Abstract for: Estimating System Dynamics Models Using Indirect Inference

System Dynamics (SD) research has not reached its potential to impact many traditional social science fields, partly because it is hard to estimate SD models using common dataset structures which include only a few data points over time, but often many units of analysis rather than a single case. Here we introduce the indirect inference, a simulation-based method which can be applied to such common data structures and is applicable to SD models which often include intractable likelihood functions. In this method, the parameters of the model are estimated in a way that simulated data and empirical data produce similar statistics. Those statistics include both common moments of the data but also parameters of auxiliary models one can estimate using traditional regressions. We also present a case study in the context of depression research where we apply the method, estimate unknown parameters and their confidence intervals, and assess the model’s fit. The overall results suggest that indirect inference can well extend the application of SD models to new application areas and datasets and provide unique insights.