Learning coalition formation under an agglomeration bonus: Impacts on coalition structure and scheme performance

Numerous theoretical and empirical studies have demonstrated the potential of coordination incentives like the agglomeration bonus for the establishment of species habitat networks in agricultural landscapes. Less well understood is the social process behind this coordination, and how it affects the performance of the instrument. In the present paper this issue is addressed by simulating the coalition formation between several landowners in a stylized but structurally realistic landscape. Rather than assuming perfectly informed rational decision makers, the landowners are modeled as learning agents. A variety of learning strategies is considered. While these affect the coalition structure they have comparatively little influence on the land-use dynamics and the scheme expenditure, suggesting that knowledge about the details of the coordination process may be less relevant for predicting the performance of an agglomeration bonus. Instead, the performance is shown to mainly depend on the economic settings, such as the spatial correlation of the conservation costs, the spatial distribution of the landowners’ properties, and the presence or absence of side payments between the landowners – where the present results largely confirm the results of previous studies. A weak relationship is observed, though, between the average size of the coalitions on the one hand and the ecological scheme performance and scheme expenditure on the other. Confirming previous studies, budget-effectiveness gains of the agglomeration bonus are limited. • Landowners under an agglomeration bonus may form coalitions • Coalitions are formed through reinforcement learning • The agents’ learning behavior affects the emerging coalition structure • The scheme performance mainly depends on the economic settings