The Importance of Being Accurate in Agent-Based Models - An Illustration with Agent Aging

Sometimes, descriptive and empirical agent-based models contain sub-models of variables whose change over time cannot be easily modelled. Due to a frequent lack of long-term data, especially on demographic change, often logical design principles are used for modeling such changes. The problem with these modeling principles is that many of them exist and the choice is a difficult one. We would like to illustrate in this article that this choice has consequences, using the example of farm agent aging sub-models in an existing land-use change model. We will show that different design principles lead to different long-term changes in age pattern. The problem here is that one cannot be sure where the change in age structure comes from, i.e. whether it is an account of realworld change and/or whether it is caused by model assumptions. Then, we will show that simulated land-use change is highly sensitive to the type of aging sub-model. As a solution, we develop an alternative aging sub-model that exactly maintains an age pattern externally prescribed, but still allows individual aging.