• agent-based models
• data-driven shock identification
• dynamic factor models
• time-series analysis
• validation

Recent developments in the VAR literature have demonstrated that it is possible to identify structural shocks by using only the distribution of reduced-form shocks and taking advantage of the information provided by its higher-order moments, making shock identification possible by relying solely on the assumptions of independence and non-Gaussianity of the structural shocks. However, the identification schemes proposed so far, which are rooted in independent component analysis, rely on additional assumptions to solve the indeterminacy of the permutation and scaling of the causal relations that this computational technique entails. After an overview of some popular identification strategies, this work introduces NGSI, a data-driven algorithm capable of performing shock identification without relying on such auxiliary assumptions. The key idea on which it is based is that it can be inferred from the data which assumptions are likely to hold and the most appropriate (and precise) identification scheme can be implemented accordingly. The performance of the algorithm is then assessed in a wide variety of settings via an extensive simulation study. Furthermore, this work proposes a new method to empirically validate simulation models that generate artificial time series data comparable with real-world data. The approach, which is based on the comparison of the causal structures estimated from the artificial and the real-world data, extends previous research by exploiting structural factor models, which, compared to standard SVARs, allow to consider a larger informative set, thereby leading to a more comprehensive validation assessment. This methodology is able to address both the problem of evaluating theoretical simulation models against the data and the problem of comparing different models in terms of their empirical reliability. Finally, an application of the validation procedure to the agent-based macroeconomic model proposed by Dosi et al. (2015) is provided.