minority groups can mitigate this problem in practice, leveraging some features of cross-ethnic party systems to achieve meaningful political recognition.

What this means in practice

The dissertation offers several takeaways for policymakers and practitioners working on post-conflict governance. Here are a few key ones:

• Designing good constitutions and electoral systems is necessary but not sufficient. Sustaining peace through power-sharing also requires daily political effort. Power-holders must continuously work to overcome their disagreements, make collective decisions, and resist the temptation to defect. Peace is hard work, not a one-time achievement.
• Consociational, centripetal, and associational power-sharing should not be seen as competing models where one universally trumps the others. They are better understood as different options for managing ethnic conflict, each suited to different contexts, and each demanding different things from those in government – whether that means bridging ethnic divides, containing hardliners, or ensuring meaningful minority representation.
• The dissertation adds associational power-sharing to the practical toolkit for post-conflict societies. Cross-ethnic party systems are possible in divided societies and can offer a path to durable peace, provided that the risk of cosmetic minority representation is taken seriously and actively guarded against.