Forecasting peace agreement content: How conflict events predict the substance of peace settlements

Abstract

Despite growing work on conflict forecasting, few studies predict conflict termination and none negotiation outcomes. We address this gap, assessing how well we can predict peace agreement content using conflict dynamics – particularly, insurgent distance from the capital. Thus, our study evaluates the predictive power of long-standing arguments in peace research, suggesting that conflict dynamics determine the prospects of negotiations. Utility theory posits that actors learn about their relative strength through conflict events and update their demands in negotiations accordingly. Ergo, actors' demands become more compatible with increasingly similar perceptions of relative capability. Scholars often proxy relative strength using insurgent distance from the capital, as it holds information on their ability to win battles. We evaluate whether battle locations predict agreement content using PA-X data. We find that a simple, theory-driven model performs almost as well as more complex, data-driven models. Nevertheless, models excluding conflict dynamics also have comparable predictive power.