Network Calibration and Metamodeling of a Financial Accelerator Agent Based Model

Abstract

We allow firms and banks to entertain multiple credit connections in a financially constrained production framework, resorting to a random network model whose parameters are calibrated with real data. The calibration is successful since the network model is able to reproduce the degree and strength (debt and loan) distributions of the Japanese credit market. We run simulations over the parameter space using an efficient design, and compare a number of alternative statistical metamodels in order to select the best specification for the relationship between the parameters and a set of endogenous variables of the model. We show that the metamodeling approach can be usefully extended to economic models in order to bridge the gap between micro and macro variables through a rigorous statistical analysis of ABMs, without imposing unrealistic restrictions on the micro model such as the representative agent hypothesis.

Agent-Based Model Calibration Using Machine Learning Surrogates

“…With respect to surrogate modelling, we extend recent contributions in the economic literature that use kriging to build a surrogate meta-model for ABMs (Salle and Yildizoglu, 2014;Dosi et al, 2017c;Bargigli et al, 2016). One of the primary challenges with kriging-based meta-models is that they cannot efficiently model more than a dozen parameters.…”

“…In the binary outcome case, we employ a two samples Kolmogorov-Smirnov (KS) test between the distribution of logarithmic returns obtained from the numerical simulation 13 We underline that the dimension of the parameter space is in line or even larger that in recent studies on ABM meta-modelling (see e.g. Salle and Yildizoglu, 2014;Bargigli et al, 2016). of the model and the one obtained from real stock market data. 14 More specifically, we rely on daily adjusted closing prices for the S&P 500 going from December 09, 2013 to December 07, 2015, for a total of 502 observations, and we compute the following test statistic: 15…”

“…Having a large number of out-of-sample, unlabelled, possibly well-spread points is fundamental to evaluate the performance of the meta-model. We use a larger evaluation set than any other meta-modelling contribution we are aware of (see, for instance, Salle and Yildizoglu, 2014;Dosi et al, 2017c;Bargigli et al, 2016).…”

“…Adding a sixth parameter would require more than 10 years. , 2017cBargigli et al, 2016) to facilitate parameter space exploration and sensitivity analyses. However, Kriging cannot be reasonably applied to large scale models with more than 20 parameters even in the linear time extensions proposed in Wilson et al (2015) and Herlands et al (2015).…”

Macroeconomic Policy in DSGE and Agent-Based Models Redux: New Developments and Challenges Ahead

“…Recent developments try to mitigate over-parameterization issues resorting to phase-diagrams (Gualdi et al, 2015), Kriging meta-modeling (Salle and Yıldızoglu, 2014;Dosi et al, 2016c;Bargigli et al, 2016), and machine-learning surrogates (Lamperti et al, 2016b). We shall briefly come back to these issues in the concluding remarks.…”