Picture the fifth election after a ceasefire. You are watching the returns come in, and the numbers look like a census. The ethnic blocs the peace deal recognised, formalised, and handed ministries to are still there, each community voting along lines the agreement itself drew in ink. Somewhere else, in a different post-conflict country that signed something that looked nearly identical on the day, those blocs have lost their grip. Cross-cutting coalitions have formed. The original categories feel almost historical. The agreement looked similar in both cases. What happened in between is the actual story.
What determines whether power-sharing provisions entrench or gradually dissolve ethnic divisions comes down to one thing: whether the agreement treats ethnic identity as a fixed input to political life or as a temporary accommodation to a specific moment of fear. That distinction sounds philosophical. Its consequences are entirely concrete.
The architecture locks in the answer before anyone votes
Consider two real models. Lebanon's confessional system, negotiated in the 1943 National Pact and reinforced by the 1989 Taif Agreement, assigns the presidency to a Maronite Christian, the prime ministership to a Sunni Muslim, and the parliamentary speakership to a Shia Muslim. Every cabinet seat, every senior civil service post, every military command runs through the same grid. To participate in Lebanese politics at all, you must present yourself as a member of a sect. The system doesn't just reflect existing identities. It manufactures the political salience of those identities in every election cycle. Eighty years on, the categories are, if anything, more politically rigid than they were when the pact was first drawn.
Northern Ireland's Good Friday Agreement works differently, and the difference is instructive. The Assembly does require members to designate as Unionist, Nationalist, or Other, and key decisions require cross-community consent. But the designation system was always framed, by its architects at least, as a transitional safeguard against majoritarian abuse, not a permanent description of political reality. That framing matters less than the institutional detail that follows from it: the agreement contained explicit review mechanisms, sunset provisions for some security measures, and a cross-border dimension that created incentives for parties to build relationships across the old divide. The centripetal pressure was built into the structure from the start. Whether it has worked is contested. That it was designed differently from Lebanon is not.
The variable political scientists return to repeatedly is whether power-sharing arrangements are centripetal or consociational in their incentive structure. Consociationalism, associated with the Dutch political scientist Arend Lijphart, guarantees each community a proportional share of power and a mutual veto. It stabilises post-conflict societies by giving every group a reason to accept the deal. The cost is that it also gives every group a permanent institutional reason to remain a group. Centripetal designs try instead to reward politicians who appeal across ethnic lines, for instance by requiring presidential candidates to win a threshold of votes in multiple regions, as Nigeria's constitution has done since 1979. The incentive runs the other way: build a coalition, blur the boundary, or lose.
Neither model is clean in practice. Nigeria's cross-regional requirement coexists with intense ethnic mobilisation at the state level. Lebanon's confessionalism has produced genuine cross-sectarian business and civil society ties even while the formal political class remains sectarianised. Real agreements are hybrids, like a building with load-bearing walls on one side and open-plan on the other, and the question is always which set of incentives dominates the floor plan.
What people get wrong about the timeline
The common assumption is that time heals: give a power-sharing agreement two decades and identities will naturally soften as the original grievances fade. Bosnia-Herzegovina makes that assumption look naive. The Dayton Agreement ended a war. It did so by drawing internal borders that mapped almost perfectly onto ethnic cleansing lines, creating two entities, the Federation of Bosnia and Herzegovina and Republika Srpska, whose political elites have had every incentive since then to sustain the ethnic narrative that justifies their own power. Nearly thirty years later, the entities remain. Politicians who try to build genuinely multi-ethnic platforms routinely find that the institutional ground has been salted against them.
Time doesn't dissolve divisions. Incentives do, or don't.
Take two hypothetical politicians, call them Mirela and Dragan, both born after the war, both representing the same mid-sized city, one from each entity. Mirela's electoral base is constitutionally defined by her ethnicity. Every vote she wins comes from within that category. Dragan's is the same. Neither has a structural reason to appeal to the other's voters. They might personally be pragmatists, but the architecture makes pragmatism electorally irrelevant. That is not a failure of political will. It is the system working exactly as designed.
And here is the question worth sitting with: if the incentive structure rewards ethnic outbidding more reliably than it rewards compromise, why would any rational politician choose compromise?
What actually shifts the trajectory over time is whether economic integration, civil society, and informal cross-community institutions grow faster than the formal ethnic architecture can contain them. South Africa's post-apartheid settlement, imperfect as it is, allowed a non-racial constitutional identity to develop alongside the ANC's political dominance, partly because the constitution explicitly prohibited ethnic parties. The formal design nudged identity politics toward class politics, which has its own pathologies but a different shape. That nudge was not accidental, and it was not cheap: the ANC's architects made a deliberate, costly choice to foreclose the ethnic route to power.
The honest caveat is that no designer of a peace agreement controls the economic conditions that follow. A deal signed into a growing, integrating economy behaves differently from the same text signed into stagnation. When jobs are scarce and the state is the main distributor of resources, ethnic categories become economic categories, and no constitutional clause reliably breaks that link. Watch the unemployment figures in the decade after a settlement, and you will learn more about whether the agreement holds than you will from reading the text itself.
Power-sharing provisions entrench divisions when they make ethnic identity the only viable ticket to political participation and public resources. They soften when they treat ethnicity as one axis among several, build in review mechanisms, and reward leaders for crossing the lines the agreement drew. The signing ceremony is not the story. What matters is who gets paid to keep the lines visible, and whether that payment ever stops.