Experiences from using behavioural economic methods to parameterize an ABM simulating the adoption of sustainable farming practices

Modelling farmers behaviour patterns can contribute to the assessment of policy, technology, or environmental changes in agriculture. In this context, agent-based models have become an important simulation tool in agricultural systems analysis. However, the availability of empirical data e.g., on farmer preferences and goal functions, represents a challenge when implementing a detailed representation of farmers’ individual decision-making in agent-based models. We here present and discuss an approach how to parameterize an agent-based model with census data, surveys, network analysis and choice experiments using the modelling framework FARMIND. This agent-based modelling framework provides a link between behavioural factors such as risk tolerance, social comparison and farming objective, and bio-economic models of agricultural production. We illustrate our approach with two case studies. In the first case study, we present the integration of an empirically measured social network into FARMIND. We analyse the extent to which knowledge exchange among socially connected farmers can contribute to farmers’ adoption of climate change mitigation measures in a Swiss region. In the second case study, we describe the implementation of information from a choice experiment on farmers’ willingness to pay for precision agriculture technologies into the same agent-based framework. The simulation analysis shows how perceived costs and benefits of the technology can influence the impact of different policy measures on the uptake of variable rate technologies in Switzerland. The examples show how tools and methods of behavioural environmental economics can be integrated into a simulation framework and how they can improve the accuracy of specific case studies of policy impact assessments. We discuss the approaches in the two examples with respect to general validity and robustness in agent-based modelling studies in agriculture. We specifically examine the role of purposeful model selection and sensitivity analyses as prerequisite for increasing model validity. However, equifinality, the generation of the same model output under different parameterizations, remains a key challenge in simulating farmers’ behavioural characteristics in FARMIND. Thus, the collection of representative data, not only for parameterization but International Congress on Environmental Modelling & Software iEMSs also for validation of model outcomes i.e., behavioural patterns, would be necessary to provide results that are generalizable beyond the case study context.