Multi-agent simulation for the targeting of development policies in less-favored areas

Complex combinations of biophysical and socio-economic constraints characterize the less-favored rural areas in developing countries. More so, these constraints are diverse as they vary considerably between households even in the same community. We propose multi-agent systems as a modeling approach well suited for capturing the complexity of constraints as well as the diversity in which they appear at the farm household level. Given that empirical multi-agent models based on mathematical programming share the characteristics of bio-economic farm models plus some additional features, one may interpret bio-economic farm models as a special case of multi-agent models without spatial dimension and direct interaction. Evidently, spatially explicit, connected multi-agent models have higher requirements in terms of development costs, empirical data and validation. Therefore, we see them as a complement, and not a substitute, to existing bio-economic modeling approaches. They might be the preferred model choice when heterogeneity and interactions of agents and environments are significant and, therefore, policy responses cannot be aggregated linearly. We illustrate the strength of empirical multi-agent models with simulation results from Uganda and Chile and indicate how they may assist policymakers in prioritizing and targeting alternative policy interventions especially in less-favored areas.