• model validation
• data-driven SVAR
• causality

The rise to the top of macroeconomic agent-based models can be easily explained by their ability to endogenously generate business cycles fluctuations and financial crisis as emergent properties of a complex adaptive system populated by heterogenous bounded-rational agents. However, this fexibility in accounting for out-of-equilibrium dynamics and non-trivial processes makes the empirical validation of macro-ABMs (especially if policy oriented) a very challenging operation. Thus, the aim of this work is to tackle the issue of empirical validation in a structural manner, by applying the validation method proposed by Guerini and Moneta (2017) to the SFC-AB model by Caiani et al.,(2016). This is done by estimating a structural vector autoregressive (SVAR) model on both real-world and model-generated time series and identified by means of causal search algorithms and independent component analysis. The two causal structures so inferred are then compared to assess whether the macroeconomic model under scrutiny is good enough to perform counterfactual and policy analysis.