ETD

• ABMs
• calibration
• DSGE
• estimation
• model confidence set
• nonlinear local projections
• simulation models
• validation

When economic interactions are conceived as pervaded by nonlinearities, evaluation procedures for simulation models should exploit these nonlinearities in order to accurately capture the underlying structural relationships among economic variables. I therefore propose a new calibration and validation protocol that explicitly incorporates nonlinearities. In line with the literature on calibration of macroeconomic simulation models, the procedure relies on minimizing the distance between the impulse response functions estimated using real-world data and those obtained using data simulated from the theoretical model. However, instead of estimating a single optimal configuration of parameters, it exploits the model confidence set to select a set of statistically equivalent configuration of parameters. The procedure is applied to an agent-based model that naturally features nonlinear dynamics. Calibration is performed under both linear and nonlinear frameworks, to assess how allowing for nonlinearities affects the results. The findings show that accounting for nonlinearities leads to a different set of admissible configuration of parameters compared to the linear case. Furthermore, according to the validation measure, incorporating nonlinearities improves the model’s performance from 76.1% to 92.9%. Overall, the results highlight the importance of explicitly modeling nonlinearities in calibration and validation procedures, particularly for models designed to capture complex economic dynamics.