• Empirical validation
• Causality
• Agent-Based models
• DSGE
• Causal search
• Graphical models
• VAR-LiNGAM
• PCMCI algorithm
• PC algorithm

Any macroeconomic model attempting to explain real world phenomena must be empirically reliable, especially if the aim is to inform economic policy. Simulation models are often validated in terms of their ability to replicate a set of stylized facts. This approach is common not only for Agent-Based models, but also for the popular DSGE models, even when they are estimated by means of sophisticated techniques. However, the mere replication of stylized facts is a not a sufficiently severe test, since different causal mechanisms may generate the same empirical regularities. Guerini and Moneta (2017) address this need for a more rigorous criterion by comparing Structural Vector Autoregressive models identified by means of graphical-based causal search algorithms. Building on their work, the present study provides a comparative application of a method of empirical validation based on graphical models to a DSGE and an Agent-Based model, and adds a contribution in that it applies a causal search algorithm which allows for the presence of non-linearities in the data.