Pathways to Truth: Comparing Different Upscaling Options for an Agent-Based Sector Model

Simulation results can be highly sensitive to the way agents are upscaled to a larger organizational and spatial level. This paper tests an ex-post validation method for forecasting models by using old base years and forecasting into recent years for which observed data is already available. Our case in point is a comparison between different upscaling methods in the agent-based agricultural sector model SWISSland. It is shown that individual-farm extrapolation factors strongly enhance alignment with the total population in the base year. However, they may cause inconsistencies in those agent-based models in which relations between the farms are an important part. Therefore, an adjustment of the sample by making almost no use of some farms whilst making highly disproportionate use of others turned out to be the most suitable method for the SWISSland model.